Increasing Mental Health Care Access, Continuity, and Efficiency for Veterans Through Telehealth With Video Tablets

Many veterans experience geographic and health-related barriers to accessing care within the Veterans Health Administration (VHA). Of the approximate 9 million enrolled veterans receiving VHA care, nearly a third live in rural, highly rural, and insular (island) areas (1), and many others experience transportation and financial challenges that are deterrents to VHA health care utilization (2). These barriers are further compounded for the 2.1 million veterans with mental health conditions (3), some of whom avoid care because of perceived stigma or privacy concerns (4).

To improve access to mental health services, VHA has adopted telehealth interventions. Early models involved veterans traveling to their closest VHA outpatient clinic and using secure telephone lines to connect with mental health specialists at other facilities (5). Such clinic-based telehealth has helped expand access to mental health services (5); however, travel distance remains a deterrent to receiving mental health care for many veterans (6). Furthermore, clinic-based telehealth does not facilitate access for veterans who experience other barriers, such as perceived stigma or scheduling constraints because of work or caregiving responsibilities. Consequently, veterans’ access to and continuity of mental health care remains restricted, leading to missed opportunities for care and potentially increased risk of use of costly acute care services.

In 2016, VHA initiated a program to distribute video-enabled tablets to veterans with geographic, clinical, or social access barriers to in-person care so that they could receive services in their homes or other convenient locations (7). Seventy-six percent of tablet recipients had a diagnosis of a mental disorder (7), providing a unique opportunity to assess the effectiveness of this national dissemination of video-enabled telehealth tablets among veterans with mental health conditions. In this study, we evaluated whether distribution of video-enabled tablets to veterans with mental health conditions and access barriers was associated with utilization and continuity of mental health care, missed opportunities for care, and emergency department (ED) and urgent care use.

Methods

In 2016, VHA purchased 5,000 tablets with data plans to distribute to veterans with barriers to in-person access. Providers could refer patients at their discretion if the patients were enrolled in the VHA, were physically and cognitively able to operate the tablet (or had a caregiver who could assist), and had a barrier to in-person access, such as geographical barriers, health limitations, or transportation challenges. Tablets were supposed to be reserved for veterans who did not have their own device or who lacked their own source of broadband or cellular Internet service.

Once issued a device, patients could receive telehealth services during scheduled clinical video-to-home visits with their providers (7). We conducted a retrospective matched cohort study to assess how receipt of a video-enabled tablet was associated with utilization, continuity of care, and clinic efficiency outcomes. We constructed our analytic cohort from the VHA patients who received tablets and a matched cohort of VHA patients who did not receive tablets over the same time frame. The evaluation was designated as a nonresearch quality improvement initiative by the local institutional review board and by U.S. Department of Veterans Affairs (VA) research and development committees.

Tablet Cohort Selection

Individuals were selected into the analytic cohort if they received a tablet between May 1, 2016, and September 30, 2017; were diagnosed as having a mental health condition in the 12 months prior to tablet receipt [an online supplement to this article lists the ICD-9 diagnosis codes]; and had at least one psychotherapy or medication management encounter in the 12 months prior to tablet receipt. Individuals were excluded if they had prior home-based clinic video telehealth encounters, as identified through a unique VHA stop code, in the 12 months prior to tablet receipt. The tablet receipt date was assigned as the index date for individuals in the tablet cohort. These procedures resulted in a cohort of 5,074 individuals.

Control Cohort Selection

Tablet recipients were matched with individuals at the same facilities who had a diagnosis of a mental disorder and at least one mental health encounter between May 1, 2015, and September 30, 2017, and who received mental health care from clinicians not providing clinical video-to-home telehealth care. The latter criterion was applied to avoid bias from selecting control patients who did not meet the inclusion criteria for tablet receipt. We randomly selected control patients at a five-to-one ratio relative to the tablet cohort and assigned an index date identical to the matched tablet recipient’s tablet receipt date. Individuals were excluded from the analysis if they had prior clinic-to-home video telehealth encounters. This resulted in a cohort of 25,370 individuals.

Matching Characteristics

We used coarsened exact matching (CEM) to match tablet recipients and nonrecipients on baseline characteristics: demographic factors (age and sex), geography (rurality and distance to the closest source of VHA primary care), mental health utilization (number of psychotherapy visits 3 and 12 months prior to the index date), diagnoses of mental disorders (posttraumatic stress disorder [PTSD] and depression), complexity (presence of two or more diagnoses of mental disorders), whether the individual was newly diagnosed as having a mental health condition (i.e., present in the previous 2 years), and high-speed wireless (LTE) coverage in the patient’s zip code of residence.

Data Sources

Tablet distribution data for May 1, 2016, to September 30, 2017, were obtained from VA’s Denver Acquisitions and Logistics Center (DALC). These data were merged with tablet usage data from the tablet vendor (Iron Bow Technologies, Herndon, Virginia). We obtained patient sociodemographic and clinical characteristics (i.e., age, sex, race-ethnicity, marital status, rural or urban status, and clinical diagnoses) from VA’s National Patient Care Database, and we used VA’s Planning System Support Group data to document patient distance from VHA primary care. We obtained data about mental health care utilization and ED and urgent care use from VA’s Corporate Data Warehouse (CDW). Data about tablet utilization were obtained from CDW, the DALC, and the tablet vendor. Patients’ primary care provider data were obtained from the VA’s Patient-Centered Management Module. Data used to identify LTE coverage were obtained from the LTE provider (Verizon).

Outcome Measures

We assessed changes in outcome measures 6 months before and after each individual’s index date.

Outpatient mental health encounters.

We calculated the total number of psychotherapy and medication management encounters (i.e., in-person and video-enabled encounters). Psychotherapy encounters were identified by using Current Procedural Terminology (CPT) codes [see online supplement]. Medication management encounters were defined as encounters with a psychiatrist or other mental health clinician qualified to prescribe medications [see online supplement].

Continuity of care.

We assessed continuity through an adapted measure from VHA’s Strategic Analytics for Improvement and Learning (SAIL) quality monitoring system (8). The SAIL metric is a population-level continuity-of-care measure assessing whether patients with PTSD, substance use disorder, serious mental illness, or depression have received at least three psychotherapy visits in a 6-week period. For each individual, we constructed an indicator that reflected whether the patient had received at least three psychotherapy visits in a 6-week period during the 6 months before and the 6 months after the index date.

Missed opportunity rate.

To examine clinic-level efficiency, we adapted VHA’s missed opportunity rate (MOR) measure, which reflects the missed opportunity for care that occurs when a patient does not attend a scheduled appointment (9). The numerator is the sum of the missed (“no show”) mental health appointments and mental health appointments canceled after the scheduled time. The denominator is the completed mental health appointments plus the missed opportunities (9).

ED and urgent care visits.

We calculated the probability of an ED or urgent care visit and the total number of visits, which were identified by using CPT codes.

Statistical Analysis

We conducted univariate and bivariate analyses, summarizing baseline characteristics and outcomes in the 6 months before and after each individual’s index date. We used t tests and chi-square statistics to test for significant differences between the tablet recipients and the control group for continuous and categorical variables, respectively. Some factors that affected the probability of receiving a tablet, such as whether a veteran owned a tablet device, were unobservable in our administrative data. To account for pretreatment differences between the two groups (i.e., selection bias), difference-in-differences estimators were used to assess the impact of tablet receipt on recipients’ outcomes. This estimator accounted for observed and unobserved time-invariant factors unique to the tablet recipient group and the control group for which we were unable to account with our matching strategy. The base case analysis used linear probability models for binary outcomes because of ease of interpretation of the interaction terms (10) and ordinary-least-squares regressions for continuous outcomes. All analyses were adjusted for CEM weights, and standard errors were clustered by individual to account for autocorrelation.

Sensitivity Analyses

To retain a larger sample, alternate analyses used a less restrictive matching approach, matching only on age, sex, rurality, baseline psychotherapy visits, and number of mental health conditions. In alternate analyses, we also used logistic regressions for binary outcomes and generalized linear models with a log link and gamma distribution for skewed count and continuous outcomes (i.e., psychotherapy and medication management encounters, the MOR, and ED and urgent care encounters). We included additional matching characteristics (i.e., 90-day risk of hospitalization and baseline hospitalizations). Alternate analyses also subdivided individuals by mental health conditions and, because of the older age distribution of VA users, by whether or not the veteran was under age 45. Finally, we separated tablet users from nonusers and compared each group with the control cohort to determine whether outcomes were similar across the two groups.

Results

Our matched sample consisted of 728 tablet recipients and 1,020 individuals from the control cohort [a flow diagram of sample sizes for all analyses is included in the online supplement]. Prior to applying the CEM weights (Table 1), there were significant differences between the tablet and control cohorts across all characteristics. After matching, the only significant differences in the matching characteristics were race (non-Hispanic blacks made up 16.4% of the control group versus 11.9% of the tablet group, p<0.05) and number of psychotherapy visits in the 3 months before the index date (0.38 for the tablet group versus 0.22 for the control group, p<.05). After matching, tablet recipients were on average 58.8 years of age, and the age distribution was similar to that of overall VHA users (i.e., 19.1% under age 45 and 45.2% over age 65) (11). Tablet recipients were predominantly non-Hispanic white (79.3%), and 42.4% resided in rural or highly rural locations. Among tablet recipient, 5.8% resided 40 or more miles from the closest VHA primary care facility. Almost all tablet recipients (99.4%) lived in a zip code where at least 80% of residents had LTE coverage. Baseline diagnoses among tablet recipients were as follows: PTSD, 41.8%; substance use disorder, 10.7%; depression, 52.1%; and serious mental illness, 9.5%; 31.6% had two or more diagnoses of mental disorders. Tablet recipients had 8.3 mental health encounters in the 12-month period prior to the index date.

Outpatient Mental Health Encounters

Between the 6 months before the index date (baseline) and the 6 months after the index date (follow-up) (Table 2), tablet recipients experienced an increase in psychotherapy visits and medication management visits (from 0.80 to 2.42 sessions and from 1.60 to 1.80 sessions, respectively). Over the same period, the control group experienced a decline in these outcomes (from 0.62 to 0.30 psychotherapy sessions and from 1.21 to 0.36 medication management visits). In the difference-in-differences analyses, tablet recipients experienced an increase of 1.94 (p<0.001) psychotherapy visits and 1.05 (p<0.001) medication management visits during the 6 months after the index date, compared with the 6 months before the index date, relative to the control group.

Continuity of Care

At baseline, no significant differences were noted between the proportion of veterans in the tablet and control groups meeting VHA’s continuity-of-care measure (3.16% and 3.64%, respectively). In the follow-up period, tablet recipients were significantly more likely than control group participants to meet VHA’s continuity-of-care measure (20.60% versus 2.55%). In difference-in-differences analyses, tablet recipients were 18.54 percentage points (p<0.001) more likely to meet VHA’s continuity-of-care measure in the 6 months after tablet receipt, compared with the baseline period, relative to the control cohort.

Missed Opportunity Rate

At baseline, tablet recipients and control group patients had similar MORs (28.40% and 30.13%, respectively). Six months after tablet receipt, recipients’ MOR was similar to their MOR at baseline (28.04%), whereas the control cohort had a significantly higher rate (50.01%) (p<0.001). In the difference-in-differences analysis, the MOR was 20.24 percentage points lower for tablet recipients in the 6 months after tablet receipt, compared with the 6 months before tablet receipt, relative to the control cohort (p<0.001). The decrease in the MOR was driven by the significant relative increase in completed mental health encounters for tablet recipients (an increase of 5.573, p<0.001), which more than offset their slight increase in canceled mental health appointments (0.771, p<0.001).

ED and Urgent Care Visits

At baseline, tablet recipients had higher rates of ED and urgent care utilization, compared with the control group (11.67% versus 8.09%). At follow-up, both groups experienced higher rates of ED or urgent care use (tablet group, 17.17%; control group, 11.51%), but in the difference-in-differences analyses, no significant differences were found in these probabilities or the number of visits between the two groups.

Sensitivity Analyses

In analyses that relaxed the matching criteria, we retained 72% of the 5,074 tablet recipients (N=3,635) and 37% of the 25,370 patients in the control cohort (N=9,382), and results were qualitatively similar to results of our base case analysis. Tablet receipt was associated with 3.5 (p<0.001) more psychotherapy visits, 1.4 (p<0.001) more medication management visits, a 24.7 percentage point (p<0.001) increase in the likelihood of meeting VHA’s continuity-of-care measure, and a decrease in the MOR of 23.4 percentage points (p<0.001) compared with the control cohort in the 6 months after tablet receipt. In alternate models using logistic regressions for binary outcomes and generalized linear models for continuous variables, we found similar results. We did not find that including additional matching variables altered our results. We also found qualitatively similar results among different diagnostic groups and age subgroups, with the exception of serious mental illness where small sample sizes may have hampered the interpretability of the results.

Finally, when we examined the subgroup of tablet recipients who did not use their tablets, we found that, similar to the control group, these patients experienced a decline in psychotherapy encounters from baseline to follow-up (from 1.14±5.31 to 0.79±2.03 encounters) [see online supplement]. Over the same period, those who used their tablets experienced an increase in psychotherapy encounters (from 0.71±1.68 to 2.82±6.99 encounters). Both users and nonusers experienced an increase in medication management visits (from 1.72±2.19 to 1.93±2.52 and from 1.12±1.36 to 1.24±3.43, respectively). Although both users and nonusers experienced an increase in the likelihood of meeting the VHA continuity-of-care measure, the increase was smaller for nonusers (from 2.07% to 6.21%), compared with tablet users (from 3.43% to 24.19%). Similar to the control group, tablet nonusers experienced an increase in the MOR (from 25.85%±30.44% to 30.70%±34.10%), whereas tablet users experienced a decline (from 29.00%±30.10% to 27.30%±28.04). All subgroups exhibited similar trends in ED and urgent care use.

Discussion

In this evaluation, we found that veterans with mental health conditions who received video-enabled tablets experienced increased psychotherapy visits and medication management encounters and were significantly more likely to meet VHA’s continuity-of-care measure for psychotherapy, compared with similar patients who did not receive tablets. From a system efficiency perspective, tablet receipt was associated with a decline in missed opportunities for mental health care. These findings demonstrate that providing access to a multipurpose technology, such as a tablet, can facilitate and enable the delivery of mental health services.

Our findings build on previous evaluations of video and telephone home-based telehealth modalities for mental health care (12–16). This evidence, often in the context of controlled clinical trials for conditions such as PTSD, points to the noninferiority of mental health services delivered through home-based video telehealth, relative to in-person care (12, 17, 18). There has been comparatively less emphasis on real-world evidence outlining patient- and system-level effects once a health system adopts this type of intervention. This is an important oversight because several factors are likely to modify the expected impact of these types of interventions outside a controlled trial. These factors include data coverage and technical issues, physician uptake, and patient preferences and experiences (7). Indeed, our analyses suggest that among the broader cohort of tablet recipients, approximately 18% did not use their tablets at all, and 19% used them only once (7). This study addressed gaps in the literature by focusing on a nationwide roll-out, including a cohort that was heterogeneous with respect to diagnosis, and examining both behavioral and pharmacological treatment.

Our results should be interpreted with study limitations in mind. First, the results constitute the sample average treatment effect of the treated. Because a matching approach was used, these results are generalizable only to the subset of treated patients for whom we could find appropriate control patients (19). However, in sensitivity analyses that relaxed our matching criteria and retained 72% of tablet recipients, our findings were similar. Second, although the matching approach accounted for observed baseline differences and the difference-in-differences estimator accounted for time-invariant unobserved heterogeneity, our overall approach could not account for time-varying factors that remained unobserved and may have affected the outcomes of interest. Our analyses that compared tablet recipients who used their tablets with recipients who did not use their tablets, however, suggested that our findings were not entirely driven by a selection effect related to the veterans who were chosen to receive tablets. We did not observe the same trends in outcomes after tablet receipt for recipients who did not use their tablets. Finally, we note that although our outcomes included important service delivery and clinic efficiency measures, they did not include patient experience and objective health measures because we were unable to observe these in our administrative data. We were also unable to observe utilization of services that were not covered by the VHA.

Conclusions

Our findings indicate that distributing video-enabled tablets to veterans with mental health conditions appeared to improve access to and continuity of mental health services while also improving clinical efficiency. To our knowledge, this study represents the largest implementation evaluation of video telehealth technology for diverse mental health conditions. The results suggest that VHA’s telehealth tablet initiative successfully improved access to mental health services for veterans with access barriers and may serve as a model for other large integrated health care systems aiming to address access barriers by providing technology-assisted virtual treatment.