Comparing Service Use and Costs of Individual Placement and Support With Usual Vocational Services for Veterans With PTSD

Numerous studies have demonstrated the effectiveness of supported employment that utilizes the individual placement and support (IPS) model (1, 2). IPS focuses on a rapid job search in a competitive work setting and individualized follow-up support to ensure job sustainment and successful transition. An IPS specialist works with a mental health treatment team to share in treatment decisions, discuss the client’s employment status, and maximize clinical recovery plans, thereby providing the client with integrated support in vocational and clinical domains (3). According to a meta-analysis of randomized controlled trials (RCTs), people receiving IPS services were 2.4 times more likely to be employed than those receiving traditional vocational services (4, 5). Although most studies of IPS’s effectiveness have focused on populations with serious mental illness (i.e., schizophrenia, bipolar disorder, or major depression), rigorous studies have also shown the effectiveness of IPS for veterans with posttraumatic stress disorder (PTSD). A Veterans Health Administration (VHA) multisite RCT found that unemployed veterans with PTSD who received IPS were twice as likely to attain steady employment and had significantly higher cumulative earnings from competitive jobs compared with those who received usual care in the form of a transitional work (TW) program (6).

PTSD has profound effects on occupational functioning and is associated with increased unemployment rates (7–10). It may also substantially affect health care utilization. In a sample of primary care patients in an urban area, patients with a PTSD diagnosis had more hospitalizations, inpatient days, and mental health visits than patients without this diagnosis (11). In a study examining VHA health utilization data, patients with comorbid depression and PTSD incurred more frequent mental health visits, more outpatient visits overall, greater use of antidepressant medications, and higher costs for mental health care compared with patients with depression but without PTSD (12).

Employment is an essential component of recovery for people with a serious mental illness (13), and evidence suggests that it may be associated with lower use and costs of health services (14, 15). IPS has been shown to be cost-effective and to lead to less inpatient and outpatient utilization and lower costs compared with usual care vocational services (16–22). These findings may be the result of greater improvements in psychosocial functioning and life satisfaction among IPS clients (23) due to the integrated treatment approach of IPS or to the direct benefits of the work itself. Whether IPS could confer a similar benefit on the use and costs of health services incurred by veterans with PTSD remains unexplored.

In this article, we describe the utilization and costs of health services in a cohort of unemployed veterans diagnosed as having PTSD who participated in the aforementioned VHA RCT comparing IPS with TW (6). This post hoc secondary analysis tested three a priori hypotheses: compared with those enrolled in usual care (i.e., TW), veterans receiving IPS would accrue significantly fewer inpatient days, have lower total costs for high-intensity services (e.g., hospitalizations and emergency department [ED] or urgent care visits), and have higher total costs for health care services. Additionally, utilization and costs during the 18-month intervention period were compared with an 18-month postintervention period. Finally, a breakdown of the use and cost of vocational services was examined.

Methods

The parent VIP-STAR (Veterans Individual Placement and Support Toward Advancing Recovery) study was a multisite RCT that compared the effectiveness of IPS (N=271 patients) and TW (N=270 patients). The inclusion criteria were U.S. military veterans ages ≤65 years, lifetime diagnosis of PTSD, current unemployment, eligibility for usual care vocational services, interest in competitive employment, and residence in the catchment area of one of the participating study sites. Individuals were excluded if they had received a diagnosis of schizophrenia, schizoaffective, or bipolar I disorder; were in immediate need of detoxification or inpatient care; or were suicidal or homicidal. Veterans across 12 study sites who provided informed consent and met eligibility criteria were randomly assigned to either IPS or TW. Their work status, PTSD symptoms, and functioning were assessed during the 18-month study period. The TW services offered veterans temporary work assignments, predominantly in minimum-wage, entry-level jobs within the VHA facility, on the premise that these employment experiences would help prepare them for competitive work. In contrast, IPS included a rapid search for a competitive, well-matched community-based job and individualized support from an IPS specialist who collaborated with the PTSD treatment team. In keeping with the practices of IPS and usual care, IPS services continued for the entire 18-month study period, regardless of whether the veteran obtained a competitive job, whereas the TW services ended soon after the veteran obtained a competitive job. All participants given a random assignment in the parent study were included in the analyses for the present health services study. This study was approved by the Birmingham, Hines, and Tuscaloosa U.S. Department of Veterans Affairs (VA) institutional review boards.

Sources and Measures of Cost Data

We linked the multisite study data with health services utilization and cost data from the VHA Corporate Data Warehouse by using VINCI (the VHA Informatics and Computing Infrastructure). All VHA utilization data, including on general medical and psychiatric hospitalizations, domiciliary or residential care, ED or urgent care visits, medical and mental health outpatient visits, vocational rehabilitation and housing services, ancillary (e.g., chaplaincy and bereavement counseling) and administrative services, and outpatient medications, were obtained from national VHA databases, which capture utilization from electronic medical records at local VHA facilities (24, 25). The direct costs of IPS, TW, and other VHA-based health services utilization were obtained from the VA Managerial Cost Accounting (MCA) National Data Extracts (NDEs) (26). MCA draws information from the VA’s accounting and payroll system and combines these data with workload information to produce cost estimates. Each IPS or TW service that a veteran receives has a unique clinic stop code (568 for IPS and 574 for TW). Consequently, the cost data in the MCA NDEs were used to estimate the costs of the intervention and costs associated with all other VHA services utilized. Costs of prescriptions from VHA pharmacies were based on the VHA’s acquisition and dispensing costs (27).

Analytic Methods

We aggregated the cost data into 11 categories: general medical hospitalizations, psychiatric hospitalizations, domiciliary or residential stays, ED or urgent care visits, outpatient medical encounters, outpatient mental health encounters, pharmacy, vocational rehabilitation services (which include IPS and TW), housing services, ancillary services, and administrative services (e.g., prescription processing). We then added all costs to calculate total costs for the 18-month intervention and 18-month postintervention periods. Annual per-person costs for each cost category were calculated separately for IPS and TW participants and were assessed by using summary statistics. Costs were annualized and standardized to 2019 U.S. dollars. We used t tests to compare the mean annual per-person cost of IPS and TW for each cost category during the intervention and postintervention periods. We used chi-square tests to compare the number of users of each health service type between groups. We also used t tests to compare differences in the mean number of encounters between groups. All analyses were conducted with R, version 4.0.0 (28).

Results

Baseline Characteristics

The IPS and TW groups did not significantly differ in demographic or clinical characteristics at baseline (Table 1). Of all study participants (N=541), 82% (N=442) were men, 51% (N=274) were White, 42% (N=225) Black, 17% (N=90) Latinx, 13% (N=68) other races, and 82% (N=444) had more than a high school education. The mean±SD age was 42.1±10.9 years. Two-thirds of the sample (N=356) had experienced a past major depressive episode, and nearly one-third (N=170) were experiencing a major depressive episode at baseline. Mean PTSD duration was 13.3±11.4 years. Other sample characteristics not used in this analysis are reported elsewhere (6).

Utilization, Inpatient Days, and Cost of High-Intensity Services

The two groups did not significantly differ in the mean number of inpatient days for psychiatric (IPS, 0.9±4.2; TW, 0.8±2.9) or general medical hospitalizations (IPS, 0.7±2.6; TW, 0.6±2.4) during the 18-month intervention period. The two groups also did not differ in ED or urgent care visits, general medical hospitalizations, or psychiatric hospitalizations (Table 2). More than 60% of the study participants had an ED or urgent care visit, 12% had a medical hospitalization, and 11% had a psychiatric hospitalization during the intervention period (Table 2). During the postintervention period, 9% of participants had a general medical hospitalization, 4% had a psychiatric hospitalization, and 38% had ED or urgent care use (Table 3). There were no significant differences in costs for other categories of high-intensity services (ED or urgent care, general medical hospitalization, or psychiatric hospitalization) between the groups during the intervention or postintervention period (Tables 4 and 5).

Total Cost and Utilization of Health Services

As hypothesized, the overall costs of health services utilization were higher for veterans in the IPS group than for those in the TW group during the 18-month intervention period (Table 4). Unadjusted annual mean costs per person were $29,691 (95% confidence interval [CI]=$26,427–$32,955) for the IPS group and $24,781 (95% CI=$22,002–$27,560) for the TW group, a difference of $4,910 (95% CI=$622–$9,198), or a 19.8% higher annual cost for IPS participants compared with TW participants. These cost differences were driven by significantly higher utilization of vocational services in the IPS group compared with the TW group, as detailed below. Outpatient medical and nonvocational mental health utilization and costs were not significantly different between the two groups during the intervention period (Tables 2 and 4). More than 96% (N=520) of participants had at least one mental health outpatient visit, general medical outpatient visit, and prescription filled.

During the postintervention period, the IPS and TW groups did not significantly differ in outpatient service use, outpatient service costs, or overall costs (Tables 3 and 5). Unadjusted annual mean costs per person were $20,821 (95% CI=$16,143–$25,499) for the IPS group and $18,292 (95% CI=$14,826–$21,757) for the TW group, a difference of $2,530 (95% CI=−$3,295 to $8,354). Outpatient general medical costs (IPS, mean=$5,884, 95% CI=$4,823–$6,945; TW, mean=$5,397, 95% CI=$4,517–$6,277) and outpatient mental health costs (IPS, mean=$3,383, 95% CI=$2,738–$4,028; TW, mean=$3,652, 95% CI=$2,835–$4,468) were the largest cost categories.

In both study groups, overall costs significantly declined from the 18-month intervention period to the postintervention period. Among IPS participants, mean overall costs dropped by $8,870 (95% CI=$3,166–$14,574). Among TW participants, mean overall costs dropped by $6,490 (95% CI=$2,048–$10,932). The unadjusted difference in the decline between the two groups was not statistically significant (mean=$2,380, 95% CI=−$3,436 to $8,197).

Outpatient mental health costs also significantly declined from the 18-month intervention period to the postintervention period in both study groups. Mean outpatient mental health costs dropped by $3,594 (95% CI=$2,449–$4,738) in the IPS group and by $2,669 (95% CI=$1,562–$3,777) in the TW group. The unadjusted difference in the decline between the two groups was not statistically significant (mean=$924, 95% CI=−$123 to $1,972).

Utilization and Costs of Vocational Services

During the 18-month intervention period, IPS participants had a mean of 30.9 vocational service encounters, which was more than twice as many encounters as for TW participants (mean=13.6, p<0.001) (Table 2). On the basis of our a priori consensus definition of clinically meaningful engagement in services, “engaged in vocational services” was defined as having at least four IPS encounters for the IPS arm and at least four TW encounters or TW income >$0 in the TW arm, because income alone would be sufficient to demonstrate TW program engagement. A significantly higher proportion of IPS users (N=222, 82%) were engaged in vocational services, compared with TW users (N=143, 53%) (χ²=52.8, p<0.001; data not shown). During the postintervention period, although vocational service use in both groups was lower during this period than during the intervention period, IPS participants were more than twice as likely as TW participants to use vocational services (54% vs. 22%, p<0.001) and had three times as many vocational service encounters (3.1 vs. 1.1, p<0.001) (Table 3).

The only significantly different cost category between the two groups during the intervention period was vocational services (Table 4): unadjusted annual mean cost per person was $6,388 (95% CI=$5,512–$7,264) for the IPS group and $2,549 (95% CI=$2,148–$2,950) for the TW group, a difference of $3,839 (95% CI=$2,874–$4,803). Although substantially lower than during the intervention period, mean costs of postintervention vocational services (Table 5) were still significantly higher in the IPS group ($966, 95% CI=$756–$1,176) than in the TW group ($385, 95% CI=$203–$567). The IPS group had a significantly greater reduction in vocational service costs from the intervention period to the postintervention period compared with the TW group (mean difference=$3,258, 95% CI=$2,351–$4,165).

Discussion

Among veterans with a diagnosis of PTSD treated in VHA settings, we found that participants receiving IPS incurred about $4,900 more in overall annual health care costs per person during the intervention period (∼20% higher) than participants receiving usual care (TW), which declined to a nonsignificant mean difference of $2,530 annually per person in the postintervention period. This cost difference was due to the greater use of vocational services among those in the IPS group and the corresponding cost (mean difference=$3,839 per person annually). Previous research has found that use of these services leads to a significantly greater number of weeks worked among IPS recipients compared with control groups (29). The magnitude of vocational services use among the IPS participants in our study (mean=30.9 encounters per year) was consistent with the number of IPS contacts in a non-VHA research study that reported a strong association between IPS service intensity and better employment outcomes (29). During the intervention period, IPS participants were significantly more likely than TW participants to engage in vocational services (82% vs. 53%, respectively). Although vocational service encounters decreased for both groups during the postintervention period, IPS participants remained more likely than those receiving TW to use vocational services.

The 2019 annual cost of IPS vocational services estimated from VHA MCA data in our study—$6,388 per person per year—falls within the range estimated in earlier reports ($4,000–$8,000 per person per year) by investigators in non-VHA settings (30). A simple calculation of the average cost of an IPS specialist in 2013 divided by the program’s low client caseload of 14 patients per specialist yielded an estimate of $5,100 per client per year (30). It is worth noting that the cost of IPS is highest during the first 12 months of service and declines as the intensity of IPS services diminishes over time. It is important to limit IPS specialists’ caseloads to 25 clients, because higher caseloads can reduce IPS fidelity and negatively affect employment outcomes.

In a separate analysis, we also evaluated the cost-effectiveness and return on investment of these interventions by using these data (31). Similar to the findings of Dixon et al. (32), Hoffmann et al. (16), and Zheng et al. (33), we found that IPS was more cost-effective and provided a better return on investment than TW. In addition, we found that when TW income (a cost typically incurred by the VHA facilities) was included in the total health care costs, total annual costs per patient for IPS and TW were very similar ($29,828 vs. $26,772, respectively, p=0.17), thus neutralizing the cost differences that were due to IPS participants’ greater use of vocational services.

This study did not find between-group differences in the number of inpatient days or use or costs of high-intensity services. This finding differs from those of studies with people diagnosed as having serious mental illness (16–18), which reported that IPS reduces the number of psychiatric hospitalizations, inpatient days, and ED visits compared with usual care vocational services. Among VHA patients who incur the greatest costs, those with a serious mental illness have higher adjusted rates of mental health services utilization (including almost twice as many hospitalizations, inpatient days, and ED visits) and a greater proportion of costs from mental health care compared with veterans with PTSD (34). Studies of populations with serious mental illness, which have high rates of high-intensity service utilization, are more likely to report an impact of IPS on such utilization and associated costs compared with studies of populations with lower service utilization.

This study had several limitations. The intervention trial was conducted within VHA facilities; therefore, these results may not fully generalize to freestanding vocational services organizations, where integrated health care services may be lacking. Additionally, these findings may not generalize to persons who access vocational services with mental health conditions other than PTSD. These limitations were partially offset by several strengths, including the completeness of data on health services utilization and cost within the VHA administrative data resources and the ability to track service use over 36 months.

Although IPS was not associated with short-term cost savings related to the use of health services among persons with PTSD, the superior vocational outcomes confer a long-term benefit that can be transformative (6). Future studies could explore whether the higher short-term costs associated with IPS relative to usual care result in long-term cost savings or higher quality of life. Research that follows up with persons with PTSD who have received vocational services over a longer period is needed to confirm these hypotheses. It is also worth exploring whether employment itself is associated with a reduction in utilization or costs of mental health services.

Conclusions

Compared with TW, the IPS intervention in VHA settings incurred predictably greater overall health care costs of about $4,900 annually per person, largely driven by the greater use of vocational services inherent in the intensive case management that is fundamental to IPS. Otherwise, IPS did not notably differ from TW in costs associated with high-intensity or general outpatient services among veterans with PTSD. Similar to other more intensive psychosocial rehabilitative interventions, IPS appears to be both more effective and more costly than usual care. Considering previous reports showing that IPS is twice as effective as TW in improving employment outcomes for veterans with PTSD (6) and leads to significantly better functional outcomes (23), these cost differences are well justified.