Facility and state-level factors associated with telemental health (TMH) adoption among mental health facilities in the United States

Abstract

Introduction:

Telemental health (TMH) is a promising approach to increase access to mental healthcare. This study examined the TMH adoption rates and associations with facility- and state-level factors among US mental health (MH) facilities.

Methods:

This retrospective, cross-sectional study used linked data for 2016 from the National Mental Health Services Survey (N=11,833), Area Health Resources File, and national reports for broadband access and telehealth policies. The associations of facility and state-level characteristics with TMH adoption were examined with multi-level logistic regressions.

Results:

Overall, 25.9% had used TMH. Having veteran affiliation [Adjusted Odds Ratio (AOR)=18.53, 95% Confidence Interval (95%CI): 10.66–32.21] and greater Information Technology (IT) capacity [AOR(95%CI): 2.89 (2.10–3.98)] were the strongest correlates of TMH adoption. Other facility characteristics associated with higher likelihood of TMH adoption were: public ownership, high patient volumes, having comprehensive MH treatments or Quality Improvement practices, having private or non-Medicaid public payers, and treating elderly patients (AORs: 1.16–2.41). TMH adoption was less likely among facilities treating more African Americans or patients with substance abuse disorders. TMH adoption varied substantially across states, with adoption more likely in states issuing special telehealth licences and those with more rural counties.

Discussion:

One in four MH facilities adopted TMH in 2016. TMH adoption varied by multiple facility- and state-level factors. Our findings suggest that: legal/regulatory burden and lower facility IT capacity may discourage TMH adoption; significant racial disparities exist in TMH adoption; and there is a need to increase TMH use for substance abuse disorders.

Introduction

Mental health conditions (MHCs) affect 44.7 million adults in the US;¹ one in five adults experience an MHC in a given year.¹ Although MHCs can be highly disabling, they can be managed with proper treatments.² Despite policy changes to increase mental health parity and integrate mental and physical healthcare,^3,4 50% of adults with MHCs remain untreated.¹ Untreated MHCs can impose a heavy burden on individuals, their families, payers, and society.^5,6 The shortage of mental health providers remains a significant barrier to mental health treatments.^7,8 As of 2017, there were 5042 designated Mental Health Care Health Professional Shortage Areas in the US.⁹

Telemental health (TMH), the delivery of mental health services through remote technologies (e.g. video-based conferencing), has emerged as a promising solution to the shortage of providers.¹⁰ Existing studies have demonstrated the efficacy of TMH in removing geographic barriers to mental health treatment^11,12 and facilitating collaboration between primary care and mental health providers.¹³ Given that many patients do not seek mental health treatment due to the stigma associated with in-person visits to mental health facilities, TMH may also increase treatment seeking by ‘bringing’ mental health providers to primary care settings.^14–16 Furthermore, a recent review of eight systematic reviews of TMH interventions indicated that TMH was comparable to in-person care in assessing and treating various MHCs.¹⁷ As TMH has great potential for improving mental healthcare delivery, it is important to understand how TMH is implemented in real-world healthcare settings.

The first step in the widespread implementation of TMH is the adoption of TMH in mental health facilities. Adoption of TMH may be determined by multiple factors. However, rigorous studies of TMH adoption are lacking. Likewise, although several studies have investigated factors associated with the adoption of the Electronic Medical Record (EMR) system and other Health Information Technologies (HITs),^18–21 little is known about the factors associated with TMH adoption. Such knowledge is needed to inform healthcare policy and allocation of healthcare resources as well as to improve healthcare outcomes of patients with MHCs through TMH-affiliated collaborative care. Therefore, using linked population-based data this study aimed to (1) estimate the prevalence of TMH adoption and (2) examine facility and state-level factors associated with TMH adoption among mental health facilities in the US.

Method

Study design and data sources

This retrospective cross-sectional study used linked data from five sources: (1) 2016 National Mental Health Services Survey (N-MHSS); (2) the 2016–2017 Area Health Resources File (AHRF); (3) the 2016 Broadband Progress Report from the Federal Communications Commission;²² (4) the State Telemedicine Gaps Analysis: Coverage & Reimbursement from the American Telemedicine Association (ATA);²³ and (5) the State Telehealth Laws and Reimbursement Policies from the Center for Connected Health Policy (CCHP).²⁴ The N-MHSS is an annual survey conducted by the Substance Abuse and Mental Health Services Administration (SAMHSA) of all known public and private mental health treatment facilities throughout all 50 states, the District of Columbia, and other US jurisdictions. It is the only source of national and state-level data on the mental health services delivery system for publicly and privately operated mental health care specialty facilities.²⁵ N-MHSS collects comprehensive information on: facility type; primary treatment focus; mental treatment characteristics; patient demographics; and management characteristics such as computerized functionality, payer mix, licensing, certification, and accreditation.²⁵ The survey universe was identified from the database produced after fielding the 2010 and 2014 N-MSS.²⁵ The 2016 N-MHSS was conducted from March 2016 through January 2017 in 13,983 facilities across the nation with a response rate of 91%. A total of 12,172 facilities were eligible for reporting in the final dataset.²⁵

The AHRF for 2016 was used to identify state-level population characteristics and healthcare resources. The AHRF data provide county, state, and national-level data on healthcare professions, health facilities, population characteristics, and the general environment from over 50 valid data sources, including the American Medical Association, the American Community Survey, and the US Census Bureau.²⁶

Study population

The study population included 11,883 mental health facilities located in the 50 states and the District of Columbia that responded to the query about TMH use and patient characteristics in the 2016 N-MHSS. Facilities in American Samoa, Guam, Puerto Rico, and the Virgin Islands were excluded due to insufficient information on telehealth policies (N=99).

Measures

Dependent variable: TMH adoption (Yes/No).

The dependent variable was TMH adoption by mental health facilities. A binary variable (yes/no) was created based on the responses of the facility director to the question ‘which of these mental health treatment approaches are offered at this facility, at this location?’ The survey provided a checklist of items regarding mental health treatment and treatment delivery; facilities that checked the item of ‘telemedicine therapy’ were identified as facilities with TMH adoption, whereas those that did not check this item were identified as facilities without TMH adoption.

Explanatory variables.

The explanatory variables were selected based on the common constructs and factors that have been examined in previous studies on telehealth, EMR, and other HITs.^18–21 Consistent with the findings of these studies, we hypothesized that the adoption of TMH by a mental health facility may be associated with multiple state-and facility-level factors, including environmental resources, policy, market, and facility-level characteristics (e.g. ownership, patient volumes, payer mix, etc.).

Facility-level characteristics were measured in five domains: (1) facility type that included care setting, ownership, affiliation, and volume of patients; (2) comprehensiveness of mental health treatment, including treatment focus, the number of mental health treatment approaches, services, and special programmes as well as the availability of non-English services; (3) quality improvement (QI) practice; (4) Information Technology (IT) capacity; and (5) payer mix [Medicare/Medicaid/Veteran Health Administration (VHA)] and case mix, which included patient demographic composition (e.g. age, sex, race/ethnicity) and percentages of high-need patients [e.g., % veteran patients and % patients with co-occurring mental health and substance abuse (MHSA) disorders]. The IT capacity was defined as the level of technology involvement in 14 routine facility activities. The level of technology involvement was measured by the mean score of all activities rated on a 3-point scale with ‘1’ denoting no technology involvement (i.e. exclusive reliance on paper records), ‘2’ indicating partial involvement (i.e. use of both electronic and paper records), and ‘3’ representing full involvement (i.e. exclusive reliance on electronic records).

State-level characteristics included environmental resources, telehealth policy, and market characteristics. Environmental resources were measured as the percentage of counties with broadband access (25Mbps/3 Mbps service) based on the 2016 Broadband Progress Report.²² The telehealth policy environment was measured by three factors: (1) reimbursement and coverage of telehealth; (2) licensure policies that require providers to have special licences for cross-state telehealth practice; and (3) consent policies that require informed consent by patients before receiving any telehealth services. We used the ATA grading (A/B/C/F) to represent the overall reimbursement and coverage environment of each state.²³ The summary grades were developed based on 13 indicators that included telehealth parity, Medicaid policies, and innovative payment models.²³ The licensure and consent policy for each state were obtained from CCHP’s State Telehealth Laws and Reimbursement Policies Report.²⁴ We categorized the licensure policy of a state into five groups: (1) requiring individual state licence for medical practice, (2) adopting the Federation of State Medical Boards (FSMB)’s Interstate Medical Licensure Compact policy that allows for an expedited application for licences in participating states; (3) issuing special telehealth licence for the provision of telehealth services across state lines (such licences are easier to obtain relative to individual standard state licences for medical practice); (4) having exemptions for acceptable practice or states; and (5) not requiring licences or policy not defined. The consent requirement of a state was dichotomized as ‘yes (i.e. consent needed)’ and ‘no (i.e. consent not needed)’.

Market characteristics were derived from the 2016–2017 AHRF and included factors that influence the demand for TMH: the level of rurality represented by the percentage of rural counties (2013 rural–urban continuum code of 8 and 9) and shortage of primary care providers (PCPs) and mental health providers (MHPs).²⁶ We categorized provider shortage into three groups based on the percentage of counties that were designated as shortage areas in the state: (1) high (states ranked in the 4th quartile for PCPs or MHPs were considered to have a high level of shortage); (2) moderate (states ranked in the 2nd or 3rd quartiles for either PCPs or MHPs); and (3) low (states ranked in the 1^st quartile for both PCPs and MHPs).

Statistical analysis

The prevalence of TMH adoption was calculated for all mental health facilities and for mental health facilities stratified by specific facility and state-level characteristics (environmental resource, policy and, market characteristics). The adjusted association of each explanatory variable and TMH adoption was examined using multi-level logistic regressions because facilities (Level 1) were nested within states (Level 2). The level-1 variables included all the facility characteristics, and the level-2 variables included the state-level environmental resource, telehealth policy environment, and market characteristics. Wald chi-square tests were used to evaluate state variance in TMH adoption. Variance Partition Coefficients (VPCs) were calculated to quantify variance due to differences between states. Several 2-level logistic regressions were constructed to identify significant factors associated with TMH adoption. The model-building process is outlined in Table 1. Model 1 was an unconditional model where only the intercept was fitted. Model 2 was a random intercept model including all level-1 explanatory variables, Models 3–5 included the stepwise addition oflevel-2 variables. The likelihood ratio test was used to assess the significance of 2-level models versus (vs.) 1-level models. Data management and analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC) and Stata 14 (StataCorp LLC, College Station, TX).

Table 1.

Model-building process for multi-level multivariable logistic regressions on Telemental health (TMH) adoption among mental health facilities in the United States.

	Model 1	Model 2	Model 3	Model 4	Model 5
Model description	Unconditional random intercept model with state	level-1 variablesa	level-1 variablesa & level-2 variables (policy)	level-1 variablesa & level-2 variables (policy & market)b	level-1 variablesa & level-2 variables (policy & market & environmental resource)b
Level-2 residual variance (Standard error)	0.584 (0.129)	0.624 (0.141)	0.541 (0.124)	0.249 (0.062)	0.230 (0.059)
Wald Chi-squared	4.527*	4.425*	4.363*	4.016*	3.905*
VPCe	0.151	0.159	0.141	0.070	0.065
Model fit (−2LL)f	898.73***	714.00 ***	542.39***	266.24***	233.00***

Note: The study population included mental health facilities located in the 50 states and the District of Columbia that responded to the query about telemedicine use and patient characteristics in the 2016 N-MHSS. Facilities in American Samoa, Guam, Puerto Rico, and Virgin Islands were excluded due to insufficient information on telehealth policies.

^alevel-1 variables included all facility characteristics: the type of facility setting, ownership, religious affiliation; the number of mental health treatment approaches, mental health services, mental health programmes, as well as quality improvement practice, the availability of non-English service, the IT capacity, the acceptance of Medicare, Medicaid, or Veteran Health Administration patient, licensing, certification, or accreditation of the facility, average daily number of patients, the number of mental health admissions in last year, the percentage of veteran patients in last year, and daily patient composition [e.g. % co-occurring mental health and substance abuse patients, % female, % elderly (>65 years old), % Hispanic/Latino, % African American, % Minority (American Indian/Asian/Pacific Islanders)].

^blevel-2 variables included environmental-, policy-, and market characteristics measured at the state-level: %counties with broadband access, the American Telemedicine Association telehealth coverage & reimbursement grading, patient consent policy, interstate telehealth licensure policy, and % rural counties.

^cThe random slope model included all the variables in Model 3 as fixed-effects and % female patients per day as random-effects.

^dWald Chi-square values were calculated as (level-2 residual variance/standard error), compared against the critical value of Chi-square with 1 degree of freedom (3.841).

^eThe Variance Partition Coefficient (VPC) was calculated with the formula: level-2 residual variance/(level-1 residual variance+level-2 residual variance), where level-1 residual variance for logistic model 3.29.

^fModel fit was assessed by the Likelihood ratio test=Log likelihood from 2-level model - Log likelihood from single-level model. A significant Likelihood ratio test indicates the necessity of a 2-level model.

^*:p<0.05;

^***:p<0.001.

Results

Description of mental health facilities

As illustrated in Table 1, most facilities included outpatient settings (84.8%), were private non-profit (64.0%), did not have religious affiliations (93.3%), and had high patient volume (54.6%, >100 patients daily). Most focused on mental health care only (67.5%), offered comprehensive mental health treatment (56.1%–83.4%), implemented four or more QI practices (71.2%), and had moderate to high IT capabilities (78.9%). Most facilities had Medicaid (88.5%), Medicare (68.3%), and private insurers (80.0%) in their payer mix. The patient mix in most facilities included at least some elderly (64.9%) and racial/ethnic minority (i.e. African American, Hispanic/Latino, Asian, American Indian, Pacific Islanders) patients (95.2%); high-need patients such as veterans and patients with MHSA comprised less than 50% of the patient population in most facilities (54.3%–77.5%). On the state level, most facilities were in states that had grade A/B telehealth reimbursement and coverage (82.4%), required patient consent for telehealth services (68.4%), and required some forms of licences for cross-state telehealth practice (62.4%). More than one-third (38.2%) of facilities were in states with a high shortage of PCPs and MHPs; almost two-fifths were in states with a high percentage (i.e. ≥20%) of rural counties. Approximately half of all facilities were in states with poor broadband access.

Description of TMH adoption among mental health facilities

Overall, 25.9% of mental health facilities reported adopting TMH as one of their treatment approaches. The adoption of TMH varied widely by facility-level characteristics, ranging from 10.7% in residential facilities to 92.6% in Veteran Administration Health Centers (VAHCs) (Table 2). TMH adoption was the highest in Veteran-majority facilities (87.7%) and facilities with high IT capacities (55.0%). At the state level, adoption varied from as low as 3.4% in Connecticut to as high as 68.8% in North Dakota. The adoption prevalence varied by environmental resource, policy, and market characteristics. TMH adoption was highest in states with high broadband access (42.1%), and lowest (9.7%) in states with low levels of provider shortage (Table 2).

Table 2.

Descriptions of mental health facilities and Telemental health (TMH) adoption among mental health facilities in the United States.

	All facilities		Facilities with TMH
Total	N 11,883	Column % 100.0	N 3082	% 25.9
Level-1: Facility characteristics
Type of facility
Care setting
Psychiatric inpatient	1804	15.2	356	19.7
Residential	1633	13.7	174	10.7
VAMC	377	3.2	349	92.6
CMHC	2587	21.8	1079	41.7
Outpatient	5088	42.8	1030	20.2
Multi-setting	394	3.3	94	23.9
Ownership
Private-for-profit	2069	17.4	380	18.4
Private-non-profit	7,601	64.0	1,714	22.5
Public	2213	18.6	988	44.6
Religious affiliation
Yes	763	6.4	105	13.8
No	11,091	93.3	2965	26.7
Not reported	29	0.2	12	41.4
Annual mental health admissions
0–100	3323	28.0	549	16.5
100–250	1472	12.4	329	22.4
250–500	1474	12.4	395	26.8
500–1000	3684	31.0	1192	32.4
>1000	1930	16.2	617	32.0
Daily volume of patientsa
1–100	5398	45.4	938	17.4
100–250	4340	36.5	1320	30.4
250–500	1311	11.0	488	37.2
500–1000	388	3.3	123	31.7
>1000	446	3.8	213	47.8
Comprehensiveness of mental health treatments
Treatment focus
Mental health only	8019	67.5	1662	20.7
Mental health & Substance Abuse	3499	29.4	1217	34.8
General health	365	3.1	203	55.6
# of mental health treatment approachesb
0–3	1967	16.6	237	12.0
4–5	9916	83.4	2845	28.7
# of mental health servicesc,l
0–4	6668	56.1	1128	16.9
≥5	5212	43.9	1954	37.5
# of special mental care programsd
0–3	10,541	88.7	2538	24.1
4–5	1342	11.3	544	40.5
Non-English services
Yes	7197	60.6	2064	28.7
No	4685	39.4	1018	21.7
Quality improvement practicee
0–3	3377	28.4	696	20.6
4	3708	31.2	915	24.7
5	4757	40.0	1462	30.7
IT capabilityyf,l
Low	2509	21.1	310	12.4
Moderate	9006	75.8	2577	28.6
High	342	2.9	188	55.0
Payer mix
Medicaid
Yes	10,515	88.5	2635	25.1
No	1155	9.7	334	28.9
Not reported	213	1.8	113	53.1
Medicare
Yes	8120	68.3	2339	28.8
No	3454	29.1	630	18.2
Not reported	309	2.6	113	36.6
Veteran Health Administration
Yes	2613	22.0	1131	43.3
No	6770	57.0	1350	19.9
Not reported	2500	21.0	601	24.0
Any private insurance
Yes	9501	80.0	2694	28.4
No	2158	18.2	306	14.2
Not reported	224	1.9	82	36.6
Case mix
Any elderlya
Yes	7714	64.9	2372	30.7
No	4169	35.1	710	17.0
% Femalesa
0–10%	567	4.8	199	35.1
11–50%	5872	49.4	1399	23.8
>50%	5444	45.8	1484	27.3
% African Americana
0–20%	9068	76.3	2520	27.8
>20%	2815	23.7	562	20.0
Any Hispanic/Latino patientsa
Yes	5993	50.4	1617	27.0
No	5890	49.6	1465	24.9
Any minority patientsa
Yes	11,308	95.2	2895	25.6
No	575	4.8	187	32.5
% Mental health & substance abuse patientsa
0–20%	4327	36.4	933	21.6
20–40%	2127	17.9	611	28.7
40–60%	1256	10.6	344	27.4
60–80%	1536	12.9	354	23.0
Not reported	2637	22.2	840	31.9
Annual % of veteran patients
0–50%	9209	77.5	2,030	22.0
>50%	302	2.5	265	87.7
Not reported	2,372	20.0	787	33.2
Level-2 variables: Policy characteristicsg
ATA telehealth coverage & reimbursement gradingh
A	1362	11.5	427	31.4
B	8425	70.9	2098	24.9
C	2096	17.6	557	26.6
Patient consenti
Yes	8127	68.4	2058	25.3
No	3756	31.6	1024	27.3
Licensurei
Individual state	1398	11.8	447	32.0
FSMB Compact	2670	22.5	707	26.5
Special telehealth	1450	12.2	485	33.4
With exemptions	1900	16.0	641	33.7
Not required/defined	4465	37.6	802	18.0
Level-2 variables: Market characteristicsh
Provider shortagej
Low	3779	31.8	298	9.7
Moderate	3566	30.0	1170	38.0
High	4538	38.2	1614	52.4
% rural countiesj
0–10%	6114	51.5	1097	17.9
10–20%	3510	29.5	1172	33.4
≥20%	2259	19.0	813	36.0
Level-2 variables: Environmental resourcesh
% counties with broadband accessk
0–10%	6112	51.4	1082	17.7
10–20%	4391	37.0	1419	32.3
≥20%	1380	11.6	581	42.1

Note: The study population included mental health facilities located in the 50 states and the District of Columbia that responded to the query about telemedicine use and patient characteristics in the 2016 N-MHSS. Facilities in American Samoa, Guam, Puerto Rico, and Virgin Island were excluded due to insufficient information on telehealth policies.

^aN-MHSS measured total number of patients and patient mix of a facility on a specific day (April 29, 2016).

^bMental health treatment approaches evaluated in N-MHSS included individual psychotherapy, couples/family/group therapy, cognitive behavioural/dialectical behavioural therapy/behaviour modification, integrated dual disorders treatment, trauma therapy, activity therapy, electroconvulsive therapy, and psychotropic medication.

^cMental health services evaluated in N-MHSS included intensive case/case management/chronic disease/illness management, integrated primary care services, any counselling, family psychoeducation, education services, psychosocial rehabilitation services, psychiatric emergency walk-in services, suicide prevention services, peer support services, screening for tobacco use, smoking cessation services.

^dSpecial mental health programmes included mental health programmes that is dedicated or designed exclusively for serious mental illnesses, co-occurring mental and substance abuse disorders, post-traumatic stress disorder, senior/elderly patients or veterans.

^eQuality improvement practice evaluated in N-MHSS included continuing education, case review, outcome follow-up after discharge, utilization review, and satisfaction surveys.

^fThe IT capacity was defined as the level of technology involvement in 14 routine facility activities, including mental health intake, scheduling appointments, assessment/evaluation, treatment planning, client progress monitoring, discharge, referral, issue/receive lab results, prescribing/dispensing medications, checking medication interactions, health records, provider collaboration, billing, and satisfaction surveys. The level of technology involvement was measured by the mean score of all activities rated on a 3-point scale, where ‘1’ denotes no technology involvement (i.e. rely on paper), ‘2’ indicates partial involvement (i.e. both electronic and paper), and ‘3’ represents fully involvement (i.e. only electronic).

^genvironmental-, policy-, and market characteristics were measured for the state where the facility was located.

^hthe American Telemedicine Association (ATA) has given each state a grade (A, B, C, or F) for the coverage and reimbursement policies for telehealth, basing on health plan parity and Medicaid conditions of payment.

ⁱthe patient consent policy that requires informed consent before telehealth services and the licensure policy that requires licence for interstate telehealth practice were obtained from the Centers for Connected Health Policy’s (CCHP) annual State Telehealth Laws and Reimbursement Policies Report.

^jThe percentage counties with primary care physician (PCP) shortage, mental health professional (MHP) shortage, and the percentage of rural counties were obtained from the Area Health Resources File (AHRF).

^kBroadband access was defined as having access to 25Mbps/3 Mbps service; % counties with broadband access in the locating state were derived from the 2016 Broadband Progress Report.

^lMissing data category is not presented due to small sample size (n<10).

Facility-level factors associated with TMH adoption

Results from the fully adjusted analyses (Model 5) indicated that facility characteristics in all five domains (facility type, comprehensiveness of mental health treatment, QI practice, IT capability, and payer and case mix) were significantly associated with TMH adoption (Table 3). It is noteworthy that VAHCs were 18 times more likely to adopt TMH as psychiatric inpatient facilities [Adjusted Odds Ratio (AOR)=18.53, 95% Confidence Interval (95%CI): 10.66–32.21]. Facilities with high IT capacities were almost three times more likely than those with low capacities to have TMH (AOR=2.89, 95%CI: 2.10–3.98). Other facility characteristics associated with higher likelihood of TMH adoption included: inclusion of outpatient settings; high patient volume; public ownership; availability of comprehensive mental health treatment; use of QI practices; private or non-Medicaid public payers; and treatment of elderly patients or higher percentages of veterans (AORs: 1.16–2.41). On the other hand, facilities that were affiliated with religious organizations [AOR (95%CI): 0.72 (0.57–0.91)] or treated higher percentages of African American patients [AOR (95%CI): 0.74 (0.65–0.85)] or patients with MHSA (AORs: 0.81–0.83) were significantly less likely to have adopted TMH (Table 3).

Table 3.

Adjusted odds ratios (AOR) and 95% confidence intervals (95% CI) of facility- and state-level characteristics on Telemental health (TMH) adoption among mental health facilities in the United States, results from adjusted multi-level logistic regressions.

Variables	Model 2 AOR [95% CI]	Model 4 AOR [95% CI]	Model 5 AOR [95% CI]
Level-1: Facility characteristics
Type of facility
Care setting
Residential	1.27 [0.98, 1.63]	1.27 [0.99, 1.64]	1.27 [0.99, 1.64]
VAMC	18.30 [10.52, 31.82]***	18.38 [10.57, 31.97]***	18.53 [10.66, 32.21]***
CMHC	2.46 [2.03, 2.98]***	2.46 [2.03, 2.98]***	2.41 [1.99, 2.93]***
Outpatient	1.57 [1.30, 1.89]***	1.57 [1.31, 1.89]***	1.56 [1.30, 1.88]***
Multi-setting	1.94 [1.43, 2.62]***	1.94 [1.43, 2.63]***	1.94 [1.43, 2.63]***
Psychiatric inpatient	Reference
Ownership
Private-non-profit	1.10 [0.94, 1.29]	1.11 [0.95, 1.29]	1.10 [0.95, 1.29]
Public	1.67 [1.39, 2.01]***	1.67 [1.39, 2.01]***	1.66 [1.38, 2.00]***
Private-for-profit	Reference
Religious affiliation
Yes	0.72 [0.57, 0.91]**	0.72 [0.57, 0.91]**	0.72 [0.57, 0.91]**
No	Reference
Annual mental health admissions
100–250	1.10 [0.92, 1.32]	1.10 [0.92, 1.32]	1.10 [0.92, 1.32]
250–500	1.27 [1.05, 1.52]*	1.27 [1.05, 1.52]*	1.26 [1.05, 1.52]*
500–1000	1.33 [1.12, 1.59]**	1.33 [1.12, 1.58]**	1.33 [1.12, 1.59]**
>1000	1.39 [1.15, 1.67]**	1.39 [1.15, 1.67]**	1.39 [1.15, 1.68]**
0–100	Reference
Total patientsa
100–250	1.11 [0.97, 1.27]	1.11 [0.97, 1.27]	1.12 [0.98, 1.28]
250–500	1.10 [0.91, 1.32]	1.09 [0.91, 1.32]	1.11 [0.92, 1.33]
500–1000	0.86 [0.64, 1.15]	0.86 [0.64, 1.14]	0.87 [0.65, 1.16]
>1000	1.41 [1.08, 1.83]*	1.41 [1.08, 1.83]*	1.42 [1.09, 1.84]*
1–100	Reference
Comprehensiveness of mental health treatments
Treatment focus
Mental health & Substance Abuse	1.24 [1.10, 1.39]***	1.24 [1.10, 1.39]***	1.23 [1.10, 1.38]***
General services	1.38 [1.00, 1.90]*	1.38 [1.01, 1.90]*	1.38 [1.01, 1.90]*
Mental health only	Reference
# of mental health treatment approachesb
4–5	2.00 [1.69, 2.37]***	2.00 [1.69, 2.37]***	2.00 [1.69, 2.36]***
0–3	Reference
# of mental health servicesc
≥5	1.80 [1.61, 2.00]***	1.80 [1.61, 2.00]***	1.79 [1.61, 2.00]***
0–4	Reference
# of special mental care programmesd
4–5	1.15 [0.99, 1.34]	1.15 [0.99, 1.34]	1.15 [0.99, 1.34]
0–3	Reference
Non-English services
Yes	1.20 [1.07, 1.35]**	1.20 [1.08, 1.35]**	1.21 [1.08, 1.36]**
No	Reference
Quality improvement practice
# of Quality Improvement practicee
4	1.13 [0.99, 1.29]	1.13 [0.99, 1.29]	1.13 [0.99, 1.29]
5	1.35 [1.19, 1.53]***	1.35 [1.19, 1.53]***	1.35 [1.18, 1.53]***
0–3	Reference
IT capabilityf
Moderate	1.91 [1.65, 2.21]***	1.91 [1.65, 2.22]***	1.90 [1.64, 2.20]***
High	2.93 [2.13, 4.03]***	2.93 [2.13, 4.04]***	2.89 [2.10, 3.98]***
Low	Reference
Payer mix
Medicaid
Yes	1.13 [0.91, 1.41]	1.13 [0.91, 1.41]	1.12 [0.90, 1.40]
No	Reference
Medicare
Yes	1.25 [1.08, 1.45]**	1.25 [1.08, 1.45]**	1.26 [1.09, 1.47]**
No	Reference
Veteran Health Administration
Yes	1.37 [1.21, 1.56]***	1.37 [1.21, 1.56]***	1.36 [1.20, 1.55]***
No	Reference
Private
Yes	1.51 [1.27, 1.79]***	1.50 [1.27, 1.78]***	1.47 [1.24, 1.74]***
No	Reference
Case mix
% of female patientsa
11–50%	0.77 [0.57, 1.05]	0.77 [0.57, 1.05]	0.78 [0.57, 1.06]
>50%	0.83 [0.61, 1.13]	0.83 [0.61, 1.13]	0.83 [0.61, 1.14]
0–10%	Reference
Having elderly patientsa
Yes	1.17 [1.03, 1.32]*	1.16 [1.03, 1.32]*	1.16 [1.02, 1.31]*
No	Reference
% of African American patientsa
>20%	0.74 [0.65, 0.85]***	0.74 [0.65, 0.85]***	0.74 [0.65, 0.85]***
0–20%	Reference
Having Hispanic/Latino patientsa
Yes	1.09 [0.98, 1.22]	1.10 [0.98, 1.22]	1.10 [0.98, 1.23]
No	Reference
Any minority patientsa
Yes	1.25 [1.00, 1.57]	1.24 [0.99, 1.56]	1.24 [0.99, 1.56]
No	Reference
% of MHSA patientsa
20–40%	1.03 [0.89, 1.19]	1.03 [0.89, 1.19]	1.03 [0.89, 1.19]
40–60%	0.95 [0.80, 1.14]	0.95 [0.80, 1.14]	0.96 [0.81, 1.14]
>60%	0.83 [0.70, 0.98]*	0.83 [0.70, 0.98]*	0.83 [0.70, 0.98]*
Not reported	0.81 [0.67, 0.98]*	0.81 [0.67, 0.98]*	0.81 [0.67, 0.98]*
0–20%	Reference
Annual % of veteran patients
>50%	2.74 [1.65, 4.56]***	2.75 [1.65, 4.57]***	2.75 [1.65, 4.57]***
Not reported	1.24 [1.02, 1.50]*	1.24 [1.02, 1.50]*	1.24 [1.02, 1.50]*
0–50%	Reference
Level-2: Policy characteristicsg
ATA telehealth coverage & reimbursement gradingh
A	–	1.02 [0.47, 2.21]	0.57 [0.31, 1.05]
B	–	0.91 [0.49, 1.68]	0.74 [0.46, 1.19]
C	Reference
Patient consenti
Yes	–	0.82 [0.52, 1.29]	0.84 [0.61, 1.17]
No	Reference
Licensurei
Individual state	–	1.70 [0.82, 3.56]	1.54 [0.88, 2.72]
FSMB Compact	–	1.63 [0.93, 2.86]	1.05 [0.69, 1.59]
Special telehealth	–	2.28 [1.02, 5.06]*	2.24 [1.16, 4.32]*
With exemptions	–	1.37 [0.73, 2.58]	1.09 [0.67, 1.78]
Not required/defined	Reference
Level-2: Market characteristicsg
Provider shortagej
High	–	–	0.90 [0.53, 1.54]
Moderate	–	–	1.41 [0.83, 2.39]
Low	Reference
% rural countiesj
10–20%	–	–	2.09 [1.33, 3.29]**
≥20%	–	–	2.36 [1.34, 4.18]**
0–10%	Reference
Level-2: Environmental resourcesg
% counties with broadband accessk
10–20%	–	–	1.02 [0.65, 1.62]
≥20%	–	–	1.63 [0.92, 2.87]
0–10%	Reference

Note: The study population included mental health facilities located in the 50 states and the District of Columbia that responded to the query about telemedicine use and patient characteristics in the 2016 N-MHSS. Facilities in American Samoa, Guam, Puerto Rico, and Virgin Islands were excluded due to insufficient information on telehealth policies.

^aN-MHSS measured total number of patients and patient mix of a facility on a specific day (April 29, 2016).

^bMental health treatment approaches evaluated in N-MHSS included individual psychotherapy, couples/family/group therapy, cognitive behavioural/dialectical behavioural therapy/behaviour modification, integrated dual disorders treatment, trauma therapy, activity therapy, electroconvulsive therapy, and psychotropic medication.

^cMental health services evaluated in N-MHSS included intensive case/case management/chronic disease/illness management, integrated primary care services, any counselling, family psychoeducation, education services, psychosocial rehabilitation services, psychiatric emergency walk-in services, suicide prevention services, peer support services, screening for tobacco use, smoking cessation services.

^dSpecial mental health programmes included mental health programmes that is dedicated or designed exclusively for serious mental illnesses, co-occurring mental and substance abuse disorders, post-traumatic stress disorder, senior/elderly patients or veterans.

^eQuality improvement practice evaluated in N-MHSS included continuing education, case review, outcome follow-up after discharge, utilization review, and satisfaction surveys.

^fThe IT capacity was defined as the level of technology involvement in 14 routine facility activities, including mental health intake, scheduling appointments, assessment/evaluation, treatment planning, client progress monitoring, discharge, referral, issue/receive lab results, prescribing/dispensing medications, checking medication interactions, health records, provider collaboration, billing, and satisfaction surveys. The level of technology involvement was measured by the mean score of all activities rated on a 3-point scale, where ‘1’ denotes no technology involvement (i.e. rely on paper), ‘2’ indicates partial involvement (i.e. both electronic and paper), and ‘3’ represents fully involvement (i.e. only electronic).

^genvironmental-, policy-, and market characteristics were measured for the state where the facility was located.

^hthe American Telemedicine Association (ATA) has given each state a grade (A, B, C, or F) for the coverage and reimbursement policies for TH, basing on health plan parity and Medicaid conditions of payment.

ⁱthe patient consent policy that requires informed consent before telehealth services and the licensure policy that requires licence for interstate telehealth practice were obtained from the Centers for Connected Health Policy’s (CCHP) annual State Telehealth Laws and Reimbursement Policies Report.

^jThe percentage counties with primary care physician (PCP) shortage, mental health professional (MHP) shortage, and the percentage of rural counties were obtained from the Area Health Resources File (AHRF).

^kBroadband access was defined as having access to 25Mbps/3 Mbps service; % counties with broadband access in the locating state were derived from the 2016 Broadband Progress Report.

^*p<0.05;

^**p<0.01;

^***p<0.001.

State-level factors associated with TMH adoption

As presented in Table 1, results of Wald Chi-square tests indicated significant variation in TMH adoption across states. Adding environmental, policy, and market characteristics (level-2 variables) into the model (Model 5) decreased the proportion of variance in TMH adoption due to state variance from 15.9% to 6.5% (Table 1). Market characteristics contributed most to the variance reduction (14.1% in Model 3 to 7.0% in Model 4). In the fully adjusted model (Table 3), licensure policy and rurality were significantly associated with TMH adoption. Facilities located in states having a special application process for licences used in cross-state telehealth practice were more than twice as likely to have adopted TMH as compared with those located in states not requiring or defining licensure [AOR=2.24, 95%CI: 1.16–4.32]. The odds of TMH adoption across states increased with increasing percentage of rural counties (AORs: 2.09–2.36, p < 0.01). Other policy factors, provider shortages, and broadband access were not associated with TMH adoption.

Discussion

In this large population-based study of US mental health facilities, only 26% reported using TMH as a treatment approach in 2016. This adoption was much lower than the prevalence reported for general telehealth (42%) in acute care hospitals (American Hospital Association Annual Survey of Hospitals²⁰), but parallel to findings from the American Medical Association’s 2016 Physician Practice Benchmark Survey indicating that 27.8% of psychiatrists had used telemedicine to interact with patients.²⁷ One reason for the discrepancy in adoption prevalence may be the high need in acute care hospitals for telehealth services such as teleradiology, teledermatology, teleophthalmology, and telecardiology.^28,29

Our results suggested that the adoption of TMH by a mental health facility could be influenced by multiple factors, including the type of facility, mental health treatment capacities, IT capability, QI practice, and payer and case mix of the facility. For example, TMH adoption was significantly higher in Veteran Administration Medical Centers (VAMCs) and facilities that had higher proportions of veteran patients. As the nation’s largest health care system, the VA has been the leader in telehealth in the US since 2003.³⁰ In addition to the VA’s continuous commitment to telehealth and mental health, the integrated nature of the VA system may also contribute to the adoption of TMH.^20,31 Mental health facilities that included outpatient settings were also more likely to have TMH, likely reflecting the fact that most mental health services adapted to TMH are provided in outpatient settings.^32,33 Consistent with previous reports,^18–20 public ownership was also strongly associated with TMH adoption. In addition, public mental facilities were more likely to adopt TMH as compared with private for-profit facilities, perhaps due to differences in fiscal incentives.^34,35 Given that most mental health facilities are private non-profit, efforts to identify and address barriers to TMH adoption among non-profit facilities are clearly warranted.

In addition to facility type, higher IT capacity was associated with higher TMH adoption. This association may in part reflect differences in infrastructure. For example, facilities with higher IT capacities are likely to be those with a patient referral, health records, and/or provider collaboration system (i.e. part of the definition of IT capacity in our study). These systems are often required for effective and sustainable TMH implementation.³⁶ Given that sufficient IT capacity serves a precondition for TMH,³⁶ facilities that recognized a need for TMH might be more willing to invest in IT infrastructure as compared with facilities that did not recognize such a need. TMH adoption was also more likely among facilities with comprehensive mental health treatments. It is plausible that facilities that adopted TMH can get access to remote specialists, thereby offering a variety of mental health services. As this was a cross-sectional study, we are unable to parse out whether TMH led to comprehensive services.

Facility payer type and patient mix were also significantly associated with TMH adoption. Facilities treating patients insured by Medicare, VHA, and private payers were more likely to adopt TMH. Facilities treating elderly patients were more likely to have TMH as compared with those who did not. The often complex health care needs of elderly patients with MHCs may contribute to TMH adoption, as TMH may facilitate the integrated management of mental and physical chronic conditions.³⁷ While the current study did not explore the reasons for this finding, we speculate that facilities offering TMH may also have comprehensive geriatric services attracting elderly patients.

TMH adoption was significantly less likely among facilities serving higher percentages of African Americans, possibly due to lack of trust in new health care innovations in this population.³⁸ George and colleague reported that African Americans expressed more concerns about confidentiality, privacy, and care quality of telemedicine than did Hispanic controls.³⁹ As African Americans tend to experience more severe MHCs due to access barriers,⁴⁰ demonstration and educational programmes are needed to increase the ‘buy-in’ of TMH in African American patients. As this was a cross-sectional study, it is also plausible that mental health facilities serving more African American patients chose to not to offer TMH services for many other reasons (e.g. lack of funding, lack of training). Future research efforts are clearly needed to elucidate the factors underlying the racial disparities observed in this study to determine causal relationships.

Facilities that primarily served patients with co-occurring MHSA disorders were significantly less likely to adopt TMH relative to those serving fewer MHSA patients. TMH has been considered a promising tool for combating the US substance abuse crisis by improving access to substance abuse experts and mental health specialists.⁴¹ However, TMH benefits may be limited due to current legal restriction on prescribing Medication-Assisted Treatments via TMH.⁴² Policy initiatives such as the Improving Access to Remote Behavioral Health Treatment Act of 2018, which aims to boost availability of Medication-Assisted Treatment prescribers and ease legal restrictions, may help fully realize the potential of TMH in behavioural health.⁴²

TMH adoption varied substantially across states, and the variation was primarily driven by differences in telehealth policies and level of rurality. TMH adoption was more likely in states with a higher percentage of rural counties, a finding consistent with those of previous research on telehealth adoption among acute hospitals.²⁰ The higher telehealth/TMH adoption in rural areas might be a result of financial incentives provided by federal and foundation funding that focuses on rural health.^20,43,44 Mental health facilities located in states that had a special application process for licensure for providing interstate telehealth services were more likely to have TMH than were those in states without specific licensure requirements. As of 2017, medical boards in nine states had established processes to issue telehealth-specific licences to allow out-of-state providers to provide telehealth services.²⁴ Twenty-two states had adopted the Federation of State Medical Board’s Interstate Medical Licensure Compact to allow an expedited process of applying for licences in other states.²⁴ Although adopting the Compact policy was not associated with TMH adoption in our study, it could nonetheless help increase future implementation of TMH adoption, as more states adopt this policy and facilities acclimatize themselves to this new policy. In contrast to the observed association of TMH adoption to state licensure policies, telehealth reimbursement and consent policy were unrelated to TMH adoption, likely reflecting the improvement in overall telehealth coverage and reimbursement in recent years.²³

To our knowledge, this is the first study to investigate the prevalence and multi-level correlates of TMH adoption in US mental health facilities. Additional strengths include the large, representative sample and the use of comprehensive linked data from all 50 states and the District of Columbia. Our findings not only provide robust estimates of overall TMH adoption prevalence in US mental health facilities but also offer insight regarding the factors that may influence facility-level decisions to invest in TMH, as well as potential strategies for boosting adoption. However, several limitations should be noted. First, we estimated TMH adoption based on reported telemedicine use in mental health facilities, raising the possibility of misclassification bias. Thus, we may have over-estimated TMH adoption and use. Second, our data did not allow determination of how telemedicine was used or when the uptake of TMH occurred. Third, limited data availability only allowed us to measure environmental resources and market characteristics at the state level; thus, we were unable to capture within-state variation in these factors. In addition, we lacked information on the financial profile of mental health facilities or potential market competition, which could also affect decisions to invest in TMH. Furthermore, N-MHSS did not adjust for non-response, potentially introducing selection bias; the results of our study may thus not be generalizable to facilities (9.9%) that did not participate in the 2016 N-MHSS. Finally, as the study relied on cross-sectional data, the temporal relationships between facility-level factors and TMH adoption cannot be ascertained. All observed relationships should thus be considered associative rather than causal.

In summary, although TMH appears to offer a promising tool for facilities for expanding access to care and serving high-need patients, only one in four mental health facilities had adopted TMH as of 2016. TMH adoption was associated with multiple facility and state-level factors. Collectively, our findings suggest that TMH adoption may be discouraged by the legal or regulatory burden on providers seeking licensure for cross-state practice and by low facility IT capacity; TMH adoption of mental health facilities with more African American patients warrants further attention; and policy initiatives that facilitate prescription of substance use disorder medications via TMH are needed to increase access to care for patients with substance use disorders.