How Does Telehealth Expansion Change Access to Healthcare for Patients With Different Types of Substance Use Disorders?

Abstract

Background:

Patients with substance use disorders (SUDs) exhibit low healthcare utilization despite high medical need. Telehealth could boost utilization, but variation in uptake across SUDs is unknown.

Methods:

Using Wisconsin Medicaid enrollment and claims data from December 1, 2018, to December 31, 2020, we conducted a cohort study of telemedicine uptake in the all-ambulatory and the primary care setting during telehealth expansion following the COVID-19 public health emergency (PHE) onset (March 14, 2020). The sample included continuously enrolled (19 months), nonpregnant, nondisabled adults aged 19 to 64 years with opioid (OUD), alcohol (AUD), stimulant (StimUD), or cannabis (CannUD) use disorder or polysubstance use (PSU). Outcomes: total and telehealth visits in the week, and fraction of visits in the week completed by telehealth. Linear and fractional regression estimated changes in in-person and telemedicine utilization. We used regression coefficients to calculate the change in telemedicine utilization, the proportion of in-person decline offset by telemedicine uptake (“offset”), and the share of visits completed by telemedicine (“share”).

Results:

The cohort (n = 16 756) included individuals with OUD (34.8%), AUD (30.1%), StimUD (9.5%), CannUD (9.5%), and PSU (19.7%). Total and telemedicine utilization varied by group post-PHE. All-ambulatory: total visits dropped for all, then rose above baseline for OUD, PSU, and AUD. Telehealth expansion was associated with visit increases: OUD: 0.489, P < .001; PSU: 0.341, P < .001; StimUD: 0.160, P < .001; AUD: 0.132, P < .001; CannUD: 0.115, P < .001. StimUD exhibited the greatest telemedicine share. Primary care: total visits dropped for all, then recovered for OUD and CannUD. Telemedicine visits rose most for PSU: 0.021, P < .001; OUD: 0.019, P < .001; CannUD: 0.011, P < .001; AUD: 0.010, P < .001; StimUD: 0.009, P < .001. PSU and OUD exhibited the greatest telemedicine share, while StimUD exhibited the lowest. Telemedicine fully offset declines for OUD only.

Conclusions:

Telehealth expansion helped maintain utilization for OUD and PSU; StimUD and CannUD showed less responsiveness. Telehealth expansion could widen gaps in utilization by SUD type.

Introduction

The majority of individuals with substance use disorders (SUDs) receive insufficient medical care, including primary care, over their lifetime.^1–5 Yet, SUDs substantially contribute to the development of health problems such as heart, liver, and lung disorders; depression, anxiety, and suicidality; and infectious diseases like HIV, hepatitis C, and COVID-19.^6–10 Consequently, increasing receipt of ambulatory care, especially primary care, may be one approach to improving the health of individuals with SUDs.^11–14

Telehealth (the provision of healthcare services via telecommunication platforms)¹⁵ presents a potential strategy to increase SUD treatment utilization by reducing costs and geographic barriers, simplifying logistics, and enhancing confidentiality.^15–20 The COVID-19 public health emergency (PHE) prompted dramatic expansion of telehealth service utilization including in the Wisconsin Medicaid program creating an opportunity to examine telehealth uptake among patients with SUDs.^21,22 However, the ability of telehealth expansion to promote telemedicine (synchronous visits between patients and clinicians) with ambulatory and primary care providers may differ by SUD type.¹⁵ If true, expansion of telehealth services may do more to support treatment access for patients with certain SUDs over others.

Reasons for potential variation in telemedicine uptake by SUD type reflect clinical and treatment characteristics. For example, primary care providers perform a substantial proportion of treatment for opioid use disorder (OUD) particularly leveraging 2 prescribed medications for OUD (MOUD), buprenorphine and naltrexone.^23,24 Methadone, a third form of MOUD, is a key intervention in treating OUD but can only be prescribed out of federally regulated opioid treatment programs (OTP).²⁴ During the PHE, federal regulations were loosened to allow initiation of controlled substances, including buprenorphine but not methadone, via telemedicine.²² Research has subsequently demonstrated both the rise and effectiveness of telemedicine MOUD treatment.^25,26 The exact mechanisms explaining the success of telemedicine MOUD management have not yet been identified, but may include reduced access barriers while leveraging patient motivation to avoid medication withdrawal on discontinuation.^27–34 Thus, it seems likely that telehealth expansion would promote increased telemedicine among patients with OUD.^35,36

In contrast with OUD’s emphasis on medications, treatment of stimulant use disorder (StimUD) emphasizes structured behavioral health programming, specifically contingency management.^37,38 Contingency management programs are rarely delivered via telehealth platforms or in primary care settings.^26,39 Moreover, rates of ambulatory treatment are particularly low among patients with StimUD,⁴⁰ while rates of emergency department^41,42 and inpatient⁴³ utilization are rising.^37,44,45As such, telehealth expansion may be less likely to promote increased ambulatory telemedicine, especially primary care, among patients with StimUD relative to OUD.

In contrast with both OUD and StimUD, treatment of alcohol use disorder (AUD) can include both medications and behavioral health.⁴⁶ However, while several medications are indicated to treat AUD (MAUD),⁴⁷ these medications may not change treatment trajectories as profoundly as MOUD. In addition, medication receipt is low, in part due to limited access to care.⁴⁸ These features of AUD treatment suggest that patients with AUD may exhibit lower telemedicine responsiveness to ambulatory and primary care telehealth expansion than OUD. Still, a substantial portion of AUD treatment occurs in the primary care setting.^49–51 Moreover, heavy alcohol use is associated with substantial medical complications requiring outpatient (including primary care based) chronic disease management.^52,53 As such, preexisting patient-provider relationships as well as ongoing health needs could promote telemedicine uptake among patients with AUD. Notably, research suggests the feasibility and effectiveness of using telemedicine to support patients with AUD from screening to treatment.^54,55

Cannabis use disorder (CannUD) again differs from other SUDs with regard to treatment components and treatment receipt in ambulatory medical settings such as primary care. There are no medications indicated for CannUD,⁵⁶ and while psychosocial interventions can promote reduced use, there are few widespread CannUD-focused interventions designed for the primary care setting.⁵⁷ Thus, individuals with CannUD lack the medication incentive offered by MOUD or MAUD to seek healthcare services via platforms like telehealth. As such, telehealth expansion is unlikely to substantially increase telemedicine receipt for patients with CannUD.

Finally, many patients with SUDs use more than 1 substance. Polysubstance use (PSU) is associated with worse health outcomes and greater social complexity due to housing instability, unemployment, and poverty.^10,58–60 Patients with PSU may therefore experience greater medical need but more barriers to care than patients with 1 SUD, and consequently exhibit more modest responsiveness to telehealth expansion.

Given these unique clinical, treatment, and social factors characterizing distinct SUD types, telehealth expansion may not symmetrically promote ambulatory and primary care telemedicine use across SUDs. This article examines total and telemedicine visit utilization changes in all-ambulatory and primary care settings for patients with SUDs. We leverage Wisconsin Medicaid policy changes put into place at the onset of the COVID-19 PHE.^21,22 We hypothesized that individuals with OUD would exhibit the greatest increases in telemedicine utilization while individuals with StimUD would exhibit the smallest increases in telemedicine utilization. We hypothesized that patients with AUD, CannUD, and PSU would exhibit moderate increases in telemedicine utilization.

Methods

Data Source

This study used Wisconsin Medicaid enrollment and claims data from December 1, 2018, to December 31, 2020. We used claims data to identify SUD diagnoses, as well as psychiatric comorbidities. Enrollment data confirmed eligibility basis and provided information about baseline sociodemographic characteristics: age, sex, race, ethnicity, education, income, and geography. This study was determined exempt from review and informed consent by the University of Wisconsin Institutional Review Board (common rule, category 5).

Population

The study cohort included adults aged 19 to 64 years enrolled in Wisconsin Medicaid as parents/caretakers or childless adults. Both age and eligibility pathway were assessed at baseline (June 1, 2019). We required continuous enrollment during the outcome ascertainment period (June 1, 2019-December 31, 2020) to minimize likelihood of misclassification by SUD type and increase likelihood of identifying all healthcare utilization. Cohort inclusion required at least one claim during a 6-month look back from December 1, 2018, to May 31, 2019, in the outpatient, inpatient, or emergency department setting for a diagnosis of AUD (F10), OUD (F11), CannUD (F12), sedative use disorder (SedUD; F13), or StimUD (F14-F15) using the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10). Continuous enrollment was not required during this look back period. Individuals with 2 or more SUD diagnoses (of OUD, AUD, StimUD, CannUD, and SedUD) were classified as PSU. Due to small cell size, we excluded individuals with SedUD and no other SUDs (N = 71). We classified individuals with both SedUD and another use disorder as PSU. All individuals were sorted into 1 of 5 mutually exclusive groups: OUD, AUD, StimUD, CannUD, and PSU. See Online Supplemental Material 1 to 2 for more information about cohort construction.

Outcome Assessment and Covariates

Outcome measures included number of all-ambulatory and number of primary care visits at the person-week level (total and telemedicine) and fraction of visits completed by telemedicine at the person-week level. We performed analyses for all-ambulatory visits and primary care clinician visits. Analyses include methadone dose administration visits. We used provider specialty codes and rendering provider taxonomy to identify primary care clinician visits (allowing physicians and advanced practice providers but excluding counselors, social workers, and nurses). We identified telemedicine visits using (1) procedure codes or (2) the presence of either a place of service code or modifier indicating telemedicine. See Online Supplemental Material 3 for more information about identifying primary care and telemedicine visits. We coded the post period as the first day of the complete person-week, March 14, 2020, containing the Governor of Wisconsin’s PHE declaration.⁶¹ Covariates included age, sex, race, ethnicity, income (as percentage of the federal poverty level [FPL]), geography (residing in a rural or urban county), and presence of a major psychiatric disorder (diagnoses involving psychotic symptoms, such as schizophrenia or bipolar disorder) using ICD-10 codes (F200 through F319, F323, F333, F340) in the 6-month look back period. Due to small cell sizes, Asian and Pacific Islander are reported together in the demographics table but separate in regression analyses. The pre- and post-PHE periods were defined as June 1, 2019, to March 13, 2020, and March 14, 2020, to December 31, 2020, respectively.

Statistical Analysis

Sociodemographic characteristics were summarized for the full cohort and each SUD subgroup. Subgroup characteristics were compared to the population mean using single sample t tests. Visit rates were estimated as average visit count in the week by any modality and telemedicine, specifically. We conducted linear regression to test for differences in the change in in-person and telemedicine visits per person per week by SUD type pre- and post-PHE (Models 1 and 2). We conducted fractional regression to test for differences in the share of visits completed by telemedicine by SUD type after the PHE (Model 3). All models adjusted for covariates. See Online Supplemental Material 4 for technical details. Regression coefficients were used to calculate 5 measures for each SUD type: total and percent change in overall visit utilization, total and percent change in telemedicine visit utilization, the proportion of the decrease in in-person visits offset by the increase in telemedicine utilization (herein, telehealth offset), and the change in the share of visits completed by telemedicine (herein, telemedicine share). Analyses were conducted using Stata statistical software (version 17; Stata Corp, LLP) and SAS 9.4 (SAS Institute). The statistical significance level was set at .05. Analyses were conducted from August 2021 to January 2023.

Results

Cohort Characteristics

We identified 16 756 individuals who met the inclusion criteria related to age, eligibility pathway, and enrollment continuity. OUD and AUD were the most prevalent SUDs in the cohort (34.8% and 30.1%, respectively) followed by PSU (19.7%; Table 1). The ordered prevalence of SUDs among individuals with PSU (n = 3293) was OUD (75.2%), StimUD (75.1%), AUD (63.2%), followed by CannUD (36.8%). The average age was 38.9 years; a majority were female (53.8%). The largest racial categories included American Indian (4.7%), Black (14.5%), and White (71.2%). A minority of the cohort was Hispanic (6.7%). Most had finished high school or more (59.1%) but over 80% reported income ≤50% FPL. Nearly two-thirds lived in urban counties. Finally, 18.1% exhibited a major psychiatric comorbidity diagnosis.

Table 1.

Characteristics of Continuously Enrolled Wisconsin Medicaid Beneficiaries With OUD, AUD, CannUD, StimUD, or PSU.

Characteristics	All, n (%)	OUD, n (%)	AUD, n (%)	CannUD, n (%)	StimUD, n (%)	PSU, n (%)
Unique subjects	16 756 (100.0%)	5837 (34.8%)	5049 (30.1%)	1593 (9.5%)	984 (9.5%)	3293 (19.7%)
Eligibility category
Childless adults	11 452 (68.3%)	3427 (58.7%)***	3802 (75.3%)***	1002 (62.9%)***	706 (71.7%)	2515 (76.4%)
Parents/caretakers	5304 (31.7%)	2410 (41.3%)	1247 (24.7%)	591 (37.1%)	278 (28.3%)	778 (23.6%)
Sex
Female	8002 (47.8%)	3141 (53.8%)***	2087 (41.3%)***	734 (46.1%)	487 (49.5%)	1553 (47.2%)
Male	8754 (52.2%)	2696 (46.2%)	2962 (58.7%)	859 (53.9%)	497 (50.5%)	1740 (52.8%)
Age
Mean age (SD), years	38.9 (10.3)	37.4 (8.8)***	43.4 (10.9)***	33.6 (10.2)***	38.4 (10.5)	37.6 (9.5)***
Race
American Indian	780 (4.7%)	245 (4.2%)	243 (4.8%)	58 (3.6%)*	38 (3.9%)	196 (6.0%)***
Asian or Pacific	114 (0.7%)	29 (0.5%)	41 (0.8%)	22 (1.4%)*	13 (1.3%)*	9 (0.3%)**
Islander
Black	2431 (14.5%)	464 (7.9%)***	827 (16.4%)***	527 (33.1%)***	217 (22.1%)***	396 (12.0%)***
Multiracial	354 (2.1%)	121 (2.1%)	74 (1.5%)**	52 (3.3%)**	23 (2.3%)	84 (2.6%)
White	11 928 (71.2%)	4585 (78.6%)***	3503 (69.4%)**	819 (51.4%)***	643 (65.3%)***	2378 (72.2%)
Race missing	1149 (6.9%)	393 (6.7%)	361 (7.1%)	115 (7.2%)	50 (5.1%)*	230 (7.0%)
Ethnicity
Hispanic	1118 (6.7%)	424 (7.3%)	263 (5.2%)***	125 (7.8%)	57 (5.8%)	249 (7.6%)*
Not Hispanic	15 436 (92.1%)	5374 (92.1%)	4680 (92.7%)	1453 (91.2%)	915 (93.0%)	3014 (91.5%)
Missing	202 (1.2%)	39 (0.7%)***	106 (2.1%)***	15 (0.9%)	12 (1.2%)	30 (0.9%)
Education
High school or more	9910 (59.1%)	3602 (61.7%)***	2712 (53.7%)***	968 (60.8%)	603 (61.3%)	2025 (61.5%)**
Less than high School	3324 (19.8%)	1205 (20.6%)	880 (17.4%)***	372 (23.4%)***	228 (23.2%)***	639 (19.4%)
Missing	3522 (21.0%)	1030 (17.6%)***	1457 (28.9%)***	253 (15.9%)***	153 (15.5%)***	629 (19.1%)**
Income
≤50 FPL	14 066 (83.9%)	4869 (83.4%)	4109 (81.4%)***	1280 (80.4%)***	878 (89.2%)***	2930 (89.0%)***
50%–100% FPL	2690 (16.1%)	968 (16.6%)	940 (18.6%)	313 (19.6%)	106 (10.8%)	363 (11.0%)
Geography
Urban	10 881 (64.9%)	3770 (64.6%)	3205 (63.5%)*	1042 (65.4%)	619 (62.9%)	2245 (68.2%)***
Rural	3424 (20.4%)	1190 (20.4%)	1084 (21.5%)	305 (19.1%)	229 (23.3%)*	616 (18.7%)*
Missing	2451 (14.6%)	877 (15.0%)	760 (15.1%)	246 (15.4%)	136 (13.8%)	432 (13.1%)*
Major psychiatric comorbidity
Diagnosis	2551 (18.1%)	467 (8.0%)***	703 (13.9%)***	335 (21.0%)**	226 (23.0%)***	820 (24.9%)***

Abbreviations: OUD, opioid use disorder; AUD, alcohol use disorder; StimUD, stimulant use disorder; CannUD, cannabis use disorder; PSU, polysubstance use; SD, standard deviation; FPL, federal poverty level.

^*P < .05.

^**P < .01.

^***P < .001

Table 1 shows variation in sociodemographic characteristics by SUD type. For example, compared to the sample average, individuals with OUD were more often female (53.8%) while individuals with AUD were more often male (58.7%). Individuals with AUD tended to be older (mean: 43.4 years) while individuals with CannUD tended to be younger (mean: 33.6 years). Individuals with OUD were more likely to be White (71.2%) and less likely to be Black (7.9%); the inverse was true for StimUD (White: 65.3%; Black: 22.1%). Individuals with StimUD and PSU were more likely than average to be ≤50% FPL (89.2% and 89.0%, respectively) and have a major psychiatric comorbidity (23.0% and 24.9%, respectively).

Trends in Total and Telemedicine Visits

Average visit counts are shown in the all-ambulatory (Figure 1) and primary care setting (Figure 2) by modality (total and telemedicine). In the all-ambulatory setting, visits initially dropped sharply and then rose for OUD and PSU post-PHE. This rise is explained by a jump in telemedicine visits for these 2 SUD categories. In contrast, overall visit counts remained relatively stable for AUD, CannUD, and StimUD, a substantial proportion of which were completed via telemedicine. In the primary care setting, visits for OUD and PSU again dropped sharply but did not recover completely. Primary care visit levels for AUD, StimUD, and CannUD dropped modestly. Primary care telemedicine visits rose for all SUD groups but waned over time relative to the ambulatory telemedicine trends. In-person visits remained the dominant primary care visit type across SUDs.

Figure 1.

Trends in all-ambulatory visits at the person-week level among continuously enrolled Wisconsin Medicaid beneficiaries with substance use disorders before and after the public health emergency.

Black vertical line indicates the declaration of the public health emergency. Trend lines indicate visit rates in total and for telemedicine visits by type of substance use disorder. Trend lines indicate visit rates for any modality and just telemedicine by type of substance use disorder.

Abbreviations: OUD, opioid use disorder; AUD, alcohol use disorder; StimUD, stimulant use disorder, CannUD, cannabis use disorder; PSU, polysubstance use.

Figure 2.

Trends in primary care visits at the person-week level among continuously enrolled Wisconsin Medicaid beneficiaries with substance use disorders before and after the public health emergency.

Black vertical line indicates the declaration of the public health emergency. Trend lines indicate visit rates for any modality and just telemedicine by type of substance use disorder.

Abbreviations: OUD, opioid use disorder; AUD, alcohol use disorder; StimUD, stimulant use disorder, CannUD, cannabis use disorder; PSU, polysubstance use.

All-Ambulatory Visit Rates by SUD Type

Table 2 presents measures of total and telemedicine utilization for each SUD type in the all-ambulatory setting. All SUD groups exhibited a decrease in cohort percent with a visit in the week (Panel 2A) but an increase in average total ambulatory visits in the week (Panel 2B). Of these, OUD and PSU exhibited the greatest absolute and proportionate changes in total visits per person-week (OUD: 0.298 [139.5%]; P < .001; PSU: 0.221 [22.8%]; P < .001; Panel 2C, columns 1 and 2). These visits for patients with OUD and PSU translate into over 15 and 11 visits per person per year on average, respectively. The changes observed for StimUD and CannUD were not significant. Relative to the change observed for OUD, the other SUD groups exhibited fewer visits post-PHE (AUD: 0.278 fewer, P < .001; StimUD: 0.273 fewer, P < .001; CannUD: 0.275 fewer, P < .001; PSU: 0.077 fewer, P = .020).

Table 2.

Total and Telehealth Utilization in the All-Ambulatory Setting for Continuously Enrolled Wisconsin Medicaid Beneficiaries With SUDs, Overall and Relative to OUD.

	2A				2B				2C
	pre-PHE	post-PHE	pre-PHE	post-PHE	pre-PHE	post-PHE	pre-PHE	post-PHE	1	2		3	4		5	6
	Total visits		Telehealth		Total visits		Telehealth		Total visits			Tele visits			Offset	Share
SUD type	% any visit in the week				Average weekly visit count				Change	% change	P	Change	% change	P	(3)/(3–1)		P
OUD	58.5%	53.6%	0.6%	14.9%	2.164	2.462	0.025	0.863	0.298	139.5%	<.001	0.489	12 530.8%	<.001	2.556	0.203	<.001
AUD	23.9%	20.1%	0.2%	7.6%	0.371	0.390	0.006	0.195	0.019	4.8%	.024	0.132	1393.7%	<.001	1.171	0.322	<.001
Relative to OUD									−0.278		<.001	−0.356		<.001		0.120	<.001
StimUD	25.4%	21.2%	0.4%	8.5%	0.432	0.457	0.015	0.242	0.025	5.3%	.322	0.160	3467.4%	<.001	1.185	0.334	<.001
Relative to OUD									−0.273		<.001	−0.329		<.001		0.131	<.001
CannUD	22.4%	18.6%	0.3%	7.0%	0.32	0.342	0.009	0.165	0.022	6.1%	.119	0.115	1489.6%	<.001	1.242	0.324	<.001
Relative to OUD									−0.275		<.001	−0.374		<.001		0.121	<.001
PSU	39.6%	35.8%	1.0%	13.9%	0.99	1.211	0.044	0.557	0.221	22.8%	<.001	0.341	42 587.5%	<.001	2.834	0.31	<.001
Relative to OUD									−0.077		.02	−0.148		<.001		0.108	<.001

Abbreviations: PHE, public health emergency; SUD, substance use disorder; OUD, opioid use disorder; AUD, alcohol use disorder; StimUD, stimulant use disorder; CannUD, cannabis use disorder; PSU, polysubstance use. Bold values indicates significant.

Post-PHE, all groups exhibited increased telemedicine utilization to differing degrees. Specifically, all groups exhibited an increase in the cohort percentage utilizing telemedicine (Panel 2A) as well as an increase in average telemedicine visits per person-week (Panel 2B). OUD exhibited the greatest increase (0.489 [12 530.8%]; P < .001, Panel 2C, columns 3 and 4). PSU and OUD exhibited the largest telemedicine offset, while AUD and StimUD exhibited the smallest offset (PSU: 2.834; OUD: 2.556; CannUD: 1.242; StimUD: 1.185; AUD: 1.171; Panel 2C, column 5). StimUD completed the greatest telemedicine share (0.334; P < .001), while OUD exhibited the smallest telemedicine share (0.203, P < .001).

Exploratory analyses aimed to understand the unexpected increases in treatment receipt for OUD and PSU. To this end, we examined the distribution of telehealth services by modality and provider type for these groups. This work revealed that the majority of observed increases was explained by phone-only billing codes used by addiction treatment counselors as well as a smaller number of family medicine, internal medicine, and psychiatry clinicians.

Primary Care Visit Rates by SUD Type

Table 3 presents measures of total and telemedicine primary care utilization for each SUD type. All SUD groups exhibited a decrease in cohort percent with a visit in the week (Panel 3A). Only OUD exhibited an increase in total visits post-PHE (Panel 3B) that was not significant (Panel 3C, column 1). The greatest proportionate decreases in visits were observed for PSU (−26.3%) and StimUD (−23.3%; Panel 3C, column 2).

Table 3.

Total and Telehealth Utilization in the Primary Care Setting for Continuously Enrolled Wisconsin Medicaid Beneficiaries With SUDs, Overall and Relative to OUD.

	3A				3B				3C
	pre-PHE	post-PHE	pre-PHE	post-PHE	pre-PHE	post-PHE	pre-PHE	post-PHE	1	2		3	4		5	6
	Total visits		Telehealth		Total visits		Telehealth		Total visits			Tele visits			Offset
SUD type	% any visit in the week				Average weekly visit count				Change	% change	P	Change	% change	P	(3)/(3–1)	Share	P
OUD	9.3%	7.6%	0.1%	1.8%	0.108	0.110	0.001	0.018	0.002	1.8%	.713	0.019	4675.0%	<.001	1.118	0.235	<.001
AUD	5.7%	4.7%	0.0%	1.0%	0.059	0.049	0.000	0.01	−0.01	−17.0%	<.001	0.010	5150.0%	<.001	0.505	0.206	<.001
Relative to OUD									−0.012		.029	−0.008		<.001		−0.030	<.010
StimUD	5.6%	4.3%	0.2%	0.9%	0.058	0.044	0.002	0.009	−0.014	−23.3%	<.001	0.009	4500.0%	<.001	0.393	0.204	<.001
Relative to OUD									−0.016		.01	−0.010		<.001		−0.031	>.050
CannUD	5.1%	4.3%	0.0%	1.0%	0.052	0.047	0.000	0.01	−0.005	−10.0%	.08	0.011	216000.0%	<.001	0.666	0.237	<.001
Relative to OUD									−0.007		.233	−0.008		<.001		0.002	>.050
PSU	10.2%	7.2%	0.1%	2.0%	0.112	0.083	0.001	0.02	−0.029	−26.3%	<.001	0.021	8480.0%	<.001	0.419	0.282	<.001
Relative to OUD									−0.031		<.001	0.003		<.050		0.047	<.001

Abbreviations: PHE, public health emergency; SUD, substance use disorder; OUD, opioid use disorder; AUD, alcohol use disorder; StimUD, stimulant use disorder; CannUD, cannabis use disorder; PSU, polysubstance use. Bold values indicates significant.

Post-PHE, all groups exhibited increased telemedicine utilization to differing degrees. The greatest absolute increases were observed for OUD (0.019, P < .001) and PSU (0.021, P < .001; Panel 3C). The smallest absolute increase was for StimUD (0.009; Panel 3C, column 3). Telemedicine only fully offset the drop in in-person visits for OUD (1.118; Panel 3C, column 5). StimUD exhibited the smallest offset (0.393) and share (0.204, P < .001; Panel 3C, column 6).

Discussion

In this study of Wisconsin Medicaid beneficiaries with SUD diagnoses, we observed a decline in the proportion of individuals receiving healthcare post-PHE for all SUDs in the all-ambulatory setting. At the same time, we observed an increase in visits for patients with OUD and PSU, as well as AUD to a lesser extent. These increases were driven by telemedicine expansion. In line with study hypotheses, individuals with OUD exhibited the greatest increase in utilization largely explained by use of newly approved telephone codes by addiction counselors. The unique increases in telemedicine utilization among individuals with OUD may, in part, reflect federal requirements to complete counseling as a condition of methadone treatment through OTPs.⁶² Further investigation exploring the impact of telemedicine counseling on utilization is warranted since prior literature has also demonstrated how counseling requirements may pose barriers to care.⁶³

The same telemedicine advantage was not observed for individuals with StimUD or CannUD in the all-ambulatory setting, matching study hypotheses. However, these 2 groups did demonstrate substantial growth in the share of visits completed by telemedicine. In other words, telehealth expansion was associated with a shift in the distribution of care toward telemedicine for these groups. These findings may signal potential telehealth responsiveness if offered focused outreach.

In the primary care setting, telemedicine expanded less than in the all-ambulatory setting, mirroring the literature on telemedicine expansion by medical specialty.²¹ In addition, total primary care visits dropped post PHE and never fully recovered for AUD, StimUD, and PSU. CannUD and OUD exhibited stable utilization without the overall increases observed in the all-ambulatory setting for OUD, again possibly hinting at the possible influence of counseling requirements on utilization specifically in the OTP setting. In line with study hypotheses, patients with StimUD exhibited the smallest telemedicine offset and telemedicine share. As such, telehealth is unlikely to boost overall receipt of primary care for most patients with non-OUD SUDs, especially StimUD.

In both settings, patients with OUD exhibited higher telemedicine uptake relative to other SUDs. These findings suggest particular telehealth responsiveness among patients with OUD. Concerted effort to leverage telehealth may improve health outcomes for patients with OUD: evidence demonstrates that integrating OUD treatment into primary care improves health outcomes.^64,65 Moreover, loosened regulations around prescribing MOUD in office-based settings has increased capacity for OUD treatment in primary care.⁶⁶ Assuming that telemedicine offers comparable quality to in-person care for OUD treatment,^34,67 telehealth expansion in primary care settings may offer one strategy to improve the health of patients with OUD.

Divergent findings by SUD type suggests something unique about OUD patients relative to other SUD subgroups. One possibility is that MOUD treatment may motivate care.³⁶ However, AUD can also be treated with medications and did not exhibit telemedicine uptake on par with OUD. Factors such as the acuity of overdose risk, potential for medication withdrawal, and prevalence of medication use distinguish OUD from AUD, and may explain greater telemedicine uptake among individuals with OUD relative to AUD. In addition, loosened federal regulations during the PHE obviated the need for in-person examination prior to initiating a controlled substance including some MOUD.²² This shift may have preferentially roused engagement among individuals in need of OUD treatment. Two findings from our study, however, suggest against this hypothesis. First, the percentage of individuals with OUD receiving treatment went down, not up. Second, individuals with PSU (of whom 75% have OUD) did not demonstrate the same level of telemedicine buffering as OUD. To this latter point, individuals with PSU may have more complex social circumstances limiting the ability of telehealth to buffer care disruptions.^22,58 Alternatively, PSU may moderate the impact of MOUD on care utilization because comorbid SUDs compete with the ability of these agents to stabilize symptoms.

Our findings expose a structural disparity in access to care for patients with non-OUD SUDs (particularly StimUD) exacerbated by telehealth expansion. These findings have major health implications given the high burden of morbidity and mortality across all SUD types.^6,7 The reason for these disparities is unclear but may reflect the prevalence of OUD-specific treatment programs such as federally regulated OTPs and MOUD clinics with far fewer similarly focused programs for other SUDs.⁶⁸ These OUD programs frequently offer integrated behavioral health and nurse care management to promote treatment engagement, including through telephone communication.^69,70 In this way, OUD treatment programs may have been better poised to leverage telehealth expansion, again reflecting regulations requiring counseling in OTP settings. In contrast, while research has begun to examine the feasibility of remote contingency management, such as for StimUD, these programs are rare.³⁹ Closing utilization gaps across SUD types will require the development of programs that bring a comparable level of resources to patients with non-OUD SUDs.

Expanding treatment utilization for non-OUD SUDs will require change beyond the immediate practice landscape to include research production and the culture of addiction practice. Substantial research and media attention has been given to OUD due to the unprecedented rise in fatal opioid overdose.⁷¹ This climate has likely motivated health systems to emphasize services to patients with OUD, specifically. Yet, rising rates of overdose mortality attributable to cocaine and methamphetamine demonstrate the need for increased focus on treatment for patients with StimUD, as well.^40,72 Moreover, the disproportionate representation of people who are Black among individuals with StimUD, including those suffering fatal overdose, exposes the life-threatening role that structural racism continues to play in the distribution of healthcare resources for patients with SUDs.^73–79 Our results parallel these broader trends with the greatest representation of White individuals among those with OUD and greatest representation of Black individuals among those with CannUD and StimUD. Further attention to the intersection of race, racism, and substance use in the epidemiology of healthcare utilization and SUD treatment outcomes is needed to advance health equity for patients with SUDs.

It is important to note that, for all SUD types, the initial spike in telemedicine subsided over the first 9 months of the PHE. These findings match published trends on primary care telehealth utilization for behavioral health conditions.⁸⁰ Maximally harnessing telehealth expansion to increase treatment receipt may require prolonged systemwide effort. In addition, health systems will need to identify strategies for ensuring that key SUD treatment services only available on-site, such as urine drug testing⁸¹ and serologic infectious disease monitoring,^8,9 remain available for patients.

There were several limitations. It is unclear whether increases in phone-only encounters for patients with OUD represent the provision of new quantities of care or new ways of documenting preexisting services. By requiring 19 months of continuous enrollment, findings may not represent trends for beneficiaries with more frequent disenrollment. However, this limitation is likely softened by federal legislation that prevented Medicaid disenrollment during the PHE.⁸² We did not include individuals with eligibility due to pregnancy or disability. We relied on ICD-10 diagnoses to identify SUD subgroups, a commonly used approach but with known limitations, specifically misclassification.^83,84 We were not able to describe trends for less common SUDs. Additional work is needed to assess disparities in utilization for these groups. Research has demonstrated the effectiveness of delivering behavioral health treatment via telehealth^33,34,85 including for OUD^26,35 and AUD,⁵⁵ but the literature on telehealth for CannUD and StimUD, specifically, is more limited.^39,86 The degree to which differential access to telehealth might influence health outcomes for CannUD and StimUD is, therefore, an area of needed research. Notably, this study used data through 2020 and trends may have shifted after the study period. Finally, results may not be generalizable outside of Medicaid, or outside of Wisconsin, which, uniquely, has not expanded Medicaid post-Affordable Care Act but does offer Medicaid coverage up to 100% FPL.⁸⁷

Conclusions

In this cohort study of Wisconsin beneficiaries with SUDs, individuals with OUD exhibited stable or increased treatment utilization during the PHE while individuals with StimUD, AUD, and CannUD experienced greater care disruption. The advantages observed for OUD were driven by telehealth expansion, though all groups exhibited substantial growth in telemedicine utilization. Findings suggest that telehealth expansion could support efforts to combat the opioid epidemic by increasing access to care. However, careful attention is needed to ensure that telehealth promotion does not deepen disparities in treatment receipt by SUD type, particularly for StimUD. When tuning telehealth programming, health systems should evaluate utilization trends by SUD type to ensure equitable access across the spectrum of addiction.

Highlights.

• Uptake of telemedicine varied significantly post public health emergency by type of substance use disorder.

• Patients with opioid use disorder exhibited the greatest telemedicine uptake.

• Patients with stimulant or cannabis use disorder exhibited the lowest telemedicine uptake.

• Telehealth expansion may widen utilization gaps by substance use disorder type.