Correlates of mild, moderate, and severe Alcohol Use Disorder (AUD) among adults with problem substance use: Validity implications for DSM-5 AUD

Abstract

Background:

The DSM-5 definition of Alcohol Use Disorder (AUD) has been well validated, but information is lacking on the extent of alcohol use, comorbidity, and impairment of the three DSM-5-defined AUD severity levels: mild, moderate, and severe. This study examined clinical and functional characteristics as predictors (validators) of these severity levels.

Methods:

Participants age ≥18 years reporting current problem substance use (N=588) were recruited between 2016–2019 and were assessed for DSM-5 AUD and a set of potential validators: indicators of alcohol use severity (i.e., craving, binge drinking frequency, problematic and harmful drinking), psychiatric disorders, and functional impairment. Multinomial logistic regression models examined the association between the predictors and the three AUD severity levels (mild, moderate, severe) vs the reference group, no AUD, controlling for sociodemographic characteristics and other substance use.

Results:

All alcohol use validators were associated with greater likelihood of all three AUD severity levels compared to the no-AUD group. However, psychiatric disorders were associated with severe AUD only and participants with major depression (aOR=2.44), post-traumatic stress disorder (aOR=1.65) borderline personality disorder (aOR=1.99), and antisocial personality disorder (aOR=1.78) had greater likelihood of severe AUD compared to the no-AUD group. Functioning validators were also associated with severe AUD only and participants with social (aOR=1.87), physical (aOR=1.62), or mental (aOR=1.84) impairment had greater likelihood of severe AUD relative to the no-AUD group. Many alcohol, psychiatric, and functioning validators remained associated with greater odds of severe AUD compared to the mild and moderate AUD groups.

Conclusion:

This study lends support to the criterion validity of the DSM-5 tri-categorical measure of AUD. Specifically, results fully supported the validity of severe AUD by its associations with all predictors, while the validity of mild and moderate AUD was partially supported since predictors were limited to alcohol use variables. Findings inform severity-specific interventions utilizing the DSM-5 AUD.

Introduction

In the United States, 66% of adults consume alcohol and 6% report heavy or high-risk drinking (Boersma, Villarroel, & Vahratian, 2018; Grant et al., 2017; Hasin et al., 2017). Excess alcohol use remains among one of the leading causes of preventable morbidity and mortality, and continues to account for a large proportion of substance-related fatalities (Probst et al., 2020; Rehm et al., 2017). Heavy drinking is a risk factor for developing Alcohol Use Disorders (AUD), which is characterized by a problematic and persistent pattern of alcohol use that contributes to significant impairment and distress (American Psychiatric Association [APA], 2013; Hasin et al., 2013). AUD is associated with a multitude of medical and psychiatric comorbidities (Castillo-Carniglia, Keyes, Hasin, & Cerdá, 2019; Xi et al., 2017) and also increases the risk of mortality as a result of injury, cancer, and cardiovascular disease, among other causes (Roerecke, & Rehm, 2014).

The diagnostic criteria for AUD have undergone considerable revision since the first Diagnostic and Statistical Manual of Mental Disorders (DSM; APA, 1952). The third edition, DSM-III (APA, 1980), was the first version to create an abuse/dependence classification using specific diagnostic criteria (Hasin et al., 1996). Published in 1987, the DSM-III-R (revised) further modified the AUD criteria for substance abuse and substance dependence based on empirical research and World Health Organization recommendations (APA, 1987; Edwards, Arif, & Hadgson, 1981; Kosten et al., 1987; Rounsaville, Kosten, Williams, & Spitzer, 1987), referring to dependence as a syndrome as first described in the 1980s (Edwards et al., 1986; Kosten et al., 1987). DSM-IV, published in 1994, maintained the AUD differentiation between abuse and dependence (Hasin & Grant, 1994; Hasin et al., 2006), but no longer referred to dependence as a syndrome but rather as a “cluster,” and removed the overlap between abuse and dependence criteria that had existed in DSM-III-R (Hasin & Grant, 1994). Since publication of DSM-IV, a large evidence base of epidemiological and clinical research supported modifying the AUD criteria to replace the distinction between abuse and dependence with a single disorder in DSM-5 (Dawson, Saha, and Grant, 2010; Hagman,& Cohn, 2011; Hasin & Beseler, 2009; Hasin et al, 2013). This disorder was defined by the same 11 AUD criteria formerly used to define abuse and dependence in DSM-IV, except that craving was added and legal problems removed. While previous studies have characterized DSM-IV and earlier AUD severity in terms of criteria counts (Hasin & Click, 1992), DSM-5 was the first version to formally define severity in three levels, with specific numbers of criteria used to define each one: mild (2–3 criteria), moderate (4–5 criteria), and severe (6+ criteria; Hasin et al., 2013). Although this modification was intended to identify individuals with varying AUD severity and differences in the extent of impairment and disability, the validity and clinical utility of the 3-level severity categorization has been debated (Fazzino et al., 2014a; Fazzino 2014b), including the mild AUD category with a threshold of just two symptoms (Borges et al., 2011; Goldstein et al., 2015).

Up to now, little empirical research has been available to inform the debate over the utility of the 3-level severity classification. Recently, clinical trial data suggested that DSM-5 AUD severity categories have clinical relevance due to their relationships with harmful levels of alcohol consumption and severity of alcohol-associated behavioral consequences (Kiluk et al., 2018). However, while important, this study only included 68 participants and covered AUD symptoms over the past 30 days, not the 12 months required for DSM-5 diagnosis. Thus, a knowledge gap exists on the validity and potential utility of the DSM-5 AUD severity differentiations. Specifically, information is needed on whether clinical characteristics e.g., the extent of alcohol use, impairment, comorbidity, and functioning, differentiate between the three-level AUD severity groups (mild, moderate, severe). Therefore, the purpose of this study was to examine the association between a set of external validators and the DSM-5 AUD severity classifications, including alcohol craving and consumption, functional impairment, and psychiatric conditions in a sample of 588 adults with some evidence of problematic substance use. We hypothesized that the validators would be associated with each AUD severity category, with stronger association with more severe disorder.

Materials and Methods

Participants and Procedures

Methods of the study have been described in detail elsewhere (Hasin et al., 2020). Briefly, the sample consisted of 588 adults age 18 years or older recruited between 2016–2019 at a suburban inpatient addiction treatment program (n=150) or at research offices in an urban medical center (n=438). Participants recruited at the medical center were informed of the study through flyers posted in the medical center, and also through newspaper and Facebook advertisements. All advertisements offered a brief explanation of the study and provided contact information for a master’s level research coordinator who screened for eligibility. Medical center participants were therefore not necessarily engaged in care. Eligibility requirements included aspects of substance use in the past 30 days or in the 30 days prior to inpatient substance use admission: binge drinking [i.e., ≥5 drinks or ≥4 drinks for men and women respectively] and/or illicit substance use (i.e., cannabis, cocaine, non-prescription opioids, etc.), and at least one DSM-5 substance use disorder (SUD) criterion. Adults were excluded from the study if they were non-English speaking, had a hearing, visual, or cognitive impairment that prevented participation, planned to relocate from the area (since it involved a follow-up component), or were currently psychotic, suicidal, or homicidal. After obtaining written informed consent, participants were assessed for demographic characteristics, substance use, mental health, and functional impairment, and compensated $50 for the baseline interview that provided the data presented below. The Institutional Review Boards of New York State Psychiatric Institute and South Oaks Hospital approved all procedures for the study.

Measures

Outcome

Mild, moderate and severe DSM-5 AUD

The Psychiatric Research Interview for Substance and Mental Disorders- 5 (PRISM-5), a semi-structured, computer-assisted interview developed specifically for adults with problem substance use, was administered by trained Master’s level clinicians. The PRISM-5 was used to assess the 11 DSM-5 AUD criteria among those who used alcohol at least 6 times within the past 12-months. DSM-5 AUD criteria include: 1) withdrawal, 2) tolerance, 3) increasing quantity or frequency of use/longer episodes of use, 4) persistent desire or unsuccessful attempts to decrease/control use, 5) lots of time spent obtaining, using or recovering from effects of alcohol, 6) social, occupational, or recreational activities given up or reduced because of use, 7) continued drinking despite knowledge of use causing or exacerbating a medical condition, 8) recurrent difficulties to fulfill major role obligations, 9) recurrent use in hazardous situations, 10) craving/strong desire to consume alcohol, 11) continued use despite interpersonal problems caused by alcohol consumption. The PRISM-5 produces DSM-5 AUD diagnoses via computer algorithm, and classifies participants based on disorder severity: no disorder (0–1 criteria), mild (2–3 criteria), moderate (4–5 criteria) and severe (≥6 criteria) disorder. PRISM-5 AUD diagnostic measures demonstrate good to excellent reliability and validity among adults who consume alcohol (Hasin et al., 2020), and have been used as validators of non-clinician interviews used in a national survey (Hasin et al., 2015). Further information about the PRISM-5 and procedures used in this study is found elsewhere (Hasin et al., 2020).

Validators

Alcohol-specific validators: Craving, problematic use, harmful use, and binge drinking frequency

Craving:

Alcohol craving was assessed using the Alcohol Urge Questionnaire (AUQ), which includes eight statements about the participant’s current thoughts and feelings about drinking, with responses forming a 7-item Likert scale ranging from “strongly disagree” to “strongly agree”. Alcohol craving scale scores were derived as the mean response value to the 8 statements, and ranged from 1–7 (Bohn, Krahn, & Staehler, 1995). The alcohol craving scale was assessed among those who either used alcohol in the past month or used alcohol regularly (3 or more times a week) for at least a year; thus, 44 individuals without such use were excluded from the analysis of alcohol craving (N=544). To ensure that association with the craving scale was not driven by the craving criterion, we adapted the AUD diagnostic set by excluding the craving criterion, and re-applied the diagnostic scheme to define adapted AUD severity using the set of 10 items with participants categorized into mild (2–3 criteria), moderate (4–5 criteria) and severe (≥6 criteria) disorder. In the adapted severity measure with craving removed, 12 respondents who had mild AUD no longer met criteria for the disorder.

Problematic use:

A binary variable indicating whether participants felt that alcohol was a major problem for them based on a question from the Addiction Severity Index (ASI; McLellan et al., 1992), which is widely used in research and clinical settings in self-administered format, as it was in the present study (Butler et al., 2001; Rosen et al., 2000).

Harmful use:

The Alcohol Use Disorders Identification Test (AUDIT), a questionnaire that assesses the usual number of drinks, frequency of six indications of risky drinking over the past year on a 5-point scale (never; less than monthly; monthly; weekly; and daily/almost daily), including binge drinking, difficulty controlling drinking, drinking upon waking, difficulties to fulfill major role obligations, feelings of guilt or remorse due to alcohol use, black-outs or memory lapses because of drinking, and two additional questions on alcohol associated injuries and familial/peer concern related to alcohol use (Allen et al., 1997; Saunders et al., 1993), was used to measure harmful use. Participant responses were summed, with a score of ≥8 indicating harmful alcohol use in the past year, consistent with AUDIT scoring guidelines. Participants were dichotomized based on this designation. The AUDIT is one of the most widely used measures of alcohol-associated difficulties and demonstrates strong psychometrics across clinical populations (Daeppen et al., 2000; de Meneses-Gaya et al., 2009; Higgins-Biddle & Babor, 2018).

Binge drinking frequency:

The AUDIT also assessed frequency (0= never; 1=<monthly; 2= monthly; 3=weekly; 4=daily/almost daily) of binge alcohol use (≥ 5 drinks on one occasion). Frequency of binge alcohol use was analyzed as count variable (0–4) corresponding with respondent answer choices.

Psychiatric validators

Five past 12-month disorders were included: Major Depressive Disorder (MDD), MDD determined not to be substance induced (MDD-NS), Post-traumatic stress disorder (PTSD), Borderline Personality Disorder (BPD), and Antisocial Personality Disorder (ASPD). The PRISM-5 was used for these assessments, with a full module assessing DSM-5 criteria for each disorder. For MDD, participants must have reported five or more symptoms during a 2-week period, with at least one of the symptoms either (1) depressed mood or (2) loss of interest or pleasure that caused significant impairment or distress. MDD-NS was defined as MDD not occurring during a period of chronic intoxication/binge use (using 4+ days a week for a month, intoxicated 3+ consecutive days, or intoxicated at least half the days within a month) or within one month of stopping or cutting down such use . For PTSD, at least one symptom was required from each of the following: Criterion A (i.e., direct exposure, witnessing event, learning of event, indirect exposure via traumatic details), Criterion B (i.e., upsetting memories, nightmares, flashbacks, emotional distress, physical reactivity), Criterion C (avoidance of external reminders, thoughts, or feelings); and two symptoms from Criterion D (anhedonia, isolation, negative affect, self-blame, difficulty recalling memories related to trauma) and Criterion E (aggressiveness or irritability, reckless behavior, hypervigilance, exaggerated startle response, problems with concentration, sleep disturbance); for at least one month, causing significant impairment or distress. BPD required five of the nine criteria across the domains of emotional dysregulation, impulsive behaviors, distorted perceptions of self, and unstable relationships over lifetime, with at least one symptom occurring in the past 12 months. ASPD required stable impairment in personality functioning (i.e., egocentrism or absence of prosocial behavior), interpersonal functioning (i.e., lack of empathy or incapacity for mutually intimate relationships), and pathological personality traits in the following domains: antagonism (i.e., manipulativeness, deceitfulness, callousness, hostility), and disinhibition (irresponsibility, impulsiveness, risk taking) since the age of 18. The PRISM-5 has been used to assess psychiatric conditions in multiple studies, and demonstrates strong reliability and validity among adults reporting substance use (Hasin et al., 1996; Hasin et al., 2006; Hasin et al., 2015; Morgello et al., 2006; Ramos-Quiroga et al., 2012; Torrens et al., 2004).

Depressive symptoms

The Patient Health Questionnaire (PHQ-9) was used as an alternative measure of depression. The PHQ-9 assesses self-reported depressive symptoms over the past two weeks, is widely used and has excellent reliability and validity, including in SUD populations (Kroenke, Spitzer, & Williams, 2001; Lowe et al., 2004; Delgadillo et al., 2011). Each of the nine items ranges from 0 “not at all” to 3 “nearly every day.” Responses were summed (total score 0–27), and the summary score was converted to a 5-level count variable: 1=0–4 points (minimal or no problems); 2=5–9 points (mild problems); 3=10–14 points (moderate problems); 4=15–19 points (moderately severe problems); and 5=20+, (severe problems; Kroenke, Spitzer, & Williams, 2001).

Functioning validators

Social impairment:

The Social Adjustment Scale Self-Report (SAS-SR) was used to assess social functioning. This widely used, reliable, self-report measure with assesses functioning in the prior two weeks across areas of work/school, social/leisure time, and familial relationships (Weissman et al., 2001; Weissman & Bothwell, 1976; Weissman et al., 1978). Each item is assessed on a 5-point scale, with higher scores indicating greater impairment. Consistent with scoring guidelines, the scale was scored as the mean of the responses within the domains of (1) employment and/or educational responsibilities (2) social and leisure activities, (3) relationships with extended family, (4) role as a marital partner, (5) parental role, and (6) role within the family unit, including perceptions about economic functioning (range, 1–5). The role areas not relevant to the respondent’s situation were skipped (Weissman & Bothwell, 1976).

Physical and mental impairment:

The Medical Outcomes Study Short Form 12-Item (SF-12) measures functioning in eight subdomains (i.e., general health, physical functioning, role physical, bodily pain, vitality, social functioning, role emotional, and mental health; Ware, Kosinski, & Keller, 1996). Using SF-12 scoring guidelines, composite Physical Component Summary (PCS) and Mental Component Summary (MCS) scores were calculated from the subdomains to indicate the extent to which physical, work, and social activities or accomplishments were negatively impacted by physical or mental health challenges in the prior 4 weeks. PCS and MCS scores are valid, reliable, widely used, and have been shown to be associated with DSM-5 AUD in several general population studies (Kirouac et al., 2017; Grant et al., 2015; Rubio et al., 2013). To facilitate interpretation for each composite score, functional impairment was defined as scoring in the bottom 25^th percentile, similar to other studies examining alcohol use (Aharonovich et al., 2017).

Control variables

Sociodemographic characteristics included age, sex (male; female); race/ethnicity (non-Hispanic White; Hispanic, non-Hispanic Black, Asian, Native Hawaiian/Pacific Islander, American Indian or Native Alaskan); education (no college; at least some college), current employment (unemployed, any employment [full or part-time]), marital status (unmarried; married/living with partner), housing status (homeless or group home; stably housed), and recruitment setting (inpatient; community). Other substance use (tobacco, cannabis, crack/cocaine, heroin, stimulants, sedatives, non-prescription opioids, and sedatives) was indicated for each substance other than alcohol by items in the PRISM-5, with variables indicating past month use.

Statistical analysis

Bivariate statistics were used to indicate potential differences in sociodemographic characteristics and the validators by AUD severity levels using Chi-Square or ANOVA tests for categorical and continuous variables, respectively. Separate multinomial logistic regression models were then fit for each validator to assess the relationship between validators and the four-level outcome of AUD severity (mild, moderate, severe, vs none), adjusting for age, sex, race/ethnicity, education, and recruitment setting. Multinomial regression was used rather than ordinal regression because the proportional odds assumption was rejected for each model (Brant, 1990). The models were used to indicate adjusted odds ratios (aOR) with 95% confidence limits for each severity level compared to the reference of no AUD. Odds ratio are interpreted as the change in likelihood (or odds) of AUD severity level for those with (as compared to without) a binary predictor, or for a one unit increase in a count predictor. The “test” statement in the logistic regression analyses was used to determine if the association was in a dose-dependent manner, i.e., if the association strength (regression coefficient) was greater for severe vs. none than moderate vs. none, for moderate vs. none than mild vs. none, and for severe vs. none than mild vs. none. For predictors that were not alcohol specific (i.e., functioning [social, mental, and physical] and psychiatric disorders), to ensure that associations were not due to co-occurring substance use, regression models were re-run, also adjusting for use of the other substances (Supplementary Table 2). Additional models were run to indicate aOR for severe AUD vs. mild, severe vs. moderate, and moderate vs. mild (Supplementary Table 3). Analyses were conducted using SAS 9.4.

Results

Sociodemographic characteristics and validators

Sample characteristics were as follows: male (69.7%), Black (47.7%), 50 and older (42.1%), unemployed (74.1%), unmarried (80.1%), and high school education or less (54.7%). Approximately 19% of the sample reported daily/almost daily binge drinking in the past month. Nearly half (48.0%) felt they had a major problem with alcohol, and 61.4% met the AUDIT-C threshold for harmful alcohol use. Twelve-month prevalence of mild, moderate, and severe AUD was 12.2%, 13.4%, and 40.3%, respectively. The most commonly reported substances used in the past month (other than alcohol) were tobacco (64.3%), cannabis (59.0%), crack/cocaine (48.6%) and heroin (24.3%). Prevalence’s of the psychiatric conditions were as follows: MDD (39.3%), MDD-NS (25.8%), BPD (37.9%), ASPD (26.0%), and PTSD (20.1%).

Bivariate Association between Sociodemographic Characteristics, Validators, and AUD Severity

Distribution of AUD severity differed by race/ethnicity (p=0.006), employment (p=0.018) and recruitment setting (p<0.0001; Table 1). AUD status was associated with all alcohol validators, including craving (p<0.0001), binge drinking frequency (p<0.0001), problematic use (p<0.0001) and harmful use (p<0.0001). Moreover, AUD severity was associated with illicit substance use, including use of non-prescription opioids (p=0.045) and heroin (p=0.001; Supplementary Table 1). AUD severity also varied by social (p<0.0001), physical (p=0.022) and mental (p<0.0001) impairment as well as psychiatric conditions such as MDD (p<0.0001), MDD-NS (p<0.0001), ASPD (p=0.006), and BPD (p<0.0001; Table 2).

Table 1:

Demographic Characteristics by DSM-5 AUD Severity Level (N=588)

	DSM-5 AUD Severity
	No Disorder (n=200)	Mild (n=72)	Moderate (n=79)	Severe (n=237)	Total (N=588)	Test Statistica	p
Demographic characteristics	% or M (SD)	% or M (SD)	% or M (SD)	% or M (SD)	% or M (SD)
Age (18–74)	41.5 (13.76)	42.7 (14.82)	46.5 (12.87)	44.91 (11.99)	43.7 (13.19)	3.80	0.010
Sex						2.72	0.843
Male	68.0	73.6	68.4	70.0	69.7
Female	31.5	26.4	31.6	30.0	30.3
Race/ethnicity						22.90	0.006
White	32.0	19.4	13.9	30.8	27.6
Black	39.5	58.3	63.3	46.4	47.8
Hispanic	22.0	13.9	20.3	17.7	19.0
Other	6.5	8.3	2.5	5.1	5.6
Education						2.27	0.519
High School or less	55.0	54.2	62.0	52.3	54.8
College or more	45.0	45.8	38.0	47.7	45.2
Marital status						3.61	0.307
Not Married	81.0	87.5	78.5	77.6	80.1
Living together/ Married	19.0	12.5	21.5	22.4	19.9
Employment						10.03	0.018
Unemployed	73.0	62.5	84.8	75.1	74.1
Any employment	27.0	37.5	15.2	24.9	25.9
Housing						2.53	0.4694
Group home or homeless	9.0	5.6	5.1	9.7	8.3
Stably housed	91.0	94.4	94.9	90.3	91.7
Recruitment setting						27.82	<0.0001
Inpatient	34.0	9.7	10.1	28.3	25.5
Community	66.0	90.3	89.9	71.7	74.5

Bold values indicate p<0.05

^achi square value for categorical predictors and F-value for continuous predictors

Table 2:

Bivariate Relationship of Validators and DSM-5 AUD Severity Levels (N=588)

	DSM-5 AUD Severity
	No Disorder (n=200)	Mild (n=72)	Moderate (n=79)	Severe (n=237)	Total (N=588)	Test Statisticb	p
Alcohol, psychiatric, and functioning validators	% or M (SD)	% or M (SD)	% or M (SD)	% or M (SD)	% or M (SD)
Alcohol validators
Cravinga, c	2.0 (0.9)	2.55 (1.3)	3.17 (1.3)	3.25 (1.6)	2.78 (1.5)	30.43	<0.0001
Problematic used	17.5	40.3	62.0	71.3	48.0	134.08	<0.0001
Harmful usee	18.0	54.2	79.7	94.1	61.4	278.62	<0.0001
Daily/almost -daily binge alcohol usef	3.7	12.7	16.5	32.9	19.3	188.04	<0.0001
Psychiatric validators
MDD	34.0	27.8	24.1	52.3	39.3	30.91	<0.0001
MDD-NSg	16.0	18.1	13.9	32.1	25.8	21.47	<0.0001
PTSD	16.0	19.4	22.8	22.8	20.1	3.53	0.316
ASPD	24.0	13.9	21.5	32.9	26.0	12.61	0.006
BPD	35.0	33.3	20.3	47.7	37.9	21.43	<0.0001
Depressive symptomsh						47.51	<0.0001
None-Minimal	42.5	56.9	41.8	23.6	36.6
Mild	27.0	20.8	36.7	34.2	30.4
Moderate	18.0	12.5	6.3	19.4	16.3
Moderately severe	8.5	9.7	11.4	13.5	11.1
Severe	4.0	0.0	3.8	9.3	5.6
Functioning validators
Social impairmenti	2.26 (0.7)	2.06 (0.7)	2.17 (0.6)	2.51 (0.7)	2.32 (0.7)	11.96	<0.0001
Physical impairmentj	19.0	22.2	34.2	29.1	25.5	9.61	0.022
Mental impairmentk	24.5	11.1	10.1	34.6	25.0	28.40	<0.0001

Bold values indicate p<0.05; MDD= Major depressive disorder; PTSD= Posttraumatic stress disorder; ASPD= Antisocial personality disorder; BPD= Borderline personality disorder; PHQ-9=Patient Health Questionnaire-9

^aN was 544 due to exclusion of those who did not use alcohol in the past month or regularly in the past year

^bchi square value for categorical predictors and F-value for continuous predictors

^ccontinuous measure from 1–7, higher values indicate more severe craving

^dbinary variable of perceived problematic alcohol use

^eAUDIT score ≥8

^fcount predictor with five response levels: never; less than monthly; monthly; weekly; and daily/almost daily

^gMDD not occurring during a period of chronic intoxication/binge substance use or within one month of stopping or cutting down use.

^hPHQ-9 assesses depressive symptoms over the past two weeks: 1=0–4 (minimal or no problems), 2=5–9 (mild problems), 3=10–14 (moderate problems), 4=15–19 (moderately severe problems), and 5=20+ (severe)

ⁱcontinuous measure from 1–5, higher values indicate worse functioning

^jimpairment is indicated by scoring in the lowest quartile of the SF-12 physical component summary score

^kimpairment is indicated by scoring in the lowest quartile of the SF-12 mental component summary score

Adjusted Associations of Alcohol-specific Validators with Mild, Moderate, and Severe AUD (vs no AUD)

After controlling for covariates, each of the alcohol-specific validators were associated with all three AUD severity groups (versus no AUD). Specifically, for each additional increase in craving (1–7, higher values indicate more severe craving), participants demonstrated greater odds of mild (aOR= 1.53), moderate (aOR= 2.14), or severe (aOR= 2.48) AUD versus the no AUD group and there was a stronger magnitude of association for severe AUD and moderate AUD than for mild AUD. Similarly, participants endorsing problematic alcohol consumption had greater odds of mild (aOR= 3.25), moderate (aOR= 7.89), or severe (aOR= 11.87) AUD versus the no-AUD group, with stronger association for severe AUD and moderate AUD than for mild AUD. Participants with harmful alcohol use demonstrated greater odds of mild (aOR= 4.92), moderate (aOR= 16.68), or severe (aOR= 76.70) AUD versus no AUD, with a dose-dependent association (stronger for severe AUD than moderate AUD and mild AUD, and stronger for moderate AUD than mild AUD). Lastly, for each increase in binge drinking frequency (never, <less than monthly, monthly, weekly, daily/almost daily) participants were more likely to have mild (aOR= 1.63), moderate, (aOR= 2.39), or severe (aOR= 3.49) AUD versus no AUD, with a dose-dependent association. In direct comparisons, participants with moderate or severe AUD demonstrated greater odds of all alcohol validators compared to the mild AUD group. Participants with moderate and severe AUD did not differ significantly on measures of craving and problematic use (Supplementary Table 3).

Adjusted Associations of Psychiatric Validators with Mild, Moderate, and Severe AUD (vs no AUD)

Diagnoses of psychiatric disorders and depressive symptoms were not associated with mild or moderate AUD. However, participants with MDD (aOR=2.44), MDD-NS (aOR=2.84), PTSD (aOR=1.65), BPD (aOR=1.99), or ASPD (aOR=1.78) had greater odds of severe AUD relative to the no-AUD group. Similarly, each additional point in depressive symptom severity category (i.e., minimal, mild, moderate, moderately severe, or severe), conferred greater odds of severe AUD (aOR=1.41) versus no AUD (Table 4). In direct comparisons, the severe AUD group had greater likelihood of most psychiatric conditions compared to the mild and moderate AUD groups. Participants with moderate and mild AUD did not differ significantly on measures of psychiatric disorders.

Table 4:

Association of Psychiatric and Functioning Validators with DSM-5 AUD, by Severity Level (N=588)^a

				Difference in association strength by severity levelb
	Mild	Moderate	Severe	Severe compared to moderate	Moderate compared to mild	Severe compared to mild
	AOR (95% CI)	AOR (95% CI)	AOR (95% CI)	Wald x2, p	Wald x2, p	Wald x2, p
Psychiatric validators c
ASPD	0.59 (0.28, 1.28)	1.11 (0.57, 2.13)	1.78 (1.14, 2.77)	2.22, 0.1360	1.92, 0.1654	8.37, 0.0038
BPD	1.17 (0.64, 2.15)	0.62 (0.32, 1.19)	1.99 (1.32, 3.00)	13.22, 0.0003	2.69, 0.1010	3.17, 0.0750
MDD	0.96 (0.51, 1.78)	0.79 (0.42, 1.46)	2.44 (1.62, 3.66)	13.77, 0.0002	0.25, 0.6143	9.41, 0.0022
MDD-NS	1.41 (0.67, 2.97)	1.10 (0.51, 2.39)	2.84 (1.75, 4.62)	6.57, 0.0104	0.28, 0.5996	3.92, 0.0477
PTSD	1.44 (0.70, 2.97)	1.91 (0.97, 3.77)	1.65 (1.00, 2.71)	0.21, 0.6455	0.47, 0.4923	0.15, 0.6974
Depressive symptomsd	0.77 (0.59, 1.02)	0.98 (0.77, 1.25)	1.41 (1.20, 1.67)	9.53, 0.0020	2.15, 0.1428	18.97, <0.0001
Functioning validators
Social impairmente	0.73 (0.46, 1.15)	0.97 (0.63, 1.47)	1.87 (1.39, 2.51)	10.07, 0.0015	1.08, 0.2991	16.82, <0.0001
Physical impairmentc, f	1.17 (0.58, 2.34)	1.78 (0.96, 3.30)	1.62 (1.01, 2.60)	0.10, 0.7473	1.17, 0.2795	0.93, 0.3336
Mental impairmentc, g	0.45 (0.20, 1.04)	0.44 (0.19, 1.01)	1.84 (1.18, 2.87)	12.17, 0.0005	0.00, 0.9670	11.52, 0.0007

Bold values indicate p<0.05; MDD= Major depressive disorder; PTSD= Posttraumatic stress disorder; ASPD= Antisocial personality disorder; BPD= Borderline personality disorder; PHQ-9=Patient Health Questionnaire-9

^aestimated from multinomial logistic regression analysis, adjusted for gender, age, race/ethnicity, education and recruitment setting

^bin the logistic regression model, difference in association was tested by determining if the regression coefficients differed significantly from each other, based on Wald χ2 test with 1 degree of freedom

^creference group = no

^dPHQ-9 assesses depressive symptoms over the past two weeks,1=0–4 (minimal or no problems), 2=5–9 (mild problems), 3=10–14 (moderate problems), 4=15–19 (moderately severe problems), and 5=20+ (severe)

^econtinuous measure from 1–5, higher values indicate worse functioning

^fimpairment is indicated by scoring in the lowest quartile of the SF-12 physical component summary score

^gimpairment is indicated by scoring in the lowest quartile of the SF-12 mental component summary score

Adjusted Associations of Functioning Validators with Mild, Moderate, and Severe AUD (vs no AUD)

A one-unit increase (worse) in social impairment resulted in greater odds of severe AUD (aOR=1.87) compared to the no-AUD group. Moreover, participants with physical (aOR=1.62) or mental (aOR=1.84) impairment were more likely to have severe AUD versus participants without AUD. In direct comparisons, mental impairment and a one-unit increase in social impairment resulted in greater odds of severe AUD compared to the moderate and mild AUD groups. Moderate and mild AUD groups did not differ significantly on measures of functioning.

Associations of Alcohol, Psychiatric and Functioning Validators with Mild, Moderate, and Severe AUD vs no AUD, Adjusted for Other Substance Use

After also controlling for other substance use, the mild AUD group demonstrated lower odds of mental impairment (aOR=0.42) compared to the no-AUD group. All other associations between validators and DSM-5 AUD severity levels were essentially the same: alcohol validators were associated with greater odds of all three AUD severity groups, and all psychiatric and functioning validators were associated with greater odds of severe AUD compared to the non-AUD group (Supplementary Table 2).

Discussion

This study examined the association between diagnostic validators and DSM-5 AUD severity classification (i.e., mild, moderate, severe) in a sample of 588 adults reporting problematic past 30-day alcohol and illicit substance use. Our hypotheses were partially supported, as DSM-5 validators of alcohol use severity, e.g., craving, problematic and harmful use, and binge drinking frequency were associated with all classifications of AUD compared to no AUD, in a generally dose-dependent manner. However, a different pattern emerged for impaired physical, mental, and social functioning and for psychiatric disorders, which were only associated with severe AUD. This study lends support to the criterion validity of the DSM-5 tri-categorical measure of AUD. Specifically, findings fully supported the validity of severe AUD by its associations with all predictors, while the validity of mild and moderate AUD was partially supported since predictors were limited to alcohol use variables. These findings provide diagnostic and treatment implications specific to each AUD severity group, and suggest that more intensive interventions are needed for severe AUD, as this group demonstrated the greatest alcohol use, comorbidity, and impairment.

All alcohol variables were associated with all AUD severity groups. These results are in line with those of previous studies showing that measures of alcohol use have high predictive validity of AUD (Borges et al., 2011; Dawson, Saha, and Grant, 2010). As expected, the aOR was generally stronger for all alcohol validators as AUD severity increased. This is consistent with early thinking about alcohol use disorders (Edwards & Goss, 1976) that people with more severe AUD should endorse greater difficulty across a spectrum of alcohol-associated symptoms. However, in this study, further analysis showed that participants with moderate and severe AUD did not differ significantly on measures of alcohol craving and perceived problematic use. Moreover, the strikingly strong relationship between AUDIT-C harmful alcohol use (aOR=76.70) and severe AUD generates additional support for use of the AUDIT as an important screening tool for AUD, particularly for identifying adults with severe AUD. This finding is not surprising, given the overlap between some AUDIT items and DSM-5 AUD criteria. However, notably, the magnitude of effect for all alcohol validators was significantly lower among the mild AUD group compared with the moderate and severe group. Given these findings, an assessment of mild DSM-5 AUD may be a good indicator for brief intervention focused on alcohol psychoeducation. Such brief intervention may be a sufficient treatment to elicit reductions in alcohol use among adults with mild AUD, and should be considered as a routine part of healthcare appointments, particularly in primary care (Knox et al., 2019). In contrast, adults with more severe AUD would likely benefit from a medically supervised alcohol detoxification followed by use of medication assisted treatment (e.g., disulfiram, naltrexone, topiramate; Kranzler & Soyka, 2018) and evidence-based behavioral therapies. Specifically, cognitive behavioral therapies, motivational interviewing techniques, and acceptance and mindfulness based interventions are among the most effective behavioral treatments for AUD (Witkiewitz, Litten, & Leggio, 2019) and are particularly important for enhancing recovery among patients in residential treatment (de Andrade et al., 2019). Moreover, social rehabilitation strategies such as linkage to employment and stable housing may be beneficial for this group as socioeconomic issues were prevalent among participants with severe AUD.

In our sample, psychiatric comorbidity and pronounced mental health impairment were strong clinical validators of severe AUD, only. These findings are in line with a large national study demonstrating higher rates of MDD, ASPD, and BPD among adults with severe AUD (Grant et al., 2015) and also have implications for AUD diagnosis. The etiological link between psychiatric conditions and AUD is complex and comorbid mental health symptoms can perpetuate AUD (Castillo-Carniglia, Keyes, Hasin, & Cerdá, 2019). It is possible that psychiatric functioning has less clinical salience among adults with mild and moderate AUD, in part, due to one’s intact ability to sustain relationships, fulfill obligations, and utilize adaptive coping strategies to manage psychosocial stressors at lower AUD severity thresholds. Mental health may begin to deteriorate exponentially with greater alcohol consumption, which was demonstrated by high rates of daily binge alcohol use among the severe AUD group in this study. MDD is particularly prevalent among adults with severe AUD, and therefore a combination of behavioral and pharmacological treatments that simultaneously improve mental health and AUD symptoms may provide additional benefit and assist with long-term recovery in this group (Ray et al., 2020). mHealth interventions have also shown promise and their accessibility may be useful to improve AUD outpatient treatment retention and engagement among adults with severe AUD (Riper et al., 2014; Suffoletto & Scaglione, 2018).

Social and physical functioning validators were associated with severe AUD, only, which is similar to a finding from another study demonstrating that “social and interpersonal problems” was the strongest indicator of DSM-5 severe AUD (Preuss, Watzke, & Wurst, 2014). Social functioning among adults with mild or moderate AUD may be largely unaffected, with impairment only apparent at a severe acuity threshold where greater time is spent engaging in alcohol-related behaviors. Further, heavy alcohol use is also associated with impaired communication skills and social cognition (Bora & Zurlu, 2017), deficits that can compound relationship difficulties. Given these findings, adults with severe AUD may benefit most from intensive group treatments (e.g., social skills training) to build social connectedness and support (Brooks et al., 2017; Ray et al., 2019). Community support groups, such as Alcoholics Anonymous or other 12-step facilitation, are another method for enhancing social support and subsequent recovery, and may be particularly useful as a supplementary intervention for AUD after completion of residential treatment (Kelly, Humphreys, & Ferri, 2020). In addition, physical impairment was only associated with severe AUD. Alcohol misuse is a robust correlate of accident/injury (White et al., 2018), missed routine healthcare appointments (McQueenie et al., 2019), and medical complications (Molina & Nelson, 2018), which are negative consequences that may be contributing to the high rates of physical impairment experienced by members of the severe AUD group in this study. Multidisciplinary treatments that improve physical health and functioning may be useful for those with severe AUD, especially since physical impairment is a barrier to AUD treatment utilization (Prangley et al., 2018).

Study limitations are noted. First, our study utilized self-report measures of substance use and mental health and participants may have under-reported their symptoms. Moreover, reporting on indications of problematic or harmful alcohol use necessitates awareness, recognition, and vulnerability that may be lacking among participants unable or unwilling to acknowledge issues with alcohol use. Second, the study utilized a cross-sectional design; thus, we cannot determine the temporality of the observed relationships. Third, approximately one-fourth of the participants were currently in treatment for a SUD, and thus were likely to currently abstain from using alcohol. These individuals were also likely receiving mental health services, which may have improved symptomatology and subsequent scores on measures of depressive symptoms (PHQ-9), social functioning (SAS), and mental health impairment (SF-12). However, the examined validators in this study were associated with the corresponding AUD severity groups even after controlling for recruitment setting (i.e., inpatient treatment vs a convenience non-hospitalized community sample recruited at a major medical center) and subsequent treatment utilization. Fourth, the sample size of the mild and moderate AUD groups was smaller than the severe AUD group, which may have limited statistical power to detect associations between validators and less severe AUD levels. Despite these limitations, this study highlights important correlates of varying AUD severity levels, and elucidates potential diagnostic challenges with mild and moderate AUD. This study benefited from use of PRISM-5 diagnoses for AUD and psychiatric conditions, and also from use of a large cohort of adults with problematic substance use.

Our study demonstrated that the alcohol use validators were associated with all AUD severity groups, while functional impairment and psychiatric comorbidity predicted only severe AUD. The minimum clinical cutoff of DSM-5 AUD and its operationalization has been debated, which has led to attempts to modify and optimize AUD diagnostic criteria (Martin et al., 2010; Fazino et al., 2014a), though there is no current universal consensus for a new diagnostic definition of AUD. Considering the extent of impairment, comorbidity, and harmful alcohol use patterns among the severe AUD group, this is suggestive of a distinct clinical group with more pronounced symptomology, while the mild and moderate severity thresholds provide useful information when the clinical or research purpose is to designate less severe alcohol-related disorders for prevention or treatment. Specifically, brief interventions may be efficient for those with mild or even moderate AUD, while referral to more intensive treatment programs is likely necessary for those with a severe disorder, though treatment engagement following referral remains a challenge for this group (Knox et al., 2019). This study also highlights the importance of examining AUD as a three-level construct instead of a dichotomous measure indicating presence of any AUD vs. none, an ongoing limitation in empirical studies (Takahashi et al., 2017; Wakefield & Schmitz, 2014). Future investigations should seek to examine the criterion validity of the DSM-5 AUD three-level severity distinction by using longitudinal designs to evaluate change in mental health and functioning over time, along with their association with AUD severity classification.

Table 3:

Association of Alcohol Validators with DSM-5 AUD, by Severity Level (N=588)

	Association between validator and each severity levela			Difference in association strength by severity levelb
	Mild	Moderate	Severe	Severe compared to moderate	Moderate compared to mild	Severe compared to mild
Alcohol validators	AOR (95% CI)	AOR (95% CI)	AOR (95% CI)	Wald x2, p	Wald x2, p	Wald x2, p
Cravingc	1.53 (1.20, 1.96)	2.14 (1.68, 2.72)	2.48 (2.00, 3.06)	2.14, 0.1438	7.01, 0.0081	18.71, <0.0001
Problematic used, e	3.25 (1.74, 6.07)	7.89 (4.29, 14.53)	11.87 (7.37, 19.12)	2.07, 0.1507	6.60, 0.0102	19.64, <0.0001
Harmful used, f	4.92 (2.68, 9.05)	16.68 (8.52, 32.65)	76.70 (39.42, 149.24)	14.26, 0.0002	10.71, 0.0011	52.77, <0.0001
Binge drinking frequencyg	1.63 (1.27, 2.10)	2.39 (1.86, 3.06)	3.49 (2.80, 4.35)	11.73, 0.0006	8.00, 0.0047	39.95, <0.0001

Bold values indicate p<0.05

^aestimated from multinomial logistic regression analysis, adjusted for gender, age, race/ethnicity, education, and recruitment setting.

^bin the logistic regression model, difference in association was tested by determining if the regression coefficients differed significantly from each other, based on Wald χ² test with 1 degree of freedom.

^ccontinuous measure from 1–7, higher values indicate more severe craving

^dreference group = no

^ebinary variable of perceived problematic alcohol use

^fAUDIT score ≥8

^gcount predictor with five response levels: 0=never; 1=less than monthly; 2=monthly; 3=weekly; and 4=daily/almost daily