University of Southern California Dissertations and Theses

An investigation of factors associated with reliability and validity of family history reports of alcohol‐related problems

(USC Thesis Other)

An investigation of factors associated with reliability and validity of family history reports of alcohol‐related problems

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content RUNNING HEAD: Reliability and Validity of Family Reports of Alcoholism 1

An Investigation of Factors Associated with Reliability and Validity of Family History Reports
of Alcohol-Related Problems

Elizabeth Chereji, MA
Department of Psychology, University of Southern California, Los Angeles CA

Dissertation Committee:
Carol A. Prescott, Ph.D., Chair
Margaret Gatz, Ph.D.
David Lavond, Ph.D.
Suzanne Wenzel, Ph.D
Reliability and Validity of Family Reports of Alcoholism 2

TABLE OF CONTENTS
Abstract…………………………………………………………………………………………... 4
Background………………………………………………………………………………………. 5
Study Aims………………………………………………………………………………………12
Methods…………………………………………………………………………………………. 13
Participants………………………………………………………………………………... 13
Measures…………………………………………………………………………………... 16
Analysis Plan……………………………………………………………………………… 19
Results…………………………………………………………………………………………... 24
Discussion………………………………………………………………………………………. 35
References………………………………………………………………………………………. 47
Acknowledgements……………………………………………………………………………... 54
Tables
Table 1: Study aims and analysis approach…………………………………………………….. 55
Table 2 Study participants, measures, and variables……………………………………………. 57
Table 3: Twin offspring samples sizes, by gender and zygosity………………………………... 58
Table 4: Lifetime prevalences of alcohol-related problems in adult male and female
offspring………………………………………………………………………………… 59
Table 5: Lifetime prevalences of alcohol-related problems in parents…………………………. 60
Table 6: Within-pair twin agreement on parents’ FH-RDC Alcoholism symptom count
and diagnosis, by twin gender and zygosity……………………………………………. 61
Table 7: FH-RDC Alcoholism diagnosis prevalences and symptom count means, by
gender of offspring informants and parent targets …………………………………….. 62
Table 8: Odds of FH-RDC Alcoholism diagnosis in parent targets based on family
history reports by affected and unaffected offspring informants……………………….. 63
Table 9: Within-pair agreement on parent FH-RDC Alcoholism symptom count and
diagnosis, by twin pair alcohol problem concordance and discordance……………….. 64
Table 10: Validity of parent FH-RDC Alcoholism based on offspring informant
reports compared to personal interview of parent target, by target gender…………….. 65
Table 11: Validity of sibling FH-RDC Alcoholism based on offspring informant
reports compared to personal interview with sibling targets, by target and
informant gender………………………………………………………………………... 66
Table 12: Validity of parent FH-RDC Alcoholism based on offspring informant
reports compared to personal interview of parent target, by informant
alcohol-related problem status…………………………………………………………. 67
Reliability and Validity of Family Reports of Alcoholism 3

Table 13: Validity of sibling FH-RDC Alcoholism based on offspring informant
reports compared to personal interview with sibling targets, by informant
alcohol-related problem status………………………………………………………….. 68
Table 14: Validity of parent FH-RDC Alcoholism based on single vs multiple
offspring informant reports compared to parent personal interview……………………. 69
Table 15: Validity of parent FH-RDC Alcoholism reported by offspring
informants based on different methods of consolidating multiple
informants responses at different diagnostic thresholds………………………………… 70
Table 16: Differences in validity of family history reports by informant-target
family relationship………………………………………………………………………. 71

Figures
Figure 1: Female-female study design………………………………………………………….. 72
Figure 2: Male-male/Male-female study design………………………………………………... 73
Figure 3: Validity of parent FH-RDC Alcoholism based on single vs multiple
offspring informant reports compared to parent personal interview, by parent gender………… 74
Figure 4: Multiple informants with varying methods of consolidating responses at
different FH-RDC diagnostic thresholds—Mothers……………………………………. 75
Figure 5: Multiple informants with varying methods of consolidating responses at
different FH-RDC diagnostic thresholds—Fathers……………………………………... 76

Appendices
Appendix A: Literature Review………………………………………………………………… 77
Appendix B: Family History Research Diagnostic Criteria—Alcoholism……………………... 83
Appendix C: Validity Indices: Definitions and Equations………………………………………84

Reliability and Validity of Family Reports of Alcoholism 4

Abstract
The family history method, which collects data from family members who serve as
informants of their relatives’ psychopathology, has been utilized as a cost-effective method of
obtaining family history data. It is widely acknowledged that there are limitations related to
collecting data via family history method; however, existing studies have yielded mixed findings
pertaining to factors that may impact reliability and validity of family history reports. Using data
from a large population-based sample of male and female twin offspring (N=7,417) and parents
(N=1,472 ) who participated in the Virginia Adult Twin Study of Psychiatric and Substance Use
Disorders (VATSPSUD), the present study evaluated a variety of factors that may influence
inter-informant agreement (i.e., reliability) and validity of family history reports. Reliability
findings indicate that offspring tended to highly agree on their reports of parent alcohol-related
problems, regardless of various informant and target characteristics. Validity differences were
found based on informant and target gender, informant diagnostic status, and family relationship
type, while use of multiple family informants versus a single informant did not substantially
improve validity. Methodological considerations are critical, as investigating differences within
(rather than between) families provides a more robust test of factors that may influence family
history reporting. These findings help inform our understanding of optimal methods for
collecting family history data, which can ultimately help researchers and treatment providers
better understand individual risk for alcohol-related problems.

Reliability and Validity of Family Reports of Alcoholism 5

A family history of mental illness is a risk factor for psychiatric disorders in biological
relatives (e.g., Brent & Mann, 2005; Hettema, Neale, & Kendler, 2001; Kendler, Davis, &
Kessler, 1997; Klein, Lewinsohn, Rohde, Seeley, & Shankman, 2003; Sullivan, Neale, &
Kendler, 2000). Because family history is a risk factor, it is of empirical and clinical relevance
that methods for gathering information on the presence of mental illness in family members are
well-established. Data on the psychiatric history of family members have been collected via
family study method and family history method. In the family study method, family members of
interest are interviewed directly. In contrast, the family history method collects data from
relatives who serve as informants on their family members’ (i.e., targets’) psychopathology.
While directly interviewing targets of interest may be preferable, it can be challenging or
impossible due to refusal to participate, death, inaccessibility, and cost. For these reasons, the
family history method has been deemed an efficient and effective way to collect data on multiple
family members from designated informants.
The family history method has been utilized and studied extensively with regards to
alcohol-related problems. Existing literature has demonstrated fair to strong levels of agreement
(inter-rater reliability) among family members serving as informants for the same relatives
(Amodeo & Griffin, 2009; Crews & Sher, 1992; Heun & Müller, 1998; Kendler et al., 1991;
Prescott et al., 2005; Rhea, Nagoshi, & Wilson, 1993; Rice et al., 1995; Roy, Walsh, Prescott, &
Kendler, 1994; Sher & Descutner, 1986; Slutske et al., 1996; Waldron et al., 2012). Studies of
consistency between informant reports and direct interviews of target relatives (often referred to
as “validity” or “accuracy”) have generally found that family informants are very good at
accurately identifying absence of drinking problems in their relatives (i.e., specificity;
Andreasen, Rice, Endicott, Reich, & Coryell, 1986; Kendler, Prescott, Jacobson, Myers, &
Reliability and Validity of Family Reports of Alcoholism 6

Neale, 2002; Kosten, Anton, & Rounsaville, 1992; Milne et al., 2009; Roy et al., 1994;
Thompson, Orvaschel, Prusoff, & Kidd, 1982) but are less able to positively identify drinking
problems in relatives when they do exist (i.e., sensitivity; Crews & Sher, 1992; Milne et al.,
2009; O’Malley, Carey, & Maisto, 1986; Smith, Przybeck, Bradford, Gogineni, & Spitznagel,
1994), which can result in underdiagnosis based on family history report alone.
Sources of variation in informant agreement and validity have been investigated based on
informant and target characteristics, including gender of the informants and targets as well as
informant’s own drinking problems. Potential effects of quantity of informants (single versus
multiple) and type of informant-target relationship (e.g., offspring-parent; sibling-sibling) have
also been studied. Findings related to many of these factors have been mixed, which hinders our
ability to understand factors that are relevant to the collection and interpretation of family history
data. To better understand discrepancies among existing studies, family history studies of
alcohol-related problems that investigate these characteristics are reviewed and discussed below.
Appendix A presents details of each study discussed in this literature review.
Impact of Informant and Target Gender on Family History Reporting
Studies investigating informant agreement based on gender of informants and/or targets
have yielded mixed findings. One study of 142 pairs of twin and nontwin adult siblings from the
United States found that brother (male-male) pairs demonstrated higher agreement about
maternal drinking problems than sister (female-female) pairs, while sister pairs demonstrated
higher agreement in ratings of fathers’ drinking problems than brother pairs (Rhea et al., 1993).
These results indicate higher agreement in same-sex siblings when reporting on their opposite
sex parent. In contrast, a study of 2,657 young adult twin pairs from Australia found no
Reliability and Validity of Family Reports of Alcoholism 7

substantial differences in comparing brother pairs’ agreement to sister pairs’ agreement. Instead,
differences in agreement were only found in female twin pair (but not male twin pair) agreement
based on parent’s gender, with higher agreement among female twin reports of maternal
compared to paternal alcohol problems. Also, like-sexed female pair’s maternal agreement was
higher than opposite sex pair maternal agreement (Slutske et al., 1996). Opposite sex pair
differences were also found in a study of 290 women from the Boston area and their male and
female siblings, with higher agreement found among opposite sex sibling pairs on ratings of
paternal alcohol-related problems compared to sister pairs (Amodeo & Griffin, 2009). In contrast
to the aforementioned studies, other studies have found no substantial gender-based differences
in agreement (Crews & Sher, 1992; Prescott et al., 2005; Searles, Alterman, & Miller, 1993).
While it is unclear why there is such variability in findings, possible explanations may be
differences in the alcohol measures used, considerable variability in populations the subjects
were drawn from [ranging from college freshman (Crews & Sher, 1992) to community dwelling
siblings (Amodeo & Griffin, 2009)], and variability in sample sizes which could impact
statistical power.
Studies of gender differences in validity (i.e., consistency of informant reports with direct
interview of target, also called ‘accuracy’) have also been inconsistent, with some studies finding
higher, lower, or no differences in accuracy of female informant reports compared to male
informant reports. In a study of parents reporting on their offspring’s psychopathology (N=959
offspring), females (i.e., mothers) did not differ from males (i.e., fathers) in the accuracy of their
reports (Milne et al., 2009). While lack of gender differences in validity has been supported by
other studies (Weissman et al., 2000), one study of family alcohol-related problems found higher
validity of female informants’ reports, including higher levels of sensitivity (Kosten et al., 1992).
Reliability and Validity of Family Reports of Alcoholism 8

Another study of various first-degree relatives serving as informants found gender-based biases
in both male and female informants’ family history reports of alcohol-related problems, with
higher rates of false-negatives among males and higher rates of false-positives among females
(Roy et al., 1994). These findings suggest that females may have a lower threshold for
identifying drinking problems and males may be less forthcoming about or perceptive to
drinking problems in relatives. While findings have been mixed and the reasons for possible
gender differences remain unclear, it is evident that gender biases are important to consider when
assessing the validity of family history reports. In addition to gender of the informants, gender of
the targets also requires additional exploration. While one study found increased false-positives
among informants when the family target was male (Kosten et al., 1992), the influence of
target’s gender on validity has not been as widely investigated in other studies.
Informant Alcohol-Related Problems and Family History Reporting
Differences in inter-rater agreement based on each informant’s own drinking status has
been a topic of interest in family history studies. A study of 2,163 female twins found no
significant differences in agreement on parental alcohol-related problems reported by pairs
discordant for lifetime history of such problems (Kendler et al., 1991). Similarly, other studies of
like-sexed male and female twin pairs (Slutske et al., 1996) and non-twin male and female
siblings (Amodeo & Griffin, 2009) did not find differences in agreement based on informants’
drinking problems.
Findings on the influence of informant diagnosis on validity have been more varied, with
studies concluding that there are higher false-positive rates among affected informants (Roy et
al., 1994), underreporting by affected informants (Smith et al., 1994), and similar accuracy
Reliability and Validity of Family Reports of Alcoholism 9

between unaffected and affected informants (Milne et al., 2009). Other studies suggest that
alcoholic informants are more accurate than family members without such problems. In a study
of 2,654 subjects from 572 families participating in The Collaborative Study on the Genetics of
Alcoholism (COGA), informants with drinking problems were better at accurately identifying
such problems in relatives than informants without drinking problems, suggesting that affected
family members may be more sensitive to alcohol-related problems in others (Rice et al., 1995).
Similar results were found in a family history study of over 300 families from New York, but
affected informants also tended to identify drinking problems in family members who did not
have them (Chapman, Mannuzza, Klein, & Fyer, 1994). This finding suggesting that affected
informants may not necessarily be “better” informants but may actually be biased, resulting in
overestimation of alcohol problems in others. Another possibility may be that targets are
underreporting problems when they are interviewed; however, honest/accurate reporting by
targets cannot be determined due to lack of objective corroborating evidence in these studies.
Family Relationship: Degree of relatedness and type of informant-target relationship
Investigating the impact of zygosity in twin studies of family history reporting is
important, as it provides a way to study how differences in degree of relatedness may influence
reporting. Twin data provides a unique opportunity to investigate whether genetic and
environmental factors are associated with differences in reporting, however very few of the twin
studies on this subject analyzed the role of relatedness. Rhea et al.’s (1993) comparison of
monozygotic (MZ), dizygotic (DZ), and nontwin sibling agreement found some differences in
level of agreement among these groups, but the pattern of findings was deemed inconsistent and
not indicative of genetically-based differences in reporting. The authors concluded that these
findings yielded no meaningful differences based on relatedness, however a lack of statistical
Reliability and Validity of Family Reports of Alcoholism 10

power attributable to small group sizes likely hindered the ability to detect meaningful group
differences in agreement (number of pairs: MZ=47; DZ=41; nontwin=54). Other twin studies of
relatives’ agreement did not investigate the relevance of zygosity (Kendler et al., 1991; Slutske et
al., 1996), with the exception of one study that found higher levels of agreement among
monozygotic twins compared to dizygotic twins in participants of European ancestry but not in
African-Americans (Waldron et al., 2012). Further investigation of zygosity differences may
help elucidate possible genetic and environmental influences on family history reporting.
The type of family relationship between informant and target (e.g., parent-offspring,
siblings) is another important factor that has been investigated in the context of family history
reporting. Three studies have tested family history reporting differences based on informant-
target family relationship; these studies primarily focused on differences in validity. One study of
differences in family history reporting based on family relationship found that offspring
informants had significantly higher sensitivity (60%) than sibling (30%), parent (28%), or spouse
(21%) informants (Kosten et al., 1992). Another study also found sensitivity levels for offspring
(57%), sibling (30%), and parent informants (24%) that were similar to Kosten et al.’s findings,
though sensitivity was found to be higher for spouses (45%; Thompson et al., 1982). In both of
these studies, specificity was found to be high across informant-target relationships (range: 94%-
99%). While offspring informants have demonstrated higher sensitivity levels in these studies,
another study found that offspring informants tend to over-report symptomology compared to
parent informants (Prescott et al., 2005). Additional research is required to better understand
potential trade-offs in accuracy and precision of family history reports based on family
relationship.

Reliability and Validity of Family Reports of Alcoholism 11

Multiple Informants
In addition to characteristics of informants, the potential increase in accuracy of family
history reports with multiple informants has also been a subject of interested to family history
researchers. Findings are varied between studies and also depend on the method employed to
combine multiple informant ratings. Use of an “Or” rule, where informant reports were counted
as positive if at least one informant reported drinking problems in the target, has been shown to
result in higher sensitivity than single informant reports (Kosten et al., 1992; Smith et al., 1994;
Thompson et al., 1982; Weissman et al., 2000). In contrast, the more stringent “And” rule, where
all informants have to concur on their target reports in order to identify a target as affected, has
been shown to decrease sensitivity (Smith et al., 1994). One study has comprehensively tested
various methods of combining multiple informant family history reports and found that there
were not substantial improvements across validity indices with additional informants (Prescott et
al., 2005). Additional research pertaining to multiple informants would help determine whether
investment of resources into collecting data from more than one informant is a cost-effective or
necessary strategy to obtain the most accurate family history data possible.
Alcohol-Related Problem Definitions
There has been considerable variability in how alcohol-related problems have been
defined and assessed in past family history studies, ranging from a single item to comprehensive
diagnostic assessments of alcohol-related problems. Some studies have found that a single global
item can yield agreement values comparable to administration of multiple items (Crews & Sher,
1992; Prescott et al., 2005; Sher & Descutner, 1986; Slutske et al., 1996), while others found that
the content of different questions may lead to variability in family history reports (i.e., higher
agreement for observable behaviors related to alcohol than inferred emotional/psychological
Reliability and Validity of Family Reports of Alcoholism 12

effects of alcohol; Rhea et al., 1993; Sher & Descutner, 1986). It remains unclear whether the
variability in diagnostic definitions used across family history studies contributes to the
inconsistencies in findings highlighted by this literature review. Testing a variety of alcohol-
related problem definitions and comparing findings across these definitions may provide more
insight into how informant agreement and validity vary as a function of designated criteria.
Study Aims
The present study helps address the inconsistencies in reliability and validity findings in
existing family history studies of alcohol-related problems, including the degree to which
informant and target characteristics, multiple informants, family relationship, and diagnostic
definitions are associated with differences in reliability and validity. Using a large population-
based sample of male and female adult twins who provided family history reports of parental and
cotwin alcohol-related problems, the present study aims to investigate the following research
questions (also presented in Table 1):
1) With regard to reliability, are informant and target characteristics (specifically
informant and target gender, informant zygosity, and informant diagnostic status)
associated with the strength of agreement (i.e., inter-rater reliability) of family history
reports of alcohol-related problems?
2)With regard to validity, are informant and target characteristics (specifically informant
and target gender and informant diagnostic status) associated with the validity of family
history reports of alcohol-related problems?
Reliability and Validity of Family Reports of Alcoholism 13

3) Do the quantity of informants (i.e., multiple versus single) and type of family
relationship between informants and targets (i.e., parent, offspring, sibling) influence
validity of family history reports of alcohol-related problems?
4) Do different definitions of alcohol-related problems result in differences in reliability
and validity findings?
METHODS
Participants
The participants in this study are from the Virginia Adult Twin Study of Psychiatric and
Substance Use Disorders (VATSPSUD), a longitudinal study of psychopathology in male and
female Caucasian twins born between 1934 and 1974. The twins were ascertained from the
Virginia Twin Registry (now part of the Mid-Atlantic Twin Registry). The VATSPSUD consists
of two parallel studies, the first of which collected data from female-female twin pairs (FF study)
and their parents (Female-Female Parent, or FFP, study), and the second collected data from
male-male and male-female twin pairs (MM/MF study). The data in the present study come from
interviews conducted as part of the FF, FFP, and MM/MF studies. Figures 1 and 2 present
diagrams of the FF and MM/MF study designs. To date, four waves of the FF interviews and two
waves of the MM/MF interviews have been completed, plus a third wave in male-male pairs.
The sources of data are designated by the study acronym followed by the wave number (e.g.,
MM/MF1, MM/MF2, FF1, FF2, etc.). An additional instance of data collection, referred to as
FF1+, consisted of female twins eligible for FF1 who: initially refused to participate, could not
be located, or became eligible (by turning 18) after the time frame of the FF1 data collection. As
indicated in Figure 1, FF1+ females did not participate in FF2 or FF3, but were eligible to
participate in FF4.
Reliability and Validity of Family Reports of Alcoholism 14

Description of relevant study waves
Table 2 summarizes study waves, measures, and variables used in the analyses. The
present study utilized data from waves one and four of the FF study (i.e., FF1 and FF4, as well as
FF1+), wave two of the MM/MF study (i.e., MM/MF2), and interviews with parents of FF twins
(i.e., FFP). FF1 data were collected from 1987 to 1989 through in-person interviews. Twins were
recruited to participate in FF1 if they were born between 1934 and 1970, were age 18 or over,
and had previously responded to a mailed questionnaire from the Virginia Twin Registry. Of the
2,352 eligible women, 2,164 (92%) were interviewed for FF1, including 1,033 complete pairs.
FF1+ interviews contributed an additional 266 women to the sample. Participant’s ages from
FF1 and FF1+ ranged from 18 to 54, with a mean age of 29.3 years (SD=7.7). FF4 data were
collected from 1995 to 1997 via telephone interviews. Female twins were eligible to participate
in FF4 if they had participated in FF1 or FF1+, were alive, and had not refused further contact
during a prior interview. Of 2,295 women eligible for FF4, 1,940 (84.5%) individuals were
interviewed.
MM/MF2 data were collected from 1994 to 1998 primarily through in-person interviews
with some exceptions: 20% of MM/MF2 were conducted by telephone due to various factors,
including participants’ relocation out of state or preference for a telephone interview. Subjects
were eligible for MM/MF2 if they were alive and participated in MM/MF1 (age range at
MM/MF1—19 to 57, Mean age=35.1, SD=9.1).
1
A total of 5,621 (82.5%) completed interviews
for MM/MF2 were obtained from eligible subjects.

1
MM/MF1 data were collected from 1993 to 1996 by telephone. Twins were eligible for MM/MF1 if they were
born between 1940 and 1974, were members of a multiple birth that included at least one male, and one member of a
pair was successfully matched to state records to obtain contact information.
Reliability and Validity of Family Reports of Alcoholism 15

Female-Female Parent (FFP) data were collected between 1990 and 1991 from living
parents of complete pairs from the FF1 sample using in-person interviews. Of the 1,698 living
parents identified, 1,472 (86.7%) were interviewed (855 mothers and 617 fathers). There were
567 families for which both parents were interviewed.
Alcohol variables used in study
The present study utilized the following data: 1) family history data obtained from twins
on their parents’ alcohol-related problems (collected during FF1, FF1+, and MM/MF2), as well
as family history data from twins reporting on their cotwins (MM/MF2) and parents reporting on
their twin offspring’s alcohol-related problems (collected during FFP), 2) twins’ own alcohol-
related problems as assessed by structured clinical interviews conducted during FF4 and
MM/MF2, and 3) parents’ own alcohol problems as assessed by structured clinical interviews
conducted during FFP.
Data on twins’ alcohol-related problems from FF4 and MM/MF2 waves were selected for
analyses based on comparability of alcohol measures and being the most recent assessment of
this information. To ensure consistency in availability and diagnostic criteria of offspring
drinking problem data (to be discussed in more detail later), female offspring were included in
the present study only if they participated in FF4 (i.e., subset of FF1, FF1+, and FFP data utilized
was exclusive to females who participated in FF4).
Table 3 presents sample sizes for complete and incomplete pairs of twins used in the
study analyses. For the present study, twins were excluded from the sample if they were missing
family history data (N= 96) or if the data were deemed poor quality (N=38; to be discussed in
more detail in the “Measures” section).

Reliability and Validity of Family Reports of Alcoholism 16

Measures
All data used in this study were collected via interviews conducted and scored by
clinically-trained interviewers. Interviewers received comprehensive and ongoing training and
had a minimum of a master’s degree in psychology or social work, or a bachelor’s degree
followed by at least two years of clinical experience. All interviews were reviewed for accuracy
and consistency by the project coordinator or project investigators.
Informant-Reported Family History of Alcohol-Related Problems
Data on family history of alcohol-related problems were obtained from twins and parents
using the Family History Research Diagnostic Criteria (FH-RDC; Endicott, Andreasen, &
Spitzer, 1978). The FH-RDC provides a stem question asking informants whether the target of
interest ever displayed alcohol-related problems that lasted at least a month (see Appendix B).
Informants who endorsed this question were asked a series of six follow-up questions related to
the target’s drinking, with answer choices of “yes”, “no”, or “don’t know.” As per FH-RDC
diagnostic criteria, informant responses were classified as consistent with an “Alcoholism”
diagnosis (dichotomous coding of 0/1) if the stem question and at least one follow-up question
were endorsed (will be referred to as “FH-RDC Alcoholism diagnosis”). In addition to use of a
categorical diagnostic variable based on the aforementioned diagnostic criteria, a symptom count
variable was used in the analyses (maximum of 6 symptoms; will be referred to as “FH-RDC
Alcoholism symptoms”). A broader definition of FH-RDC alcohol-related problems was also
tested in one of the study analyses: for this definition, parent alcohol-related problems were
based on informant endorsement of the FH-RDC Alcoholism stem item which inquired about
lifetime drinking problems (“Did your mother/father ever have a problem with drinking that
lasted at least a month? Yes/No,” coded as 0/1).
Reliability and Validity of Family Reports of Alcoholism 17

Offspring and Parent’s Alcohol-Related Problems
Offspring and parents’ own alcohol-related problems were assessed using in-person
structured clinical interviews. Parent’s alcohol-related problems were assessed during FFP with
an interview composed of items adapted from the Structured Clinical Interview for DSM-III-R
Axis I Disorders (SCID-I; Spitzer, Williams, & Gibbon, 1987). DSM III-R and DSM-IV criteria
were used to diagnose lifetime presence of the following disorders based on parent personal
interview data: DSM III-R Alcohol Dependence (AD), DSM-IV Alcohol Dependence, and a
variable representing the presence of either Alcohol Abuse or Alcohol Dependence based on
DSM III-R diagnostic criteria (AA/AD). Dichotomous scores (0/1) were assigned to each
diagnosis based on the subject’s diagnostic status. Parent diagnoses based on DSM-IV AA/AD
criteria were not utilized because FFP alcohol interview questions did not include items
pertaining to DSM IV Alcohol Abuse.
Offspring were assessed for alcohol-related problems during FF4 and MM/MF2 using
items based on a version of the SCID-R that was adapted to assess diagnostic criteria from both
the Third-Revised and Fourth editions of the Diagnostic and Statistical Manual of Mental
Disorders (DSM III-R, DSM-IV; American Psychiatric Association, 1987, 1994). Lifetime
presence of DSM-IV AD and DSM-IV AA/AD were diagnosed based on personal interview data
and assigned a dichotomous score (0/1) based on absence/presence. Test-retest reliability
estimates for FF4 and MF2 for 382 individuals re-interviewed after an average interval of 30
days were κ=0.72 (95% CI=0.61-0.82) for AD and κ=0.74 (95% CI=0.66-0.82) for AA/AD
(Kendler & Prescott, 2006).
“Problem Drinking” was assessed in both offspring and parents. Problem Drinking was
determined based on responses to items inquiring whether there had ever been a time when the
Reliability and Validity of Family Reports of Alcoholism 18

participant drank “too much” or “someone else objected” to their drinking. Endorsement of
either item by the subject resulted in coding “Problem Drinking” as 1, whereas denial of both
items resulted coding it as “0”
Defining Onset Ages for Alcohol-Related Problems
To improve data quality and accuracy, specific adjustments were made to the data to
ensure temporal consistency of when informants provided family history reports and when
personal interviews were completed by targets. Because the potential effects of informant’s own
diagnosis on family history reporting is a question of interest in this study, informant age at onset
was utilized to confirm that informants were affected before or at the time that family history
reports were provided. For offspring informants reporting on parent targets’ alcohol-related
problems, informant age at onset reported at FF4 was used to confirm that the offspring
informant was affected before or by the time family history reports were provided at FF1.
Informants with an age at onset that occurred after FF1 were coded as not having the disorder.
Similar age at onset modifications were made for parent informants reporting on offspring
targets to ensure that offspring target’s own diagnosis reported at FF4 was positive before or by
the time the family history data were collected from parents at FFP.
Age at onset adjustments were not deemed necessary for validity analyses, as FFP was
conducted in close temporal proximity to FF1 and it is unlikely that there were new onsets in
parents during the time lapse between these studies. There were no age at onset adjustments
made for twin informant reports on sibling (i.e., cotwin) targets because both sibling family
history data and sibling personal interview data were collected during the same wave of data
collection (i.e., during MM/MF2).

Reliability and Validity of Family Reports of Alcoholism 19

Zygosity
Informant zygosity was classified as identical (i.e., monozygotic, MZ) or fraternal (i.e.,
dizygotic, DZ) based on a combination of questionnaire responses, photographs, and DNA
testing. Early in the study of FF pairs, zygosity was determined through blind review of
photographs along with questions related to physical characteristics and how frequently the twins
were mistaken for each other during childhood. Genotyping of DNA from blood samples was
used in cases where it was difficult to determine zygosity. During FF4 and MM/MF2, DNA was
collected from all willing participants and was used to validate the original zygosity
determination, which yielded a correct classification rate of 95%. Zygosity determination of MM
was originally based on the algorithm used for FF pairs and was later refined by developing
discriminant function analyses based on genotyping information from 227 MM pairs.
Measurement of data quality
A family history data quality question was present only in MM/MF and required that the
interviewer select a data quality rating based on how the twin obtained information related to
parents’ alcoholism (i.e., acquired from twin’s own experience with parent or hearing it from
someone else). There were four data quality ratings, ranging from “high quality” (1) to “poor
quality” (4). Data identified as “poor quality” were coded as missing to increase data accuracy.
Analysis Plan
Table 1 presents a summary of study aims and analyses associated with these aims (the
Analysis Plan will refer back to aims as they are labeled in this table). The first two aims of this
study are intended to comprehensively assess differences in reliability (Aim 1) and validity
(Aim 2) of family history reports based on potential effects of target and informant
characteristics. The specific characteristics tested are target gender, informant gender, target
Reliability and Validity of Family Reports of Alcoholism 20

zygosity (for reliability only), and informant’s alcohol-related problem status. Potential
improvements in validity with multiple versus single informants were then assessed, as well as
differences in validity based on family relationship (Aim 3). Differences in reliability and
validity based on informant alcohol-related problem definitions (Aim 4) are incorporated into
analyses corresponding with reliability and validity from Aims 1 and 2.
Reliability Analyses: Informant Agreement for Family History Reports
First, a series of analyses were conducted to assess informant agreement for reports of
family alcohol-related problems (Aim 1). Analyses for Aim 1 utilized data on offspring
informants reporting on parent targets. The first goal of these analyses was to estimate the
strength of inter-rater reliability (i.e., offspring agreement) for reports of target alcohol-related
problems based on target gender, informant gender, and informant zygosity (Study Aim 1a;
Table 6). Inter-rater reliability was calculated within twin pairs for FH-RDC Alcoholism
Diagnosis and Alcoholism symptom count variables. Informant agreement for FH-RDC
Alcoholism Diagnosis was calculated using Cohen’s kappa (κ; Cohen, 1960), as this provides a
more robust measure of inter-rater agreement for categorical variables than other methods (e.g.,
percent agreement calculations). Polychoric correlations, which measure the strength of the
relationship between ordinal-level variables, were used to calculate inter-rater agreement for FH-
RDC Alcoholism symptom count. To learn more about differences based on specific target and
informant characteristics, inter-rater reliability analyses were stratified by informant gender,
target gender, and informant zygosity. To assess target gender differences, findings were
compared between mothers and fathers. Agreement of MM and FF pairs was compared to
examine overall informant gender differences. Opposite-sex (MF) pair agreement was also
examined, as it permits an investigation of gender differences within informants pairs who are
Reliability and Validity of Family Reports of Alcoholism 21

reporting on the same parents. Like-sexed pairs were stratified by zygosity and compared to
examine possible genetically-based differences in reporting.
The goal of the next set of analyses was to assess the influence of informants’ own
alcohol-related problems on their family history reports and on inter-rater agreement (Study Aim
1b; Tables 8, 9). First, individual-level data were utilized to investigate differences in reporting
FH-RDC Alcoholism diagnosis based on the informant’s own alcohol diagnostic status (Table
8). Odds ratios were calculated to determine the degree to which history of an alcohol diagnosis
in informants was associated with an increase in reporting FH-RDC Alcoholism in targets,
compared to informants without an alcohol diagnosis.
Next, within-pair agreement of FH-RDC Alcoholism reports among twin pairs
concordant and discordant for alcohol-related problems was calculated to further test the
influence of informant alcohol diagnosis on family history reports (Table 9). Within-pair
analyses permit an investigation of informant diagnosis-based differences in family history
reporting while minimizing possible confounding effects of other variables, such as sex, age, and
familial factors. Family history agreement among pairs where both (concordant affected), one
(discordant), or neither (concordant unaffected) of the members had a specific alcohol-related
problem was compared. Low agreement among discordant pairs and high agreement among
concordant pairs would indicate differences in family history reports based on informants’
alcohol problem status. In contrast, similar correlations and FH-RDC Alcoholism prevalences
across concordant and discordant categories would indicate that informant alcohol diagnostic
status may not play a role in informant’s perception of family members’ alcohol-related
problems. Calculation of mean symptom counts/diagnostic prevalences of family history reports
were used to examine trends in reporting based on pair concordance/discordance. Informant
Reliability and Validity of Family Reports of Alcoholism 22

DSM-IV AD, AA/AD, and Problem Drinking were tested and compared in both individual- and
pair-level analyses to test for differences in findings based on varying diagnostic definitions
(Aim 4). Varying diagnostic definitions were applied to all participants in these analyses.
Validity of informant family history reports
The next series of analyses tested the validity of family history reports by comparing
them to the diagnoses derived from structured clinical interviews conducted directly with targets
(Aim 2). The target’s personal interview-based diagnosis were used as the “gold standard,” or the
best estimate, of their true diagnostic status to which informant FH-RDC Alcoholism diagnosis
reports were compared. It is important to acknowledge that diagnoses based on self-reports can
suffer from inaccuracies (e.g., underreporting) that hinder our ability to know the target’s true
diagnosis. However, multiple past studies of validity have utilized target family members’
personal interviews as the best estimate of their problems. This issue is considered in more detail
in the Discussion.
Data from offspring informant/parent target and offspring informant/sibling target pairs
were used to conduct validity analyses (parent informant/offspring target pairing was not used
for these validity analyses due to very low rates of FH-RDC Alcoholism reporting by parents on
offspring). Validity was assessed using the following indices: sensitivity (proportion of affected
targets identified as such by informants), specificity (proportion of unaffected targets identified
as such by informants), positive predictive value (PPV; proportion of positives based on
informant reports that are true positives), and negative predictive value (NPV; proportion of
negatives based on informant reports that are true negatives; Appendix C).
First, differences in validity based on informant and target gender were assessed (Aim 2a;
Tables 10 and 11). For offspring informant/parent target validity analyses, target gender
Reliability and Validity of Family Reports of Alcoholism 23

differences (i.e., mother target vs. father target) were assessed to determine if the gender of the
target is related to the accuracy of family history reports. Informant gender differences could not
be assessed for offspring informant/parent target validity analyses because only mothers and
fathers of FF twins were interviewed. For offspring informant/sibling target analyses, differences
based on target gender (male informant/male target versus male informant/female target) and
informant gender (male informant/male target versus female informant/male target) were
assessed.
Validity analyses were then stratified by informant’s alcohol diagnostic status to test for
differences in validity based on informant’s alcohol-related problems (Aim 2b, Offspring
informant/parent target: Table 12; offspring informant/sibling target: Table 13). Because FH-
RDC Alcoholism and target interview-based alcohol diagnoses were not measured in the same
manner, a variety of target diagnoses will be tested for parent targets (DSM-IIIR AD, DSM-IV
AD, DSM IIIR AA/AD, and Problem Drinking) and sibling targets (DSM IV-AD, DSM-IV
AA/AD, and Problem Drinking). Testing a variety of diagnoses permits a comprehensive
assessment of differences in validity based on diagnostic definition (Aim 4).
Multiple Informants and Family Relation
Analyses were conducted to test potential increases in validity when using information
from multiple informants compared to a single informant (Aim 3; Tables 14 and 15). Data from
two offspring informants reporting on parent targets were used for multiple informant analyses.
First, multiple informant data were combined using an “Or” rule, where FH-RDC Alcoholism
diagnosis was endorsed if either of two informants had a positive report (Table 14). Validity
indices were then calculated and compared to single informant FH-RDC Alcoholism validity
indices. More comprehensive analyses of different consolidation methods at varying FH-RDC
Reliability and Validity of Family Reports of Alcoholism 24

Alcoholism diagnostic cutoffs were also conducted to determine if other methods yielded
optimal validity (Table 15). The other methods of consolidating data from two informants were
“Minimum” (selecting lowest symptom count reported by either informant), “Maximum”
(selecting highest symptom count reported by either informant), and “Mean” (average symptom
count across the two informants) rules. The informant-reported symptom count selected based on
the consolidation rules were then compared to different FH-RDC Alcoholism diagnostic cutoffs
that would designate a positive diagnosis (one or more symptoms, two or more, three or more,
etc.).
To learn more about the influence of relationship type on how alcohol-related problems
are perceived by informants, differences in family history reporting were also tested based on
informant-target family relationship (Aim 3, Table 16). Informant relationships parings of
interest were: offspring informant/parent target, parent informant/offspring target, and offspring
informant/sibling (i.e., cotwin) target. Validity indices were calculated for each pairing by
comparing FH-RDC Alcoholism diagnoses from informant family history reports to AA/AD
diagnoses from personal interviews with targets (AA/AD was selected as the target’s personal
interview diagnosis for validity calculations in order to avoid use of alcohol problem definitions
that were too stringent [i.e., AD] or too broad [i.e., Problem Drinking] ). Validity indices were
compared between informant-target family relationships to test differences in accuracy based on
the pairings.
RESULTS
Findings in this section are organized by study aims, as identified in Table 1. Tables 4a
and 5a present descriptive statistics related to prevalences of alcohol-related problems in
offspring and parents based on personal interviews conducted with them. These alcohol-related
Reliability and Validity of Family Reports of Alcoholism 25

problems range from the narrowest/most restrictive diagnostic category (i.e., Alcohol
Dependence) to the broadest/most inclusive category (i.e., Problem Drinking), as reflected by the
increase in prevalences as the diagnostic definition broadens. The prevalence of alcohol-related
problems is higher in male twins compared to female twins, which is consistent with gender
differences in alcohol problems in the general population. Tables 4b and 5b show FH-RDC
Alcoholism prevalences for sibling (i.e., cotwin) and parent targets based on reports from
offspring informants.
Informant Agreement on Target Alcoholism (Aim 1)
Table 6 presents results of analyses of twin pair agreement (i.e., inter-rater reliability) for
parent FH-RDC Alcoholism diagnosis and symptom count, stratified by gender and zygosity.
Polychoric correlations of twin-reported parental FH-RDC Alcoholism symptom count are
strong and positive for both mothers and fathers across gender and zygosity categories, ranging
from 0.87 to 0.91 in mothers and 0.78 to 0.86 in fathers (p<.05 for all correlations). These
findings indicate that offspring informant reports of parent FH-RDC Alcoholism symptoms are
very similar within twin pairs. Pair agreement (κ) on parent FH-RDC Alcoholism diagnosis is
moderate to strong, ranging from 0.55 to 0.74 for mothers and 0.59 to 0.69 for fathers.
Informant gender, informant zygosity, and target gender (Aim 1a): Prior to investigating
pair-level gender differences in agreement, preliminary analyses were conducted to test
individual-level gender differences in family history reports (Table 7). Chi square tests were
conducted for FH-RDC Alcoholism diagnosis and t-tests were used for symptom count. There
were significant informant gender differences for FH-RDC Alcoholism diagnosis and symptom
count reports, with female informants generally identifying higher levels of problems in target
mothers and fathers than male informants.
Reliability and Validity of Family Reports of Alcoholism 26

Comparisons of inter-rater reliability for FH-RDC Alcoholism symptoms by twin-pair
gender (Table 6, i.e., All FF pairs vs. All MM pairs vs. opposite sex pairs) show that male and
female informant pairs did not differ in agreement for mothers. However, opposite sex pair
agreement for fathers (ρ =0.78) was significantly lower than that of like-sexed male pairs
(ρ=0.84) and female pairs (ρ =0.86).
For FH-RDC Alcoholism diagnosis, opposite sex twin pairs were significantly different
from like-sex twin pairs within mothers and fathers. Within maternal diagnostic reports, opposite
sex pair agreement was significantly higher than like-sex females, but did not differ from like-
sex males (Maternal agreement--Opposite sex: κ =0.74; All female--κ =0.61). In contrast,
diagnostic reports on fathers show that opposite sex agreement was significantly lower than that
in like-sex female twins (Paternal agreement-- Opposite sex: κ =0.59; All females: κ =0.66).
Comparisons of zygosity groups (MZ vs DZ) indicate no significant informant zygosity
differences for FH-RDC Alcoholism diagnosis or for symptom count reports within mothers and
fathers.
To test target gender effects on agreement for FH-RDC Alcoholism diagnosis and
symptom count, informant agreement for reports on mothers and fathers was compared within
informant gender (i.e., All FF mother vs. All FF father; All MM mother vs. All MM father).
Based on these comparisons, no significant target gender differences were found for FH-RDC
Alcoholism diagnosis or symptom count. Findings were similar when comparing the same sex
zygosity groups between parents (e.g., MZF mother vs. MZF father). However, opposite sex
pairs had significantly higher agreement for mothers’ FH-RDC symptom count relative to
fathers’ symptom count (Mother: ρ=0.91; Father: ρ =0.78). This pattern of findings was similar
for FH-RDC Alcoholism diagnosis (Mothers: κ=0.74; Father: κ=0.59). These findings suggest
Reliability and Validity of Family Reports of Alcoholism 27

that opposite sex twin pairs’ reports are less similar when they are reporting on their fathers
compared to their mothers. To learn more about differences in opposite sex pair reporting on
mothers and fathers, follow-up analyses were conducted to compare within-pair differences in
FH-RDC Alcoholism diagnosis prevalences and symptom count means. Sisters report higher
levels of parent drinking severity than their brothers, particularly in reporting on their fathers,
however these were not statistically significant gender differences (Maternal prevalences:
Sisters=5.1%, Brothers=4.8%; Paternal symptom count means[SD]: Sisters—M=0.50 , SD=
1.12, Brothers-- M=0.44 , SD=1.12; Paternal prevalences: Sisters=21.6%, Brothers=18.5%;
Maternal symptom count means[SD]: Sisters—M=0.10, SD=0.48 , Brothers-- M=0.10 ,
SD=0.49.)
Differences in agreement based on informant’s own diagnosis (Aim 1b)
Tables 8 and 9 present differences in family history reports and agreement based on
informant’s own alcohol-related diagnoses. Findings from individual-level analyses of affected
and unaffected informants’ FH-RDC Alcoholism diagnosis reports are shown in Table 8. For
each diagnostic definition of informant alcohol-related problems (DSM-IV AD, AA/AD, and
Problem Drinker), affected informants have greater odds of identifying FH-RDC Alcoholism in
their parents compared to unaffected informants, with odds of positive reports by affected
informants being 2.0 to 2.3 times higher for mothers and 1.7 to 1.9 times higher for fathers than
unaffected informants. Differences in effect sizes by informant diagnostic definition (AD vs.
AA/AD vs. Problem Drinking) were small, suggesting little difference in outcomes based on
how informant alcohol-related problems are defined (Aim 4). Effect size comparisons based on
parent gender show a trend of higher odds of affected informants reporting FH-RDC Alcoholism
Reliability and Validity of Family Reports of Alcoholism 28

than unaffected informants in mothers compared to fathers, though these differences were only
statistically significant within the informant Problem Drinking category.
Odds ratio analyses were conducted again for a broader definition of FH-RDC alcohol
problems (i.e., ever had a problem with drinking that lasted one month) to determine if a more
flexible definition would yield different results (Table 8). Odds ratios based on this item were
similar to those based on FH-RDC Alcoholism diagnosis, indicating little difference in findings
based on broader FH-RDC alcohol problem definition.
Table 9 presents offspring informant within-pair agreement for parent FH-RDC
Alcoholism symptoms and diagnosis by pair concordance or discordance for their own alcohol-
related problems. Analyses were conducted separately by twin diagnostic definitions of AD,
AA/AD, and Problem Drinking. Matched alphabetical superscripts in Table 9 identify which
comparisons (concordant affected vs. concordant unaffected vs. discordant) within informant
diagnostic definitions were significantly different. Differences among concordant and discordant
pairs within each informant diagnostic definition were examined based on statistically significant
differences as well as differences in effect sizes.
Generally, FH-RDC Alcoholism symptoms count means and diagnostic prevalences
present a pattern of highest informant-reported values in concordant affected pairs, lowest values
in concordant unaffected pairs, and values in between the concordant groups for the discordant
pairs. All parent FH-RDC Alcoholism symptoms count correlations were strong, positive, and
statistically significant, indicating high agreement within concordant and discordant pair types.
Maternal and paternal FH-RDC diagnostic twin agreement values (κ) were moderate to strong.
FH-RDC Alcoholism symptom count and diagnostic agreement among discordant pairs is
moderate to strong, indicating that twin pairs generally provide similar family history reports
Reliability and Validity of Family Reports of Alcoholism 29

regardless of their own alcohol-related diagnostic status. Comparisons of agreement among
concordant and discordant pairs reporting on mothers and fathers show some statistically
significant differences (as indicated in Table 9). However, effect sizes were generally similar
among concordant and discordant pairs, which suggests similar levels of agreement across the
different diagnostic status concordance/discordance categories. Findings were similar across
diagnostic definitions (Aim 4).
Validity of family history reports of problems (Aim 2)
Informant and Target Gender Differences in Validity (Aim 2a)
Tables 10 and 11 present validity indices (i.e., sensitivity, specificity, PPV, and NPV)
based on comparison of informant-reported FH-RDC Alcoholism to diagnoses obtained from
direct interviews of targets. As noted in the Methods section, offspring informant/parent target
data (Table 10) only allowed for analyses of target gender differences in validity, however
offspring informant/sibling target validity analyses (Table 11) tested both informant and target
gender differences.
Target gender differences--Offspring informant/parent target validity: Target gender
comparisons were made between mothers and fathers (within diagnosis) and significant parent
gender differences in validity statistics are indicated by numeric superscripts (Table 10). There
were no significant differences in sensitivity between mothers and fathers and values were
generally low for both parents (34%-47%), indicating that informant’s FH-RDC Alcoholism
reports correctly identified only a small number of parents who actually had alcohol problems.
Correct identification of unaffected parents by informants (i.e., specificity) was high for both
parents and was significantly higher for mothers than fathers across all diagnostic definitions.
For both parents, most PPV values ranged from low to moderate. Father’s PPVs were generally
Reliability and Validity of Family Reports of Alcoholism 30

higher than mothers, but significant gender differences existing only within the AA/AD (Mother:
36.1%; Father: 52.9%) and Problem Drinking definitions (Mother: 55.4%; Father: 72.7%),
indicating that gender differences in the chances of a target actually having alcohol problems
when identified as such by informants are only detectable among the broadest/most flexible
alcohol-related diagnoses. In contrast, the chances of a target not having a diagnosis when
identified as such by informants was high for both parents (NPV range: 75.0% to 97.2%) and
maternal NPVs were significantly higher than paternal NPVs across definitions (Range across
definitions--Mother targets: 93.3% to 96.5%; Father targets: 75.0% to 87.5%).
Differences in validity indices based on diagnostic definition (Aim 4) were assessed by
comparing definitions within mothers and fathers (significant differences identified by
alphabetical superscripts). The most notable and consistent differences were primarily found
when comparing Problem Drinking (i.e., the broadest definition of alcohol-related problems) to
the other, more restrictive definitions. PPV was significantly higher for Problem Drinking than
other definitions, indicating that the chances of targets having alcohol problems when identified
as affected by informants is slightly higher when using a more flexible diagnostic definition. In
contrast, the precise identification of unaffected targets (NPV) was slightly higher with the use of
more stringent diagnostic definitions.
Target and informant gender differences--offspring informant/sibling target validity:
Informant and target gender differences for offspring informant/sibling target validity analyses
are presented in Table 11. Analyses were compared based on male informant/male targets,
female informants/male targets, and male informants/female targets (significant differences
indicated by alphabetical superscripts).
Reliability and Validity of Family Reports of Alcoholism 31

Informant gender was not associated with substantial differences in validity when the
target was male, as validity indices for male informant/male target pairs were generally similar to
female informant/male target pairs. Differences in validity indices were more pronounced when
male informants reported on female targets. Male informants reporting on female targets were
less able to correctly identify affected targets (sensitivity) and better able to identify unaffected
targets (specificity) compared to male and female informants reporting on male targets. Also, the
chances of female targets actually having alcohol problems when identified as affected by male
and female informants was lower than male targets (PPV range across definitions—Female
targets: 46.4 to 71.4; Male informants/male targets: 64.6 to 87.5; Female informants/male
targets: 68.6 to 90.4), while the probability of being unaffected when identified as such by
informants was higher for female targets than male targets (NPV range across definitions—
Female targets: 70.1 to 88.8; Male informants/male targets: 55.9 to 81.3; Female
informants/Male targets: 51.0 to 81.2).
Validity of twin reports based on informant diagnostic status (Aim 2b)
To determine whether informants with alcohol-related problems are more/less able to
identify such problems in family members, validity analyses were repeated and stratified by
informants’ diagnoses (Tables 12 and 13). Given that most of the diagnostic differences found in
prior validity analyses were based on the strictest and broadest definition of alcohol-related
problems, the most narrow (DSM-IV AD) and broad (Problem Drinking) definitions of target
alcohol-related problems were selected for these analyses. For consistency with target problem
definition, informant DSM-IV AD and Problem Drinking were used to test validity based on
informant diagnostic status. PPV and NPV were selected as indices of validity in this table.
There were no significant differences in PPVs and NPVs based on reports by affected versus
Reliability and Validity of Family Reports of Alcoholism 32

unaffected informants for target DSM-IV AD. The only significant difference found was for
informant problem drinkers reporting on parent problem drinking, as there were lower NPVs for
affected informants compared to unaffected informants (Mother targets--Unaffected informants:
94.1, Affected informants: 90.5; Father targets--Unaffected informants: 76.7; Affected
informants: 68.4).
The same analyses were conducted for offspring informants reporting on sibling targets
(Table 13). Within target AD and target Problem Drinking, there was significantly lower NPV
for affected offspring informants’ reports on siblings compared to unaffected informants for both
informant AD and Problem Drinking. There were no significant differences in PPV.
Multiple Informants and Informant-Target Family Relationship (Aim 3)
Multiple Informants (Aim 3a)
Tables 14 and 15 show findings related to differences in validity based on multiple
informants. Table 14 presents validity indices for single versus multiple (i.e., two) offspring
informants reporting on maternal and paternal FH-RDC Alcoholism diagnosis (Figure 3 also
presents findings by mothers and fathers). Selection of target diagnosis using multiple informant
reports was based on an “Or” rule, which classified a target as affected if either offspring
informant within a twin pair identified symptoms in their parent that were consistent with an
FH-RDC Alcoholism diagnosis. Based on the findings, there were some improvements in
validity indices when two informants were utilized compared to one informant, though the
differences were not substantial. For both parent targets, the identification of affected cases
(sensitivity) was slightly better with multiple informants compared to single informants (Mothers
48.5% vs 39.5%; Fathers: 53.2% vs 44.0%), and differences were small for all other validity
indices.
Reliability and Validity of Family Reports of Alcoholism 33

Additional analyses were conducted to determine if other methods of consolidating
multiple informant reports (i.e., Minimum, Maximum, and Mean) resulted in differences in
validity indices compared to single informant reports (Table 15, Figures 4 and 5). These
consolidation methods were tested at varying FH-RDC Alcoholism diagnostic cutoffs (i.e.,
varying symptom thresholds) to determine optimal cutoffs that would increase validity. Overall,
there were few improvements in validity indices across multiple informant consolidation
methods compared to single informants. Combining multiple informant reports with the
‘Maximum’ rule resulted in slightly better detection of affected cases compared to single
informants across most diagnostic cutoffs for mothers (Sensitivity for ‘Maximum rule’ vs single:
48.5 vs. 39.5 for 1 symptom cutoff, 39.4 vs. 30.3 for 2 symptom cutoff, etc.) as well as fathers
(53.2 vs. 43.8 for 1 symptom cutoff, 34.0 vs. 28.9 for 2 symptom cutoff, etc.). In fathers, precise
identification of affected cases was highest for the ‘Minimum’ rule (PPV for ‘Minimum rule’ vs.
single: 64.0 vs. 52.9 for 1 symptom cutoff, 74.1 vs. 60.0 for 2 symptom cutoff, etc.). Differences
in PPV for mothers were most apparent for ‘Minimum’ and ‘Mean’ rules, but only as diagnostic
cutoffs increased (particularly for cutoffs of 3 or higher). Specificity and NPV values across
multiple informant consolidation methods were similar to single informant indices for both
mothers and fathers.
Family Relationship Between Targets and Informants
Table 16 presents prevalences and validity of family history reports of Alcoholism based
on informant-target family relationship (i.e., offspring informant/parent target; parent
informant/offspring target; offspring informant/sibling target). For parent informant/offspring
target family relationship, both mothers and father informants reported lower prevalences of
alcohol-related problems in their offspring than the offspring reported about themselves (FH-
Reliability and Validity of Family Reports of Alcoholism 34

RDC Alcoholism diagnosis by mother informant: 2.0%, AA/AD reported by offspring target:
13.5%; FH-RDC Alcoholism diagnosis by father informant: 1.3%, AA/AD reported by offspring
target: 13.8%). In contrast, offspring informant reports on parent targets were more comparable
to parent’s reports of their own alcohol-related problems (Mother target: AA/AD reported by
mother target: 5.5%, FH-RDC diagnosis by offspring informant: 6.0%; Father target: AA/AD
reported by father target: 20.2%, FH-RDC diagnosis by offspring informant: 16.8%). Offspring
informant/sibling target family relationship shows that offspring informants endorse lower levels
of alcohol-related problems than endorsed by sibling targets (AA/AD reported by sibling target:
35.0%; FH-RDC diagnosis by sibling informant: 10.7%).
Based on validity findings, all three family relationship groups are good at accurately
identifying unaffected informants (specificity range: 90.1% to 99.5%), with parent
informant/offspring target pairs having the highest specificity. There were more pronounced
differences across family relationship types with regard to informant detection of affected family
members, as correct identification of affected targets was significantly higher in offspring
informant/parent target pairs (Sensitivity: Mother targets: 39.5%; Father targets: 44.0%) than
other informant-target family pairs. The parent informant/offspring target pairing had the lowest
sensitivity levels (Mother informant: 11.3%; Father informant: 6.5%).
While sensitivity was highest for the offspring informant/parent target relationship, the
chances of actually being affected when identified as such by informants was generally lowest in
this family pairing compared to other family relationships (PPV-- Mother targets: 36.1; Father
targets: 52.91), which suggests that offspring informants tend to over-report problems in their
parents. PPV values were highest in offspring informant/sibling target pairs (PPV=80.3%), but
this finding was only statistically significant in comparison to offspring reports on parents
Reliability and Validity of Family Reports of Alcoholism 35

(Offspring reports on mothers: PPV=36.1 ; Offspring reports on fathers: PPV=52.9). Precision of
identifying unaffected targets was generally high across all family relationships: NPV was
significantly higher for offspring informants reporting on mothers (NPV=96.5) and significantly
lower for offspring reporting on siblings (NPV=70.4) than all other family relation groups. NPV
for offspring informants reporting on fathers resembled values for parent informant reports on
offspring.
DISCUSSION
Summary of findings
The present study assessed differences in reliability and validity of family history reports
based on various informant, target, and diagnostic variables. Reliability findings indicate that
informants generally demonstrated moderate to high agreement in their reports of mother’s and
father’s drinking problems. With regard to informant and target characteristics, target gender
were associated with differences in agreement, but informant zygosity was not. Opposite sex
agreement analyses were critical to clarifying the role of target gender in agreement, as brothers
and sisters within opposite sex pairs were found to have similar reports of parent alcoholism.
While individual-level findings show that affected informants identify targets as having alcohol
problems more than unaffected informants do, findings within informant twin pairs reporting on
the same family members indicate that there are few differences in reporting based on informant
alcohol-related problem status.
Overall, validity findings indicated that informants are better at identifying family
members who are unaffected rather than affected by alcohol-related problems. There were target
gender differences in validity, with higher accuracy and precision of non-case identification (i.e.,
specificity and NPV) in female and male informants when reporting on female targets compared
Reliability and Validity of Family Reports of Alcoholism 36

to male targets. There were significant differences in validity between affected and unaffected
informants, with higher chances of a target not meeting criteria for alcohol-related problems
when identified as such by unaffected informants compared to affected informant reports.
Reliability and validity findings did not vary substantially with the use of different diagnostic
definitions, however there were some notable differences in findings when comparing the
narrowest and broadest definitions of alcohol-related problems.
Findings pertaining to multiple informants indicated that there are some improvements in
validity when using multiple informants compared to a single informant, but differences were not
substantial. Family relationship analyses show that offspring informant/parent target pairs have
higher sensitivity than offspring informant/sibling target and parent informant/offspring target
pairs, however there was also evidence indicating that offspring informants tend to over-report
alcohol-related problems in their parents. Parent informants demonstrated the highest level of
specificity levels compared to the other informant-target family relationships.
Informant Agreement Findings
Informant gender, target gender, and informant zygosity
The moderate to high levels of agreement for family history reports found in this study
are comparable to other studies investigating twin and non-twin sibling agreement (Amodeo &
Griffin, 2009; Crews & Sher, 1992; Heun & Müller, 1998; Kendler et al., 1991; Prescott et al.,
2005; Rhea et al., 1993). Collectively, these findings indicate that family history reporting tends
to be very similar in siblings. This information is particularly useful in the presence of limited
resources or lack of accessibility to other family members, as researchers/clinicians who only
have access to one offspring informant could expect similar reports from other offspring
informants within the same family.
Reliability and Validity of Family Reports of Alcoholism 37

There were significant informant and target gender-related differences in agreement in
the present study, which were primarily found in comparisons that involved opposite sex twin
pairs. Opposite sex pairs tended to agree less than like-sex pairs on reports of father’s alcohol-
related problems. Also, agreement about mothers was significantly higher in opposite sex pairs
compared to same-sex females for FH-RDC Alcoholism diagnosis. With regard to parent gender
differences, opposite sex pair agreement for fathers was significantly lower than opposite sex
pair agreement for mothers. Further investigation of prevalences of family history reports in
opposite sex pairs showed that while sisters tend to report more alcohol problems in parents,
these reports were not significantly different from their brother’s family history reports on
mothers or fathers. These findings suggest that while agreement may vary in terms of strength
based on informant and target gender, there are still considerable similarities in family history
reporting regardless of gender.
The use of opposite sex twin data in this study was key to elucidating gender differences
in agreement found in this and other studies. The tendency for female informants to report higher
levels of relatives’ drinking problems than male informants has been found in other family
history studies (Kosten et al., 1992; Roy et al., 1994). However, as noted earlier, family history
prevalences based on reports within opposite sex twin pairs (i.e., reports made by brothers versus
sisters) did not yield significant informant gender differences. Based on these discrepancies,
findings related to informant gender differences may be attributable to methodology rather than
actual gender differences. Unlike past studies that have investigated gender differences in
informants who were reporting on different parents (e.g., individual-level comparisons), this
study presents the unique advantage of investigating gender differences in opposite sex twins
reporting on the same parents. Gender differences/similarities found within the context of
Reliability and Validity of Family Reports of Alcoholism 38

opposite sex pairs are particularly informative, as they present sex effects while naturally
controlling for other factors that vary between non-twin males and females (e.g., born from same
mother at the same time, raised in same family, reporting about the same parents, etc.).
Therefore, testing gender differences within opposite sex twin pairs allows for a more
methodologically robust investigation of gender differences.
Informant diagnostic status and agreement
Individual-level analyses show that affected twins are more likely than unaffected twins
to report problems in their parents. While individual-level findings may appear to indicate
differences in reporting based on informant diagnostic status, these findings alone cannot
distinguish whether the results are attributable to affected informant reporting biases or actual
presence of more drinking problems in family members of affected informants. Pair-level
analyses helped clarify the likely origin of these findings: because discordant pairs’ reports on
the same family members did not differ substantially based on their diagnostic status (affected
versus unaffected), the increased odds of positive reports by affected twins at the individual-level
are more likely attributable to the actual presence of drinking problems in the family than to
informant reporting biases. Findings from these pair-level analyses support other studies that
have also found similarities in informant family history reports, regardless of diagnostic status
(Amodeo & Griffin, 2009; Kendler et al., 1991; Slutske et al., 1996). Also, multiple studies
provide very strong evidence for genetic and family environmental factors that confer risk of
alcohol-related problems in families (for reviews see Cotton, 1979; Dick, Prescott, & McGue,
2009),which supports the concept that affected informants have family members that are more
likely to be affected and are simply identifying them as such. These findings highlight the
importance of investigating family history reporting differences within rather than between
Reliability and Validity of Family Reports of Alcoholism 39

families in order to thoroughly test factors that may affect how informants describe their family
members.
Validity Findings
Validity findings in this study indicate that informants are better able to identify non-
cases versus cases in targets, which is consistent with validity findings in many other family
history studies (e.g., Kendler et al., 2002; Kosten et al., 1992; Milne et al., 2009; Thompson et
al., 1982). One reason for low sensitivity levels in this and other studies is underreporting by
informants (i.e., high false-negative rates), which has long been considered one of the
weaknesses of the family history method (Andreasen et al., 1986). However, it is unclear
whether underreporting is attributable to informants being unaware of or reluctant to disclose
targets’ problems.
While it is not typical that individuals are completely unaware of alcohol-related
problems in their family members, there may be specific factors contributing to informant’s lack
of knowledge about target’s drinking. For example, age gap between targets and informants may
affect informant’s knowledge of alcohol-related problems in their family members. This could
particularly apply to informants who are considerably younger that targets (e.g., offspring
reporting on parents), as targets may not have witnessed alcohol-related problems in parents
when their parents were at higher risk for experiencing these problems (i.e., during parents’
younger years). However, analyses related to informant-target family relationship indicated that
sensitivity was also low for siblings who are the same age (i.e., twins). Age differences between
informant and target merits additional research in order to better understand its potential effects
on informants’ understanding and knowledge of alcohol-related problems in family members.
Reliability and Validity of Family Reports of Alcoholism 40

Reluctance of informants to disclose family member’s drinking problems could also
result in underreporting. For individuals with alcohol and other substance abuse problems,
perceived stigma has been shown to interfere with disclosure of problems and can serve as a
barrier to treatment-seeking behavior (Copeland, 1997; Cunningham, Sobell, Sobell, Agrawal, &
Toneatto, 1993; Grant, 1997). Such stigma has been shown to have an impact on family
members of individuals with psychopathology, as they may become concerned about the impact
that stigmatization can have on their affected relatives (Corrigan & Miller, 2004; Struening et al.,
2001; Wahl & Harman, 1989). Family member’s perception of and concerns about stigma
experienced by mentally ill relatives has been referred to as ‘vicarious stigma’(Corrigan &
Miller, 2004). While there is not much research on vicarious stigma in family members of
individuals with alcohol-related problems, it may be possible that perceived stigma serves as a
barrier to informants disclosing family members’ drinking problems. Additional research is
required to better understand motives for and barriers to informant disclosure of alcohol-related
problems in family members.
Another explanation for possible underreporting of alcohol-related problems by family
informants could be targets’ non-disclosure of their own drinking problems to family members.
Shame and concern about repercussions of disclosure may hinder honest communication of
presence or extent of substance abuse to family members (Dearing, Stuewig, & Tangney, 2005;
Meehan et al., 1996; O’Connor, Berry, Inaba, Weiss, & Morrison, 1994). This can also result in
underreporting by targets to researchers. However, honest reporting by targets is difficult to
determine, as family history studies of alcohol problems rarely employ objective measures of
alcohol use. For this reason, the use of targets’ structured clinical interview-based diagnoses as
Reliability and Validity of Family Reports of Alcoholism 41

the ‘gold standard’ must be recognized as a common but imperfect method of assessing validity
of family history reports.
Informant and Target Gender Differences in Validity
Specificity and NPV values were high for both mothers and fathers in offspring
informant/parent target validity. However, there were some target gender differences, with
maternal specificity and NPV values being somewhat higher than paternal values. While these
findings may be due in part to the high number of unaffected females, informant identification of
true negatives in mother targets may be easier for other reasons. Time spent with parents may
partly account for gender differences in parent targets. Mothers typically spend more time with
their children while engaged in a variety of childcare activities (Craig et al., 2006; Yeung et al.,
2004). This may provide more opportunities for children to observe mother’s lack of alcohol use,
whereas father’s behaviors may not be as evident. Considering that alcohol problems are more
likely to be identified and agreed on by family members when they are linked to observable
behaviors (O’Farrell & Maisto, 1987; Rhea et al., 1993; Sher & Descutner, 1986), less time spent
with fathers may mean fewer opportunities to observe father’s drinking behaviors.
Sibling validity findings provide additional insight into gender differences in validity.
Male sibling informants’ reports of female sibling targets yielded significantly lower sensitivity
levels compared to male and female sibling informant reports on male sibling targets. For parent
target validity findings, there were no significant differences between sensitivity levels for
female offspring informant reports on mothers and fathers. These findings indicate that males
may be less perceptive to or forthcoming about alcohol-related problems when it comes to
identifying them in females, as false-negative rates were decreasing sensitivity levels.
Underreporting of alcohol problems in females by males may be attributable to gender biases in
Reliability and Validity of Family Reports of Alcoholism 42

beliefs about female drinking behavior. Past studies have shown that valuing traditional feminine
roles have been associated with lower alcohol use, while traits perceived as consistent with male
gender roles are associated with more problematic use (Chomak & Collins, 1987; Ricciardelli,
Connor, Williams, & Young, 2001; Snell, Belk, & Hawkins, 1987). Gender biases in beliefs
related to female alcohol use may result in misperceptions of actual levels of use by family
informants. Further investigation of informants’ beliefs related to family members’ gender roles
may help clarify the gender differences in validity observed in this study.
Informant Diagnostic Status and Validity
Validity findings provide some evidence for differences based on informant diagnosis, as
affected informants were less likely than unaffected informants to identify unaffected targets.
Unlike previous studies (Chapman et al., 1994; Rice et al., 1995), there was no evidence that
affected informants are better able to identify affected targets. These findings may indicate that
while affected informants have more difficulty precisely identify targets who do not have alcohol
problems, this does not conversely make them “better” able to identify affected informants.
Therefore, conclusions pertaining to accuracy and precision of informants’ family history reports
based on their own diagnostic status must be approached with caution. Future studies can
provide insight into informants’ perception of drinking problems in others by directly assessing
how they reach conclusions about which family members have alcohol-related problems (e.g.,
assessing the kind of evidence/information affected and unaffected informants utilize to reach
these conclusions).
Multiple informants and Family Relation
Findings from this study indicate that reports from multiple informants result in few
improvements in validity. Furthermore, these improvements did not result in considerably higher
Reliability and Validity of Family Reports of Alcoholism 43

validity indices. Given that there were few differences in validity, these findings suggest that it
may not be cost-effective to collect data from multiple informants. While another study also
found few differences in validity with multiple sibling informants (Prescott et al., 2005), others
have found marked improvements in validity with more informants (Kosten et al., 1992; Smith et
al., 1994; Thompson et al., 1982; Weissman et al., 2000). These differences in findings may be
explained by the type of informant reports being combined. While the present study tested
changes in validity by adding an additional twin offspring informant, many of the past studies
with increased validity combined informant reports from non-sibling informants (e.g., combining
reports from any first-degree relatives). Agreement analyses from the present study indicate that
twin informants tend to agree on family history reports of parental alcohol problems, which may
indicate that using multiple twin informants may not add new information that was not already
provided by a single informant.
Based on these findings, quantity of informants alone may not successfully increase
validity indices, as the type of family members to combine may also be important to take into
account. While the present study could not test a variety of multiple informant combinations
given temporal differences in when data were collected, future studies may want to test costs and
benefits of combining informant reports from different kinds of family members. It is also
important that future studies investigate a range of multiple informant consolidation rules to help
optimize utilization of family informant data, as most of the previous studies did not test a
variety of informant consolidation rules.
Findings from this study also show that family relationship type can play a role in the
validity of family history reports. Although actual sensitivity values were somewhat lower than
those found in previous studies of informant-target family relationship type, patterns in
Reliability and Validity of Family Reports of Alcoholism 44

sensitivity were similar to past studies: offspring informants were found to have the highest
sensitivity and parents were found to have the lowest sensitivity. However, we also found higher
specificity in parent informants compared to offspring informants, indicating that there may be
trade-offs in certain validity indices depending on which family members are assessed. These
findings highlight potential strengths and weaknesses of information provided by specific family
members, which can guide more informed decisions about which family members should serve
as informants.
Agreement and Validity Findings Based on Diagnostic Definitions
While there were some statistically significant differences based on diagnostic definitions
tested, agreement and validity did not vary considerably across definitions. Some differences
were found when using the broadest definition of alcohol problems (i.e., Problem Drinking),
however this was not a consistent finding across analyses. These findings suggest that alcohol
problem definition may not have a substantial impact on findings and various definitions of
alcohol-related problems can be considered suitable for use in family history studies. However, it
may be important for researchers to consider the costs and benefits of utilizing certain diagnostic
definitions; for example, diagnostic definitions that vary in severity (i.e., broad vs strict
definitions) may warrant careful selection based on the impact they could have on false-positive
or false-negative rates. It is recommended that future studies remain attentive to possible
differences in reliability and validity based on diagnostic definition, particularly in light of the
recent changes introduced by the Diagnostic and Statistical Manual of Mental Disorders, Fifth
edition (DSM-V), which places more of an emphasis on spectrum of severity in the diagnosis of
Alcohol Use Disorder.

Reliability and Validity of Family Reports of Alcoholism 45

Limitations and Strengths
The present study has some limitations. The sample‘s homogeneous composition (i.e.,
families of Caucasian twins born in Virginia) may impact the generalizability of the results to
other populations. Also, diagnostic criteria were not uniform across informants and targets in
validity analyses, as FH-RDC Alcoholism diagnosis reported by informants was compared to
DSM-based diagnoses from personal interviews with targets. However, the inclusion of multiple
diagnostic definitions in the analyses helped provide evidence for findings that were not
definition-dependent. Another limitation is the use of personal target interviews data as the “gold
standard” for validity analyses, as this method of data collection suffers from a variety of
problems including underreporting. This can be particularly problematic when asking individuals
to recall problems/events that occurred in their lifetime (i.e., lifetime history of drinking
problems).
There are many strengths of the present study. The attention to both informant and target
characteristics is an important feature of this study. While many family history studies of
reliability and validity of alcoholism only focused on one gender in targets or informants, this
study includes both male and female targets and informants. The use of opposite sex twin pairs
was particularly informative and a robust method of testing gender differences, as opposite sex
pairs permit an investigation of gender differences in siblings who were conceived, born, and
reared at the same time with the same parents. The opposite sex twin pair design is a unique
feature of this study that helped clarify gender differences in reliability and validity found in this
and other studies. Another strength of this study is the comprehensive testing of differences in
reliability and validity based on diagnostic definitions, which had not been investigated in detail
in prior family history studies of alcohol problems. While it is not possible to conduct an
Reliability and Validity of Family Reports of Alcoholism 46

exhaustive test of all diagnostic definitions used in past family history studies, the present study
provides strong evidence for consistency in findings regardless of diagnostic definitions used.
Also, the twin sample allowed for both individual-level and pair-level analyses which aided the
investigation of possible biases in reporting.
Conclusions
The present study highlights the importance of informant and target characteristics in
assessing agreement and validity of family history reports. This study identifies a range of
factors that are important to take into consideration when using family history data, including
informant and target gender, informant’s own diagnostic status, and the family relationship of the
informant and target. Future studies can conduct a more comprehensive investigation of
mechanisms underlying certain findings in this study, such as potential sources for gender
differences in validity. Findings from the present study can help researchers and treatment
providers better understand and utilize valuable information provided by relatives in order to
learn more about risk for alcohol-related problems in families.

Reliability and Validity of Family Reports of Alcoholism 47

References
American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders
(3rd ed., text rev.). Washington DC: Author.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental
disorders (4th ed.). Washington DC: Author.
Amodeo, M., & Griffin, M. L. (2009). Sibling agreement on retrospective reports of parental
alcoholism and other childhood events. Substance Use & Misuse, 44(7), 943–964.
Andreasen, N. C., Rice, J., Endicott, J., Reich, T., & Coryell, W. (1986). The family history
approach to diagnosis: How useful is it? Archives of General Psychiatry, 43(5), 421–429.
Bierut, L. J., Dinwiddie, S. H., Begleiter, H., Crowe, R. R., Hesselbrock, V., Nurnberger, J. I., …
Reich, T. (1998). Familial transmission of substance dependence: Alcohol, marijuana,
cocaine, and habitual smoking. Archives of General Psychiatry, 55, 982–988.
Brent, D. A., & Mann, J. J. (2005). Family genetic studies, suicide, and suicidal behavior.
American Journal of Medical Genetics Part C: Seminars in Medical Genetics, 133C(1),
13–24.
Byrne, M. (2002). Family history of psychiatric disorders and age at first contact in
schizophrenia: An epidemiological study. The British Journal of Psychiatry, 181(43), 19–
25.
Chapman, T. F., Mannuzza, S., Klein, D. F., & Fyer, A. J. (1994). Effects of informant mental
disorder on psychiatric family history data. American Journal of Psychiatry, 151, 574–
579.
Reliability and Validity of Family Reports of Alcoholism 48

Coelho, H. F., Cooper, P. J., & Murray, L. (2007). A family study of co-morbidity between
generalized social phobia and generalized anxiety disorder in a non-clinic sample.
Journal of Affective Disorders, 100(1), 103–113.
Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological
Measurement, 20, 37–46.
Copeland, J. (1997). A qualitative study of barriers to formal treatment among women who self-
managed change in addictive behaviours. Journal of Substance Abuse Treatment, 14(2),
183–190.
Corrigan, P. W., & Miller, F. E. (2004). Shame, blame, and contamination: A review of the
impact of mental illness stigma on family members. Journal of Mental Health, 13(6),
537–548.
Cotton, N. S. (1979). The familial incidence of alcoholism: A review. Journal of Studies on
Alcohol and Drugs, 40(1), 89–116.
Crews, T. M., & Sher, K. J. (1992). Using adapted short MASTs for assessing parental
alcoholism: Reliability and validity. Alcoholism: Clinical and Experimental Research,
16(3), 576–584.
Cunningham, J. A., Sobell, L. C., Sobell, M. B., Agrawal, S., & Toneatto, T. (1993). Barriers to
treatment: Why alcohol and drug abusers delay or never seek treatment. Addictive
Behaviors, 18(3), 347–353.
Dearing, R. L., Stuewig, J., & Tangney, J. P. (2005). On the importance of distinguishing shame
from guilt: Relations to problematic alcohol and drug use. Addictive Behaviors, 30(7),
1392–1404.
Reliability and Validity of Family Reports of Alcoholism 49

Dick, D. M., Prescott, C. A., & McGue, M. (2009). The genetics of substance use and substance
use disorders. In Handbook of behavior genetics (pp. 433–453). New York: Springer.
Endicott, J., Andreasen, N., & Spitzer, R. L. (1978). Family history research diagnostic criteria.
New York: Biometrics Research, New York State Psychiatric Institute.
Grant, B. F. (1997). Barriers to alcoholism treatment: Reasons for not seeking treatment in a
general population sample. Journal of Studies on Alcohol, 58, 365–371.
Hettema, J. M., Neale, M. C., & Kendler, K. S. (2001). A review and meta-analysis of the
genetic epidemiology of anxiety disorders. American Journal of Psychiatry, 158(10),
1568–1578.
Heun, R., & Müller, H. (1998). Interinformant reliability of family history information on
psychiatric disorders in relatives. European Archives of Psychiatry and Clinical
Neuroscience, 248(2), 104–109.
Kendler, K. S., Davis, C. G., & Kessler, R. C. (1997). The familial aggregation of common
psychiatric and substance use disorders in the National Comorbidity Survey: A family
history study. The British Journal of Psychiatry, 170(6), 541–548.
Kendler, K. S., & Prescott, C. A. (2006). Genes, environment, and psychopathology. New York:
The Guilford Press.
Kendler, K. S., Prescott, C. A., Jacobson, K., Myers, J., & Neale, M. C. (2002). The joint
analysis of personal interview and family history diagnoses: Evidence for validity of
diagnosis and increased heritability estimates. Psychological Medicine, 32(5), 829–842.
Kendler, K. S., Silberg, J. L., Neale, M. C., Kessler, R. C., Heath, A. C., & Eaves, L. J. (1991).
The family history method: Whose psychiatric history is measured. American Journal of
Psychiatry, 148(11), 1501–1504.
Reliability and Validity of Family Reports of Alcoholism 50

Klein, D. N., Lewinsohn, P. M., Rohde, P., Seeley, J. R., & Shankman, S. A. (2003). Family
study of co-morbidity between major depressive disorder and anxiety disorders.
Psychological Medicine, 33(4), 703–714.
Kosten, T. A., Anton, S. F., & Rounsaville, B. J. (1992). Ascertaining psychiatric diagnoses with
the family history method in a substance abuse population. Journal of Psychiatric
Research, 26(2), 135–147.
Meehan, W., O’Conner, L., Berry, J., Weiss, J., Morrison, A., & Acampora, A. (1996). Guilt,
shame, and depression in clients in recovery from addiction. Journal of Psychoactive
Drugs, 28(2), 125–134.
Miles, D. R., Stallings, M. C., Young, S. E., Hewitt, J. K., Crowley, T. J., & Fulker, D. W.
(1998). A family history and direct interview study of the familial aggregation of
substance abuse: The adolescent substance abuse study. Drug and Alcohol Dependence,
49(2), 105–114.
Milne, B. J., Caspi, A., Crump, R., Poulton, R., Rutter, M., Sears, M. R., & Moffitt, T. E. (2009).
The validity of the family history screen for assessing family history of mental disorders.
American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, 150(1), 41–
49.
Newman, S. C., & Bland, R. C. (2006). A population-based family study of DSM-III generalized
anxiety disorder. Psychological Medicine, 36(09), 1275.
O’Connor, L. E., Berry, J. W., Inaba, D., Weiss, J., & Morrison, A. (1994). Shame, guilt, and
depression in men and women in recovery from addiction. Journal of Substance Abuse
Treatment, 11(6), 503–510.
Reliability and Validity of Family Reports of Alcoholism 51

O’Farrell, T. J., & Maisto, S. A. (1987). The utility of self-report and biological measures of
alcohol consumption in alcoholism treatment outcome studies. Advances in Behaviour
Research and Therapy, 9(2), 91–125.
O’Malley, S. S., Carey, K. B., & Maisto, S. A. (1986). Validity of young adults’ reports of
parental drinking practices. Journal of Studies on Alcohol and Drugs, 47, 433.
Prescott, C. A., Sullivan, P. F., Myers, J. M., Patterson, D. G., Devitt, M., Halberstadt, L. J., …
Kendler, K. S. (2005). The Irish Affected Sib Pair Study of Alcohol Dependence: Study
methodology and validation of diagnosis by interview and family history. Alcoholism:
Clinical & Experimental Research, 29(3), 417–429.
Rhea, S. A., Nagoshi, C. T., & Wilson, J. R. (1993). Reliability of sibling reports on parental
drinking behaviors. Journal of Studies on Alcohol, 54, 80–84.
Rice, J. P., Reich, T., Bucholz, K. K., Neuman, R. J., Fishman, R., Rochberg, N., … Begleiter,
H. (1995). Comparison of direct interview and family history diagnoses of alcohol
dependence. Alcoholism: Clinical and Experimental Research, 19(4), 1018–1023.
Roy, M.-A., Walsh, D., Prescott, C. A., & Kendler, K. S. (1994). Biases in the diagnosis of
alcoholism by the family history method. Alcoholism: Clinical and Experimental
Research, 18(4), 845–851.
Searles, J. S., Alterman, A. I., & Miller, S. M. (1993). Comparability of self-report of familial
alcoholism among male and female college students. Journal of Studies on Alcohol, 54,
730–732.
Sher, K. J., & Descutner, C. (1986). Reports of paternal alcoholism: Reliability across siblings.
Addictive Behaviors, 11(1), 25–30.
Reliability and Validity of Family Reports of Alcoholism 52

Slutske, W. S., Heath, A. C., Madden, P. A. ., Bucholz, K. K., Dinwiddie, S. H., Dunne, M. P.,
… Martin, N. G. (1996). Reliability and reporting biases for perceived parental history of
alcohol-related problems: Agreement between twins and differences between discordant
pairs. Journal of Studies on Alcohol, 57, 387–395.
Smith, E. M., Przybeck, T. R., Bradford, S. E., Gogineni, A., & Spitznagel, E. L. (1994). Adult
offspring of alcoholic women as family history informants. Alcoholism: Clinical and
Experimental Research, 18(6), 1354–1360.
Spitzer, R. L., Williams, J. B., & Gibbon, M. (1987). Structured Clinical Interview for DSM III-
R. New York: New York State Psychiatric Institute Biometrics Research Department.
Struening, E. L., Perlick, D. A., Link, B. G., Hellman, F., Herman, D., & Sirey, J. A. (2001).
Stigma as a barrier to recovery: The extent to which caregivers believe most people
devalue consumers and their families. Psychiatric Services, 52(12), 1633–1638.
Sullivan, P. F., Neale, M. C., & Kendler, K. S. (2000). Genetic epidemiology of major
depression: Review and meta-analysis. American Journal of Psychiatry, 157(10), 1552–
1562.
Thompson, W. D., Orvaschel, H., Prusoff, B. A., & Kidd, K. K. (1982). An evaluation of the
family history method for ascertaining psychiatric disorders. Archives of General
Psychiatry, 39(1), 53.
Wahl, O. F., & Harman, C. R. (1989). Family views of stigma. Schizophrenia Bulletin, 15(1),
131–139.
Waldron, M., Madden, P. A. F., Nelson, E. C., Knopik, V. S., Glowinski, A. L., Grant, J. D., …
Heath, A. C. (2012). The interpretability of family history reports of alcoholism in
Reliability and Validity of Family Reports of Alcoholism 53

general community samples: Findings in a midwestern U.S. twin birth cohort.
Alcoholism: Clinical and Experimental Research, 36(6), 590–597.
Weissman, M. M., Wickramaratne, P., Adams, P., Wolk, S., Verdeli, H., & Olfson, M. (2000).
Brief screening for family psychiatric history: The family history screen. Archives of
General Psychiatry, 57(7), 675.

Reliability and Validity of Family Reports of Alcoholism 54

Acknowledgements
We acknowledge the contribution of the Virginia Twin Registry, now part of the Mid-Atlantic
Twin Registry, to ascertainment of subjects for this study. We would like to thank Rubin
Khoddam and Drs Margaret Gatz, David Lavond, Suzanne Wenzel, Kelly Young-Wolff, Lewina
Lee, Nicole Sintov, and Kevin Petway for their guidance and contributions.
Reliability and Validity of Family Reports of Alcoholism 55

Aim of interest Analysis Approach
Aim 1: Estimate the strength of inter-informant
agreement on family alcohol-related problems
based on the target and informant characteristics
of:
a. target and informant gender, as well as
zygosity of twin informants
b. informant’s alcohol-related problem status
[Both aims 1a and 1b were conducted for twin offspring reports (informants) on their parents (targets).]

Aim 1a: Assess inter-rater agreement within informant twin pairs by calculating polychoric correlations for
target (parent) symptom count and Cohen’s kappa for target categorical diagnosis. Stratified by informant
gender, target gender, and twin zygosity.
Aim 1b: Individual-level differences in informant family history reports based on informant’s own
diagnostic status (0/1) were tested using odds ratios.
Within-pair differences in informant family history agreement based on informant’s own diagnostic status
were tested by stratifying inter-rater agreement (i.e., polychoric correlations and Cohen’s kappa) by pair
diagnostic status (both, one, or neither twin has diagnosis).
Aim 2: Estimate the validity of informant
reports of family alcohol-related problems based
on the target and informant characteristics of:
a. target and informant gender
b. informant’s alcohol-related problem status
[Both aims 2a and 2b were conducted separately for offspring informant reporting on parent target and
offspring informant reporting on twin sibling target.**]
Aim 2a: Compare informant family history reports to target’s own reports of alcohol-related problems by
calculating sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Test
target and informant gender differences for each of these validity indices.
Aim 2b: Validity analyses from Aim 2a were stratified by informant’s alcohol-related diagnostic status.
Aim 3: Test impact of (a)multiple informants
and (b) target and informant family relationship
on family history reporting.

Aim 3a: Compare whether combining family history reports from multiple informants versus using a single
informant results in differences in validity—utilize different methods of combining multiple informant
reports (i.e., combined based on maximum, minimum, and mean ‘rules’) and calculate sensitivity,
specificity, positive predictive value, and negative predictive value for comparison to single-informant
validity indices.

Aim 3b: Test differences in validity of family history reporting based on family relation between target and
informant. Relationships to be compared are: Child-parent relationships (offspring reporting on parent, and
parent reporting on offspring) and sibling relationship (twin reporting on cotwin).
Table 1: Study Aims and Analysis Approach
Reliability and Validity of Family Reports of Alcoholism 56

** Data on parent informant-offspring target pairings were not used in validity analyses due to very low rates of parent reports of family history of
alcoholism in offspring
-AD=Alcohol Dependence; AA/AD=Alcohol Abuse or Dependence
-DSM-IV= Diagnostic and Statistical Manual of Mental Disorders Fourth edition (American Psychiatric Association, 1994); DSM-IIIR= Diagnostic
and Statistical Manual of Mental Disorders Third edition, Revised (American Psychiatric Association, 1987)

Aim 4: In agreement and validity analyses, test
differences in findings based on varying
definitions of alcohol-related problems.

Investigation of Aim 4 was incorporated into aforementioned analyses from Aims 1 and 2. The manner in
which different definitions were tested depended on diagnostic data available, as follows:

For agreement analyses:
-Agreement analyses related to informant diagnosis (Aim 1b) were stratified and compared by the following
informant interview-based diagnoses: DSM-IV AD, AA/AD, and Problem Drinking

For validity analyses:
-Validity analyses (Aim2a) were compared based on the following target diagnoses--For parents targets’
diagnoses based on personal interview: DSM-IIIR AD, DSM IV-AD, DSM-IIIR AA/AD, Problem Drinking;
for sibling targets’ diagnoses based on personal interview: DSM-IV AD, DSM-IV AA/AD, Problem Drinking

-Validity analyses related to informant diagnosis (Aim 2b) were compared based the following informant
and target diagnoses: informant DSM-IV AD and Problem Drinking.
(continue: Table 1: Study Aims and Analysis Approach)
Reliability and Validity of Family Reports of Alcoholism 57

-Twins are from complete and incomplete pairs
-AD=Alcohol Dependence; AA/AD=Alcohol Abuse or Dependence
-DSM-IV= Diagnostic and Statistical Manual of Mental Disorders Fourth edition (American Psychiatric Association, 1994); DSM-IIIR= Diagnostic and Statistical
Manual of Mental Disorders Third edition, Revised (American Psychiatric Association, 1987)
-FH-RDC= family history reports of alcohol-related problems based on Family History Research Diagnostic Criteria (Endicott, Andreasen, and Spitzer, 1978)
Study Wave Participants Measures Variables used in analyses
FF1 and FF1+ Twins from
female-female pairs
- Family history reports: Family History
Research Diagnostic Criteria (FH-RDC)
- FH-RDC Alcoholism Diagnosis and Symptom Count,
based on twin reports of parents (i.e., offspring
informant/parent target)

FF4 Twins from
female-female pairs
-Own diagnosis: Structured diagnostic
interview adapted from the Structured
Clinical Interview for DSM Disorders
-Twin’s own alcohol-related problem status (lifetime):
DSM-IV AD, DSM-IV AA/AD, and Problem Drinking

MF2 Twins from
male-male and
male-female pairs
- Family history reports: Family History
Research Diagnostic Criteria (FH-RDC)

-Own diagnosis: Structured diagnostic
interview adapted from the Structured
Clinical Interview for DSM Disorders
-FH-RDC Alcoholism Diagnosis and Symptom Count,
based on twin reports of parents (i.e., offspring
informant/parent target)

-FH-RDC Alcoholism Diagnosis, based on twin reports of
cotwin (i.e., offspring informant/sibling target)

-Twin’s own alcohol-related problem status (lifetime):
DSM-IV AD, DSM-IV AA/AD, and Problem Drinking

FFP Mothers and fathers
of female-female
twins
- Family history reports: Family History
Research Diagnostic Criteria (FH-RDC)

-Own diagnosis: Structured diagnostic
interview adapted from the Structured
Clinical Interview for DSM Disorders
- FH-RDC Alcoholism Diagnosis reported by parents on
twin offspring
(i.e., parent informant/offspring target)

-Parent’s own alcohol-related problem status (lifetime):
DSM-IIIR AD, DSM-IIIR AA/AD, DSM-IVAD, and
Problem Drinking

Table 2: Study participants, measures, and variables
Reliability and Validity of Family Reports of Alcoholism 58

Table 3: Twin offspring samples sizes, by gender and zygosity

Gender and pair
type
Number of
individuals
a

Number of
Complete Pairs
All females
b
3279
All males
b

4138
Female-Female
(FF)
1920 821
c

MZ=495
DZ=323
Male-Male
(MM)
2870 1162
MZ=695
DZ=467
Male-Female
(MF)
2627 1019

a
Includes individuals from complete and incomplete pairs.
b
“All male/female” categories based on male and female offspring across the different pair types (FF, MM, MF)
c
3 female-female pairs have unknown zygosity and were not included in the MZ or DZ complete pair counts

Reliability and Validity of Family Reports of Alcoholism 59

Table 4: Lifetime prevalences of alcohol-related problems in adult male and female offspring

a) Lifetime prevalences of alcohol-related problems in adult offspring based on personal interviews

b) Lifetime prevalences of alcohol-related problems in adult offspring based on family history reports by
cotwins

FH-RDC Alcoholism
b

Target Informant Alcohol Symptom Count Alcoholism Diagnosis
Male Male cotwin
(N=2360)
0.28 (0.89) 12.2%
Female cotwin
(N=1034)
0.06 (0.44) 2.8%
Female Male cotwin
(N=1033)
0.38 (1.06) 15.2%

a
All female categories include females from FF and MF pairs; All male categories include males in MM and MF
pairs.
b
Family history reports of parent alcoholism based on Family History Research Diagnostic Criteria (FH-RDC)
for alcoholism (Endicott, Andreasen, and Spitzer, 1978).

Type of alcohol-
related problem:
Overall
Sample
(N=7417)
All
Females
a

(N=3279)
Females
from FF
pairs
(N=1920)
Females
from MF
pairs
(N=1357)
All
Males
a
(N=4138)
Males
from MM
pairs
(N=2870)
Males
from MF
pairs
(N=1268)
DSM-IV Alcohol
Dependence (AD)
18.5% 9.2% 6.6% 12.8% 25.9% 24.9% 25.9%
DSM-IV Alcohol
Abuse or Dependence
(AA/AD)
30.0% 15.8% 13.4% 19.2% 41.4% 39.6% 41.4%
Problem Drinking 41.4% 27.6% 25.5% 30.9% 52.3% 50.2% 52.5%
Reliability and Validity of Family Reports of Alcoholism 60

Table 5: Lifetime prevalences of alcohol-related problems in parents

a) Lifetime prevalences of alcohol-related problems in parents
a
based on personal interviews

b) Lifetime prevalences of alcohol-related problems in parent based on family history reports by adult
offspring
b

a
Based on parents of FF twin pairs who completed FFP interviews
b
Family history reports of parent alcoholism based on Family History Research Diagnostic Criteria (FH-RDC) for
alcoholism (Endicott, Andreasen, and Spitzer, 1978).
Mothers
N=852
Fathers
N=615
Prevalence (%)
DSM-IIIR Dependence 5.3% 18.2%
DSM-IV Dependence 4.7% 15.9%
DSM-IIIR Abuse or Dependence 5.4% 19.4%
Problem Drinking 9.5% 32.9%
Parent FH-RDC Alcoholism

Family history reports from all
offspring in sample
Family history reports from female
offspring whose parents provided
personal interviews
Mothers
(N=7365 twins)
Fathers
(N=7235 twins)
Mother
(N=1394 twins)
Father
(N=1027 twins)
“Alcoholism “ diagnosis 4.9% 20.9% 6.0% 16.7%
Symptom count
[M(SD)]
0.10 (0.54) 0.51 (1.18) 0.12 (0.56) 0.37 (0.99)
Reliability and Validity of Family Reports of Alcoholism 61

Table 6: Within-pair twin agreement on parents’ FH-RDC Alcoholism symptom count and diagnosis, by
twin gender and zygosity

- κ= kappa coefficient
- ρ= polychoric correlation coefficient
-CI=confidence interval
-Alphabetical superscripts - significant informant gender differences, p<.05
-Numeric superscripts - significant target gender differences, p<.05
-No significant differences based on informant zygosity

Parental Symptom Count
Based on Twin Report
Parental Alcoholism Diagnosis Based
on Twin Report
Mother
ρ (95% CI)
Father
ρ (95% CI)
Mother Father
κ (95% CI) κ (95% CI)
All female pairs
(821 pairs)
0.89
(0.84-0.95)
0.86
a
(0.82-0.90)
0.61
b
(0.50-0.72)
0.66
c
( 0.60-0.72)
MZF
(495 pairs)
0.91
(0.85-0.97)
0.86
(0.81-0.91)
0.55
( 0.39-0.72)
0.65
( 0.57-0.74)
DZF
(323 pairs)
0.87
(0.79-0.96)
0.86
(0.80-0.91)
0.67
( 0.51- 0.82)
0.67
( 0.57-0.76)

All male pairs
(1162 pairs)
0.89
(0.83-0.95)
0.84
a
(0.80-0.87)
0.63
(0.51-0.75)
0.65
( 0.59- 0.71)
MZM
(695 pairs)
0.88
(0.79-0.97)
0.86
(0.81-0.90)
0.64
(0.48-0.80)
0.69
( 0.62- 0.76)
DZM
(467 pairs)
0.89
(0.80-0.98)
0.81
(0.74-0.87)
0.62
(0.43-0.80)
0.60
( 0.50- 0.69)

Opposite sex pairs

(1019 pairs)
0.91
1
(0.87-0.96)
0.78
a, 1
(0.73-0.83)
0.74
b, 2
( 0.64-0.84)
0.59
c, 2
( 0.52-0.65)
Reliability and Validity of Family Reports of Alcoholism 62

Table 7: FH-RDC Alcoholism diagnosis prevalences and symptom count means, by gender of offspring
informants and parent targets

Target
(Parent)
Informant
Gender
Alcoholism Symptom
Count
M (SD)
t P Alcoholism
Diagnosis
(%)
Chi
Square
p
Mother Male
(N=4102)
0.09 (0.49)
-2.8 0.005
4.1%
10.9 0.001
Female
(N=3263)
0.12 (0.59) 5.8%
Father Male
(N=4011)
0.48(1.16)
-2.0 0.044
19.9%
5.4 0.02
Female
(N=3224)
0.54(1.20) 22.2%
Reliability and Validity of Family Reports of Alcoholism 63

Table 8: Odds of FH-RDC Alcoholism diagnosis in parent targets based on family history reports by affected and unaffected offspring informants

a
“Informant diagnostic status”-- whether informant is affected or unaffected for specified diagnosis/problem
b
“Prevalence” is target prevalence of alcohol-related problems based on informant family history report
-Different diagnostic definitions were applied to all participants in these analyses—presence or absence of diagnoses were coded for each
definition based on participant’s diagnostic status.
Mother Alcoholism Diagnosis Mother One Month
Problem
Father Alcoholism Diagnosis Father One Month
Problem
Informant diagnostic
status
a

N
informant
Prevalence
b

Odds Ratio
(95% CI)
Prevalence Odds Ratio
(95% CI)
N
informant
Prevalence Odds Ratio
(95% CI)
Prevalence Odds Ratio
(95% CI)
Alcohol
Dependence
Affected

1355 8.7% 2.28
(1.81- 2.86)
10.7% 2.10
(1.71-2.58)
1324 30.2% 1.86
(1.63-2.13)
35.7% 1.79
(1.58-2.04)
Unaffected

6009 4.0% 5.4% 5910 18.8% 23.6%
Abuse or
dependence
Affected

2208 7.4% 2.04
(1.65 -2.53)
9.6% 2.00
(1.66-2.42)
2161 27.1% 1.66
(1.48-1.87)
33.1% 1.67
(1.50-1.87)
Unaffected

5156 3.8% 5.0% 5073 18.3% 22.8%
Problem
drinker
Affected

3041 7.0% 2.20
(1.77-2.73)
9.6% 2.31
(1.92-2.80)
2975 26.5% 1.76
(1.57-1.97)
32.3% 1.70
(1.53-1.90)
Unaffected 4323 3.3% 4.4% 4259 17.0% 21.9%
Reliability and Validity of Family Reports of Alcoholism 64

Table 9: Within-pair agreement on parent FH-RDC Alcoholism symptom count and diagnosis, by twin pair alcohol problem concordance and discordance
-Alphabetical superscripts – significant differences (p<.05)
a
Number of twin pairs reporting on mothers and fathers within each diagnostic category
-Different diagnostic definitions were applied to all participants in these analyses—presence or absence of diagnoses were coded for each definition based on participant’s
diagnostic status
Twin concordance/discordance for
specific diagnosis
Alcoholism symptom count for parent target Alcoholism Diagnosis for parent target
Mother Father Mother Father
M(SD) ρ (95% CI)
a
M(SD) ρ (95% CI )
a
Prevalence
(%)
κ(95% CI ) Prevalence
(%)
κ(95% CI )
Alcohol Dependence vs No Diagnosis
Both AD

a
(N=187 for mothers; N=183 for fathers)
0.26
(0.81)
0.89
(0.81-0.97)
0.84
(1.47)
0.83
(0.75-0.90)
12.4% 0.73
(0.58-0.88)
33.1% 0.67
(0.55-0.78)
Discordant
(N=681 for mothers; N=651for fathers)

1
0.15
(0.67) 0.92
(0.88-0.97)
1
0.65
(1.3) 0.77
a

(0.72-0.83)
1 6.5%
0.72
e

(0.61-0.83)
1 27.6%
0.60
(0.53-0.67)
0
0.12
(0.61)
0
0.59
(1.27)
0 5.7% 0 24.4%
Neither AD
(N=2100 for mothers; N=2059 for
fathers)
0.07
(0.43)
0.88
(0.84-0.93)
0.42
(1.09)
0.84
a

(0.81-0.87)
3.7% 0.61
e

(0.51- 0.70)
17.5% 0.63
(0.59- 0.68)
Abuse or Dependence vs No Diagnosis
Both AA/AD
(N=402 for mothers; N=389 for fathers)
0.21
(0.74)
0.87
(0.80-0.94)
0.69
(1.34)
0.84
b

(0.79-0.89)
9.4% 0.62
f

(0.49-0.76)
28.0% 0.69
h

(0.61- 0.77)
Discordant
(N=932 for mothers; N=896 for fathers)

1
0.13
(0.61) 0.93
(0.89-0.97)
1
0.57
(1.26) 0.78
b,c

(0.74-0.83)
1 5.3%
0.76
f,g

(0.66-0.85)
1 23.0%
0.59
h

(0.52-0.65)
0
0.12
(0.57)
0
0.59
(1.26)
0 5.4% 0 24.1%
Neither AA/AD
(N=1634 for mothers; N=1608 for
fathers)
0.06
(0.39)
0.88
(0.83-0.94)
0.40
(1.05)
0.85
c

(0.81-0.88)
3.4% 0.59
g

(0.48-0.70)
16.7% 0.64
(0.59-0.69)
Problem Drinking vs No problem Drinking
Both Problem Drinkers
(N=660 for mothers; N=633 for fathers)
0.20
(0.76)
0.92
(0.88-0.96)
0.70
(1.36)
0.83
(0.78-0.88)
8.6% 0.71
(0.60-0.82)
28.3% 0.67
i

(0.60-0.74)
Discordant
(N=1099 for mothers; N=1070 for
fathers)

1
0.11
(0.52) 0.87
(0.81-0.93)
1
0.57
(1.24) 0.80
d

(0.75-0.84)
1 5.5%
0.64
(0.53-0.74)
1 23.5%
0.60
i

(0.54-0.65)
0
0.10
(0.49)
0
0.54
(1.20)
0 5.5% 0 22.1%
Neither Problem Drinkers
(N=1209 for mothers; N=1190 for
fathers)
0.05
(0.37)
0.90
(0.84-0.96)
0.35
(1.00)
0.85
d

(0.81-0.88)
2.7% 0.63
(0.49-0.76)
14.8% 0.63
(0.57- 0.69)
Reliability and Validity of Family Reports of Alcoholism 65

Table 10: Validity of parent FH-RDC Alcoholism based on offspring informant reports compared to personal interview of parent target, by target
gender

-Effective sample size calculated to account for correlated observations attributable to twins reporting on the same parents. Effective sample size for twin reports
on mothers is N=823 and fathers is N=613.
-Statistically significant differences between parent gender (within diagnostic definitions) are identified with numeric superscripts and significant differences in
diagnostic definitions (within parent gender) are identified by alphabetical superscripts.
Informant FH-
RDC Alcoholism
Diagnosis reports
versus:
Mothers (N=823) Fathers (N=613)
Sensitivity (%) Specificity (%) PPV (%) NPV (%) Sensitivity (%) Specificity (%) PPV (%) NPV (%)
Parent DSM IIIR
AD
39.5
(25.7-54.7)
96.0
1

(94.3-97.2)
36.1
a

(23.3-50.7)
96.5
7,d

(95.0-97.7)
44.8
(35.9-54.1)
89.8
1,h

(86.9-92.2)
50.6
k

(40.8-60.3)
87.5
7,n

(84.3-90.2)
Parent DSM IV
AD
44.8
(29.4-60.9)
96.0
2

(94.5-97.3)
36.1
b

(23.3-50.7)
97.2
8,e

(95.8-98.2)
47.4
g

(37.6-57.2)
89.4
2,i

(86.5-91.8)
47.1
l

(37.4-56.9)
89.5
8,o,q

(86.6-91.9)
Parent DSM IIIR
AA/AD
39.5
(25.7-54.7)
96.0
3

(94.3-97.2)
36.1
5,c

(23.3-50.7)
96.5
9,f

(95.0-97.7)
44.0
(35.2-52.9)
90.1
3,j

(87.2-92.5)
52.9
5,m

(43.1-62.6)
86.4
9,p,q

(83.2-89.2)
Parent Problem
Drinking
34.3
(24.3-45.6)
97.1
4

(95.6-98.1)
55.4
6,a,b,c

(40.9-69.2)
93.3
10,d,e,f

(91.4-95.0)
36.9
g

(30.4-43.8)
93.2
4,h,i,j

(90.4-95.4)
72.7
6,k,l,m

(63.3-80.7)
75.0
10,n,o,p

(71.1-78.6)
Reliability and Validity of Family Reports of Alcoholism 66

Table 11: Validity of sibling FH-RDC Alcoholism based on offspring informant reports
compared to personal interview with sibling targets, by gender of target and informant
Offspring informant FH-RDC
Alcoholism Diagnosis reports
versus:
Sensitivity(%) Specificity(%) PPV(%) NPV(%)
Male informant - Male target
(N=2352 complete pairs)
Sibling Alcohol Dependence 32.5
(28.7-36.5)
94.3
(93.1-95.3)
64.6
(58.7-70.1)
81.3
(79.5-82.9)
Sibling Alcohol Dependence
or Abuse (AA/AD)
25.3
(22.6-28.3)
96.0
(94.9-97.0)
80.2
(75.0-84.6)
67.0
(64.9-69.0)
Sibling Problem Drinking 21.7
(19.4-24.2)
97.00
(95.8-97.8)
87.5
(83.0-91.0)
55.9
(53.7-58.0)
Male informant - Female Target
(N=1027 complete pairs)
Sibling Alcohol Dependence 10.4
(5.9-17.5)
98.3
(97.2-99.0)
46.4
(28.0-65.8)
88.8
(86.6-90.6)
Sibling Alcohol Dependence
or Abuse (AA/AD)
8.4
(5.0-13.5)
98.6
(97.4-99.2)
57.1
(37.4-75.0)
82.5
(80.0-84.8)
Sibling Problem Drinking 6.3
(4.0-9.7)
98.9
(97.7-99.5)
71.4
(51.1-86.1)
70.1
(67.1-72.9)
Female informant - Male Target
(N=1025 complete pairs)
Sibling Alcohol Dependence 39.6
(33.8-45.8)
93.5
(91.5-95.1)
68.6
(60.6-75.7)
81.2
(78.5-83.8)
Sibling Alcohol Dependence
or Abuse (AA/AD)
30.0
(25.8-34.6)
95.9
(93.9-97.3)
84.6
(77.8-89.7)
64.6
(61.3-67.7)
Sibling Problem Drinking 24.9
(21.4-28.7)
96.7
(94.5-98.1)
90.4
(84.4-94.3)
51.0
(47.6-54.4)
Reliability and Validity of Family Reports of Alcoholism 67

Table 12: Validity of parent FH-RDC Alcoholism based on offspring informant reports compared to personal
interview of parent target, by informant alcohol-related problem status

-Matching alphabetical superscripts – significant differences (p<.05)
-Effective sample size for twin reports on mothers is N=823 and fathers is N=613.

Mother (N=823) Father (N=613)
PPV(%) NPV(%) PPV(%) NPV(%)
Target DSM IV Alcohol Dependence
Overall validity 36.1
(23.3-50.7)
97.2
(95.8-98.2)
47.1
(37.4-56.9)
89.5
(86.6-91.9)
Informant DSM-IV
AD status
No AD 39.3
(24.3-56.1)
97.2
(95.8-98.3)
45.3
(35.1-55.8)
89.9
(87.0-92.3)
AD 27.3
(10.6-51.9)
96.0
(87.5-99.1)
58.3
(35.9-78.1)
81.4
(65.3-91.5)
Informant problem
drinking status
No PD 36.1
(18.2-57.6)
97.4
(95.9-98.4)
44.4
(32.5-56.9)
90.5
(87.4-92.9)
PD 36.2
(21.6-53.2)
96.5
(93.1-98.4)
50.7
(37.6-63.7)
85.6
(78.7-90.8)
Target Problem Drinking
Overall Validity 55.4
(40.9-69.2)
93.3
(91.4-95.0)
72.7
(63.3-80.7)
75.0
(71.1-78.6)
Informant DSM-IV
AD status
No AD 60.7
(43.9-75.7)
93.5
(91.6-95.1)
71.6
(61.5-80.3)
75.4
(71.4-79.0)
AD 40.9
(20.3-64.7)
89.3
(78.9-95.3)
79.2
(56.2-92.7)
67.4
(50.6-81.1)
Informant problem
drinking status
No PD 55.6
(34.6-75.1)
94.1
a

(92.0-95.7)
69.7
(57.4-80.1)
76.7
b

(72.5-80.4)
PD 55.3
(38.6-71.1)
90.5
a

(85.8-93.8)
76.7
(63.9-86.4)
68.4
b

(60.1-75.8)
Reliability and Validity of Family Reports of Alcoholism 68

Table 13: Validity of sibling FH-RDC Alcoholism based on offspring informant reports compared to personal
interview with sibling targets, by informant alcohol-related problem status

-Matching alphabetical superscripts – significant differences (p<.05)

PPV(%) NPV(%)
Target DSM IV Alcohol Dependence
Overall validity 64.8
(60.3-69.1)
83.2
(82.0-84.3)
Informant DSM-IV AD
status
No AD 59.7
(53.4-65.7)
85.6
a

(84.3-86.8)
AD 71.0
(64.4-76.9)
72.8
a

(69.5-76.0)
Informant problem
drinking status
No PD 58.7
(50.5-66.5)
87.6
b

(86.1-88.9)
PD 67.8
(62.3-72.9)
77.5
b

(75.5-79.5)
Target Problem Drinking
Overall validity 87.5
(84.1-90.3)
58.4
(56.8-59.9)
Informant AD diagnostic
status
No AD 84.9
(79.8-88.9)
61.5
c

(59.7-63.2)
AD 90.7
(85.7-94.1)
45.4
c

(41.8-49.1)
Informant problem
drinking status
No PD 85.2
(78.4-90.2)
64.8
d

(62.8-66.8)
PD 88.6
(84.5-91.8)
50.2
d

(47.8-52.6)
Reliability and Validity of Family Reports of Alcoholism 69

Table 14: Validity of parent FH-RDC Alcoholism based on single vs multiple
a
offspring informant reports
compared to parent personal interview

a
Multiple informant family history reports combined based on an “Or” rule: if either informant report was
consistent with an endorsement of parent FH-RDC Alcoholism diagnosis, then the parent target was coded as
having Alcoholism.
Validity Index Mother (N=823) Father (N=613)
One Informant Two
Informants
One Informant Two
Informants
Sensitivity(%) 39.5 48.5 44.0 53.2
Specificity(%) 96.0 93.9 90.1 86.5
PPV(%) 36.1 30.8 52.9 50.0
NPV(%) 96.5 97.0 86.4 87.9
Reliability and Validity of Family Reports of Alcoholism 70

Table 15: Validity of parent FH-RDC Alcoholism reported by offspring informants based on different methods
of consolidating multiple informants responses at different diagnostic thresholds

- Bold values show findings based on FH-RDC diagnostic threshold of one symptom

Mothers (N=823) Fathers (N=613)
FH-RDC
Symptom
Cutoff
Sensitivity
(%)
Specificity
(%)
PPV
(%)
NPV
(%)
Sensitivity
(%)
Specificity
(%)
PPV
(%)
NPV
(%)
Multiple offspring informants
Min 1 27.3 97.6 39.1 96.0 34.0 95.1 64 85.0
2 18.3 99.0 50.0 95.6 21.3 98.1 74.2 83.1
3 9.2 99.7 60.0 95.1 13.8 98.7 72.2 81.8
4 3.0 100 100 94.8 4.3 100 100 80.4
5 3.0 100 100 94.8 0 100 100 79.7
6 3.0 100 100 94.8 0 100 100 79.7

Max
1 48.5 93.9 30.8 97.0 53.2 86.5 50.0 87.9
2 39.4 96.8 40.6 96.6 34.0 92.7 54.2 84.7
3 15.2 98.8 41.7 95.4 25.5 96.5 64.9 83.6
4 9.1 99.7 60 95.1 19.2 98.1 72.0 82.7
5 6.1 100 100 95.0 12.8 99.2 80.0 81.7
6 3.0 100 100 94.8 0 100 100 79.7
Mean 1 39.4 96.1 36.1 96.6 38.3 92.2 55.4 85.5
2 21.2 99.0 53.9 95.7 27.7 96.5 66.7 84.0
3 12.1 99.7 66.7 95.3 17.0 98.7 76.2 82.4
4 3.0 100 100 94.8 9.6 99.2 75.0 81.2
5 3.0 100 100 94.8 0 100 100 79.7
6 3.0 100 100 94.8 0 100 100 79.7
Single informant
1 39.5 96.0 36.1 96.5 43.8 90.1 52.9 86.5
2 30.3 98.0 46.0 96.1 28.9 95.1 60.0 84.1
3 13.2 99.2 50.0 95.2 21.6 97.7 70.3 83.1
4 7.9 99.8 66.7 95.0 11.5 99.3 77.4 81.6
5 5.3 100 100 94.8 6.3 99.6 81.3 80.8
6 4.0 100 100 94.8 0 100 100 79.8
Reliability and Validity of Family Reports of Alcoholism 71

Table 16: Differences in validity of family history reports by informant-target family relationship

-Matching alphabetical superscripts – significant differences (p<.05)
-Significant gender differences within pairings (e.g.,offspring reporting on fathers versus mothers) are not identified in this table, as only significant
differences in family relation are discussed. Refer to Tables 10 and 11 for informant/target gender differences in reporting.

Target Informant Target prevalence
based on FH-RDC
Dx from informant
(0/1); % affected
Target
prevalence
based on target
interview (0/1);
% affected
Sensitivity
(%)
Specificity
(%)
PPV
(%)
NPV
(%)
Parents’ reports of offspring
Offspring
Mother 0 1347
2.0%
1188
13.5%
11.3
a

(7.3-17.0)
99.5
e

(98.9-99.8)
77.8
i

(57.3-90.6)
87.8
m

(85.9-89.4)
1 27 186
Father
0
992
1.3%
866
13.8%
6.5
b

(3.2-12.3)
99.5
f

(98.7-99.9)
69.2
j

(38.9-89.6)
86.9
n

(84.6-88.9)
1 13 139
Offspring’s reports of parents
Mother
Offspring
0 1309
6.0%
1316
5.5%
39.5
a,b,c

(28.7-51.4)
96.0
e,f

(94.73-96.94)
36.1
i,j, k

(26.1-47.5)
96.5
m,n,o

(95.3-97.3)
1 83 76
Father 0 855 16.8% 820 20.2% 44.0
a,b,d

(37.1-51.0)
90.1
e,f, g

(87.8-92.0)
52.9
i, l

(45.2-60.5)
86.4
p

(83.9-88.6)
1 172 207
Twin’s reports of cotwins
Sibling Offspring
0 3932
10.7%
2861
35.0%
24.6
a,b,c,d

(22.5-26.8)
96.8
e,f,g

(96.0-97.4)
80.3
k,l

(76.4-83.7)
70.4
m,n,o,p

(68.9-71.8)
1 472 1543
Reliability and Validity of Family Rep
Figure 1: Female-female study design

--Figure from: Kendler, K. S., & Prescott, C. A. (2006).
The Guilford Press.
-FF-- Female-female twin study
-Numbers next to FF (1 through 4)-- study waves
-FF1+ --
-IV--Interview
-FFP--Female-female parent study
Reliability and Validity of Family Reports of Alcoholism
esign
Kendler, K. S., & Prescott, C. A. (2006). Genes, environment, and psychopathology
study waves

72

Genes, environment, and psychopathology. New York:
Reliability and Validity of Family Rep
Figure 2: Male-male/Male-female study d

-Figure from: Kendler, K. S., & Prescott, C. A. (2006).
The Guilford Press.
-MM/MF – Male-Male/Male-Female study
-Numbers after MM/MF study (1 and 2)
-IV--Interview
Reliability and Validity of Family Reports of Alcoholism
female study design

Kendler, K. S., & Prescott, C. A. (2006). Genes, environment, and psychopathology
Female study
Numbers after MM/MF study (1 and 2)—study waves
73

Genes, environment, and psychopathology. New York:
Reliability and Validity of Family Reports of Alcoholism 74

Figure 3: Validity of parent FH-RDC Alcoholism based on single vs multiple
a
offspring informant
reports compared to parent personal interview, by parent gender
a) Mothers

b) Fathers

a
Multiple informant information combined based on an “Or” rule: if either informant report was consistent with
an endorsement of parent FH-RDC Alcoholism diagnosis, then the parent target was coded as having Alcoholism.
0
20
40
60
80
100
Sensitivity Specificity PPV NPV
One Informant
Two Informants
0
20
40
60
80
100
Sensitivity Specificity PPV NPV
One Informant
Two Informants
Reliability and Validity of Family Reports of Alcoholism 75

Figure 4: Multiple informants with varying methods of consolidating responses at different FH-RDC diagnostic thresholds—Mothers

a) Sensitivity
b) Specificity
c) Positive Predictive Value d) Negative Predictive Value
Reliability and Validity of Family Reports of Alcoholism 76

Figure 5: Multiple informants with varying methods of consolidating responses at different FH-RDC diagnostic thresholds—Fathers

a) Sensitivity b) Specificity
c) Positive Predictive Value d) Negative Predictive Value
Reliability and Validity of Family Reports of Alcoholism 77

Authors Reliability
or Validity
Study
Sample description Outcomes of interest Measurement/Definition of
Alcoholism used
Results Conclusions
Amodeo & Griffin
(2009)
Reliability N=143 pairs of
female siblings;
age 21-60

Sibling reports of maternal
and paternal alcoholism.
Agreement between sibling
reports assessed.
Measure used: Children of
Alcoholics Screening Test
(30 items). Assesses child’s
lifetime experience of
alcohol problems in parents.
Score ≥6 indicate parent is
more than likely an
alcoholic. Agreement on
individual items re whether
or not sib believed parent
was alcoholic was also
assessed.
Based on CAST score: Sib
agreement for mothers: r=0.80,
p<.01;
Sib agreement for fathers: r=0.79,
p<.001). Agreement for selected
individual items ranged 84.4%-
89.4% for fathers, 89.2%-90.7%
for mothers.
Siblings
demonstrate high
agreement on
reports related to
their parents
having problems
with alcohol.
Chapman et al.,
1994
Validity Informants and
targets N=2193
related individuals
who had
alcoholism, had
some diagnosis
other than
alcoholism, or did
not have any
diagnosis
Sensitivity and specificity of
alcoholism reports from
informant-target family
pairs (non-twin).
Measure used:
Target reports on self—
alcohol data collected using
Schedule for Affective
Disorders and
Schizophrenia (used
Research Diagnostic
Criteria to diagnose
alcoholism).

Informant family history
reports—Family Informant
Schedule and Criteria
(similar to FH-RDC)
Sensitivity was significantly
higher among informants with
alcoholism compared to
informants with no dx of any
kind(alcoholism affected: 69.9%;
no dx: 33.3%). Specificity was
significantly lower only among
informants with alcoholism
compared to those with a different
(non-alcohol) diagnosis, though
differences in effect sizes are not
large (alcoholism affected:
95.0%; other dx: 98%).
Informants’
family history
reports may be
associated with
their own
alcoholism
status.
Crews & Sher
(1992)
Reliability;
Validity
Informants were
N= 216 sibling
pairs (male-male,
male-female, and
female-female).
Targets were their
mothers and
fathers.
Reliability: Sibling
agreement on mother and
father alcoholism based on
sibling responses to F-
MAST and M-MAST.
Validity: Sensitivity and
specificity of sibling reports
compared to F-MAST and
M-MAST completed by
fathers and mothers.
Measures used: Short
Michigan Alcoholism
Screening Test (SMAST),
adapted to separately assess
Fathers (F-MAST) and
Mothers (M-MAST). These
are 13 item measures
(yes/no response) asking a
variety of questions related
to parent’s alcohol use
(intended to assess
“alcoholism.” F-MAST and
M-MAST were completed
by siblings for reliability
analyses, and by parents for
Different thresholds for
alcoholism diagnosis (0/1) were
used to test variation in sibling
agreement. Kappa values
generally show moderate to
strong sibling agreement for
maternal and paternal MAST
scores (particularly for cutoff
score of 5). Validity analyses
conducted separately for different
score thresholds indicate good
sensitivity and specificity for
parents (with some exceptions for
mothers). Validity findings were
as follows: Fathers—Sens=(0.77-
Different
alcoholism
diagnostic
thresholds
present different
levels of
agreement and
validity.
Maternal validity
(sens and spec)
was lower than
fathers, possibly
due to low
maternal base
rate of
Appendix A: Literature Review
Reliability and Validity of Family Reports of Alcoholism 78

validity analyses. Tested
agreement and validity
based on different cut off
scores (e.g., score ≥3, score
≥4, etc).
0.94), Spec=(0.77 to 0.94);
Mothers—Sens=(0.18-0.68);
Spec=0.78-1.00).
alcoholism.
Heun & Müller
(1998)
Reliability Targets were
N=1306 first
degree relatives
and spouses of
patients and control
subjects from an
existing study.
Informants were
two relatives of
targets .
Family agreement (from
two informant relatives) on
psychiatric problems of
targets including FH-RDC
alcoholism
Measure used: Family
History Research Diagnostic
Criteria—used to collect
family history data from
informants on Alcoholism
and other psychiatric
disorders in targets.
Interinformant reliability for
alcoholism was: k=0.41; CI=0.14-
0.38
Reliability of
family history of
alcoholism was
moderate.
Kendler et al.
(1991)
Reliability Informants were
twin pairs from the
Virginia Twin
Register who were
discordant for
DSM IIIR Alcohol
Dependence. 86
twin informants
discordant for AD
reported on
mothers. 85
discordant AD
twins reported on
fathers.
Assessed twin agreement
for parental alcoholism in
twins discordant for alcohol
dependence (based on
DSM-IIIR criteria).
Measure used: Family
History Research Diagnostic
Criteria--used to collect
family history reports from
twins on maternal and
paternal alcoholism and
other psychiatric problems.
DSM IIIR criteria used to
assess AD in twins.
There were not significant
differences in likelihood of
reporting parental FH-RDC
Alcoholism based on twin’s own
affected/unaffected status
(Mothers: χ2=0.17; p=0.68;
Fathers: χ2=0.75; p=0.39).
Twin’s alcohol-
related problems
were not
associated with
how they
reported parent’s
alcoholism.
Kosten et al.
(1992)
Validity N=341 subjects
from a larger study
of probands who
were seeking
treatment for opiate
addition. Sample
composed of 107
probands and 234
of their family
members
Compared family history
reports from informants
with direct interviews
conducted with targets.
Direct interview and family
history interviews
conducted with all subjects.
Alcohol-related problems
measured as follows:
-Direct interview of targets:
Collected data with
Schedule for Affective
Disorders and
Schizophrenia (SADS) and
reached diagnosis based on
Research Diagnostic
Criteria (RDC).
-Family history data: Also
collected with SADS and
used family history Family
History Research Diagnostic
Criteria (FH-RDC) to reach
-Overall, specificity was high but
sensitivity was low.
--Validity based on:
-Gender of informant: Female
informants had higher sensitivity
(39%) than male informants
(21%). Specificity was similar for
both genders.
-Relationship of informant:
Offspring informant had
significantly higher sensitivity
(60%) than parent (28%), sibling
(30%), or spouse (21%)
informants. No differences in
specificity.
Family history
reports have low
sensitivity and
high specificity.
Informant
gender,
relationship of
informant to
target, and
multiple
informants are
associated with
differences in
sensitivity levels
but not
Reliability and Validity of Family Reports of Alcoholism 79

diagnosis. -Multiple informants (one vs more
than one): Significantly higher
sensitivity when more than one
informant (61%) vs only one
informant (18%). No significant
differences in specificity.
specificity levels.
Milne et al. (2009) Validity N=959 adult child
probands of
Dunedin Study;
N=933 maternal
guardians and
N=850 paternal
guardians of
probands
Mother and/or father reports
of alcohol dependence in
their adult children. Validity
of parent reports assessed by
comparing to adult child’s
self-reports.
-Family hx measure:
Family-History Screen.
-Proband diagnosis: DSM-
III or DSM IV current
alcohol dependence
assessed during three data
collection waves data
collection (at ages 21, 26, or
32). Proband considered to
have alcohol dependence if
it was diagnosed during any
of these waves.
-Validity findings (were not
distinguished by mom/dad):
Sensitivity: 21.7%; Specificity:
94%. Female gender of informant
associated with reporting of
reporting alcohol problems in
proband (OR: 1.59[1.04-2.44]).
No differences in accuracy of
informant reports based on
informant gender. No sig
association of informant’s own
history of alcohol dependence
with their FH reports of proband
alcoholism.
Sensitivity of
informant
identification of
alcohol problems
in probands is
low, while
specificity is
high. Female
informants more
likely to report
that proband has
an alcohol
problem.
O’Malley et al.
(1986)
Validity Informants: N=49
college students.
Target: Informants’
mothers and
fathers.
Parents’ drinking practices
based on informant report
and target report.
-Measure was a
questionnaire (no specific
name/title) inquiring about
parents’ drinking behavior.
Analyses used responses to
single items pertaining to
different drinking domains
(1 item for each of the
following: quantity,
frequency, drinking more
heavily). Specific answer
choices were selected by
informants/targets (e.g.,
Average frequency of
alcohol consumption:
(answer choices) (1)rarely,
(2) about once a month, (3)
2 to 3 times a month, (4) 1
or 2 times a week, etc.
-Validity findings: Correlations
between informant and target
responses for quantity and
frequency ranged from 0.37 to
0.62 for the different maternal
and paternal drinking behaviors.
Validity assessed by comparing
informant and target responses to
individual items (total of 3 items
related to quantity, frequency, and
heaviest drinking used in
analyses). Analyses indicated %
of informants who selected the
same, lower, or higher answer
choice compared to parents. % of
informants who chose same
answer choice as parent was
moderate to high (39%-71%).
Percentages were lower for
informants who chose lesser
answer choice than targets (range:
17%-35%), and greater answer
choice (range: 4%-27%)
Correlation
between
informant and
target reports of
drinking
behavior were
moderate to
strong.
Rhea et al. (1993) Reliability Informants: N=142 Sibling agreement on Measure used: Informants -There was lower agreement for Same-sex
Reliability and Validity of Family Reports of Alcoholism 80

pairs of twin (MZ
and DZ) and
nontwin (biological
and non-biological)
siblings. Pairs
included female,
male, and opposite
sex siblings.
Target: Pair’s
mothers and
fathers.
reports of mother’s and
father’s alcohol-related
problems.
completed Family
Alcoholism History
Questionnaire (FHQ)—20
item measure assessing
alcohol use problems that
could predict development
of alcoholism (response
format: yes/no)
items that require inference (e.g,
how parent feels) compared to
observable behaviors. There were
gender differences in informant
agreement, with male offspring
agreed more about mothers’
drinking behavior, while female
offspring agreed more about
fathers’ drinking behavior. There
were no significant differences in
responses based on zygosity.
informants tend
to agree more on
reports of parent
that is opposite
gender.
Rice et al. (1995) Validity N=2654
individuals,
composed of
probands and their
family members
who were part of
The Collaborative
Study on the
Genetics of
Alcoholism.
Informant family history
reports on their relatives.
All subjects provided self-
reports and family history
reports, thus # of
informants per family
member varied. Because
family members were
reporting on each other,
each individual could be an
informant and a target.
-Measures used: reports of
own alcohol problems
measured by Semi-
Structured Assessment for
the Genetics of Alcoholism
(SSAGA). Family history
data collected using Family
History Assessment Module
(FHAM). Measures were
used to compare informant
and family history reports
indicative of DSM IIIR
Alcohol Dependence.
Sensitivity levels were low and
sensitivity levels were high
(values depended on cut offs).
Gender of target significant in
predicting history diagnosis, but
informant gender is not.
Informants with drinking
problems better at identifying
problems in family members than
unaffected informants.
Target gender,
cut off used, and
informant
diagnostic status
are factors
associated with
validity of family
history reports.
Roy et al. (1994) Validity N=1459 probands
and their family
members recruited
as part of the
Roscommon
Family Study.
Compare informants’ family
history reports to ‘Best
Estimate’ (BE) of the targets
actual alcohol-related
diagnostic status (BE based
on personal interview,
hospital records, and review
of family history
information). All
participants provided both
direct interviews and family
history information.
-Family history data
collected used Family
History Research Diagnostic
Criteria (FH-RDC).
-Target diagnosis
determined through BE
-Rates of false negatives was
increased by: young age of target,
if informant was male,
informant’s comorbid diagnosis
of nonaffective psychosis. False
negative rates were decreased by:
target’s history of previous
hospitalization. False positive risk
increased by: older age of target,
if target was a male, if informant
was a female, and informant had
alcoholism, nonaffective
psychosis.
Various
informant and
target
characteristics
are associated
with rates of
false-positives
and false
negatives when
comparing
family history
reports to

Searles et al.
(1993)
Gender
differences
in family
hx
reporting
Informants: N=427
men and N=607
women (college
age) reporting on
their mothers,
fathers, and other
biological relatives.
Differences between males
and females in endorsement
of alcohol abuse in their
first and second degree
relatives
Measures: Correlates of
Drinking Behavior Project
Screening Questionnaire—
assessed symptoms
considered indicative of
alcohol abuse (e.g., job loss,
arrest, health problems due
Percent of males and females (not
related to each other) reporting
alcohol abuse in their parents and
other 1
st
and 2
nd
degree relatives
were compared—no significant
gender differences were found.
No gender
differences in
rates of family
history of alcohol
abuse reported
among females
and males who
Reliability and Validity of Family Reports of Alcoholism 81

Participants were
not related to each
other.
to drinking) are not related to
each other.
Sher & Descutner
(1986)
Reliability Informants: N=88
siblings. Targets:
Fathers.
Sibling agreement on
reports of father’s
alcoholism.
Measures used: Short
Michigan Alcoholism
Screening Test (SMAST),
adapted to specifically
assess Fathers (F-MAST)--
13 item measures (yes/no
response) asking a variety
of questions related to
parent’s alcohol use. Tested
agreement based on
different cut off scores (e.g.,
score ≥3, score ≥4, etc).
-Informant agreement ranged
from κ=0.54 to 0.79 based on
different cut off scores, with
cutoff of ≥6 displaying strongest
informant agreement. For specific
F-MAST items: agreement was
lower for items requiring
inference (e.g., father’s feeling
guilty for drinking) and higher for
items related to observable
behavior (e.g., being arrested,
driving under influence).
Sibling
informants
display adequate
levels of inter-
rater reliability
for their reports
of father’s
alcoholism (as
defined by F-
MAST).
Informant
agreement can
vary based on
whether measure
items are related
to inference vs
observable
behaviors.
Slutske et al.
(1996)
Reliability Informants:
N=2657 twin pairs
(1,444 female-
female pairs, 626
male-male pairs,
and 587female-
male pairs).
Targets were
twins’ mothers and
fathers.
Twin pair agreement on
reports of parent drinking
problems.
Family history of alcohol-
related problems was
assessed by asking twins
single item: “Has drinking
ever caused your
father/mother to have
problems with health,
family, job or police, or
other problems?”
-Informant agreement findings
were as follows: agreement
(kappa) for female-female, male-
female, and male-male pairs
ranged from 0.61 to 0.70 for
fathers and 0.39 to 0.65 for
mothers (with lowest value of
0.39 for male- female twins’
maternal reports).
Agreement may
vary depending
on gender of
informants.
Single item
assessment of
family alcohol
problems was
able to yield
agreement values
comparable to
more detailed
assessments.
Smith et al. (1994) Validity N=74 mothers with
hx of
hospitalization for
alcohol problems;
N=164 offspring
-Determine whether certain
informant and offspring
characteristics are
associated with offspring
informants’ endorsement of
drinking problems in their
mothers.
-Family hx measure:
Offspring’s response to an
item regarding whether their
mother “ever had a drinking
problem” was used
-Target diagnosis: Mother’s
hospitalization for alcohol
problems (all mother in
study had been hospitalized
eight years prior to study).
-There were multiple significant
predictors of offspring informant
reporting, including severity of
mothers problems, informant
gender, mother’s history of
suicide attempts, and gender and
alcohol status of the informant
(female alcoholic offspring
underreported mother’s alcohol
problems).Combining multiple
-Multiple
informant and
target
characteristics
are associated
with informant
identifying
drinking
problems in
family history
Reliability and Validity of Family Reports of Alcoholism 82

offspring reports helped increase
sensitivity of informant reports.
reports. Multiple
informants help
improve validity.
Thompson et al.
(1982)
Validity N=696 subjects
from a large
family-genetic
study of affective
disorders. All
subjects provided
direct interview
and family history
reports
-Compare family history
reports to direct interview of
target.
-Measure for direct
interview of alcohol
problems: Schedule of
Affective Disorder and
Schizophrenia –Lifetime
version (SADS-L).
-Measure for family history
of alcohol problems: Asked
whether family members
(i.e., first degree relatives
and spouses) had any of the
problems discussed in direct
interview using SADS-L.
Asked additional SADS-L
diagnostic questions for any
possible family psychiatric
problems reported by
informant.
For alcoholism reports:
Sensitivity was generally low
(range 24%-57%) , but specificity
was high (96%-99%). Varied
based on relationship of family
member to informant—sensitivity
was highest for offspring
informant FH reports (57%),
followed by spouse (45%),
sibling (30%), and parent (24%);
specificity was similar across
relationships 96%-99%).
Sensitivity increased with
multiple informants using “or”
rule, but remained the same for
specificity.
Sensitivity levels
are low and
specificity levels
are high for
family history
reports of
alcoholism.
Multiple
informants can
help increase
sensitivity.
Weissman et al.
(2000)
Validity N=594 probands
and their relatives
-Test validity of proband’s
family history reports on
relatives against “best
estimate” (diagnosis of
relatives reached by
clinicians)
-Measures used--Family
History Screen: asks
questions regarding multiple
psychiatric problems in
different family members,
including drinking
problems. At least one
question was used to assess
each psychiatric problem.
Best estimate dx—clinicians
reviewed various measures
completed by participants
and reached a diagnosis
based on this information.
- For alcohol-related problems,
proband family history reports
demonstrated lower sensitivity
and PPV and higher specificity
and NPV. Validity values for
proband’s FHS reports for
relatives compared to best
estimates are as follows:
Sensitivity: 46.4 (10.1),
Specificity: 96.4 (1.1), Positive
Predictive Value: 56.5 (10.7),
Negative Predictive Value: 94.7
(1.4). There were no differences
in reports by proband or relative
age and sex for drinking
problems. Use of multiple
informants resulted in substantial
increases in sensitivity but not
specificity.
Family members
may be better
able to identify
unaffected rather
than affected
family members.
Findings do not
indicate
differences in
reports based on
target or
informant sex or
age.
Reliability and Validity of Family Reports of Alcoholism 83

Appendix B: Family History Research Diagnostic Criteria—Alcoholism*

Did (TARGET) ever have a problem with drinking that lasted at least a month
*

(DEFINE WORST EPISODE) During that time, did (TARGET) have…
1. Legal Problems associated with his/her drinking?
2. Health problems associated with his/her drinking?
3. Marital or family problems associated with his/her drinking?
4. Social problems associated with his/her drinking?
5. Problems at work (or loss of work) associated with drinking?
6. Treatment for his/her drinking problem?

*This FH-RDC Alcoholism measure was used to obtain family history data from 1) parent informants on offspring targets (in FFP),
and from 2) offspring informants on parent and sibling targets (in FF1 and MM/MF2)
**”Don’t Know” was offered as a response choice for stem item in FFP but not FF1 and MM/MF2. However, twins in FF1 and
MM/MF2 did indicate when they did not know, and procedures with handling such a response were similar to FFP (i.e., module was
skipped). All other content (e.g., questions, response choices) were identical across interviews.

YES NO DON’T KNOW
YES NO DON’T KNOW
YES NO DON’T KNOW
YES NO DON’T KNOW
YES NO DON’T KNOW
YES NO DON’T KNOW
YES NO DON’T KNOW
**
Reliability and Validity of Family Reports of Alcoholism 84

Appendix C: Validity Indices: Definitions and Equations

0 1
0 TN FN
1 FP TP

True Positive (TP)=Affected targets correctly identified as such by informants.
False Positive (FP)= Unaffected targets incorrectly identified as affected by informants.
True Negative (TN)=Unaffected targets correctly identified as such by informants.
False Negative (FN)=Affected targets incorrectly identified as unaffected by informants.

Validity Indices Definition Equation
Sensitivity Proportion of affected targets
identified as such by informants

Specificity Proportion of unaffected targets
identified as such by informants

Positive Predictive Value (PPV) Proportion of positives based on
informant reports that are true
positives

Negative Predictive Value (NPV) Proportion of negatives based on
informant reports that are true
negatives

Classification
based on
Informant Report
Classification based on Target Report

Abstract (if available)

Abstract The family history method, which collects data from family members who serve as informants of their relatives’ psychopathology, has been utilized as a cost‐effective method of obtaining family history data. It is widely acknowledged that there are limitations related to collecting data via family history method

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Examining the effect of fertility on lifespan using data from the Swedish Twin Registry

PDF

Patterns of EEG spectral power in 9-10 year old twins and their relationships with aggressive and nonaggressive antisocial behavior in childhood and adolescence

PDF

Drinking alcohol to improve mood partially mediates the relation between major depression and alcohol dependence

PDF

Effects of estradiol on cortisol response, working memory, and emotional memory during stress in young and post-menopausal women

PDF

Different genetic and environmental structures for the overlap of three antisocial behavior factors with alcohol initiation

PDF

Parental military service and adolescent mental and behavioral health: the role of adolescent-civilian parent interactions

PDF

Longitudinal relationships between bereavement and physical health‐related quality of life in middle‐ to older‐aged women

PDF

An ecological mixed-methods analysis of homeless students’ school experience at the state, district, and school levels

PDF

Construct validity of psychopathic personality traits in a cohort of young twins

PDF

Psychosocial adjustment among patients with metastatic colorectal cancer

PDF

Predicting alcohol use: a nonlinear model with expectancies and memory associations

PDF

Family conflict, negative mood, and adolescents' daily problems in school

PDF

Family environment as a moderator of the association between theory of mind and social functioning in people with schizophrenia

PDF

Experimental evaluation of a new intervention designed to promote dementia caregivers' acceptance and empathic responding towards care -recipients

PDF

Later life success of former college student-athletes as a function of retirement from sport and participant characteristics

PDF

Attrition in longitudinal twin studies: a comparative study of SEM estimation methods

PDF

Youth homelessness and gang involvement: ties, trajectories, and timelines

PDF

Predicting cognitive decline and dementia in elderly twins from indicators of early life oral health

PDF

The amygdala and conditioned taste aversion

PDF

Engaging homeless men and shelter providers to adapt an existing evidence-based HIV prevention intervention

Asset Metadata

Creator Chereji, Elizabeth (author)

Core Title An investigation of factors associated with reliability and validity of family history reports of alcohol‐related problems

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Psychology

Publication Date 07/29/2014

Defense Date 06/22/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Alcohol,Alcoholism,family history reports,OAI-PMH Harvest,reliability,validity

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Prescott, Carol A. (committee chair), Gatz, Margaret (committee member), Lavond, David (committee member), Wenzel, Suzanne L. (committee member)

Creator Email chereji@usc.edu,elizabeth.chereji@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-451015

Unique identifier UC11286862

Identifier etd-CherejiEli-2753.pdf (filename),usctheses-c3-451015 (legacy record id)

Legacy Identifier etd-CherejiEli-2753-0.pdf

Dmrecord 451015

Document Type Dissertation

Format application/pdf (imt)

Rights Chereji, Elizabeth

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA