Development and Psychometric Analysis

Development and Psychometric Analysis of the Brief DSM–5 Alcohol Use Disorder Diagnostic Assessment: Towards Effective Diagnosis in

College Students

Brett T. Hagman National Institute of Alcohol Abuse and Alcoholism, Bethesda, Maryland

The Diagnostic and Statistical Manual of Mental Disorders (5th edition) Alcohol Use Disorder (DSM–5 AUD) criteria have been modified to reflect a single, continuous disorder. It is critical that we develop brief assessment measures that can accurately assess for DSM–5 AUD criteria in college students to assist in screening, referral, and brief intervention services implemented on college campuses. The present study sought to develop and assess for the psychometric properties of a brief 13-item measure designed to capture the full spectrum of the DSM–5 AUD criteria in a sample of college students. Participants were past-year drinkers (N � 923) between the ages of 18 to 30 enrolled at 3 universities. Respondents completed a 30-min anonymous battery of questionnaires online. The Brief DSM–5 AUD Assessment consisted of 13 items designed to reflect the DSM–5 AUD criteria. Results indicated a high degree of internal consistency reliability with high item-to-scale correlations. Confirmatory factor analyses indi- cated that a dominant single factor emerged with good model fit. The Item Response Theory (IRT) analyses indicated that the difficulty parameters for each criterion were intermixed along the upper portion of the underlying AUD severity continuum, and the discrimination parameters were all high. Additional analysis indicated that those with a DSM–5 AUD had greater levels of alcohol and other drug use and problem severity in comparison to those without a DSM–5 AUD. Study findings provide empirical support for the reliability and validity of the Brief 13-item DSM–5 Assessment. It should be routinely included into research and clinical practice efforts.

Keywords: college students, AUD, alcohol use, screening, assessment

The college years constitute as a critical developmental period wherein alcohol use and risky drinking practices significantly increase (Windle, 2003). As such, people in this critical period experience the highest rates of heavy alcohol use compared to any other at-risk groups of drinkers (Campbell & Demb, 2008; Daw- son, Grant, Stinson, & Chou, 2004). This high-risk level of alcohol involvement is associated with a plethora of alcohol-related con- sequences that are specific (i.e., poor academic functioning) to this important life transition (Beck et al., 2008; Kahler, Strong, Read, Palfai, & Wood, 2004). More importantly, research has consis- tently indicated that rates of alcohol use disorders (AUDs) also peak during the college years (Dawson et al., 2004; Hagman, Cohn, Schonfeld, Moore, & Barrett, 2014). Epidemiological evi-

dence has shown that prevalence estimates of AUDs for college students range up to approximately 30% under the Diagnostic and Statistical Manual (4th edition; DSM–IV) and DSM–5 diagnostic systems (Dawson et al., 2004; Hagman et al., 2014; Hasin & Grant, 2004; Knight et al., 2002). These high rates of AUDs are partic- ularly disconcerting because if an AUD in college is left undiag- nosed, then it has the potential to lead to a more hazardous form of AUD severity (Campbell & Demb, 2008). Thus, it is critical that college treatment providers and administrators develop brief as- sessment tools that provide reliable and accurate diagnostic infor- mation to identify individuals who may be “at risk” or in need of treatment/referral to deter risky levels of alcohol use and/or pre- vent a more severe course of problematic alcohol use from devel- oping in later adulthood.

The DSM–IV has been the primary taxonomic system used to diagnose someone with an AUD (DSM–IV–TR; American Psychi- atric Association [APA], 2000). Under the former DSM–IV AUD diagnostic system, alcohol abuse and dependence were represented as separate diagnoses with a hierarchical structure posited between them (i.e., alcohol dependence criteria set were considered more severe than abuse criteria; Hasin, Hatzenbuehler, Keyes, & Og- burn, 2006; Hasin, 2003; Martin, Chung, & Lagenbucher, 2008). While the DSM–IV AUD criteria have been used extensively in research and clinical practice, several limitations have consistently been identified: (a) factor analytic and Item Response Theory (IRT) analyses have indicated a dominant single factor with the abuse and dependence criteria intermixed at the upper portion of

Parts of the manuscript have been presented at the annual Research Society on Alcoholism’s annual research conference in Denver, Colorado. This study was funded by contract LD966 from the Florida Department of Children and Families. The contents of this article only reflect the views of the authors and not those of the National Institute on Alcohol Abuse and Alcoholism (NIAAA) or National Institutes of Health. I thank Lawrence Schonfeld, from the Department of Mental Health Law and Policy at the University of South Florida for his consultation to this project from which these data were derived.

Correspondence concerning this article should be addressed to Brett T. Hagman, Division of Treatment and Recovery Research, National Institute of Alcohol Abuse and Alcoholism, 5635 Fishers Lane, Room 2044, Bethesda, MD 20892. E-mail: brett.hagman@nih.gov

Psychology of Addictive Behaviors In the public domain 2017, Vol. 31, No. 7, 797–806 http://dx.doi.org/10.1037/adb0000320

797

mailto:brett.hagman@nih.gov
http://dx.doi.org/10.1037/adb0000320
the underlying AUD severity continuum, suggesting no hierarchy among the DSM–IV criteria; (b) the legal problem criterion dem- onstrates poor item fit in factor analytic analyses, and (c) a craving criterion should be incorporated into the DSM–5 AUD criteria given that it is a pertinent indicator of the AUD severity diagnostic syndrome (APA, 2013; Hagman & Cohn, 2011; Hasin, Fenton, Beseler, Park, & Wall, 2012). As a result of these limitations, the DSM–5 Substance Use Task Force made the following changes to the AUD diagnostic criteria in the DSM–5 manual: (a) eliminate the alcohol abuse and dependence distinction by combining the DSM–IV criteria into a single disorder; (b) add a new diagnostic threshold whereas endorsement of two or more of any AUD criteria reflect an AUD; (c) create a severity qualifier that reflect a minimal AUD (2 to 3 criteria), moderate AUD (4 to 5 criteria), or severe AUD (�6 criteria); and (d) exclude the legal problems criterion and incorporate a new craving criterion into the DSM–5 criteria set (APA, 2013).

The process of reliable and valid screening and assessment for detecting AUD symptoms has become a routine procedure within screening, referral, and brief intervention protocols implemented across college campuses and universities (Bien, Miller, & Tonigan, 1993; Monti, Tevyaw, & Borsari, 2004/2005). Several brief as- sessment and alcohol screening measures (i.e., Alcohol Depen- dence Scale [ADS]; Short Alcohol Dependence Data questionnaire [SADD]; Severity of Alcohol Dependence Questionnaire [SADQ]; Alcohol Use Disorders Identification Test [AUDIT]) have been developed to detect at-risk problem drinking, identify individ- uals at-risk for an AUD or to determine the presence and severity of AUD symptomatology within these protocols (Ba- bor, Higgins-Biddle, Saunders, & Monterio, 2001; Raistrick, Dunbar & Davidson, 1983; Skinner & Allen, 1982; Stockwell, Murphy, & Hodgson, 1983). A primary limitation associated with these assessment-based measures is that each was designed for a specific purpose and do not fully capture the range of AUD criteria conceptualized in the DSM–5. For example, while the Short Alcohol Dependence Data questionnaire (SADD) was de- signed to measure the severity of alcohol dependence, it only includes items that reflect behavioral and subjective changes as- sociated with problem drinking, and therefore it has greater sen- sitivity in identifying drinkers who are not experiencing with- drawal symptoms (Raistrick et al., 1983). Along these lines, the SADQ is focused on assessing withdrawal symptoms and does not include items that reflect the development of tolerance and the subjective awareness of the compulsion to drink, thereby providing greater sensitivity to individuals experiencing withdrawal symp- toms (Stockwell et al., 1983). With respect to the AUDIT, while the items are used to screen for being at-risk for an AUD, three of the 10 items only reflect alcohol consumption and do not capture the full range of diagnostic criteria, thereby requiring additional follow-up assessment to make a clinical diagnosis. More importantly, under the new DSM–5 diagnostic guidelines, a crav- ing criterion has been added to the diagnosis, but none of these measures include item(s) that assess for craving. A final limitation is that each of these measures focuses on assessing the nature and severity of symptoms of alcohol dependence and has not been validated for obtaining a DSM–5 AUD diagnosis. In light of these shortcomings, it is critical to develop brief assessment measures that accurately capture the full spectrum of the AUD continuum as

well as more validly reflect the criteria outlined in the newly implemented DSM–5 AUD criteria.

In sum, college students represent a distinct group of drinkers at elevated risk for developing an AUD in comparison to other populations of drinkers. The DSM–5 AUD criteria have been modified to reflect a single, continuous disorder with the removal of the legal problems criterion and the addition of a craving criterion. As such, it is critical that we develop brief assessment measures that can accurately assess for and directly capture the DSM–5 AUD criteria in college students as well in other popula- tions of drinkers. The development of such a measure will assist in detecting an AUD diagnosis more quickly compared to most alcohol screening measures within our alcohol screening, referral, and brief intervention protocols via directly assessing the DSM–5 AUD criteria, thereby permitting more expedient patient referrals to an appropriate level of intervention. A brief assessment measure of DSM–5 AUD criteria can also cut down on the time and costs of undergoing a thorough, rigorous, standardized clinical assess- ment that requires a trained clinician to conduct, and can easily be self-administered to clients and research participants without un- dergoing the stigma that can result from undergoing a face-to-face clinical assessment. In addition, a brief assessment of DSM–5 AUD criteria has the potential to enhance epidemiological, needs assessment, and program planning efforts across college and uni- versity settings by providing a cost-effective method to conduct mass screenings across a college campus in order to obtain campus-specific prevalence rates of DSM–5 AUDs. Lastly, such a measure could be routinely included into university health settings as part of their formal intake and assessment procedures. Based on this background, the present study focused on the development and measurement of the Brief DSM–5 AUD Assessment, which is designed to capture the full spectrum of the DSM–5 AUD criteria in a sample of college students. The present study utilizes methods from Classical Test Theory (e.g., Cronbach’s alpha) and IRT to evaluate the psychometric properties of the Brief DSM–5 AUD Assessment.

Method

Participants and Procedure

This study is a secondary data analysis of the Core Alcohol and Drug Use survey, which was implemented at several universities (Presley, Meilman, & Lyerla, 1994). The data for this study sample (N � 923) were collected at three public universities located in the southeastern United States, with enrollment occurring during the Spring and Fall, 2014 semesters. Participants were invited to participate via e-mail in an online anonymous assessment of their drug and alcohol use as part of a larger effort to understand more about the etiology and prevalence of alcohol use and problems among college students. Participants were included in this study if they were between the ages of 18 to 30 years of age, an under- graduate attending college either full- or part-time, and consumed alcohol in the prior year. For all participants, after providing informed consent to the study, respondents completed a 30-min anonymous battery of questionnaires online. Due to anonymity of responses, all procedures were considered exempt for review by the current Institutional Review Board (IRB). There was no com- pensation given to participants for their participation.

798 HAGMAN

Measures

Development of the Brief DSM–5 AUD assessment. As shown in the Appendix, a total of 13 questions were developed to reflect the DSM–5 AUD criteria and included as part of the administration of the Core survey. These diagnostic questions were developed by the author and not routinely included in the Core survey. For this measure, two separate questions (4 total) were used to obtain diagnostic information for the tolerance cri- terion (i.e., diminished effect with continued use and need to drink more to get desired effect) and withdrawal criterion (i.e., experi- ence withdrawal symptoms from not drinking and drink to avoid withdrawal symptoms). Endorsement of either question or both for each criterion reflected presence of that criterion. The questions paralleled wording from the Diagnostic and Statistical Manual’s DSM–5 AUD criteria (APA, 2013; see the Appendix for specific wording of each diagnostic criterion). The DSM–5 craving ques- tion was included by asking participants the following: “During the past year, as a result of your alcohol use, did you have a strong desire or craving to drink?” This item has been used in prior research (Casey, Adamson, Shevlin, & McKinney, 2012), which has indicated it to be a reliable and valid indicator of craving. Participants were asked to report if the occurrence (yes) or absence (no) of each criterion occurred more than once within the past year.

Alcohol, other drug use, and negative consequences. A series of questions from the Core survey were developed to collect alcohol and other drug use data from each participant. With respect to the alcohol use module, participants were asked to report their frequency of alcohol use (1 � did not use to 9 � every day) in the prior year and during the prior 30-days (1 � 0 days to 7 � all 30 days) on Likert-type scales. With respect to binge drinking, par- ticipants were asked to report on a 6-point Likert-type scale the number of times (1 � none to 6 � 10 or more times) they consumed five or more drinks in a sitting in the prior 2 weeks. Quantity of alcohol use was assessed by asking participants to report the average number of standard drinks consumed per week. Lastly, participants reported their age of first alcohol use (1 � did not use to 9 � 26 or older) on a 9-point Likert-type scale. Pearson correlations between each of the alcohol use measures were high and ranged from .595 to .803 providing evidence of their validity.

In regard to the illicit drug use module, three sets of questions were of interest. Participants were asked to report their frequency of drug use in the prior year (1 � did not use to 9 � every day) by reporting whether they had used each of 11 specific types of drugs (marijuana, cocaine, amphetamines, sedatives, hallucinogens, opi- ates, inhalants, designer drugs, steroids, and other types not listed) on a 9-point Likert-type scale. Similarly, participants reported their frequency of using (1 � none to 6 � 10 or more times) 11 specific types of drugs in the prior 30 days using a 6-point Likert-type scale. Lastly, participants indicated how often (1 � never to 6 � 10 or more times) they experienced 19 different types of conse- quences (e.g., “performed poorly on a test or important project”; “been hurt or injured”) as a result of their drug or alcohol use in the last year based on a 6-point Likert-type scale.

Construction of drug use indices. Three indices of illicit drug use were constructed based on self-reports from the illicit drug use module. First, a drug use in the prior year frequency index was created by summing items specific to the frequency of illicit drug use in the prior year. Second, a drug use in the prior 30

days frequency index was created by summing the response items specific to the frequency of illicit drug use in the prior 30 days. Lastly, an alcohol and drug use consequences frequency index was constructed by summing the items specific to the types of alcohol and drug use consequences that occurred in the prior year. To ensure that each index is approximately unidimensional, principal components analyses were performed. Results indicated that each index had an approximately unidimensional structure associated with it and accounted for approximately 63.42%, 50.43%, and 31.28% of the common variance for the drug use in the prior 30 days index, drug use in the prior year index, and alcohol and drug use consequences index, respectively. Cronbach’s coefficient al- phas for each index was high and ranged from .732 for the frequency of drug use in prior year index to .871 for the alcohol and other drug use consequences index.

Classification of AUD status. For the classification of the DSM–5 AUD diagnostic system, we used the guidelines set forth by DSM–5 Substance Use Task Force (APA, 2013). Participants who did not endorse any criteria were classified as no AUD; those who endorsed one of any criteria were classified as DSM–5 diag- nostic orphans/DO; and those who endorsed two or more of any criteria were classified as DSM–5 AUD�.

Data Analytic Plan

Classical test theory analyses: Reliability and validity analyses. Classical test theory (CTT) analyses were conducted to evaluate the overall reliability and validity of the Brief DSM–5 AUD Assessment. To evaluate the reliability of the DSM–5 AUD crite- ria, internal consistency reliability was assessed by calculating Cronbach’s coefficient alpha with each item removed, and addi- tional reliability analyses examined the item-to-total scale corre- lations for each criterion.

To assess the validity of the Brief DSM–5 AUD Assessment, several analyses were conducted. Convergent validity analyses were conducted by performing Pearson correlations between total number of DSM–5 AUD symptoms endorsed with several external validators of alcohol and other drug use (average drinks per week, binge drinking in prior 2 weeks, age of onset of drinking, fre- quency of alcohol use in prior year, frequency of alcohol use in prior 30 days, drug use in the prior 30 days index, drug use in the prior year index, and the alcohol and drug use consequences index). To evaluate differences between those with and without a DSM–5 AUD, a Hoetellings T2 test was performed. For the Hoe- tellings T2 analysis, we evaluated differences across several exter- nal validators of alcohol and illicit drug use using the DSM–5 AUD system (i.e., two groups: No DSM–5 vs. DSM–5 AUD) as a primary independent variable in the analysis. The following eight alcohol use and illicit drug use external validators were included as dependent variables: average drinks per week, binge drinking in prior 2 weeks, age of onset of drinking, frequency of alcohol use in prior year, frequency of alcohol use in prior 30 days, drug use in the prior 30 days index, drug use in the prior year index, and the alcohol and drug use consequences index. Post hoc t tests were conducted if multivariate significance was achieved. To control for Type I error inflation, we set alpha at p � .01 to achieve statistical significance.

Item Response Theory analysis. Item Response Theory anal- yses were also conducted on the DSM–5 AUD criteria to derive

799DEVELOPMENT OF A BRIEF DSM–5 AUD ASSESSMENT

item difficulty and discrimination parameters for each criterion. Prior to conducting the IRT analyses, a confirmatory factor anal- ysis (CFA) on the 11 DSM–5 AUD criteria was conducted to ensure that assumptions (i.e., items reflect a single factor solution) were met. A single factor solution and tetrachoric correlation matrix was specified for the CFA. The following guidelines pro- posed were used to assess for model fit in the CFA. Comparative Fit Index (CFI) � 0.95, Tucker-Lewis Index (TLI) � 0.95, and a root mean square error of approximation (RMSEA) � 0.06 (Hu & Bentler, 1999). A robust unweighted least squares estimation was specified to derive parameter estimates. Next, an IRT analysis was conducted on the 11 DSM–5 AUD criteria. Two-parameter logistic models were specified estimating item difficulty (location) and discrimination (slope) parameters for each criterion. A high diffi- culty parameter indicates that a greater level of alcohol problem severity is necessary to endorse that criterion. The discrimination parameter provides a numerical value (typically ranges from 0 to 3) of the magnitude of the relationship between each AUD crite- rion and the underlying latent-trait continuum. A high discrimina- tion value indicates that a specific AUD criterion is able to accu- rately classify individuals with various levels of the latent-trait of AUD severity. Item characteristic curves (ICCs) were then plotted for all criteria. ICCs provide a graphical depiction of the proba- bility that a specific criterion is endorsed as a function of the value of the purported underlying latent-trait. The typical ICC indicates

that the probability of endorsing a specific item increases mono- tonically as the latent-trait continuum increases. Lastly, a total information curve was plotted for the 11 DSM–5 criteria. This curve provides information about the point along the continuum where the DSM–5 AUD criteria are most reliable. All IRT models were analyzed using Parscale IRT software (Scientific Software International, 2003), which estimates criterion parameters via a Bayesian expectation-maximization (EM) equation. The conver- gence criterion for the EM equation was set to .005 for all IRT analyses.

Results

Demographic Characteristics of Current Sample

As shown in Table 1, participants were between the ages of 18 to 30 (M � 19.64; SD � 1.19). The sample was fairly represen- tative of college students with respect to race and ethnicity with 70.7% (n � 653) Caucasian, 16.1% (n � 149) Hispanic, 4.9% (n � 45) African American, 3.1% (n � 29) Asian/Pacific Islander, and 4.8% (n � 44) representing other racial/ethnic groups. With respect to class rank, 28.9% (n � 267) were freshman, 28.4% (n � 262) sophomores, 28.2% (n � 260) juniors, and 14% (n � 129) seniors. The majority of the sample were female (68.7%; n � 634),

Table 1 Demographics of Current Study Sample Across DSM–5 AUD Status

Demographic classification variable

Overall NO AUD DO DSM–5 AUD

N % N % N % N %

Class rank Freshman 267 28.9 111 25.8 50 27.6 93 34.8 Sophomore 262 28.4 121 28.1 53 29.3 75 28.1 Junior 260 28.2 132 30.6 50 27.6 65 24.3 Senior 129 14.0 64 14.8 28 15.5 32 12.0

Age 18 to 20 733 79.6 335 77.7 140 77.4 230 86.2 21 to 22 175 19 89 20.6 39 21.6 33 12.4 23 or older 13 1.4 6 1.4 2 1.1 4 1.5

Ethnicity Hispanic 149 16.1 69 16.1 30 16.6 44 16.5 Asian/Pacific Islander 29 3.1 14 3.2 4 2.2 10 3.7 White 653 70.7 301 69.8 124 68.5 196 73.4 Black 45 4.9 19 4.4 15 8.3 9 3.4 Other 4.8 44 26 6.0 8 4.4 8 2.9

Gender Male 289 31.3 113 26.2 64 35.4 92 34.5 Female 634 68.7 318 73.8 117 64.6 175 65.5

Residence On campus 358 31.3 157 36.4 64 35.4 124 46.4 Off campus 634 68.7 274 63.6 117 64.6 143 53.6

GPA A average 383 41.5 205 47.6 59 32.6 99 37.1 B average 431 46.7 176 40.8 98 54.1 137 51.3 C or below average 108 11.8 50 11.6 24 13.3 30 11.6

Student status Full time 890 96.4 419 97.2 173 95.6 257 96.6 Part time 32 3.5 12 2.8 8 4.4 9 3.4

Note. No AUD � no DSM–5 AUD diagnosis; DO � DSM–5 diagnostic orphans; AUD � DSM–5 AUD diagnosis; DSM–5 AUD � Diagnostic and Statistical Manual of Mental Disorders, fifth edition Alcohol Use Disorder; GPA � grade-point average.

800 HAGMAN

lived off campus (61.2%; n � 565), and reported having at least a B or higher grade-point average (GPA; n � 814; 88.2%).

Alcohol and Other Drug Use Characteristics

The overall sample reported drinking, on average, 4.24 (SD � 7.61) standard drinks per week. Approximately 52.4% (n � 483), 47% (n � 435), and 42.6% (n � 393) of the sample reported drinking at least 5 or more days per month, binge drinking on at least one occasion during the prior 2 weeks, and consumed alcohol more than once a week, respectively. The most commonly occur- ring alcohol- and drug-related consequences among the overall sample in the past year were as follows: “had a hangover” (66.9%; n � 622), “vomited” (64.5%; n � 602), “had memory loss” (42.9%; n � 406), and “latter regretted action under the influence” (36.3%, n � 349). With respect to illicit drug use, the most commonly used in the prior year were as follows: marijuana (47.4%; n � 442), tobacco (39.4%: n � 366), designer drugs (e.g., ecstasy; 10.4%; n � 97), and hallucinogens (8.7%; n � 84).

Classification of DSM–5 AUDs

Based on this classification scheme, the percentages of those with no DSM–5 AUD diagnosis, DSM–5 diagnostic orphans, and DSM–5 AUD� diagnosis were 46.7% (n � 431), 19.6% (n � 181), and 28.9% (n � 267), respectively. In addition, approxi- mately 17.8% (n � 164), 6.6% (n � 61), and 4.8% (n � 42) were classified as mild DSM–5 AUD, moderate DSM–5 AUD, and severe DSM–5 AUD, respectively.

Reliability Analyses

As shown in Table 2, Cronbach’s alpha with each item removed were conducted for each of the DSM–5 AUD criteria, which all were in the high range with relatively little variation (Cronbach’s alphas ranging from .754 to .778). Along these lines, item-to-total scale correlations were conducted with correlations ranging from .321 to .517. The overall Cronbach’s alpha for the DSM–5 AUD

criteria was .781, which indicates a high degree of internal con- sistency reliability.

Validity Analyses

To demonstrate the validity of the DSM–5 AUD criteria, Pear- son correlations between total sum of DSM–5 criteria endorsed and other meaningful variables were conducted. Results indicated that total sum DSM–5 criteria scores were significantly related to average binge drinking in the prior 2 weeks (r � .45, p � .001), average drinks per week (r � .44, p � .001), age of alcohol use onset (r � �.16, p � .001), alcohol use in prior year (r � .46, p � .001), alcohol use in prior 30 days (r � 46, p � .001), have a perceived problem with alcohol and other drugs (r � .441, p � .001), drug- and alcohol-related negative consequences (r � .69, p � .001), frequency of drug use in the prior 30 days (r � .31, p � .001), and frequency of drug use in the prior year (r � .36, p � .001).

Evaluating Differences Between Those With and Without a DSM–5 AUD

Table 3 displays results of the Hotellings T2 that examined mean differences between the DSM–5 diagnostic groups (i.e., No diag- nosis vs. DSM–5� diagnosis) across the external validator vari- ables of alcohol consumption, illicit drug use, and alcohol/drug- related negative consequences. With respect to the Hotellings T2

analysis, the overall omnibus tests was significant for the DSM–5 AUD criteria [Hotelling’s Trace � .469, F(8, 850) � 49.80, p � .001]. All follow-up univariate t tests across each external valida- tor were significant (all ps � .01). Compared to those who did not meet criteria for a DSM–5 AUD (i.e., No AUD diagnosis), those with a DSM–5 AUD diagnosis reported greater levels of alcohol use, illicit drug use, and drug/alcohol-related negative conse- quences providing support for the utility of the DSM–5 diagnostic threshold.

Table 2 Reliability Analyses (Item-to-Scale Correlations; Cronbach’s Coefficient Alpha With Each Item Missing) for the DSM–IV and DSM–5 AUD Criteria

DSM–5 AUD diagnostic criteria %

endorsed DSM–5:

Item to-scale DSM–5: Alpha

1) Unable to fulfill role obligations (abuse) 8.2 .351 .774 2) Physically hazardous situations (abuse) 19.8 .362 .778 3) Legal problems (abuse) 3.1 � �

4) Social/Interpersonal problems (abuse) 9.2 .509 .755 5) Larger/Longer amounts (dependence) 24.1 .517 .754 6) Unsuccessful efforts (dependence) 4.9 .442 .765 7) Great deal of time (dependence) 7.3 .493 .758 8) Important activities given up (dependence) 4.7 .475 .763 9) Recurrent physical/psychological problems (dependence) 7.3 .480 .760

10) Craving (DSM–5) 16.7 .483 .757 11) Tolerance (dependence) 26.9 .480 .760 12) Withdrawal (dependence) 3.9 .431 .767 Overall Cronbach’s alpha .781

Note. DSM–5 � Diagnostic and Statistical Manual of Mental Disorders, fifth edition Alcohol Use Disorder. � p � .001.

801DEVELOPMENT OF A BRIEF DSM–5 AUD ASSESSMENT

Item Response Theory Analyses

As shown in Table 5, results from the CFA indicated that a dominant single factor emerged with good model fit for the DSM–5 AUD criteria: Tucker-Lewis Index (TLI) � 0.991, Com- parative Fit Index (CFI) � 0.982, and Root Mean Square Error of Approximation (RMSEA) � 0.028. As shown in Table 2, the standardized factor loadings were high and ranged from .542 “physically hazardous (abuse)” to .836 “important activities given up (dependence).” In addition, the “alcohol craving” criterion factor loading was adequate (.737) and indicated good fit within the CFA model. Overall, findings indicated IRT assumptions were met and that the DSM–5 AUD criteria model reflects a strong, dominant single factor.

Final IRT Model Analyses

As shown in Table 4, the frequency of endorsement for each of the 11 DSM–5 AUD criteria ranged from 3.9% “withdrawal (abuse)” to 26.9% “tolerance (dependence).” The DSM–5 AUD criteria with the highest level of endorsement were “tolerance

(dependence),” “drinking in larger/longer amounts (dependence),” “physically hazardous situations (abuse),” “craving (DSM–5),” whereas the items with the lowest frequency of endorsement were “withdrawal (dependence)” and “important activities given up (dependence).” Table 5 presents IRT difficulty and discrimination parameter estimates across the DSM–5 AUD IRT model. The difficulty parameters for the IRT model that includes the DSM–5 AUD criteria indicated that the abuse and dependence criteria were intermixed along the latent-trait AUD severity continuum. The IRT difficulty parameters for the DSM–5 AUD criteria ranged from 0.81 “tolerance (dependence)” to 2.44 “unable to fulfill role obligations (abuse).” Overall, the following difficulty parameters for the DSM–5 AUD criteria were ranked the lowest and were plotted toward the middle of the latent-trait AUD severity contin- uum (values ranging from 0 to 1.5): “tolerance (dependence),” “larger/longer amounts (dependence),” “alcohol craving (DSM– 5),” and “physically hazardous (abuse).” In addition, the difficulty parameters for the DSM–5 criteria “social/interpersonal problems (abuse),” “great deal of time (dependence),” “recurrent physical and psychological problems (dependence),” and “important activ-

Table 3 DSM–5 AUD Hoetelling’s T2 Analysis Across External Validators of Alcohol and Illicit Drug Use

Alcohol and illicit drug use variable

Place this order or similar order and get an amazing discount. USE Discount code “GET20” for 20% discount