Demographic differences in the cascade of care for unhealthy alcohol use: A cross-sectional analysis of data from the 2015–2019 National Survey on Drug Use and Health

Abstract

Background:

The screening, brief intervention, and referral to treatment (SBIRT) model is recommended by the U.S. Preventive Services Task Force to improve recognition of and intervention for unhealthy alcohol use. How SBIRT implementation differs by demographic characteristics is poorly understood.

Methods:

We analyzed data from the 2015–2019 National Survey on Drug Use and Health from respondents ≥18 years old who used an outpatient clinic and had at least one alcoholic drink within the past year. Respondents were grouped into one of three mutually exclusive groups: “no binge drinking or alcohol use disorder (AUD),” “binge drinking without AUD,” or “AUD.” Outcome variables were likelihood of screening, brief intervention (BI), referral to treatment (RT), and AUD treatment. The demographic predictors on which outcomes were regressed included gender, age, race and ethnicity, sexual orientation, insurance status, and history of military involvement. Consistent with SBIRT guidelines, the entire sample was included in the screening model; screened persons with either binge drinking without AUD or with AUD were included in the BI model; screened persons with AUD were included in the RT model, and persons referred to treatment with AUD were included in the AUD treatment model.

Results:

Analyses included 120,804 respondents. Women were more likely than men to be screened, but less likely to receive BI or RT. When referred to treatment, women were more likely than men to receive it. Persons aged ≥50 were least likely to be screened about alcohol, but most likely to receive BI, while persons aged 18–25 were least likely to receive BI or AUD treatment. Racial and ethnic minorities were less likely than White persons to be screened; Asians were less likely to receive RT, and Black persons were less likely to receive treatment than White persons. Persons identifying as gay, lesbian, or bisexual were equally as likely or more likely to receive SBIRT or AUD treatment as those identifying as heterosexual. Persons without insurance were less likely to be screened than those with insurance. Persons with a history of military involvement were more likely to be screened and receive BI and RT than persons who had not served in the military.

Conclusions:

Demographic disparities in SBIRT implementation exist. Addressing the sources of these disparities and minimizing attrition from care could improve outcomes for persons with unhealthy alcohol use.

INTRODUCTION

Alcohol misuse is the leading risk factor for death among individuals aged 15–49 years (GBD 2016 Alcohol Collaborators, 2018), yet fewer than 10% of persons with alcohol use disorder (AUD) in the United States receive treatment (Mintz et al., 2021; National Institute on Alcohol Abuse and Alcoholism, 2023). The screening, brief intervention, and referral to treatment (SBIRT) model is one public health strategy to close this gap and is recommended by the U.S. Preventive Services Task Force (U.S. Preventive Services Task Force et al., 2018). In this approach, universal screening to assess for unhealthy alcohol use is recommended for adults aged 18 and older in primary healthcare settings. Unhealthy alcohol use, broadly defined, is any drinking pattern that exceeds the U.S. Dietary Guidelines for “low-risk” drinking—that is, no more than seven drinks per week or three drinks per occasion for women and no more than 14 drinks per week or four drinks per occasion for men (U.S. Department of Agriculture and U.S. Department of Health and Human Services, 2020).

For persons identified as engaging in unhealthy alcohol use, brief intervention is recommended. Brief interventions vary in time and intensity, but often include counseling on recommended alcohol use patterns and recommendations to cut down on drinking (U.S. Preventive Services Task Force et al., 2018). For persons who are identified via screening as having unhealthy drinking patterns that are not severe enough to warrant a diagnosis of AUD, brief interventions have been associated with decreased alcohol consumption (Kaner et al., 2009). Given negative health consequences associated with this type of unhealthy alcohol use including blood pressure elevations (Roerecke et al., 2017) and increased risk of injuries (Martin et al., 2016), brief interventions in this population could have a significant public health impact.

For persons with AUD—which occurs when the severity of unhealthy alcohol use causes significant functional impairment—brief intervention alone is unlikely to be effective (Saitz, 2010), and thus referral to treatment—which may include psychotherapy, medication(s), or a combination thereof—is recommended (Substance Abuse and Mental Health Services Administration, 2013).

We previously examined the degree to which SBIRT is implemented in real-world settings using a cascade of care framework, which has been used to identify gaps in care continuums for other chronic diseases (Gardner et al., 2011; Williams et al., 2019). Using data from the 2015–2019 National Survey on Drug Use and Health (NSDUH), we measured the prevalence of the following “steps” in care: AUD prevalence, healthcare utilization, screening, brief intervention, referral to treatment, and treatment. We found that among an estimated 19,251,648 U.S. adults with AUD, 80% used health care in the past year and 70% were screened, but subsequent steps of care were far less common: 12% received brief intervention, 5% were referred to treatment, and only 6% received treatment (Mintz et al., 2021).

Whether demographic characteristics influence this pattern of care is a critical question to answer so that efforts to mitigate leakage from the AUD care cascade are implemented effectively and equitably. There is a large body of epidemiologic research demonstrating associations between certain demographic characteristics and AUD prevalence. For example, it has been repeatedly shown that AUD is more prevalent among men than women (Delker et al., 2016; Grant et al., 2015; White, 2020), although the magnitude of this gender difference has narrowed in recent decades as the prevalence of AUD among women has increased (White, 2020). AUD is most common among Native Americans and White persons and less common among Black, Hispanic, and Asian persons (Delker et al., 2016; Grant et al., 2015). In general, AUD is more prevalent among younger adults than those over the age of 50 (Delker et al., 2016; Saunders & Rudowitz, 2022) with the highest prevalence of AUD occurring among adults aged 18–29 (Grant et al., 2015). In more recent years, increased attention has been paid to associations between sexual orientation and unhealthy alcohol use risk, with persons who identify as gay, lesbian, or bisexual thought to have a higher risk of problematic alcohol use than those who identify as heterosexual (Compton & Jones, 2021; McCabe et al., 2019).

However, relationships between demographic variables and subsequent steps of AUD care—particularly SBIRT and AUD treatment—are less well defined, and few data examining these relationships have been collected within the past decade. NSDUH data from 2013 to 2014 suggest that women with AUD are asked about alcohol use by healthcare professionals more commonly than men (Glass et al., 2016), but are less likely to receive subsequent intervention (Glass et al., 2016; Williams et al., 2017), including AUD treatment (Alvanzo et al., 2014; Dawson, 1996; Gilbert et al., 2019; Simpson et al., 2020). Racial and ethnic differences in AUD care have also been observed: Data from the National Epidemiologic Survey on Alcohol and Related Conditions from 2001 to 2002 and 2004 to 2005 showed that Hispanic persons are less likely to receive brief interventions about alcohol misuse than non-H ispanic White persons (Mulia et al., 2014). In addition, national survey data from the 1990s and early 2000s indicate that Black and Hispanic persons with AUD are less likely to utilize AUD treatment than their White counterparts (Chartier & Caetano, 2011).

Age has been associated with the likelihood of alcohol screening, with older adults being less likely to be screened about alcohol than young adults (McKnight-Eily et al., 2017; Sahker & Arndt, 2017). Whether age is associated with brief intervention, treatment referral, or AUD treatment, however, is not well understood. In addition, despite data showing strong associations between sexual orientation and AUD prevalence, the degree to which sexual orientation is associated with SBIRT and AUD treatment patterns is unknown.

Because SBIRT services are now reimbursable by most insurances (Zoorob et al., 2017), whether insurance status influences the relationship between demographic variables and the AUD cascade of care should be examined. In addition, given substantial efforts in recent years to improve SBIRT implementation within the Veteran’s Health Administration (VHA) healthcare system (e.g., Bradley et al., 2006; Williams et al., 2018)—where more than 25% of Veterans receive health care (U.S. Department of Veteran Affairs, 2020)—accounting for Veteran status when measuring associations between demographics and AUD cascade of care steps is an important consideration, underscored by estimates that unhealthy alcohol use is over-represented among Veterans (Agaku et al., 2020).

In addition to improving our understanding of the relationship between the above demographic variables and AUD care patterns, examining how these demographic variables influence care patterns for persons who engage in unhealthy alcohol use but whose drinking patterns are not severe enough to meet AUD criteria has important public health significance, as this population is significantly larger than the AUD population and is most likely to benefit from brief interventions (Scharer et al., 2022). Thus, examining associations between demographic variables and the care cascade for unhealthy alcohol use—both with and without AUD—is likely to yield important insights regarding where in the care continuum and for whom implementation interventions should be leveraged to create public health impact. To our knowledge, no previous study has examined the influence of these variables on each “step” of the care cascade using data from the past decade. In the current study, we used 2015–2019 NSDUH data to adapt and extend our alcohol cascade of care model to examine whether care patterns differ by key demographic characteristics. We focused specifically on the cascade steps for which clinicians have the highest likelihood of effecting change: screening, brief intervention, referral to treatment, and treatment.

MATERIALS AND METHODS

Data source

NSDUH is administered annually to U.S. civilians aged 12 and older who are not homeless or institutionalized. Interview response rates per year of data analyzed ranged from 64.9% to 71.2%, and final sample sizes ranged from 67,625 to 68,073. We analyzed NSDUH data from 2015 to 2019 because NSDUH first included questions about sexual orientation in 2015, and there were marked changes in AUD prevalence (Grossman et al., 2020) and healthcare utilization patterns (Schimmel et al., 2021) that occurred in 2020 in association with the COVID-19 pandemic, as well as changes to NSDUH’s AUD criteria in 2020.

The study was exempted from human subjects review by the Institutional Review Board at Washington University School of Medicine.

Participants

Because USPSTF recommends SBIRT for persons 18 years or older in primary healthcare settings, we limited participants to adults who indicated that they visited an outpatient clinic in the past year. Due to skip patterns for NSUDH items on SBIRT, the sample was further restricted to persons who endorsed having had at least one alcoholic drink in the past year. Table S1 provides additional details on the NSDUH items used to define our analytic sample.

Alcohol use pattern

We created three mutually exclusive groups to characterize alcohol use pattern: no binge drinking or AUD, binge drinking without AUD, and AUD. Adults who did not meet criteria for binge drinking or AUD were assigned to the “no binge drinking or AUD” category. Adults who indicated that they had engaged in at least one episode of binge drinking, a type of unhealthy alcohol use characterized by consumption of four or more drinks on one occasion (women) and five or more drinks on one occasion (men)—within the past 30 days but did not meet criteria for AUD were assigned to the “binge drinking without AUD” category. This category served as a proxy for unhealthy alcohol use that was not severe enough to qualify for AUD. Details on NSDUH items used to define this group are shown in Table S2a. We defined past-year AUD using NSDUH questions corresponding to Diagnostic and Statistical Manual (DSM) version 5 criteria as has been done previously (Mintz et al., 2021). During our time period of interest, NSDUH used DSM-IV criteria to define AUD, with 10 of the 11 DSM-5 criteria for AUD included in available NSDUH questions (see Table S2b for details). There was 96% concordance between the DSM-IV and DSM-5 AUD classifications (Cohen’s kappa 0.71, 95% CI 0.70–0.0.71; Table S3).

Outcome variables

Screening:

Adults who indicated that a healthcare provider had asked them about alcohol use in the past year—either verbally or on a written form—were coded as having been screened about alcohol use.

Brief intervention:

Adults who indicated that a healthcare provider advised them to cut down on drinking in the past 12 months were coded as having received a brief intervention.

Referral to treatment:

Adults who indicated a healthcare provider offered to give them information about alcohol treatment were coded having been referred to treatment.

AUD treatment:

Adults who indicated that they had received AUD treatment in the past year were counted as having received treatment.

Table S4a provides details on NSDUH items used to define outcome variables.

Exposure variables

We focused on the following demographic predictors: gender, race/ethnicity, sexual orientation, age, insurance status, and history of military involvement. Variables and possible responses were dictated by NSDUH and are detailed in Table S4b. As NSDUH does not survey persons in active military duty, a positive response to having a history of military involvement indicated Veteran status or current reserve duty.

Analytic plan

Analyses were conducted using SAS version 9.4. Survey weights were applied to account for NSDUH’s complex sample design and aggregation of multiple years of data.

We calculated weighted prevalence of each step of the care cascade stratified by alcohol use pattern and demographic predictor. We used logistic regression to model each step of the care cascade as a function of our demographic variables of interest. Each association was adjusted for other variables included in the model.

Denominators for each set of analyses were conditional on completion of the previous SBIRT “step” and influenced by USPSTF guidelines. Thus, for the model predicting screening, the entire sample was included. For the model predicting brief intervention, persons with binge drinking without AUD and those with AUD who reported having been screened were included. For the model predicting referral to treatment, persons with AUD who reported having been screened were included. For the model predicting AUD treatment, persons with AUD who had been referred to treatment were included. Figure 1 illustrates the logic used to determine each analytic subsample.

FIGURE 1.

Conceptual flowchart composition of regression samples.

Our primary research interest was to examine associations between demographic predictors and SBIRT and AUD treatment likelihood. However, given the large body of research demonstrating associations between medical illnesses related to unhealthy alcohol use (Engler et al., 2013; Grant et al., 2015; Roerecke et al., 2017; Ruidavets et al., 2010; Rumgay et al., 2021) which could moderate relationships between demographic characteristics and SBIRT and AUD treatment implementation, we conducted sensitivity analyses that adjusted for the following alcohol-related health conditions: hypertension, diabetes, cardiac disease, cancer, major depressive disorder, and illicit drug use disorder(s).

RESULTS

Missing data

Missing data were uncommon. Among demographic variables, there were 1001 missing responses for sexual orientation. For outcome variables, there were 1140 missing responses for the screening outcome, 449 missing responses for the brief intervention variable, and 120 missing responses for the referral to treatment variable.

Composition of the analytic sample

Of the 120,804 adults who used outpatient health and consumed at least one alcoholic drink in the past year, 62% did not engage in binge drinking in the past month or meet criteria for AUD in the past year, 27% engaged in past-month binge drinking but did not have AUD, and 11% had AUD. Women were more likely to be in the “no binge drinking or AUD” category than men and less likely to be in either unhealthy alcohol use group. More than half of persons aged 18–25 met criteria for some form of unhealthy alcohol use: 34% met criteria for binge drinking without AUD and 20% met criteria for AUD. Among racial and ethnic groups examined, Asian persons were least likely to engage in a form of unhealthy alcohol use; Hispanic persons were most likely. Persons who identified as gay, lesbian, or bisexual were more likely than those identifying as heterosexual to engage in either binge drinking or meet criteria for AUD. Individuals without insurance were more likely than those with insurance to engage in unhealthy alcohol use. Finally, Veterans were less likely than their non-Veteran peers to engage in unhealthy alcohol use. The detailed demographic composition of each alcohol use pattern is detailed in Table 1.

TABLE 1.

Demographics of the sample by alcohol use pattern.

	No binge drinking or AUD		Binge drinking without AUD		AUD
	n	Weighted % (95% CI)	n	Weighted % (95% CI)	n	Weighted % (95% CI)
Total	68,437	62.4 (61.9, 62.8)	36,353	27.0 (26.5, 27.4)	16,014	10.6 (10.4, 10.9)
Gender
Male	26,398	57.4 (56.8, 58.1)	16,603	29.6 (28.9, 30.3)	8157	13.0 (12.5, 13.4)
Female	42,039	66.7 (66.1, 67.2)	19,750	24.7 (24.2, 25.2)	7857	8.6 (8.3, 9.0)
Age
18–25	17,626	46.7 (46.0, 47.4)	12,650	33.9 (33.2, 34.5)	7208	19.5 (18.9, 20.0)
26–34	13,584	52.3 (51.5, 53.0)	8540	33.5 (32.8, 34.3)	3489	14.2 (13.6, 14.8)
35–49	19,944	59.9 (59.2, 60.5)	10,046	30.1 (29.5, 30.6)	3583	10.1 (9.7, 10.4)
50+	17,283	72.2 (71.4, 72.9)	5117	20.8 (20.1, 21.4)	1734	7.1 (6.6, 7.5)
Race and ethnicity
White	46,116	63.5 (62.9, 64.1)	24,332	26.3 (25.8, 26.8)	10,194	10.2 (9.9, 10.5)
Black	7439	59.0 (57.9, 60.2)	3959	29.3 (28.1, 30.5)	1795	11.7 (11.0, 12.4)
Asian	2992	72.0 (70.3, 73.7)	934	18.1 (16.6, 19.6)	519	9.9 (8.3, 11.6)
Hispanic	8860	56.4 (55.2, 57.6)	5468	31.8 (30.7, 32.9)	2457	11.8 (11.0, 12.7)
Othera	3030	58.8 (56.4, 61.1)	1660	26.9 (24.7, 29.1)	1049	14.3 (12.7, 15.9)
Sexual orientationb
Heterosexual	62,956	62.9 (62.5, 63.4)	33,160	26.9 (26.4, 27.4)	13,912	10.2 (9.9, 10.4)
Gay/lesbian	1474	55.6 (53.2, 58.1)	898	27.0 (24.9, 29.2)	613	17.3 (15.8, 18.9)
Bisexual	3366	50.8 (48.9, 52.7)	2055	29.4 (27.5, 31.3)	1369	19.8 (18.3, 21.2)
Insurance
Yes	63,837	63.2 (62.7, 63.7)	33,233	26.5 (26.0, 26.9)	14,432	10.3 (10.1, 10.6)
No	4600	50.3 (48.9, 51.7)	3120	34.5 (33.0, 35.9)	1582	15.2 (14.3, 16.1)
History of military involvementc
Yes	4754	68.6 (67.4, 69.9)	1985	23.0 (21.9, 24.2)	760	8.3 (7.7, 9.0)
No	63,683	61.7 (61.3, 62.2)	34,368	27.4 (26.9, 27.8)	15,254	10.9 (10.6, 11.2)

Note: The sample was comprised of n = 120,804 adult respondents from the 2015–2019 National Survey on Drug Use and Health who used an outpatient clinic and had at least one alcoholic drink in the past year.

Abbreviations: AUD, alcohol use disorder; CI, confidence interval.

^aOther race/ethnicity group was composed of persons who indicated they identified as Native American or Alaskan Native, Native Hawaiian or Other Pacific Islander, or more than one race.

^bThere were n = 1001 missing responses for sexual orientation.

^cHistory of military involvement denoted Veteran status or reserve duty.

Overall prevalence of SBIRT and AUD treatment

Table 2 shows conditional prevalences of each step of the cascade of care for unhealthy alcohol use. Most persons—83%—who used an outpatient clinic and had at least one alcoholic drink in the past year reported having been screened about alcohol. Among persons screened who endorsed binge drinking without AUD or having AUD, only 8% received a brief intervention. Among screened individuals who met criteria for AUD, 7% were referred to treatment. Among persons with AUD who were screened and referred to treatment, 23% reported receiving AUD treatment.

TABLE 2.

Conditional prevalences of alcohol screening, brief intervention, referral to treatment, and AUD treatment.

Sample included in weighted prevalence	Screeneda	Brief interventionb	Referral to treatmentc	AUD treatment
	Entire sample	Screened persons who engaged in binge drinking or had AUD	Screened persons with AUD	Screened and referred persons with AUD
	n = 119,664	n = 43,896	n = 13,742	n = 983
	Weighted % (95% CI)	Weighted % (95% CI)	Weighted % (95% CI)	Weighted % (95% CI)
Overall	82.7 (82.3, 83.0)	7.8 (7.4, 8.1)	7.1 (6.4, 7.8)	23.0 (18.6, 27.4)
Alcohol use pattern
No binge drinking or AUD	81.2 (80.8, 81.6)	-	-	-
Binge drinking without AUD	84.0 (83.5, 84.5)	4.2 (3.9, 4.6)	-	-
AUD	87.9 (87.2, 88.6)	16.3 (15.4, 17.1)	7.1 (6.4, 7.8)	23.0 (18.6, 27.4)
Gender
Male	80.9 (80.3, 81.5)	10.8 (10.1, 11.5)	8.8 (7.8, 9.9)	20.9 (15.3, 26.6)
Female	84.2 (83.8, 84.6)	4.6 (4.1, 5.0)	4.9 (4.2, 5.5)	27.5 (21.1, 34.0)
Age
18–25	81.1 (80.6, 81.6)	5.1 (4.7, 5.5)	6.5 (5.6, 7.4)	14.2 (9.8, 18.6)
26–34	86.3 (85.6, 86.9)	6.6 (5.9, 7.2)	7.8 (6.6, 9.0)	22.9 (15.7, 30.1)
35–49	86.3 (85.8, 86.7)	7.6 (7.0, 8.2)	6.7 (5.5, 7.9)	26.3 (18.9, 33.7)
50+	79.8 (79.1, 80.4)	10.1 (9.1, 11.1)	7.3 (5.3, 9.2)	26.5 (14.2, 38.8)
Race and ethnicity
White	83.7 (83.3, 84.1)	7.8 (7.4, 8.3)	7.0 (6.2, 7.8)	23.9 (19.5, 28.4)
Black	80.6 (79.7, 81.5)	7.3 (6.3, 8.2)	7.5 (5.7, 9.3)	17.7 (7.8, 27.6)
Asian	76.5 (74.4, 78.5)	6.6 (4.7, 8.5)	3.7 (1.1, 6.3)	d
Hispanic	80.4 (79.4, 81.5)	7.7 (6.6, 8.8)	7.5 (5.4, 9.6)	25.5 (7.6, 43.3)
Othere	84.1 (82.3, 85.8)	8.5 (6.2, 10.8)	8.5 (5.6, 11.5)	22.4 (7.3, 37.6)
Sexual orientationf
Heterosexual	82.5 (82.2, 82.9)	7.6 (7.2, 8.0)	6.7 (6.0, 7.4)	22.8 (17.9, 27.6)
Gay/lesbian	85.7 (83.8, 87.5)	10.5 (8.5, 12.6)	7.0 (4.3, 9.7)	d
Bisexual	85.5 (84.2, 86.8)	8.5 (6.9, 10.2)	10.2 (7.4, 13.0)	20.8 (10.4, 31.2)
Insurance
Yes	83.1 (82.8, 83.4)	7.8 (7.3, 8.2)	7.0 (6.2, 7.7)	23.4 (18.6, 28.2)
No	76.0 (74.7, 77.4)	7.7 (6.3, 9.1)	8.2 (6.4, 9.9)	18.4 (10.4, 26.4)
History of military involvementg
Yes	82.8 (81.6, 84.0)	13.8 (11.8, 15.8)	12.9 (9.5, 16.3)	23.2 (18.9, 27.5)
No	82.7 (82.3, 83.0)	7.2 (6.8, 7.6)	6.5 (5.8, 7.4)	21.6 (11.9, 31.3)

Note: The sample was comprised of adult respondents from the 2015–2019 National Survey on Drug Use and Health who used an outpatient clinic and had at least one alcoholic drink in the past year.

Abbreviations: AUD, alcohol use disorder; CI, confidence interval.

^aThere were n = 1140 missing responses for screening variable.

^bThere were n = 449 missing responses for brief intervention variable.

^cThere were n = 120 missing responses for referral to treatment variable.

^dEstimate not shown due to low statistical power.

^eOther race/ethnicity group was composed of persons who indicated they identified as Native American or Alaskan Native, Native Hawaiian or Other Pacific Islander, or more than one race.

^fThere were n = 1001 missing responses for sexual orientation.

^gHistory of military involvement denoted Veteran status or reserve duty.

Prevalence and associations between demographics and cascade of care outcomes

Table 2 shows the conditional weighted prevalence of each step of the care cascade stratified by demographic predictor of interest. Table 3 shows the adjusted odds ratios associated with each predictor of interest.

TABLE 3.

Associations between demographic predictors and alcohol-related cascade of care steps.

Sample included in model	Screeneda	Brief interventionb	Referral to treatmentc	AUD treatment
	Entire sample	Screened persons who engaged in binge drinking or had AUD	Screened persons with AUD	Screened and referred persons with AUD
	n = 118,740	n = 43,665	n = 13,654	n = 972
	AOR (95% CI)	AOR (95% CI)	AOR (95% CI)	AOR (95% CI)
Alcohol use pattern
No binge drinking or AUD	Ref	-	-	-
Binge drinking without AUD	1.20 (1.15, 1.25)***	Ref	-	-
AUD	1.72 (1.58, 1,87)***	4.55 (4.10, 5.06)***	-	-
Gender
Male	Ref	Ref	Ref	Ref
Female	1.32 (1.25, 1.41)***	0.45 (0.39, 0.53)***	0.51 (0.43, 0.61)***	1.70 (1.08, 2.67)*
Age
18–25	1.10 (1.05, 1.16)***	0.44 (0.38, 0.52)***	1.07 (0.70, 1.62)	0.43 (0.20, 0.91)*
26–34	1.67 (1.56, 1.79)***	0.62 (0.53, 0.73)***	1.24 (0.85, 1.80)	0.83 (0.37, 1.86)
35–49	1.68 (1.57, 1.79)***	0.78 (0.67, 0.91)**	1.00 (0.66, 1.51)	1.04 (0.47, 2.33)
50+	Ref	Ref	Ref	Ref
Race and ethnicity
White	Ref	Ref	Ref	Ref
Black	0.76 (0.71, 0.81)***	0.96 (0.82, 1.11)	1.01 (0.73, 1.41)	0.45 (0.22, 0.92)*
Asian	0.61 (0.55, 0.69)***	0.88 (0.62, 1.25)	0.38 (0.19, 0.75)**	d
Hispanic	0.78 (0.73, 0.84)***	1.04 (0.87, 1.25)	1.04 (0.75, 1.43)	1.16 (0.39, 3.46)
Othere	0.97 (0.84, 1.12)	1.08 (0.80, 1.47)	1.23 (0.81, 1.88)	0.96 (0.37, 2.53)
Sexual orientationf
Heterosexual	Ref	Ref	Ref	Ref
Gay/lesbian	1.27 (1.08, 1.50)**	1.24 (0.96, 1.60)	1.03 (0.66, 1.61)	d
Bisexual	1.10 (0.98, 1.23)	1.39 (1.11, 1.75)**	1.89 (1.36, 2.63)***	0.72 (0.34, 1.53)
Insurance
Yes	Ref	Ref	Ref	Ref
No	0.60 (0.55, 0.65)***	1.02 (0.81, 1.29)	1.10 (0.81, 1.49)	0.84 (0.45, 1.55)
History of military involvementg
Yes	1.23 (1.12, 1.35)***	1.43 (1.17, 1.74)***	1.77 (1.11, 2.81)*	1.02 (0.56, 1.86)
No	Ref	Ref	Ref	Ref

Note: The sample was comprised of adult respondents from the 2015–2019 National Survey on Drug Use and Health who used an outpatient clinic and had at least one alcoholic drink in the past year.

Abbreviations: AOR, adjusted odds ratio; AUD, alcohol use disorder; CI, confidence interval.

^aThere were n = 1140 missing responses for screening variable.

^bThere were n = 449 missing responses for brief intervention variable.

^cThere were n = 120 missing responses for referral to treatment variable.

^dEstimate not shown due to low statistical power.

^eOther race/ethnicity group was composed of persons who indicated they identified as Native American or Alaskan Native, Native Hawaiian or Other Pacific Islander, or more than one race.

^fThere were n = 1001 missing responses for sexual orientation.

^gHistory of military involvement denoted Veteran status or reserve duty.

^*p < .05

^**p < .01

^***p < .001.

Gender

There were significant gender differences at each step in the care cascade. Women were more likely than men to be screened about alcohol (84.2% vs. 80.9%, AOR 1.32, 95% CI 1.25–1.41). Among persons screened, however, women were less likely to receive either a brief intervention (4.6% vs. 10.8%; AOR 0.45, 95% CI 0.39–0.53) or a referral to treatment (4.9% vs. 8.8%, AOR 0.51, 95% CI 0.43–0.61). Among persons referred for AUD treatment, women were more likely than men to receive treatment (27.5% vs. 20.9%, AOR 1.70 95% CI 1.08–2.67).

Age

Persons aged 50 or older were less likely to be screened about alcohol than their younger counterparts (Tables 2 and 3). Age disparities in subsequent steps of care were more pronounced, with 18–25-year-olds being particularly affected. Among persons screened, those in younger age groups were less likely than those aged 50 or older to receive a brief intervention (5.1% for persons aged 18–25 vs. 10.1% for persons 50 or older; AOR 0.44, 95% CI 0.38–0.52; 6.6% for persons aged 26–34; AOR 0.62, 95% CI 0.53–0.73; 7.6% for persons 35–49; AOR 0.78, 95% CI 0.67–0.91). Among persons with AUD who were screened about alcohol, age was not associated with the likelihood of being referred to treatment. However, among individuals referred, those aged 18–25 years were less likely than those aged 50 or older to receive AUD treatment (14.2% vs. 26.5%, AOR 0.43, 95% CI 0.20–0.91).

Race and ethnicity

Racial and ethnic differences were observed for odds of screening, referral to treatment, and AUD treatment. Specifically, Black (80.6%), Asian (76.5%), and Hispanic (80.4%) persons were less likely than White persons (83.7%) to be screened about alcohol use (Black AOR 0.76, 95% CI 0.71–0.81; Asian AOR 0.61, 95% CI 0.55–0.69; Hispanic AOR 0.78, 95% CI 0.73–0.84). Among persons with AUD who were screened about alcohol, Asian persons were less likely than White persons to be referred to treatment (3.7% vs. 7.0%, AOR 0.38, 95% CI 0.19–0.75). Among persons referred to treatment, Black persons were less likely than White persons to receive AUD treatment (17.7% vs. 23.9%, AOR 0.45, 95% CI 0.22–0.92).

Sexual orientation

Persons who identified as gay or lesbian were more likely to be screened about alcohol than those who identified as heterosexual (85.7% vs. 82.5%; AOR 1.27, 95% CI 1.08–1.50). Persons identifying as bisexual were more likely to receive brief intervention and referral to AUD treatment than heterosexual persons (brief intervention: 8.5% vs. 7.6%; AOR 1.39, 95% CI 1.11–1.75; referral to treatment: 10.2% vs. 6.7%, AOR 1.89, 95% CI 1.36–2.63). Among those referred, there was no association between sexual orientation and odds of AUD treatment.

Insurance status

Insurance status was strongly associated with odds of being screened about alcohol, such that persons without insurance were less likely to be screened about alcohol than those with insurance (76.0% vs. 83.1%, AOR 0.60, 95% CI 0.55–0.65). There were no significant associations between insurance status and subsequent steps of care for persons with unhealthy alcohol use.

History of military involvement

Veterans had increased odds of being screened about alcohol, although the magnitude of the difference in prevalence vs. non-Veterans was small. Veterans had greater odds of receiving a brief intervention (13.8% vs. 7.2%, AOR 1.43, 95% CI 1.17–1.74) and AUD treatment referral (12.9% vs. 6.5%, AOR 1.77, 95% CI 1.11–2.81) than non-Veterans. Among those referred, there was no association between history of military involvement and odds of receiving AUD treatment.

Sensitivity analyses

Additional adjustments for co-occurring alcohol-related health conditions did not meaningfully change the results of primary analyses and are shown in Table S5.

DISCUSSION

In this large epidemiologic sample of adults who visited an outpatient clinic in the past year, we identified multiple demographic disparities in the care cascade for unhealthy alcohol use that warrant comment.

Gender

Despite being more likely to be screened than their male counterparts, women were less likely to receive either a brief intervention or a treatment referral. This pattern replicates what has been previously observed with NSDUH data (Glass et al., 2016), and it has been hypothesized that women are reluctant to disclose alcohol use to providers given perceived stigma and family responsibilities that often disproportionately affect women and may conflict with time required for AUD treatment (Glass et al., 2016). Though these patient factors undoubtedly influence the low prevalence of brief intervention and treatment referral we observed for women, there is also evidence that provider bias may contribute to this gender disparity. For example, studies using VHA data have found that even when unhealthy drinking is documented in the medical record and thus theoretically known to clinicians, clinicians are less likely to provide interventions to women (Chen et al., 2020; Williams et al., 2017). Provider bias as pertains to patient’s gender has been hypothesized to affect other medical disciplines as well: Alabas et al. (2017) found that among a Swedish cohort of persons with a recent myocardial infarction, women were less likely than men to receive evidence-based treatment and had higher likelihood of mortality relative to men, but when analyses were adjusted for receipt of evidence-based care, the mortality risk was similar between women and men. Reasons for potential provider bias in implementation of alcohol-related evidence-based interventions are an important direction for future research, underscored by increases in unhealthy alcohol use among women both before (Grucza et al., 2018) and during (Pollard et al., 2020) the COVID-19 pandemic.

Among persons referred to treatment, women were more likely than men to receive AUD treatment, a finding that differs from other studies that have shown that women are less likely to utilize AUD treatment resources than men (Alvanzo et al., 2014; Dawson, 1996; Gilbert et al., 2019; Simpson et al., 2020). We posit that this is due to differences in ascertainment in that previous studies assessed treatment prevalence in the general population, and we assessed it among persons referred to treatment by a healthcare provider. Based on these findings, we hypothesize that being referred to treatment by a healthcare provider may have a greater influence on women than on men. Formal testing of this hypothesis could yield important insights into how to structure implementation efforts to optimally increase AUD treatment utilization among both genders.

Age

We found that persons ≥50 were less likely to be screened about alcohol use than younger ones, aligning with previous research that adults aged 65 or older are less likely to be asked about drinking than their younger peers (McKnight-Eily et al., 2017; Sahker & Arndt, 2017). Combined with the existing literature, our findings suggest that persons aged 65 and older have a particularly low likelihood of screening. This disparity likely reflects in part the fact that older adults are less likely to engage in unhealthy alcohol use than younger age groups, and thus, providers may be less inclined to query alcohol use in this population. In addition, as older adults are more likely to have medical issues that require clinical attention and management than their younger peers, it may be possible that clinicians have less time to implement preventive health recommendations like alcohol screening in their geriatric patients. However, given increased alcohol consumption among older adults in recent years (Grucza et al., 2018; Han et al., 2017), clinicians should be cognizant of the need to assess alcohol use patterns in this population. This is particularly important as older adults are at increased risk for negative consequences from alcohol use due to changes in metabolism and increased likelihood of co-occurring illnesses that require medication(s) which can interact negatively with alcohol (Holton et al., 2017; Moore et al., 2007).

Persons aged 18–25—more than half of whom reported engaging in unhealthy alcohol use—were least likely to receive either a brief intervention or AUD treatment, revealing a misalignment between clinical need and the likelihood of clinicians’ intervention efforts. Although many young adults who engage in unhealthy alcohol use—with or without associated AUD—decrease alcohol consumption as they age without clinical intervention (Dawson et al., 2005), unhealthy alcohol use patterns during late adolescence and early adulthood are significant predictors of developing AUD in middle age (Patrick et al., 2023). Thus, our results reveal a key missed opportunity to intervene in a substantial population of persons early in the course of the disorder and potentially prevent long-term consequences of unhealthy alcohol use.

Race and ethnicity

Black, Hispanic, and Asian persons were less likely than White persons to be screened about alcohol. At a population level, White persons have a higher prevalence of unhealthy alcohol use patterns than Black, Hispanic, or Asian persons (Delker et al., 2016; Grant et al., 2015), and thus, it is possible our findings reflect these population differences. Of note, a recent study using data from more than 700,000 Veterans demonstrated that among patients screened about alcohol use with an evidence-based screening tool, providers were more likely to diagnose AUD in Black and Hispanic persons than in White persons despite patients endorsing similar levels of alcohol consumption, suggestive of potential provider bias (Vickers-Smith et al., 2023). As Black and Hispanic persons with AUD are more likely to experience negative medical and social consequences related to alcohol use than White persons (Flores et al., 2008; Mulia et al., 2009), effective screening and accurate diagnosis of AUD within these populations is particularly important for reducing adverse effects of alcohol misuse.

Asian persons had the lowest odds of screening among any racial or ethnic group, which may reflect the relatively low prevalence of unhealthy alcohol use in this population. However, another potential contributing factor is that Asian persons are often considered part of a “model minority” stereotype, which refers to a minoritized demographic group that achieves financial success and maintains good physical health. Some have argued that this stereotype has led to Asian persons being understudied in health outcomes research (Kim et al., 2021), including alcohol-related research (Iwamoto et al., 2016). Our finding that among persons with AUD who were screened about alcohol use, Asian persons had 60% lower odds of being referred to treatment relative to White persons aligns with this concern.

Black persons with AUD were less likely than White persons to receive treatment. Our finding supports previous research demonstrating that among Veterans receiving health care within the VHA system, Black persons are less likely than White persons to receive medications for AUD (mAUD; Williams et al., 2017). Other racial and ethnic disparities in substance use disorder treatment have been previously documented (Cook et al., 2017), and a deeper understanding of AUD treatment barriers for racial and ethnic minorities remains of critical import. Recent evidence, for example, suggests that perceived discrimination by healthcare providers may delay and negatively impact addiction care for Black persons (Hall et al., 2022) and emphasizes the need for providing culturally component care to persons with addiction.

Sexual orientation

We found persons identifying as gay or lesbian were more likely to be screened about alcohol than heterosexual persons, and bisexual persons were more likely to receive brief intervention and/or be referred to treatment than heterosexual persons. Our study findings augment limited existing data characterizing associations between alcohol-related care and sexual orientation. It should be noted that a previous study examining data from the 2014 Behavioral Risk Factor Surveillance System (Lehavot et al., 2017) found that associations between sexual orientation and screening and brief intervention likelihood differed by gender.

While we did not examine interaction effects between sexual orientation and gender, Lehavot et al.’s findings suggest that there are nuances in understanding how demographic factors can influence steps of the care cascade for unhealthy alcohol use.

Insurance

Insured persons were more likely to be screened about alcohol than those without insurance. One interpretation of this difference is that insurance reimbursements for alcohol screening are associated with increased likelihood of clinicians performing such screenings. However, whereas brief interventions are also reimbursable by insurances and we observed no insurance differences in the likelihood of receiving a brief intervention, clinician compensation alone appears not to be a sufficient motivator to increase the implementation of evidence-based alcohol-related care. Furthermore, whereas our results show that persons without insurance are disproportionately affected by unhealthy alcohol use, improving the identification of unhealthy alcohol use in this group could have an important public health impact.

We found no significant differences in the likelihood of receiving AUD treatment between insured and uninsured persons referred to treatment by a provider. It should be noted that other than specific inquiries about attending self-help groups or 12-step programs, NSDUH did not query the type of AUD treatment received during our time period of interest. Thus, it is unknown whether the treatment received was evidence-based. Previous studies have suggested that the type of insurance one has may influence the type of treatment received: Disparities in access to mAUD in particular were recognized prior to implementation of the Affordable Care Act such that persons with private insurance were more likely than those with public forms of insurance to receive mAUD (Abraham et al., 2013). Though it is presumed that as an increasing number of states have adopted Medicaid expansion these insurance disparities have been mitigated, there is evidence that inequities continue to persist (Dickson-Gomez et al., 2022). Research is needed to determine the degree to which this disparity remains and, if so, how best to remedy it.

History of military involvement

We found that persons who had history of military involvement were more likely to be screened, receive a brief intervention, and be referred to treatment than those who had not. Greater prevalences of care steps in the Veteran population may result from substantial efforts by the VHA system in the past 20 years to incorporate SBIRT into primary care (Bradley et al., 2006; Williams et al., 2017, 2018). Importantly, NSDUH does not specify which healthcare systems respondents utilized, and a substantial proportion of Veterans do not receive health care through the VHA system. Thus, it is possible that our Veteran sample differs from Veterans who receive health care exclusively from the VHA. Regardless, the overall low prevalences of brief intervention, referral to treatment, and AUD treatment among Veterans highlight the need for greater implementation of alcohol-related care for this population.

Attrition from care following screening

In addition to the important demographic disparities noted throughout the unhealthy alcohol use care cascade, it should be emphasized that significant attrition from care occurred following screening regardless of demographics or alcohol use pattern. Using NSDUH’s population estimates, our findings suggest that of the approximately 44 million U.S. adults with unhealthy alcohol use (either binge drinking or AUD) who were screened about alcohol, roughly 40 million—more than 90%—are lost to subsequent intervention each year. The magnitude of this observed “leakage” from care deserves discussion on both possible reasons for these gaps in care and proposed solutions to fill them.

It has been argued that the quality of alcohol screening questions contributes both to the likelihood of accurately identifying persons with unhealthy alcohol use and subsequent intervention (Subbaraman et al., 2022). Optimal alcohol screening questions do not simply assess whether a person uses alcohol but also distinguish between low-risk and unhealthy alcohol use. The USPSTF recommends use of the 10-item Alcohol Use Disorders Identification Test (AUDIT), the three-item AUDIT-Consumption instrument, or the NIAAA-recommended Single Alcohol Screening Question for this purpose (O’Connor et al., 2018), yet in their survey of 1500 primary care providers, Tan et al. (2018) reported that fewer than half used one of the recommended tools. Whether use of recommended screening tools is associated with a higher likelihood of subsequent care, then, is an important direction for future study.

In addition, many have advocated for increased efforts to incorporate AUD treatment into primary healthcare settings as a way to minimize attrition from care (Carvalho et al., 2019; Mintz et al., 2021; Williams et al., 2018). In this model, the “referral” step is eliminated from the care cascade and point-of-care intervention—either a brief intervention or more formal psychotherapy and/or medication—are provided during the primary care visit. One frequently cited barrier to this approach is providers’ lack of knowledge about AUD treatment (Ehrie et al., 2020; Williams et al., 2018). In recent years, there have been increased efforts to improve physicians’ knowledge of addiction treatment. For example, since July 2022, the Accreditation Council on Graduate Medical Education has required that internal medicine resident physicians complete 2 weeks of addiction medicine training during residency (Accreditation Council for Graduate Medical Education, 2022). Beginning in June 2023, the Drug Enforcement Administration (DEA) has required that all physicians applying for or renewing their DEA license complete a one-time eight-hour training on substance use disorders (U.S. Drug Enforcement Administration, 2023). Finally, the National Institute on Alcohol Abuse and Alcoholism recently released its Core Resource on Alcohol developed to educate health professionals on evidence-based approaches for treating patients with alcohol-related problems (National Institute on Alcohol Abuse and Alcoholism, 2022). Assessing whether these interventions improve the implementation of intervention and treatment efforts for persons with unhealthy alcohol use within primary care settings and across demographic groups is critical to informing whether and which additional implementation efforts are needed to ensure equitable care for persons with unhealthy alcohol use.

Limitations

There are limitations to this study. During the time period studied, NSDUH did not assess alcohol craving, a criterion for DSM-5 AUD, so we may have underestimated the true prevalence of the diagnosis. We used past-month binge drinking as a proxy for unhealthy alcohol use less severe than AUD, but because there are unhealthy drinking patterns other than binge drinking and it is possible that some persons in our “no binge drinking or AUD” category engaged in other types of unhealthy alcohol use. While our results demonstrate overall prevalence patterns for SBIRT and AUD treatment, because NSDUH does not assess whether screening questions, brief interventions, or AUD treatment received are evidence-based, prevalences for implementation of evidence-based screening, brief intervention and AUD treatment are likely lower than what our estimates show. Our participant sample was limited to individuals who indicated that they visited an outpatient clinic in the past year, but we did not have access to the type of clinic visit attended, and some participants could have visited only a specialty clinic (e.g., cardiology). As the USPFTF recommendations for alcohol screening apply to primary care settings specifically, it is possible that our results would have differed if we excluded persons who did not utilize a primary care clinic. NSDUH does not assess genders other than male/female or sexual orientations other than heterosexual, gay/lesbian, and bisexual. Due to statistical power limitations, we were unable to characterize care patterns for the latter steps of the care cascade for some demographic features or to examine statistical interactions. Our analyses are specific to the use of outpatient clinics and thus may not generalize to other healthcare settings like emergency rooms or hospitals. NSDUH responses are based on self-report and thus subject to potential response biases. NSDUH excludes institutionalized, homeless, and incarcerated persons from its sample, and therefore, our findings cannot be generalized to these important subsets of the population. Finally, our time period of interest intentionally ended prior to the onset of the COVID-19 pandemic in 2020. However, given known changes in alcohol use patterns during the pandemic, examining whether disparities in the alcohol use cascade of care have changed or persisted in more recent years is an important area for future research.

CONCLUSIONS

There are demographic disparities in the alcohol cascade of care, particularly for women, young adults, and racial and ethnic minorities. Addressing the sources of these disparities is critical to both improving clinicians’ recognition of and intervention with persons with unhealthy alcohol use and minimizing attrition from care.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are publicly available and can be accessed at https://www.datafiles.samhsa.gov/dataset/national-survey-drug-use-and-health-2019-nsduh-2019-ds0001.