Characterizing unhealthy alcohol use patterns and their association with alcohol use reduction and alcohol use disorder during follow-up in HIV care

Abstract

Outcomes of PWH with unhealthy alcohol use, such as alcohol use reduction or progression to AUD, are not well-known and may differ by baseline patterns of unhealthy alcohol use. Among 1299 PWH screening positive for NIAAA-defined unhealthy alcohol use in Kaiser Permanente Northern California, 2013-2017, we compared two-year probabilities of reduction to low-risk/no alcohol use and rates of new AUD diagnoses by baseline use patterns, categorized as exceeding: only daily limits (72% of included PWH), only weekly limits (17%), or both (11%), based on NIAAA recommendations. Overall, 73.2% (95% CI 70.5%-75.9%) of re-screened PWH reduced to low-risk/no alcohol use over two years, and there were 3.1 (95% CI 2.5-3.8) new AUD diagnoses per 100 person-years. Compared with PWH only exceeding daily limits at baseline, those only exceeding weekly limits and those exceeding both limits were less likely to reduce and likelier to be diagnosed with AUD during follow-up. PWH exceeding weekly drinking limits, with or without exceeding daily limits, may have a potential need for targeted interventions to address unhealthy alcohol use.

INTRODUCTION

Persons with HIV (PWH) in the United States (US) have high rates of unhealthy alcohol use, with 32% of PWH in care in a recent national cohort study exceeding recommended limits (1-3). Unhealthy alcohol use, which includes problems ranging from excessive drinking to alcohol use disorder (AUD), can reduce antiretroviral adherence, HIV care engagement, and virologic suppression (4-7). PWH with unhealthy alcohol use have higher rates of liver and cardiovascular disease, cognitive impairment, severe pneumonia, and hospitalization, relative to PWH without unhealthy alcohol use (8-12). Unhealthy alcohol use among PWH is also associated with greater mortality and lower life expectancy (13-17). US guidelines therefore highlight the importance of addressing unhealthy alcohol use in HIV care with evidence-based screening and treatment approaches (18-20).

Important knowledge gaps exist in understanding changes in unhealthy alcohol use over time among PWH. Many cross-sectional studies have characterized unhealthy alcohol use prevalence and risk factors in HIV care (1-3, 21-23). Some studies have examined changes in alcohol use and group-based trajectories of alcohol use levels following an index measurement (24-33). Studies have usually found groups of PWH with persistently low alcohol use and a group with persistently high alcohol use (24-28); however, some have also found groups of PWH with increasing or decreasing alcohol use (24-26), while others reported groups with persistent moderate alcohol use (27, 28). Reported risk factors for persistently high or increasing alcohol use include depression, non-alcohol substance use, older age, poor control of HIV, and lower socioeconomic status (24-33). Only two studies have reported alcohol use changes for PWH with unhealthy alcohol use at index measurement. One reported that only 25% of PWH with unhealthy alcohol use reduced their consumption, while the other found 16%–31% reduced their consumption to low-risk levels (25, 32).

Randomized trials (34-39) and clinic-based studies (40, 41) have also evaluated interventions to reduce unhealthy alcohol use among PWH. In a Veterans Health Administration (VA) study, approximately 60% of PWH had a resolution of their unhealthy alcohol use after an index screen, with no association between brief intervention receipt and resolution (40). In a clinical trial of integrated stepped alcohol treatment, almost half of PWH with at-risk alcohol use reduced to no heavy drinking days by week 52; the trial did not recruit enough participants to examine treatment effect (35). Other clinical trials have found efficacious treatments for unhealthy alcohol use. Extended-release naltrexone with counseling led to a mean of 0 heavy drinking days vs. 5 in the control arm at week 52 (38). Receipt of two brief interventions led to a 52%–59% reduction in drinking days vs. 20%–33% in the control group among women with HIV (36). In another trial, receipt of a computerized brief intervention was associated with a reduction of 9 drinks per week (41).

Existing evidence on alcohol use over time in PWH is therefore limited in two ways. Most studies have examined alcohol use trajectories across all levels of alcohol use, without providing in-depth analyses of PWH with unhealthy alcohol use, while those examining specifically PWH with unhealthy alcohol use have been based on clinical trial data. In this study, we focused specifically on PWH identified as having unhealthy alcohol use during routine clinical care. Among these patients, we sought to estimate how many reduce their alcohol use to a low-risk level, how many go on to experience more severe problems like AUD, and which patient factors are associated with these outcomes. Evidence from clinical care settings on the course of unhealthy alcohol use could help identify areas for improvement and patient subgroups with unmet needs who may benefit from targeted interventions.

In particular, patterns of unhealthy alcohol use among PWH, such as exceeding daily vs. weekly recommended limits, may have important implications for clinical management and merit attention. Some studies have found that, compared to PWH who did not consume alcohol, the increase in viral non-suppression rates was greater for PWH exceeding weekly limits than for those only exceeding daily limits (32, 42). In a recent Swiss study, PWH with elevated AUDIT-C scores but not exceeding daily limits had lower rates of mortality and liver-related events than PWH exceeding daily limits (43). Certain patterns of unhealthy alcohol use might reflect different reasons for drinking (e.g., weekend social drinking vs. daily drinking to manage stress) or be linked to other concerns that require attention such as mental health and substance use comorbidities (21, 22, 44). In the general population, brief alcohol interventions have sometimes been shown to be differentially effective according to baseline alcohol use patterns (45). Given the heterogeneity of unhealthy alcohol use and its clinical implications, further investigation of the patient characteristics and health outcomes of PWH with different patterns of unhealthy alcohol use is warranted.

Investigating drinking patterns and the course of unhealthy alcohol use among PWH is essential to help inform clinical efforts to address unhealthy alcohol use and prevent its adverse consequences. In this study, among PWH who screened positive for unhealthy alcohol use in primary care, we compared patient characteristics across daily vs. weekly alcohol use patterns and estimated the association of these patterns with alcohol use reduction and new AUD diagnosis during follow-up. Our study was informed by the Gelberg-Andersen Behavioral Model for Vulnerable Populations, which identifies factors that affect behavior as pre-disposing (e.g. age or gender), enabling/impeding (e.g. barriers to change), and need factors (e.g. self-perceived health status) (46, 47). The Gelberg-Andersen model has been extensively used to investigate health care utilization, medication uptake and adherence, and other health promoting behaviors, including in PWH (48-55). Based on this model and prior evidence on unhealthy alcohol use risk factors, we hypothesized that medical comorbidities, mental health diagnoses, and non-alcohol substance use disorders (SUDs) would be more common among PWH with more severe unhealthy alcohol use, and that these PWH would be least likely to reduce their alcohol use over time, and most likely to have baseline AUD or develop AUD during follow-up.

METHODS

Study population

This study was conducted in Kaiser Permanente Northern California (KPNC), an integrated health system providing primary and specialty care to >4.5 million members, who are demographically similar to other privately insured individuals in this region (56, 57). In July 2013, KPNC implemented systematic unhealthy alcohol use screening at adult primary care visits as part of a Screening, Brief Intervention and Referral to Treatment (SBIRT) program, with automated reminders for follow-up screening at six months after a positive screen, or at one year otherwise (58).

We included adult (≥18 years) PWH in KPNC with a positive screen for unhealthy alcohol use between 1 July 2013 and 31 December 2017 (N=1299, Supplemental Fig. 1). Based on National Institute on Alcohol Abuse and Alcoholism (NIAAA) recommendations, the KPNC SBIRT program defines unhealthy alcohol use as ≥5 drinks in a day at least once in the last 90 days or ≥15 average weekly drinks for men <66 years old, and ≥4 drinks in a day or ≥8 average weekly drinks for men ≥66 years old and women of any age (59). The KPNC and University of California, San Francisco Institutional Review Boards approved this study with waivers of informed consent.

Study measures

We defined baseline as the first positive screen date for two reasons. Firstly, our study objective was to examine outcomes of PWH identified as having unhealthy alcohol use to inform possible interventions in routine clinical care, and the first positive screen represents the first opportunity to intervene. Secondly, this approach allowed us to compare our findings to published estimates of reductions of unhealthy alcohol use from studies with similar baseline definitions (25, 32, 40, 60). Unhealthy alcohol use patterns at baseline (first positive screen date) were categorized as exceeding: (1) only daily limits, (2) only weekly limits, or (3) both daily and weekly limits. Exceeding daily limits is also known as heavy episodic drinking (61, 62).

The first outcome of interest was a reduction to low-risk or no alcohol use within two years of baseline, defined as having ≥1 negative follow-up screen, i.e., reporting either no alcohol use or alcohol use within recommended daily and weekly limits. We selected two years as a follow-up period to allow sufficient opportunity for re-screening. Patients with no follow-up screen within two years (N=274) were excluded from reduction analyses and compared to those with a follow-up screen who were included (N=1025, Supplemental Fig. 1). To evaluate the impact of a less strict definition, we conducted a sensitivity analysis where we also counted as a reduction screens that were positive for unhealthy alcohol use, but showed a decrease from baseline in the number of days exceeding daily drinks limits or in the number of average weekly drinks, provided there was no increase in either measure.

Among patients without a history of AUD at baseline (N=1083, Supplemental Fig. 1), the second outcome of interest was a new AUD diagnosis, captured using International Classification of Diseases, Ninth and Tenth Revisions (ICD-9/10) codes (Supplemental Table 1). Prior AUD was defined as a diagnosis recorded by a provider in the electronic health record (EHR) any time up to baseline. A new AUD diagnosis was defined as a diagnosis recorded after baseline. Diagnoses of AUD are entered into the EHR by a primary or specialty health care provider as part of routine clinical care. Assessment of AUD may be prompted by patient report, provider decision, or a positive screen for unhealthy alcohol use. As part of the KPNC SBIRT program, patients screening positive for unhealthy alcohol use receive a brief alcohol intervention based on motivation interviewing techniques and may be referred to an addiction medicine specialist for further evaluation (58). In addition, all KPNC members can access addiction medicine care without referral.

Covariates measured at baseline were sex, race/ethnicity, age, HIV risk group, smoking status, history of depression and non-alcohol substance use disorder (SUD, including cannabis, opioid, stimulant, and other use disorders), Charlson comorbidity index (CCI) excluding AIDS, neighborhood deprivation index (NDI), receipt of a brief intervention for unhealthy alcohol use (within 60 days of baseline as per prior KPNC analyses) (63), CD4 cell count, and HIV viral load (VL). NDI was calculated according to published methodology and categorized as quartiles based on the sample distribution (64). Brief interventions and depression and SUD diagnoses were captured using ICD-9/10 codes (Supplemental Table 1). CD4 count was the closest measurement eight months before or after baseline. VL was dichotomized as <200 or ≥200 copies/mL using the closest measurement six months before or one month after baseline.

Statistical analysis

We compared patient demographic and clinical characteristics by baseline unhealthy alcohol use patterns. Among patients with ≥1 follow-up screen within two years, we estimated the probability of reduction to low-risk/no alcohol use. To evaluate the possibility of selection bias, we compared follow-up screening rates by baseline unhealthy alcohol use patterns.

Among patients without a prior AUD diagnosis at baseline, we estimated AUD diagnosis rates, calculated as the number of patients with a new AUD diagnosis divided by the person-time at risk. Because PWH might have previously had AUD that was not detected during routine care, we refer to these rates as “AUD diagnosis rates” rather than “AUD incidence rates”. Patients contributed person-time at risk from baseline until new AUD diagnosis, death, interruption of KPNC membership for >90 days, 18 months without alcohol use screening, or 31 December 2019, whichever occurred first. We included lack of alcohol use screening as a censoring criterion to minimize the possibility of missed AUD diagnoses.

We compared two-year probabilities of reduction to low-risk/no alcohol use and rates of new AUD by baseline unhealthy alcohol use patterns. We used Poisson regression models to estimate reduction risk ratios and AUD diagnosis rate ratios, adjusting for all covariates. We selected a priori covariates that have been shown to be associated with alcohol use levels cross-sectionally or with longitudinal trajectories of alcohol use (1-3, 21-33). Among the covariates, we adjusted for receipt of a brief intervention. We did not adjust for other alcohol-related treatments, as a previous KPNC study reported that only 5% of PWH with unhealthy alcohol use initiated addiction specialty treatment within a year (63). For all risk ratios, we used a robust variance estimator (65).

We also conducted a supplemental analysis to identify factors independently associated with reduction to low-risk/no alcohol use and with new AUD diagnosis, which could help highlight patient subgroups with greater need. In this analysis, we examined associations of each individual covariate with study outcomes, using separate, unadjusted models including only one variable.

Analyses were conducted in SAS v9.4 (SAS Institute Inc., Cary, NC). P values were two-sided with a pre-specified alpha of 0.05.

RESULTS

Study sample

The 1299 included patients with unhealthy alcohol use were mostly men (95%) and White (57%), with a median age of 46 (interquartile range [IQR] 35, 54). At baseline, 932 (72%) reported exceeding only daily limits, 226 (17%) only weekly limits, and 141 (11%) both.

Patients only exceeding daily limits were younger (median age 43 vs. 58 and 50 years; Kruskal-Wallis=155.81, P<0.01), more frequently men who have sex with men (MSM, 79% vs. 71% and 70%; χ²=13.71, P<0.01), and less frequently White (52% vs. 71% and 63%; χ²=32.07, P<0.01), compared with those only exceeding weekly limits and those exceeding both, respectively (Table 1). Patients only exceeding weekly limits were more frequently women (10% vs. 4% and 6%; χ²=10.03, P<0.01) and had more comorbidities (40% with CCI ≥2 vs. 22% and 33%; χ²=39.35, P<0.01), compared with those only exceeding daily limits and those exceeding both, respectively. Patients exceeding both daily and weekly limits had higher prevalence of current smoking (41% vs. 23% and 22%; χ²=34.87, P<0.01) and AUD history (35% vs. 13% and 19%; χ²=45.80, P<0.01) compared with patients exceeding only daily or weekly limits, respectively.

Table 1.

Demographic and clinical characteristics of 1299 persons with HIV screening positive for unhealthy alcohol use in Kaiser Permanente Northern California, 2013–2017.

Characteristic at index positive screen	Only exceeding daily limits a N=932	Only exceeding weekly limits b N=226	Exceeding daily and weekly limits a, b N=141	Test Statistic c	P c
Men	891 (96%)	204 (90%)	133 (94%)	10.03	<0.01
Race/ethnicity				32.07	<0.01
White	485 (52%)	161 (71%)	89 (63%)
Black	192 (21%)	30 (13%)	18 (13%)
Hispanic	127 (14%)	16 (7%)	19 (13%)
Other/unknown	128 (14%)	19 (8%)	15 (11%)
Age, years	43 (33, 51)	58 (46, 66)	50 (39, 57)	155.81	<0.01
HIV transmission risk group				13.71	<0.01
MSM	739 (79%)	161 (71%)	98 (70%)
IDU	48 (5%)	19 (8%)	15 (11%)
Heterosexual or other	145 (16%)	46 (20%)	28 (20%)
Smoking status				34.87	<0.01
Never	398 (43%)	75 (33%)	36 (26%)
Current	211 (23%)	50 (22%)	58 (41%)
Former	323 (35%)	101 (45%)	47 (33%)
History of depression d	246 (26%)	73 (32%)	34 (24%)	3.96	0.14
History of AUD d	122 (13%)	44 (19%)	50 (35%)	45.80	<0.01
History of non-alcohol SUD d	134 (14%)	34 (15%)	31 (22%)	5.48	0.06
Received a brief intervention e	330 (35%)	68 (30%)	72 (51%)	17.40	<0.01
Charlson comorbidity index f				39.35	<0.01
0	708 (76%)	129 (57%)	88 (62%)
1	17 (2%)	6 (3%)	6 (4%)
≥2	207 (22%)	91 (40%)	47 (33%)
Neighborhood deprivation index g				4.35	0.63
1 (least)	219 (23%)	64 (28%)	31 (22%)
2	232 (25%)	58 (26%)	32 (23%)
3	243 (26%)	52 (23%)	36 (26%)
4 (most)	231 (25%)	52 (23%)	41 (29%)
CD4 count, cells/μL h	643 (472, 859)	630 (468, 828)	648 (461, 872)	0.84	0.66
HIV RNA <200 copies/mL i	717 (77%)	189 (84%)	112 (79%)	3.29	0.19

All numbers are N (%) or median (interquartile range). Abbreviations: AUD, alcohol use disorder; SUD, substance use disorder.

^aExceeding daily limits was defined as reporting ≥5 drinks in a day at least once in the last 90 days for men <66 years old, and ≥4 drinks in a day at least once in the last 90 days for men ≥66 years old and all women.

^bExceeding weekly limits was defined as reporting ≥15 average weekly drinks for men <66 years old, and ≥8 average weekly drinks for men ≥66 years old and all women.

^cStatistics and P values from χ² tests (categorical variables) and Kruskal-Wallis tests (continuous variables) comparing patients in all three groups.

^dCaptured using International Classification of Diseases codes (see Supplemental Table 1) and defined as any record prior to or on baseline date.

^eCaptured using International Classification of Diseases codes (see Supplemental Table 1) and defined as a recorded code within 60 days of baseline date.

^fExcluding AIDS.

^gEstimated using the approach of Messer et al. and divided into quartiles based on the distribution of the entire sample (64).

^hClosest measurement eight months prior to or following baseline. Missing for 57 (6%), 9 (4%), and 8 (6%) patients exceeding only daily limits, only weekly limits, and both daily and weekly limits, respectively.

ⁱClosest measurement six months prior to or one month following baseline. Missing for 95 (10%), 17 (8%), and 12 (9%) patients exceeding only daily limits, only weekly limits, and both daily and weekly limits, respectively.

The number of days of excess drinking was higher for patients exceeding both daily and weekly limits at baseline compared with those only exceeding daily limits (Table 2, median 10 vs. 2 days in the last 90 days; Kruskal-Wallis=148.56, P<0.01). Average weekly drinking was similar for patients only exceeding weekly limits and those exceeding both daily and weekly limits (both medians 21 weekly drinks for men <66, Kruskal-Wallis=0.84, P=0.36; and both medians 14 weekly drinks for others, Kruskal-Wallis=2.29, P=0.13).

Table 2.

Alcohol use among 1299 persons with HIV in Kaiser Permanente Northern California screening positive for unhealthy alcohol use, 2013–2017.

Characteristic at index positive screen	Only exceeding daily limits a N=932	Only exceeding weekly limits b N=226	Exceeding daily and weekly limits a, b N=141	Test Statistic c	P c
Days exceeding daily limits in last 90 days	2 (1, 5)	0 (0, 0)	10 (4, 24)	148.56	<0.01 d
Drinking days per week	2 (1, 3)	6 (4, 7)	6 (4, 7)	476.37	<0.01
Drinks per drinking day	2 (1, 3)	3 (2, 4)	4 (3, 5)	229.68	<0.01
Average drinks per week
Men <66 years old	4 (2, 8)	21 (16, 25)	21 (16, 28)	0.84	0.36 e
Men ≥66 years old and all women	2 (0, 4)	14 (10, 14)	14 (12, 18)	2.29	0.13 e

All numbers are median (interquartile range).

^aExceeding daily limits was defined as reporting ≥5 drinks in a day at least once in the last 90 days for men <66 years old, and ≥4 drinks in a day at least once in the last 90 days for men ≥66 years old and all women.

^bExceeding weekly limits was defined as reporting ≥15 average weekly drinks for men <66 years old, and ≥8 average weekly drinks for men ≥66 years old and all women.

^cStatistics and P values from Kruskal-Wallis tests comparing patients in all three groups, unless otherwise specified.

^dTests comparing only patients who reported exceeding daily limits.

^eTests comparing only patients who reported exceeding weekly limits.

Reduction of unhealthy alcohol use

Of the 1299 patients with unhealthy alcohol use, 1025 (79%) had ≥1 follow-up screen within two years of baseline, with a median of 1 (IQR 1, 2) follow-up screens a median of 290 days apart (IQR 210, 396). The proportion of patients with ≥1 follow-up screen within two years was similar for patients exceeding only daily, only weekly, or both limits at baseline, but screening frequency varied slightly (Supplemental Table 2). Patients with ≥1 follow-up screen were older (median age 47 vs. 41; Kruskal-Wallis=46.22, P<0.01) and more likely to have prior depression (28% vs. 22%; χ²=4.23, P<0.05), CCI ≥2 (30% vs. 15%; χ²=24.19, P<0.01), least deprived NDI (25% vs. 20%; χ²=12.82, P<0.01), and VL<200 copies/mL (80% vs. 72%; χ²=15.71, P<0.01) (Supplemental Table 3).

Among patients with ≥1 follow-up screen, the probability of reducing to low-risk/no alcohol use within two years was 73.2% (95% confidence interval [CI] 70.5%, 75.9%), and this ranged from 65.8% (57.5%, 75.2%) among patients exceeding both daily and weekly limits at baseline, to 76.0% (72.9%, 79.1%) among those only exceeding daily limits (Fig. 1). In unadjusted analyses, compared with patients only exceeding daily limits at baseline, the reduction risk ratio was 0.88 (95% CI 0.79, 0.98; P<0.05) for those only exceeding weekly limits and 0.87 (0.75, 1.00; P<0.05) for those exceeding both. After adjusting for all covariates, risk ratios were 0.88 (0.77, 1.00; P<0.05) and 0.85 (0.72, 1.00; P<0.05), respectively.

Figure 1.

Number of patients, two-year probabilities, and risk ratios for reduction to low-risk or no alcohol use, among 1025 persons with HIV in Kaiser Permanente Northern California with a positive screen for unhealthy alcohol use, 2013–2017. Patients who did not have ≥1 repeat screen within two years of index positive screen were excluded. Error bars are 95% confidence intervals. Risk ratios and 95% confidence intervals (CIs) were estimated using Poisson regression models with robust variances. Models estimating adjusted risk ratios included age, race/ethnicity, HIV risk factor, smoking status, history of alcohol use disorder, history of depression, history of non-alcohol substance use disorder, Charlson comorbidity index, neighborhood deprivation index, CD4 cell count, HIV viral load, and receipt of a brief intervention.

When evaluating any reduction in alcohol use (Supplemental Fig. 2), reduction rates were slightly higher than in primary analyses, and there was no longer a difference between patients only exceeding daily limits vs. both daily and weekly limits at baseline (adjusted risk ratio 1.01 [0.90, 1.12]; P=0.92), while other estimates were similar to the main findings.

Rates of new AUD diagnosis

Of the 1299 patients with unhealthy alcohol use, 1083 (83%) without prior AUD contributed a median of 2.4 years of follow-up (IQR 1.5, 3.6). Among these, the new AUD diagnosis rate per 100 person-years was 3.1 (95% CI 2.5, 3.8), and this ranged from 2.7 (2.1, 3.6) among patients only exceeding daily limits at baseline to 5.2 (2.9, 9.3) among those exceeding both daily and weekly limits (Fig. 2). In unadjusted analyses, compared with patients only exceeding daily limits at baseline, the AUD diagnosis rate ratio was 1.36 (95% CI 0.80, 2.32; P=0.25) for those only exceeding weekly limits and 1.88 (0.99, 3.60; P=0.05) for those exceeding both. In adjusted analyses, rate ratios were 2.09 (1.08, 4.03; P<0.05) and 2.25 (1.07, 4.76; P<0.05), respectively.

Figure 2.

Number of events, diagnosis rates, and rate ratios for alcohol use disorder (AUD) among 1083 persons with HIV in Kaiser Permanente Northern California (KPNC) with a positive screen for unhealthy alcohol use, 2013–2017. Error bars are 95% confidence intervals. Patients contributed person-time at risk from baseline (first screen positive for unhealthy alcohol use) to the earliest of: AUD diagnosis, death, interruption of KPNC membership for >90 days, 18 months without an unhealthy alcohol use screen, or 31 December 2019. Rate ratios and 95% confidence intervals (CIs) are from Poisson regression models. Models estimating adjusted rate ratios included age, race/ethnicity, HIV risk factor, smoking status, history of depression, history of non-alcohol SUD, Charlson comorbidity index, neighborhood deprivation index, CD4 cell count, HIV viral load, and receipt of a brief intervention.

Independent predictors

When examining independent predictors of reduction to low-risk/no alcohol use and of new AUD diagnosis, using separate, unadjusted models (Supplemental Table 4), being younger than 40 years was associated with worse outcomes. For example, compared with patients aged 40–49, those aged 18–29 were less likely to have a reduction (risk ratio 0.86 [0.74-0.99]; P<0.05) and more likely to be diagnosed with AUD (rate ratio 2.31 [1.11, 4.78]; P<0.05). Women were more likely than men to have a reduction (risk ratio 1.18 [1.06, 1.33]; P<0.01) but had similar AUD diagnosis rates (rate ratio 1.16 [0.47, 2.85]; P=0.75). Patients with CCI ≥2 vs. 0 had a reduction risk ratio of 1.07 (0.99, 1.15; P=0.11) and AUD diagnosis rate ratio of 0.57 (0.33, 0.96; P<0.05). AUD diagnosis rates were higher for patients who were current vs. non-smokers, with history of non-alcohol SUD, and with a VL≥200 vs. <200 copies/mL. Receipt of a brief intervention was not associated with either outcome.

DISCUSSION

Among PWH in an integrated health system with unhealthy alcohol use, the majority (72%) reported exceeding daily but not weekly recommended limits. Patients only exceeding daily limits were more likely to be men, younger, and of minority race/ethnicity, and patients only exceeding weekly limits had more comorbidities. Within two years, 73% of re-screened patients reduced to low-risk/no alcohol use. Among patients without baseline AUD, we observed 3 new AUD diagnoses per 100 person-years. Patients who reported exceeding both daily and weekly limits at baseline were least likely to have a reduction and most likely to be diagnosed with AUD during follow-up. Younger age was associated with a lower probability of reduction, and younger age, smoking, having an unsuppressed viral load, and comorbid SUD were associated with higher AUD diagnosis rates.

Few studies have examined daily and weekly unhealthy alcohol use among PWH. A US multisite HIV cohort study reported that exceeding daily limits was more prevalent than AUDIT-C-defined at-risk drinking (32% vs. 16%), and that 14% of those exceeding daily limits had excessive drinking at least one day per week (1). In another study from the same cohort, 24% exceeded daily limits only, and 4% exceeded weekly limits (with or without excessive daily drinking) (32). In a Florida cohort, 25% of PWH reported exceeding daily limits only, and 9% reported exceeding weekly limits (with or without excessive daily drinking) (42). For these latter two studies, the reported estimates represent 85% and 75%, respectively, of unhealthy alcohol use being accounted for by exceeding daily limits only. Our findings are consistent in highlighting the relative contribution of excessive daily drinking in PWH with unhealthy alcohol use, and extend previous studies in two ways. First, our study further shows the variation in excessive drinking frequency, with more days of excessive drinking for patients who also reported exceeding weekly limits vs. those who did not (median 10 vs. 2 days in the last 90 days). Second, our findings highlight distinct demographic and clinical profiles associated with drinking patterns among PWH with unhealthy alcohol use, most notably a 15-year median age difference between PWH exceeding daily vs. weekly limits. This age difference is consistent with findings from the general population (66, 67).

This is also one of few studies reporting on unhealthy alcohol use reduction in HIV primary care. In a KPNC trial, PWH in the standard-of-care arm (which included SBIRT) reduced the number of excessive drinking days by half over a year (34). Among military veterans, 60% of PWH with elevated AUDIT-C scores reduced to low-risk/no alcohol use within a year (40). In a US cohort, 31% of PWH exceeding only daily limits reduced to low-risk/no alcohol use within 18 months, compared with 16% for those with excessive weekly drinking (with or without excessive daily drinking) (32). In the same cohort, 25% PWH with elevated AUDIT-C scores had a reduction over approximately two years (25). In contrast, several studies have reported stable longitudinal alcohol use levels among PWH with unhealthy, low-risk, or no alcohol use at baseline (24, 27-30, 68). While direct comparison is hampered by study differences, the frequency of reduction in our study appears higher than previous reports. This could be partly explained by population differences. PWH in KPNC are mostly privately insured individuals who may have fewer risk factors for persistent unhealthy alcohol use than those receiving care at Veterans Affairs or academic medical centers. In the KPNC general population, 63% of people with unhealthy alcohol use reduced to low-risk/no use within a year (69).

One potential explanation for the differences in alcohol use reduction and AUD diagnosis rates we observed by baseline alcohol use patterns is that these patterns reflected different reasons for drinking. PWH in a large clinic drinking due to social pressure had more excessive drinking days and drinks per drinking day, but not higher overall drinking frequency, while PWH drinking to cope with negative affect had higher alcohol use across all frequency and quantity patterns, and more alcohol dependence symptoms during follow-up (44, 70). Drinking reasons might affect alcohol use reduction or progression to AUD, with patients facing different challenges such as social pressure vs. untreated depression or anxiety. We adjusted for prior depression diagnosis in our analyses but may not have captured ongoing problems or severity of depression, and we did not consider drinking reasons, or social factors like partner drinking, which future studies should explore.

Another potential explanation for our findings is that, by definition, reducing to low-risk/no alcohol use represented a greater change for patients with more severe baseline unhealthy alcohol use. When counting any decrease in use as a reduction, we no longer observed a difference in reduction frequency between patients exceeding only daily vs. both daily and weekly limits, suggesting that as a harm reduction approach, encouraging PWH with unhealthy alcohol use to reduce any amount may be successful. Any reduction could be beneficial, including for HIV disease severity, viral suppression, and sexual risk behaviors (29, 31-33). However, PWH with only excessive weekly drinking still had lower reduction rates in this sensitivity analysis. These patients might face challenges to even small reductions in alcohol use. Additionally, alcohol use above the recommended limits can still lead to adverse health outcomes, and reduction to low-risk levels should be supported with adequate resources and treatment as needed (71).

Several characteristics were independently associated with reductions and AUD diagnoses in our study. Patients with more comorbidities were more likely to reduce to low-risk/no alcohol use and less likely to be diagnosed with AUD, which could be explained by motivation to reduce alcohol use for health reasons (67, 72, 73). Women were also more likely to reduce in our study, whereas women and men with elevated AUDIT-C scores in a trajectories study had similar probabilities of reduction (25). We found that younger patients had lower reduction rates and higher new AUD diagnosis rates, which has not previously been reported. Patients who were not virally suppressed at baseline had higher AUD diagnosis but not reduction rates. Previous work has shown that viral non-suppression was associated with persistent at-risk alcohol use, and lower rates of reduction (25, 27). We also found that injection drug use history and non-alcohol SUD were associated with lower reduction and higher AUD rates, consistent with several prior studies (24, 27-29, 68).

An important implication of our findings is that, among PWH identified as having unhealthy alcohol use, it may be clinically useful to further consider whether patients have excessive daily drinking, weekly drinking, or both. We found that these patterns were predictive of reduction and new AUD diagnosis and could help providers identify patients most in need of interventions such as reduction counseling, harm reduction messaging, medications to reduce alcohol use, and referral to specialty care. Evaluation of alcohol use patterns can also be an opportunity to assess context and reasons for drinking, motivation to reduce, and comorbid mental health and substance use problems (34). This information can help further target interventions and provide resources to patients, including depression or anxiety treatment and strategies to mitigate social pressure. Importantly, additional research is needed to understand drivers of worse outcomes among PWH with unhealthy alcohol use who are younger than 40 years old, and whether these patients require particular attention in clinical care.

This study’s strengths include a large sample size and almost seven years of data, leveraging repeated alcohol use measurements and AUD diagnoses from the EHR. Although alcohol use was measured at health encounters as part of routine care rather than investigator-specified time points of interest (e.g. entry to HIV care), analyzing screens from routine care enabled us to emulate clinical decision-making by identifying which patients at an index routine screen may be less likely to reduce by their next encounter and thus require targeted interventions. While we had a large sample size, approximately half of PWH in this study were White and most were men, mostly MSM, with a median age of 46 years. This demographic distribution is representative of PWH in Northern California but includes a greater proportion of White PWH and MSM compared with large US clinical cohorts (1, 74, 75). In addition, almost all PWH in KPNC are privately insured, compared with 40% of PWH in California and 34% nationally, and they are members of an integrated health care system (2, 76). PWH in KPNC may therefore have better access to care than PWH in academic center-based cohorts or other clinical settings. However, PWH in KPNC have viral suppression rates that are similar to PWH in care in large US cohorts (77, 78). Therefore our findings may not be generalizable to all PWH, particularly those who are not in care, or to clinical populations with a greater proportion of PWH of minority race/ethnicity, with higher barriers to care, or who are women, particularly given differences in NIAAA weekly limits for men and women.

Another strength of our analyses is that we were able to adjust for receipt of a brief intervention, which varied between comparison groups in our study and could be an important confounder. Clinical trials in the general population and among PWH have shown that brief interventions can lead to greater reductions in alcohol use than screening alone, although in our analyses brief intervention receipt was not associated with a reduction, similar to some other clinical care-based studies (36, 40, 79, 80). While we did not adjust for other alcohol-related treatments, a previous KPNC study reported that only 5% of PWH with unhealthy alcohol use initiated addiction specialty treatment (63).

One important limitation is the use of self-reported alcohol use measures, which could underestimate true consumption due to social desirability bias. Another possible limitation is that AUD went undiagnosed during follow-up as it is not systematically assessed during care, meaning AUD diagnosis rates in our study may underestimate true AUD incidence rates. The level of underdiagnosis may also have varied across patient groups. Patients with more severe unhealthy alcohol use at baseline could have received more attention from providers during follow-up, leading to AUD diagnosis. Conversely, these patients might have been less likely to return to primary care and have additional alcohol use screening. However, almost 80% of patients had a follow-up screening within two years in our study, with proportions comparable across the three alcohol use groups. Nonetheless, re-screened patients were more likely to have CCI ≥2, reside in less deprived areas, and have suppressed VLs, and this could have impacted our results. Another consideration is that most patients only had 1–2 repeat screens during follow-up, and a reduction in alcohol use at a single measurement may not represent a durable change. Future studies should examine rates of durable reductions in alcohol use in PWH. Finally, AUD diagnoses recorded in the EHR may not be 100% specific, and some PWH with a recorded ICD code may not have been adequately evaluated for and diagnosed with AUD.

CONCLUSIONS

Among PWH with unhealthy alcohol use, patients who at baseline exceeded only daily limits, only weekly limits, or both varied in demographic and clinical profiles and post-index screen outcomes, with more frequent reduction and lower AUD rates for patients only exceeding daily limits. Younger patients experienced lower rates of reduction and higher rates of AUD diagnosis. Interventions to reduce unhealthy alcohol use in HIV primary care should consider drinking patterns and potentially associated factors that could affect longitudinal alcohol use and should be addressed in clinical care.

Supplementary Material

1839879_Sup_Info_File

Supplemental Table 1. International Classification of Diseases (ICD) diagnostic codes used to identify comorbidities.

Supplemental Table 2. Screening for unhealthy alcohol use within two years of an index positive screen, stratified by use patterns at index screen.

Supplemental Table 3. Patient characteristics stratified by availability of alcohol use screens within two years of index screen positive for unhealthy alcohol use.

Supplemental Table 4. Independent predictors of reduction and alcohol use disorder (AUD) among persons with HIV (PWH) in Kaiser Permanente Northern California with a positive screen (baseline) for unhealthy alcohol use.

Supplemental Figure 1. Flowchart of patient inclusion and exclusion.

Supplemental Figure 2. Probabilities and risk ratios of reduction in unhealthy alcohol use, in a sensitivity analysis counting any decrease in the number of days exceeding daily limits or in the number of average drinks per week.