Psychometric evaluation of the Drinking Patterns Questionnaire: A measure of high-risk drinking situations

Abstract

Purpose

The Drinking Patterns Questionnaire (DPQ) is a self-report instrument designed to identify high-risk (HR) drinking situations. While prior investigation has established the preliminary psychometric properties of the DPQ, additional research is needed. The current study evaluated the construct validity of the Work-Related, Financial, Parents, and Children subscales of the DPQ as well as the internal consistency of all subscales.

Method

One hundred and thirty-four alcohol-dependent inpatients completed a questionnaire packet containing the DPQ, a demographics questionnaire, four measures used to evaluate the convergent validity of DPQ subscales, and three measures of alcohol use disorder symptoms.

Results

DPQ subscales evidenced adequate to strong internal consistency (coeffficient alphas of .691 to .921). Significant Pearson’s r correlations were found between scores on the Work-Related, Financial, and Children subscales and those on their corresponding correlate measures. Findings for the validity of the Parents subscale were mixed.

Conclusion

Study results provide support for the construct validity for the Work-Related, Financial, and Children subscales and indicate that all DPQ subscales exhibit solid internal consistency.

1. Introduction

1.1. High-risk (HR) drinking situations

Identification of high-risk (HR) drinking situations is viewed as an essential component of empirically-based treatments for alcohol use disorders (AUDs). Unique to each individual, HR situations represent an array of cognitive, affective, behavioral, interpersonal, and environmental cues that have preceded or accompanied prior drinking (Marlatt, 1996). While the Identification of HR situations may be useful at various stages of treatment, the alcohol assessment and treatment literatures suggest that the most robust application of this construct is within the domain of post-treatment relapse (Annis & Davis, 1989; Annis & Graham, 1995). Within this domain, future HR situations (also known as “contextual triggers” and “proximal precipitating factors for relapse”) are believed to be best predicted by prior drinking antecedents (Donovan, 1996; Marlatt, 1996).

Empirical support has been established for the theoretical connection between prior antecedents and future HR situations using alcohol cue-reactivity paradigms. Not only do alcohol-dependent individuals exhibit strong cue-reactivity on a variety of cognitive (e.g., coping self-efficacy), affective (e.g., guilt), and physiological (e.g., salivation, heart rate) indexes, they also report dramatic cue-elicited increases in subjective urge to drink (reviewed in Carter & Tiffany, 1999). These findings have led researchers to suggest that exposure to prior drinking antecedents may predict subsequent alcohol consumption and relapse after treatment (Drummond, 2000).

Outside of the laboratory, Identification of HR situations during treatment has shown predictive validity; AUD individuals in one study who relapsed after treatment tended to do so in situations previously Identified as high-risk antecedents (Miller, McCrady, Abrams, & LaBouvie, 1994). Relapse prevention interventions, therefore, attempt to tailor the development of coping skills to target those HR situations unique to each client (Rohsenow et al., 2001).

While Identification of HR situations may be accomplished through clinical interviewing, a more directive approach using a comprehensive list of potential high-risk situations may decrease the likelihood of omitting less easily-recalled triggers. This process may be facilitated by the use of a psychometric instrument designed specifically for the Identification of HR situations.

1.2. Measures of HR situations

Although several self-report assessment instruments have been developed to identify HR situations, most have limitations. For example, instruments such as the Reasons for Drinking Questionnaire (Zywiak, Connors, Maisto, & Westerberg, 1996) and Drinking Context Scale (O’Hare, 1997) consist of relatively few items (16 and 23, respectively) and raise concerns of inadequate comprehensiveness. While the Inventory of Drinking Situations (IDS; Annis, Graham, & Davis, 1987) is a longer measure, aspects of the instrument’s development may weaken its clinical utility. Specifically, the IDS is an empirically-derived scale and, as such, its item selection relied heavily on psychometric testing instead of perhaps more clinically-meaningful indicators. Items hypothesized to tap theoretical constructs may have less real-world application than items derived from self-reported antecedents of alcohol-dependent individuals.

1.3. Drinking Patterns Questionnaire (DPQ)

The measure examined in the current study, the Drinking Patterns Questionnaire (DPQ; Zitter & McCrady, 1979), was designed as a clinical tool to identify prior drinking antecedents and to predict potential relapse precipitants (Miller et al., 1994). A paper-and-pencil self-report instrument, the DPQ consists of 189 items representing possible drinking antecedents from nine categories of HR situations: Environmental, Work-Related, Financial, Physiological, Interpersonal, Marital, Parents, Children, and Emotional. Respondents indicate whether they have drank in each situation within the past six months, how influential the situation was to their drinking, and how important each of the cue categories is to their drinking in general.

In addition to assessing a broad range of drinking antecedents, an advantage of the DPQ is the clinical derivation of its items. In contrast to empirically-derived measures such as the IDS, the DPQ consists of items generated by the self-reported drinking antecedents of actual patients with AUDs (Zitter & McCrady, 1979). Not only were the items clinically derived, but subsequent modification and refinement of the DPQ has relied on both empirical data and expert clinical judgment in making decisions about retention of items during psychometric testing (Zweig, McCrady, & Epstein, in press).

Recent research conducted by Zweig, McCrady, and Epstein (in press) examined the construct validity of individual subscales via analyses of convergence with measures of similar constructs. Analyzing archival data from prior clinical trials, Zweig was limited to using as correlate measures only those instruments administered in the original studies. As a result, convergent validity analyses were not possible for all DPQ subscales. While Zweig provided evidence of satisfactory construct validity of the Physiological, Interpersonal, Marital, and Emotional subscales, the validity of the Work-Related, Financial, Parents, and Children subscales remains unclear (the ninth subscale, Environmental, consists of items representing multiple constructs and is not appropriate for such analyses).

1.4. Current study

The primary aim of the current study was to evaluate the construct validity of the four unvalidated subscales of the DPQ. One means of investigating construct validity is through analysis of convergence between scores on a measure or subscale and those on a previously-validated measure of the same or similar construct (Cohen & Swerdlik, 2002).

A second aim was to examine the internal consistency of DPQ subscales. Prior evaluation of internal consistency relied on data from disparate, restricted samples (i.e., three separate randomized clinical trials; Zweig et al., in press). In contrast, the current study examined internal consistency in an unselected clinical population.

2. Method

2.1. Participants

Participants were 159 adult residents of the Caron Foundation, an inpatient substance abuse treatment center located in eastern Pennsylvania. While all Caron residents were eligible for participation, only those who fully completed the Alcohol Dependence Scale (ADS; Skinner & Allen, 1982) and who scored a 9 or above on this measure were included in the statistical analyses. Prior research has shown that this cut-off score is highly predictive of a DSM-IV diagnosis of alcohol dependence (Chantarujikapong, Smith, & Fox, 1997).

Of the 159 residents who participated in the research protocol, 134 met study inclusion criteria. Among the 25 participants excluded from the statistical analyses were five participants who did not complete the ADS and 20 participants who scored less than 9 on this measure. Excluded participants did not differ Significantly from their included counterparts on any demographic variables (i.e., age, gender, race/ethnicity, education, household income, marital status, or employment status).

2.2. Measures

2.2.1. Drinking Patterns Questionnaire (DPQ; Zitter & McCrady, 1979)

The DPQ is a 189-item self-report measure of HR drinking situations. Prior psychometric evaluation has indicated that the DPQ exhibits adequate internal consistency (i.e., the majority of subscales yield coefficient alphas greater than .70; Zweig et al., in press) and appropriate gender-specific response patterns (Sell, McCrady, and Epstein, 2003; Zitter & McCrady, 1979). Evidence of its predictive validity has been garnered as well; contextual precipitants associated with relapse after treatment have been found to be Significantly associated with HR situations Identified by the DPQ during treatment (Miller et al., 1994). Also, the Physiological, Interpersonal, Marital, and Emotional subscales have also been shown to exhibit satisfactory construct validity (Zweig et al., in press).

2.2.2. Correlate measures for analysis of convergent validity

2.2.2.1. Abridged Job in General Scale (AJIG; Russell et al., 2004)

A 7-item self-report measure of overall job satisfaction, the AJIG was used to evaluate the convergent validity of the Work-Related subscale. The AJIG exhibits strong internal consistency (coefficient alpha=.85) and its scores have been shown to be highly correlated with those of other measures of the same construct, suggesting that it is a valid measure of job satisfaction (Russell et al., 2004). Higher scores on the AJIG indicate greater job satisfaction.

2.2.2.2. Money Attitude Scale (MAS; Yamauchi & Templer, 1982)

Part of a larger, 29-item measure of attitudes toward money, the Anxiety subscale of the MAS is a 9-item measure of financial anxiety that was used to evaluate the convergent validity of the Financial subscale of the DPQ. The Anxiety subscale has shown adequate internal consistency (coefficient alpha=.69), strong test–retest reliability (Pearson’s r=.88), and convergent validity with measures of a similar construct (Yamauchi & Templer, 1982). Higher scores on the Anxiety subscale indicate greater levels of financial anxiety.

2.2.2.3. Parent Adult–Child Relationship Questionnaire (PACQ; Peisah, Brodaty, Luscombe, Kruk, & Anstey, 1999)

The PACQ is a 26-item self-report measure of relationship satisfaction between adult children and their parents as perceived by the adult child and was used to evaluate the validity of the Parents subscale. The Mother and Father subscales of the PACQ evidence high concurrent validity (correlation with ratings of independent clinical assessments of relationship satisfaction yielded Pearson r’s ranging from .71 to .90), strong internal consistency (coefficient alpha’s ranging from .74 to .87), and high test–retest reliability (Pearson r’s ranging from .77 to .93) (Peisah et al., 1999). Higher scores on the PACQ indicate greater relationship satisfaction.

2.2.2.4. Parent–Child Interaction Questionnaire (PACHIQ-R; Lange, Evers, Jansen, & Dolan, 2002)

The PACHIQ-R is a 21-item measure of relationship satisfaction between parents and children as perceived by the parent. Of the two PACHIQ-R subscales, the 12-item Conflict Resolution subscale was chosen to evaluate the validity of the Children subscale (the other subscale, Acceptance, taps a construct conceptually less closely related to potential drinking antecedents and was, therefore, excluded). The Conflict Resolution subscale exhibits strong internal consistency among both mothers (coefficient alpha=.90) and fathers (coefficient alpha=.93) (Lange et al., 2002). Higher scores indicate greater relationship satisfaction.

2.2.3. Measures of participant characteristics

2.2.3.1. Demographic characteristics

A self-report questionnaire designed by the first author assessed basic demographic characteristics including age, gender, race/ethnicity, education, household income, marital status, and employment status.

2.2.3.2. Alcohol-related characteristics

The Alcohol Dependence Scale (ADS; Skinner & Allen, 1982), Alcohol Use Disorders Identification Test (AUDIT; Babor, de la Fuente, Saunders, & Grant, 1992), and Short Index of Problems (SIP; Miller, Tonigan, & Longabaugh, 1995) were used to assess alcohol dependence severity, drinking behavior, and adverse alcohol-related consequences, respectively. The strong psychometric properties of these well-established measures have been reported elsewhere (ADS: Chantarujikapong, Smith, & Fox, 1997; Ross, Gavin, & Skinner, 1990; AUDIT: Reinert & Allen, 2002; SIP: Feinn, Tennen, & Kranzler, 2003).

2.3. Procedure

Five group administrations were conducted between September 2006 and January 2007, with approximately 30 residents participating on each occasion. Having provided informed consent, participants were given a packet of questionnaires including, first, the Demographics Questionnaire and the DPQ, and then the ADS, AUDIT, AJIG, MAS, PACQ, PACHIQ-R, and SIP. The order of the questionnaires was counterbalanced after the DPQ, yielding seven versions of the questionnaire packet. Upon completion of the packet, participants were debriefed regarding the nature of the study, compensated with a Rutgers University water bottle, and thanked for their participation.

3. Results

3.1. Participant characteristics

Slightly less than half of the sample was female (48.5%), 94.8% of participants were Caucasian, 41.7% were married, and 59.7% were employed. The mean age of the sample was 40.0 years (SD=11.4) and on average participants were well educated and had high household incomes, as reflected by a mean education of 15.3 years (SD=4.3) and a median income of $80,000. Due to differences in these and other demographic variables, varying numbers of participants completed each DPQ subscale: Work-Related (n=110), Financial (n=133), Parents (n=115), and Children (n=68).

Mean scores on the ADS (M=22.5, SD=7.4), AUDIT (M=25.9, SD=6.9), and SIP (M=30.4, SD=8.4) fell within the range of clinical norms (Skinner & Allen, 1982; Babor et al., 1992; and Project MATCH Research Group, 1993, respectively), indicating that participants exhibited an expected degree of alcohol dependence severity and adverse alcohol-related consequences given their inpatient status.

3.2. Internal consistency

Internal consistency of DPQ subscales was assessed using Cronbach alpha coefficients. As shown in Table 1, nearly all DPQ subscales evidenced high internal consistency, with coefficient alphas ranging from α=.691 to α=.921.

Table 1.

Internal consistency of DPQ subscales

DPQ subscale	Sample size	Coefficient alpha
Work-Related	110	.731
Financial	133	.846
Physiological	132	.691
Interpersonal	134	.921
Marital	96	.919
Parents	115	.888
Children	68	.881
Emotional	134	.918

3.3. Data coding, scoring, and transformation

The response format for each DPQ item consists of three possible answers (“Did not drink in this situation,” “Sometimes drank in this situation,” and “Major drinking situation”) and each item was coded 0, 1, and 2, respectively. A subscale score was calculated for each DPQ subscale by summing the coded scores of its items. All correlate measures were coded and scored in accordance with their instruction manuals. It should be noted that the PACQ was scored uniquely and yielded three separate sets of scores: one for participants with two living parents, one for those with only a living mother, and one for those with only a living father.

Skewed distributions were corrected by the removal of outliers (i.e., a data point more than 1.5 interquartile ranges [IQRs] below the first quartile or above the third quartile Howitt & Cramer, 2000) and/or the application of square root transformations. Six data points were Identified as outliers and removed from subsequent analyses, including three from the Financial subscale of the DPQ, one from the Parents subscale of the DPQ, and two from the PACQ.

3.4. Convergent validity

Two-tailed Pearson’s r correlations were conducted between each subscale score and the score of its corresponding correlate measure. Table 2 summarizes these results. Significant correlations were found between scores on the Work-Related subscale of the DPQ and those on the AJIG (n=89, r=−.213, p<.05), the Financial subscale and the MAS (n=130, r=.423, p<.001), and the Children subscale and the PACHIQ-R (n=62, r=−.510, p<.001). Greater item endorsement on the Work-Related, Financial, and Children subscales was associated with lower job satisfaction, a higher degree of financial anxiety, and lower child relationship satisfaction, respectively.

Table 2.

Pearson’s r correlations between DPQ subscale scores and scores of their corresponding correlate measures

DPQ subscale/correlate measure	n	r	p
Work-Related subscale/Abridged Job in General Scale (AJIG)	89	−.213	.045
Financial subscale/Money Attitude Scale (MAS)	130	.423	.000
Parents subscale/Parent Adult–Child Relationship Questionnaire (PACQ)
Both Parents	75	−.151	.195
Mother only	23	−.537	.008
Children subscale/Parent–Child Interaction Questionnaire (PACHIQ-R)	62	−.510	.000

A Significant correlation was found between scores on the Parents subscale of the DPQ and those on the PACQ among those with only a living mother (n=23, r=−.537, p<.01). In this group of participants, greater item endorsement on the Parents subscale was associated with lower parental relationship satisfaction with one’s mother. A Significant correlation was not found between scores on the Parents subscale and those on the PACQ among those with both parents (n=75, r=−.151, p=.195), or with only a living father (n=9, r=.472, p=.200).

4. Discussion

The current study sought to evaluate the convergent validity of the Work-Related, Financial, Parents, and Children subscales of the Drinking Patterns Questionnaire (DPQ). A secondary aim was to investigate the internal consistency of all DPQ subscales in an unselected clinical population.

All eight DPQ subscales evidenced high internal consistency, with six subscales yielding coefficient alphas greater than .80. These results build upon prior evidence of strong internal consistency (Zweig et al., in press) and suggest that DPQ subscales reliably measure their respective domains of drinking antecedents.

Convergent validity analyses indicated that scores on the Work-Related, Financial, and Children subscales were Significantly associated with scores on their correlate measures. While the correlation found for the Work-Related subscale is statistically Significant, the magnitude of association is modest and should be considered as only preliminary evidence of convergent validity. These analyses of convergent validity extend prior research investigating the construct validity of DPQ subscales (Zweig et al., in press). Taken together, results from the current study and those found by Zweig (published separately due to the discrepant nature of their respective samples as well as timing and method of data collection) provide initial evidence of the construct validity of seven of the eight DPQ subscales measuring singular constructs, including the Physiological, Interpersonal, Marital, Emotional, Work-Related, Financial, and Children subscales. The DPQ purports to measure the relationship between exposure to antecedents within these domains of potential HR situations and subsequent drinking behavior; evidence of construct validity suggests that it does measure this relationship.

The remaining subscale, Parents, yielded mixed results in the current study. Although a Significant correlation was found between scores on the Parents subscale and those on the PACQ among those participants with only a living mother, this relationship was not found among participants with only a father or with two living parents. These curious results may be attributed to the potential susceptibility of the PACQ to the influence of a confounding variable. It is possible that performance on the PACQ may be influenced by relationship variables other than satisfaction alone, such as degree of parental involvement and frequency of contact. Without a direct measure of these additional relationship variables within the PACQ, this conclusion remains speculative and the construct validity of the Parents subscale remains unclear.

A primary limitation of the current study is the homogeneity of the sample. The overwhelmingly white and affluent nature of the sample suggests the use of caution when generalizing study results to other populations. Another limitation involves shortcomings associated with the correlate measure chosen to evaluate the convergent validity of the Parents subscale, the PACQ. To the degree that the PACQ is a measure of multiple relationship constructs, its utility as a correlate measure in evaluating the construct validity of the Parents subscale may be limited.

Additional research is needed to clarify the construct validity of the Parents subscale using a more heterogeneous population and a more reliable correlate measure. Next, a full confirmatory factor analysis is recommended to uncover the latent structure of DPQ items and to determine whether the existing subscales appropriately categorize HR situations. Further investigation of the predictive validity of the DPQ is warranted as well and would serve to extend preliminary evidence found in prior research (Miller et al., 1994). Lastly, this line of research would be complemented by an evaluation of what impact, if any, the DPQ has on treatment efficiency and overall treatment outcome.

The DPQ holds considerable promise as a clinical tool to aid in the Identification of drinking antecedents. Although additional research is needed, the DPQ has direct clinical applications. Important in various stages of alcohol treatment, a clear understanding of a client’s unique set of drinking antecedents is particularly useful in predicting post-treatment HR situations and in developing specific skills to cope with them. A comprehensive array of clinically-derived items, initially promising psychometric data, and strong clinical utility make the DPQ a potentially valuable asset in treating individuals with AUDs.