Differences Between VHA-Delivered and VHA-Purchased Behavioral Health Care in Service and Patient Characteristics
Abstract
Objective:
Federal legislation has expanded Veterans Health Administration (VHA) enrollees’ access to VHA-purchased “community care.” This study examined differences in the amount and type of behavioral health care delivered in VHA and purchased in the community, along with patient characteristics and area supply and demand factors.
Methods:
This retrospective cross-sectional study examined data for 204,094 VHA enrollees with 448,648 inpatient behavioral health stays and 3,467,010 enrollees with 55,043,607 outpatient behavioral health visits from fiscal years 2016 to 2019. Standardized mean differences (SMDs) were calculated for patient and provider characteristics at the outpatient-visit level for VHA and community care. Linear probability models assessed the association between severity of behavioral health condition and site of care.
Results:
Twenty percent of inpatient stays were purchased through community care, with severe behavioral health conditions more likely to be treated in VHA inpatient care. In the outpatient setting, community care accounted for 3% of behavioral health care visits, with increasing use over time. For outpatient care, veterans receiving community care were more likely than those receiving VHA care to see clinicians with fewer years of training (SMD=1.06).
Conclusions:
With a large portion of inpatient behavioral health care occurring in the community and increased use of outpatient behavioral health care with less highly trained community providers, coordination between VHA and the community is essential to provide appropriate inpatient follow-up care and address outpatient needs. This is especially critical given VHA’s expertise in providing behavioral health care to veterans and its legislative responsibility to ensure integrated care.
HIGHLIGHTS
In the VHA, more than a quarter of veterans receiving inpatient behavioral health care used some VHA-purchased community care, and severe behavioral health conditions were more frequently treated in VHA than in community inpatient settings.
Only 4% of VHA-enrolled veterans receiving outpatient behavioral health care used community care, and these veterans saw less highly trained providers, compared with veterans receiving care in VHA, and were more likely to receive individual therapy than group therapy.
Coordination between VHA and community care providers is needed to ensure continuity and quality of outpatient behavioral health care, as well as follow-up care after inpatient behavioral health stays.
In response to access concerns, the Veterans Health Administration (VHA) now purchases a considerable amount of care in the community (“community care”). Since implementation of the Veterans Choice Act of 2014 (Public Law 113-146) and MISSION Act of 2018 (Public Law 115-182), over 31% of 8.92 million VHA enrollees have received community care referrals (1). Implications of this transition are unclear for patients with behavioral health needs, who represent over 25% of veterans receiving VHA primary care (2). Historically these veterans have relied on VHA for behavioral health care, including treatment for psychiatric and substance use disorders (3–7). Not only are VHA providers trained in evidence-based therapies for behavioral health, but they are also trained in military cultural competence, which is often lacking among community providers (8–11).
Demand for behavioral health care is outpacing supply. For VHA to strategically manage its make-versus-buy decisions (i.e., provide more services in house or purchase community care) in the future, it is important to understand current behavioral health utilization patterns in VHA and community care and the relative strengths of each setting in meeting veterans’ needs. Given VHA’s expertise in behavioral health care, we hypothesized in this study that only a small fraction of veterans would utilize outpatient behavioral health services in the community but that a larger portion would utilize inpatient services, given the need to address acute behavioral health concerns quickly and close to home. We also hypothesized that treatment for more severe (high-risk, high-cost) behavioral health conditions (12) would more likely occur in VHA than in the community. Additionally, we anticipated that behavioral health care providers seen in VHA outpatient visits would be more highly trained than those seen in community outpatient visits, given VHA’s long-standing behavioral health experience and expertise. Understanding the types and characteristics of behavioral health care provided by VHA and community care offers an opportunity to help veterans receive appropriate, high-quality behavioral health care.
Methods
Study Design
We conducted a retrospective cross-sectional study from federal fiscal years (FYs) 2016 to 2019, examining VHA-delivered versus VHA-purchased behavioral health care. This study was administratively reviewed by the University of Utah Institutional Review Board and the Department of Veterans Affairs (VA) Salt Lake City Health Care System and deemed exempt from human subjects review as a quality improvement initiative. This study adhered to Strengthening the Report of Observational Studies in Epidemiology guidelines for cross-sectional studies.
Data Sources
We obtained administrative data for VHA-delivered care from the VA Corporate Data Warehouse (CDW) and claims data for community care from CDW’s Program Integrity Tool, Fee, and Fee Basis Claims System. We created separate inpatient and outpatient data sets because they are distinct care types and because inpatient stays occur infrequently, compared with outpatient visits. Drive distance to VHA primary care (for outpatient analyses) and VHA secondary care (for inpatient analyses) came from the VHA Planning Systems Support Group. We included county characteristics from the Health Resources and Services Administration’s Area Health Resources File.
Measures
Outcome variables.
The main outcome was whether behavioral health care was provided in VHA or in the community.
Inpatient behavioral health stays.
VHA’s Office of Mental Health and Suicide Prevention provided 616 ICD-10 codes grouped by behavioral health condition type (a list is included in an online supplement to this article): serious mental illness, substance use disorders, posttraumatic stress disorder (PTSD), personality disorders, mood disorders, anxiety disorders, and other behavioral health disorders. We identified inpatient behavioral health care by using the principal diagnosis on VHA inpatient stays and community care institutional claims, along with type of bill codes signifying inpatient care (11x, 41x, 42x, and 44x paired with Revenue Codes <250).
Outpatient behavioral health use.
We identified 394 Current Procedural Terminology (CPT) and Healthcare Common Procedure Coding System (HCPCS) codes used for behavioral health services (see online supplement). These codes came from the Centers for Medicare and Medicaid Services (CMS), the Agency for Healthcare Research and Quality’s Clinical Classification Software (13), and Berenson-Eggers Type of Service codes for psychiatry (14, 15). We excluded CPT codes for laboratory tests, medications, provider-to-provider training and education, and ancillary services, such as employment, child care, and legal counseling.
Provider training.
Next, we identified general CPT codes that represent behavioral health care when provided by a behavioral health specialist (see online supplement), including evaluation and management codes; neuropsychiatric procedures, tests, and therapies; assessment and treatment of pain; nonspecific procedures, treatments, or screenings (e.g., HCPCS 4065F: “therapeutic, preventive, or other interventions”); and electrocardiogram. To count as behavioral health care, codes had to be administered by a provider with a behavioral health taxonomy (see online supplement). Finally, we limited outpatient care to ambulatory settings by using specific place of service codes for community care and excluding VHA-delivered care outside specific clinics, identified by stop codes (see online supplement).
Behavioral health severity and comorbid conditions.
Because veterans commonly have more than one behavioral health condition and more than one condition can be treated in a single outpatient visit, we assigned veterans over the study period and each individual visit to the most severe behavioral health condition group, building off Hunter and colleagues’ (12) hierarchy of high-cost VHA patients with behavioral health conditions, from most severe to least severe: serious mental illness, substance use disorders, PTSD, personality disorders, mood disorders, anxiety disorders, and other behavioral health disorders. For example, a veteran treated for PTSD and anxiety in the same visit would be categorized as PTSD. If an outpatient behavioral health visit did not include a behavioral health ICD-10 diagnosis code, we labeled these visits or veterans as having diagnoses that were non–behavioral health specific. These were mostly diagnoses or conditions such as homelessness, counseling, insomnia, headache, etc., that describe services, circumstances, and symptoms frequently associated with behavioral health diagnoses.
For other comorbid conditions, we calculated a modified Gagne comorbidity score at the veteran-year level. We modified the algorithms by removing any diagnoses (alcohol abuse, drug abuse, psychoses, and depression) that overlapped with our behavioral health conditions (16).
Sociodemographic and access characteristics.
For sociodemographic characteristics, we identified the veteran’s age, sex, race, ethnicity, marital status, and history of homelessness. With respect to health care system access, we included distance to VHA primary care (outpatient analyses) or secondary care (inpatient analyses), rurality, VHA priority group assignment, and health insurance status (VHA coverage plus private, Medicare, or Medicaid insurance). VHA priority group is a composite score (1–8, highest to lowest priority, respectively) assigned to veterans receiving VHA health care. Priority group scores consider military service, percent service-connected disability, socioeconomic status, qualification for Medicaid, and other VA benefits. All else equal, veterans in higher VHA priority groups may receive care earlier than veterans in lower VHA priority groups and may pay less (if anything) for care. Although we controlled for distance, we did not control for hardship (e.g., traveling by boat), given its correlation with distance, or wait time, given that veterans usually have timely access to VHA behavioral health care; the MyVA Mental Health Initiative has stipulated same-day appointments since 2016 (17), when our study began.
We also included the following county characteristics: presence of a VA medical center (VAMC) (in outpatient and inpatient analyses) or a community-based outpatient clinic (in outpatient analyses), ratio of population to behavioral health providers, median income (in thousands), unemployment rate, and poverty rate (percentage of households with income below the federal poverty level). In outpatient visit and inpatient stay analyses, we included an index variable for FY of the visit or stay.
Data Analysis
We compared medical, sociodemographic, and access characteristics between veterans receiving all their behavioral health care in VHA and veterans receiving some behavioral health community care separately for inpatient and outpatient care. We also compared behavioral health conditions and provider type between VHA and community care, with the inpatient stay and outpatient visit as the unit of analysis. For these comparisons, we conducted t tests for continuous variables and chi-square tests for categorical variables. Given our large sample sizes, we also calculated standardized mean differences (SMDs) to assess the magnitude of differences between means and proportions in VHA and community care groups (18). An SMD of 0.2 has been described as small, 0.5 as medium, and 0.8 as large (19). For the multivariate analysis of the association between severity of behavioral health condition and care setting (VHA versus community), we used a linear probability model (LPM) because of the large sample size and included facility-level fixed effects because of consistent differences in behavioral health treatment styles and capacity at VHA facilities. Because veterans could have multiple behavioral health visits or stays, we clustered standard errors at the patient level to account for the correlation in characteristics (e.g., gender) within individuals. In sensitivity analyses, a logistic regression model (with random effects for VHA facilities and county) for inpatient stays was consistent with the LPM results, although a logistic regression model for outpatient visits did not converge.
We emphasize point estimates and confidence intervals over significance testing when interpreting results, but we also provide results of hypothesis tests on a comparisonwise basis based on a two-sided significance level of 0.05, without adjustment for multiple comparisons. Given the large number of analyses performed, some nominally significant results may reflect type 1 errors, particularly those with p values close to 0.05. All analyses were performed with SAS Enterprise Guide, version 8.2.
Results
Behavioral Heath Inpatient and Outpatient Cohort Characteristics
Of the 204,094 veterans using inpatient behavioral health care between FY2016 and FY2019, 27% used some community care (Table 1, Figure 1, and online supplement). Compared with veterans who received inpatient behavioral health care in VHA only, those who received some community care had more severe behavioral health conditions (SMD=0.25) and worse access to care—with greater distance to VHA specialty care (SMD=0.22), less health insurance coverage outside VHA (SMD=0.21), and less likelihood of having a VAMC in their county of residence (SMD=−0.23); however, even these differences were small (i.e., not much over 0.20).
| Only VHA care (N=148,284) | Any community care (N=55,810) | |||||
|---|---|---|---|---|---|---|
| Characteristic | N | % | N | % | Standardized mean difference | p |
| Veteran behavioral health condition group | .25 | <.001 | ||||
| Serious mental illness | 47,037 | 31.7 | 21,781 | 39.0 | ||
| Substance use disorders | 41,759 | 28.2 | 16,263 | 29.1 | ||
| Posttraumatic stress disorder | 13,059 | 8.8 | 3,115 | 5.6 | ||
| Personality disorders | 1,332 | .9 | 170 | .3 | ||
| Mood disorders | 38,852 | 26.2 | 13,237 | 23.7 | ||
| Anxiety disorders | 2,998 | 2.0 | 399 | .7 | ||
| Other behavioral health disorders | 3,247 | 2.2 | 845 | 1.5 | ||
| Gagne physical health score (M±SD)a | 1.26±2.14 | 1.21±1.98 | −.02 | <.001 | ||
| Veteran characteristic | ||||||
| Age (M±SD) | 51.45±14.89 | 49.97±14.34 | −.10 | <.001 | ||
| Sex | .06 | <.001 | ||||
| Male | 132,261 | 89.2 | 48,697 | 87.3 | ||
| Female | 16,023 | 10.8 | 7,106 | 12.7 | ||
| Missing | 0 | — | 7 | .0 | ||
| Race | .02 | <.001 | ||||
| American Indian or Alaska Native | 1,533 | 1.0 | 649 | 1.2 | ||
| Asian | 1,151 | .8 | 410 | .7 | ||
| Black or African American | 37,633 | 25.4 | 13,637 | 24.4 | ||
| Native Hawaiian or other Pacific Islander | 1,135 | .8 | 485 | .9 | ||
| White | 99,563 | 67.1 | 37,789 | 67.7 | ||
| Multiple races | 1,754 | 1.2 | 719 | 1.3 | ||
| Missing | 5,515 | 3.7 | 2,121 | 3.8 | ||
| Ethnicity | .04 | <.001 | ||||
| Hispanic | 9,690 | 6.5 | 4,279 | 7.7 | ||
| Non-Hispanic | 135,181 | 91.2 | 50,114 | 89.8 | ||
| Missing | 3,413 | 2.3 | 1,417 | 2.5 | ||
| Marital status | .13 | <.001 | ||||
| Married | 44,879 | 30.3 | 15,265 | 27.4 | ||
| Previously married | 62,309 | 42.0 | 24,006 | 43.0 | ||
| Never married | 38,596 | 26.0 | 14,706 | 26.4 | ||
| Missing | 2,500 | 1.7 | 1,833 | 3.3 | ||
| Homeless status | .09 | <.001 | ||||
| Homeless | 46,259 | 31.2 | 18,997 | 34.0 | ||
| Not homeless | 97,685 | 65.9 | 34,601 | 62.0 | ||
| Missing | 4,340 | 2.9 | 2,212 | 4.0 | ||
| Access to health care system | ||||||
| Distance to VHA specialty care | .22 | <.001 | ||||
| >40 miles | 38,580 | 26.0 | 19,757 | 35.4 | ||
| ≤40 miles | 108,080 | 72.9 | 34,988 | 62.7 | ||
| Missing | 1,624 | 1.1 | 1,065 | 1.9 | ||
| Rurality | .02 | <.001 | ||||
| Urban | 108,242 | 73.0 | 41,219 | 73.9 | ||
| Rural | 35,256 | 23.8 | 12,855 | 23.0 | ||
| Highly rural or island | 1,222 | .8 | 353 | .6 | ||
| Missing | 3,564 | 2.4 | 1,383 | 2.5 | ||
| VHA priority group assignment | .15 | <.001 | ||||
| 1–6 (high-priority service on the basis of service-connected disability or income) | 137,687 | 92.9 | 52,903 | 94.8 | ||
| 7–8 (lower priority; above annual income threshold) | 9,556 | 6.4 | 2,715 | 4.9 | ||
| Missing | 1,041 | .7 | 192 | .3 | ||
| Health insurance status | .21 | <.001 | ||||
| No health insurance | 85,620 | 57.7 | 36,104 | 64.7 | ||
| Health insurance (private, Medicare, or Medicaid) | 59,901 | 40.4 | 17,501 | 31.4 | ||
| Missing | 2,763 | 1.9 | 2,205 | 4.0 | ||
| County characteristic | ||||||
| VA medical center | −.23 | <.001 | ||||
| Yes | 72,212 | 48.7 | 20,885 | 37.4 | ||
| No | 76,072 | 51.3 | 34,925 | 62.6 | ||
| Ratio of population to behavioral health providers (in thousands) (M±SD) | .88±1.59 | .97±1.61 | .05 | <.001 | ||
| Median income (in thousands) (M±SD) | 58.92±14.80 | 57.13±13.06 | −.13 | <.001 | ||
| Unemployment rate (M±SD) | 4.5±1.2 | 4.6±1.2 | .02 | <.001 | ||
| Percentage of households with income below federal poverty level (M±SD) | 15.3±4.8 | 15.6±4.9 | .08 | <.001 | ||
TABLE 1. Characteristics of veterans using inpatient behavioral health care from only the Veterans Health Administration (VHA) or any community care, fiscal years 2016–2019
FIGURE.1. Percentage of veterans using any community care for behavioral health, fiscal years 2016–2019
Of the 3,467,010 veterans using outpatient behavioral health care, 4% received some community care between FY2016 and FY2019 (primarily as dual users of VHA and community care; Table 2 and online supplement), with rates increasing over time (Figure 1). Compared with veterans who received outpatient behavioral health care in VHA only, those who received some community care had more severe behavioral health conditions (SMD=0.64), were more likely to be female (SMD=0.27), and had higher VHA priority status (SMD=0.23); they also had lower comorbidity scores (SMD=−0.24), were younger (SMD=−0.46), and had less health insurance coverage outside VHA (SMD=0.24).
| Only VHA care (N=3,335,426) | Any community care (N=131,584) | |||||
|---|---|---|---|---|---|---|
| Characteristic | N | % | N | % | Standardized mean difference | p |
| Veteran behavioral health condition group | .64 | <.001 | ||||
| Serious mental illness | 407,176 | 12.2 | 26,428 | 20.1 | ||
| Substance use disorders | 428,445 | 12.9 | 20,264 | 15.4 | ||
| Posttraumatic stress disorder | 671,222 | 20.1 | 45,058 | 34.2 | ||
| Personality disorders | 14,921 | .5 | 840 | .6 | ||
| Mood disorders | 555,706 | 16.7 | 20,669 | 15.7 | ||
| Anxiety disorders | 139,192 | 4.2 | 3,673 | 2.8 | ||
| Other behavioral health disorders | 60,756 | 1.8 | 1,462 | 1.1 | ||
| Non–behavioral health diagnosisa | 1,058,008 | 31.7 | 13,190 | 10.0 | ||
| Gagne physical health score (M±SD)b | .95±2.01 | .54±1.39 | −.24 | <.001 | ||
| Veteran characteristic | ||||||
| Age (M±SD) | 58.07±17.79 | 50.34±15.53 | −.46 | <.001 | ||
| Sex | .27 | <.001 | ||||
| Male | 2,976,720 | 89.3 | 105,196 | 80.0 | ||
| Female | 358,694 | 10.8 | 25,481 | 19.4 | ||
| Missing | 12 | .0 | 907 | .7 | ||
| Race | .15 | <.001 | ||||
| American Indian or Alaska Native | 29,674 | .9 | 1,851 | 1.4 | ||
| Asian | 35,730 | 1.1 | 2,530 | 1.9 | ||
| Black or African American | 712,943 | 21.4 | 22,471 | 17.1 | ||
| Native Hawaiian or other Pacific Islander | 29,696 | .9 | 1,716 | 1.3 | ||
| White | 2,308,833 | 69.2 | 91,770 | 69.7 | ||
| Multiple races | 35,758 | 1.1 | 2,158 | 1.6 | ||
| Missing | 182,792 | 5.5 | 9,088 | 6.9 | ||
| Ethnicity | .12 | <.001 | ||||
| Hispanic | 231,318 | 6.9 | 12,502 | 9.5 | ||
| Non-Hispanic | 2,994,520 | 89.8 | 113,437 | 86.2 | ||
| Missing | 109,588 | 3.3 | 5,645 | 4.3 | ||
| Marital status | .09 | <.001 | ||||
| Married | 1,615,480 | 48.4 | 63,913 | 48.6 | ||
| Previously married | 1,096,465 | 32.9 | 39,429 | 30.0 | ||
| Never married | 566,983 | 17.0 | 23,308 | 17.7 | ||
| Missing | 56,498 | 1.7 | 4,934 | 3.8 | ||
| Homeless status | .15 | <.001 | ||||
| Homeless | 339,084 | 10.2 | 11,595 | 8.8 | ||
| Not homeless | 2,977,822 | 89.3 | 119,579 | 90.9 | ||
| Missing | 18,520 | .6 | 410 | .3 | ||
| Access to health care system | ||||||
| Distance to VHA primary care | .16 | <.001 | ||||
| >40 miles | 165,552 | 5.0 | 11,532 | 8.8 | ||
| ≤40 miles | 3,164,488 | 94.9 | 119,757 | 91.0 | ||
| Missing | 5,386 | .2 | 295 | .2 | ||
| Rurality | .15 | <.001 | ||||
| Urban | 2,292,912 | 68.7 | 85,981 | 65.3 | ||
| Rural | 956,048 | 28.7 | 40,235 | 30.6 | ||
| Highly rural or island | 35,102 | 1.1 | 1,917 | 1.5 | ||
| Missing | 51,364 | 1.5 | 3,451 | 2.6 | ||
| VHA priority group assignment | .23 | <.001 | ||||
| 1–6 (high-priority service on the basis of service-connected disability or income) | 2,925,189 | 87.7 | 122,871 | 93.4 | ||
| 7–8 (lower priority; above annual income threshold) | 359,024 | 10.8 | 7,195 | 5.5 | ||
| Missing | 51,213 | 1.5 | 1,518 | 1.2 | ||
| Health insurance status | .24 | <.001 | ||||
| No health insurance | 1,408,595 | 42.2 | 67,703 | 51.5 | ||
| Health insurance (private, Medicare, or Medicaid) | 1,856,243 | 55.7 | 59,260 | 45.0 | ||
| Missing | 70,588 | 2.1 | 4,621 | 3.5 | ||
| County characteristic | ||||||
| VA medical center or community-based outpatient clinic | .14 | <.001 | ||||
| Yes | 2,352,547 | 70.5 | 91,184 | 69.3 | ||
| No | 960,622 | 28.8 | 39,601 | 30.1 | ||
| Missing | 22,257 | .7 | 799 | .6 | ||
| Ratio of population to behavioral health providers (in thousands) (M±SD) | 1.00±1.71 | .84±1.29 | −.10 | <.001 | ||
| Median income (in thousands) (M±SD) | 58.73±14.72 | 60.27±14.57 | .10 | <.001 | ||
| Unemployment rate (M±SD) | 4.5±1.2 | 4.5±1.3 | −.01 | <.001 | ||
| Percentage of households with income below federal poverty level (M±SD) | 15.1±4.9 | 14.9±4.6 | −.05 | <.001 | ||
TABLE 2. Characteristics of veterans using outpatient behavioral health care from the Veterans Health Administration (VHA) only or using any community care, fiscal years 2016–2019
Unadjusted VHA and Community Care Behavioral Health Utilization
From FY2016 to FY2019, 80% (N=360,148) of the 448,648 inpatient behavioral health stays occurred in VHA and 20% (N=88,500) occurred in the community, and 97% (N=53,346,731) of 55,043,607 outpatient behavioral health visits occurred in VHA and 3% (N=1,696,876) occurred in the community (online supplement). The distribution of treated behavioral health conditions differed considerably within inpatient and outpatient care and across care settings. In particular, for inpatient stays, more severe conditions were treated in VHA than in the community; however, no clear pattern was seen for outpatient care (online supplement).
Provider Training
In addition, larger variability by provider type was noted for outpatient visits in VHA, compared with the community (SMD=1.06; Table 3). There was a greater presence of highly trained specialists—psychiatrists and behavioral neurologists (22% versus 10%) and psychologists (25% versus 18%)—treating veterans in VHA, compared with those treating veterans in the community, along with a greater presence of social workers in VHA (36% versus 15%). For care in the community, the dominant provider type for outpatient behavioral health visits was counselors and therapists at 40%, compared with 7% in VHA. The top two services provided in VHA outpatient visits were group psychotherapy (14%) and individual psychotherapy for 60 minutes (9%) (see online supplement). Individual psychotherapy for 60 minutes was the most common service provided in the community (47%), and methadone administration was the second most common (15%) (see online supplement).
| VHA visits | Community care visits | |||
|---|---|---|---|---|
| Provider type | N | % | N | % |
| Psychiatry and behavioral neurology | 11,769,045 | 22.1 | 172,661 | 10.2 |
| Psychology | 13,574,507 | 25.5 | 306,056 | 18.0 |
| Behavioral health advanced practice provider | 2,818,914 | 5.3 | 45,048 | 2.7 |
| Behavioral health counseling and therapy | 3,799,639 | 7.1 | 681,181 | 40.1 |
| Behavioral health social worker | 19,154,090 | 35.9 | 246,609 | 14.5 |
| Psychiatric pharmacist | 176,647 | .3 | 8,217 | .5 |
| Other | 2,053,889 | 3.9 | 237,104 | 14.0 |
TABLE 3. Type of provider seen by veterans for behavioral health outpatient visits in the Veterans Health Administration (VHA) and in the community, fiscal years 2016–2019a
Adjusted Probabilities for Behavioral Health Utilization in VHA or Community Care
For inpatient behavioral health care, veterans with PTSD, personality disorders, and anxiety disorders had a significantly lower probability of receiving community care, compared with veterans with other behavioral health disorders, and veterans with mood disorders had a significantly higher probability of receiving purchased care in the community versus VHA, compared with veterans with other behavioral health disorders (Table 4). Additionally, veterans who were older, male, Asian (versus White), closer to a VAMC, in rural or highly rural areas (versus urban), with any health insurance, or receiving services in FY2019 (versus FY2016) had significantly lower probabilities of using community care, compared with VHA, for an inpatient behavioral health stay. Veterans who were Native Hawaiian or other Pacific Islander (versus White), non-Hispanic (versus Hispanic), not homeless, and with a high VHA priority group assignment had a significantly higher probability of using community care, compared with VHA, for an inpatient behavioral health stay. These results were consistent with logistic regression results (see online supplement).
| Inpatient stay | Outpatient visit | |||
|---|---|---|---|---|
| Characteristic | Probability | 95% CI | Probability | 95% CI |
| Behavioral health condition group (reference: other behavioral health disorders) | ||||
| Serious mental illness | .00 | −.01, .01 | −.02* | −.03, −.02 |
| Substance use disorders | .00 | −.01, .01 | −.00 | −.01, .00 |
| Posttraumatic stress disorder | −.05* | −.07, −.04 | −.00* | −.01, −.00 |
| Personality disorders | −.12* | −.13, −.10 | −.02* | −.02, −.01 |
| Mood disorders | .04* | .03, .05 | −.00 | −.00, .00 |
| Anxiety disorders | −.07* | −.08, −.05 | −.00* | −.01, −.00 |
| Non–behavioral health diagnosis | na | na | −.03* | −.03, −.03 |
| Gagne physical health score | −.00* | −.00, −.00 | −.00* | −.00, −.00 |
| Sociodemographic characteristic | ||||
| Age | −.03* | −.05, −.02 | −.00 | −.00, .00 |
| Male (reference: female) | −.03* | −.05, −.02 | −.01* | −.02, −.01 |
| Race (reference: White) | ||||
| American Indian or Alaska Native | −.02 | −.04, −.01 | −.01* | −.01, −.01 |
| Asian | −.02* | −.04, −.01 | −.01* | −.01, −.00 |
| Black or African American | −.00 | −.02, .01 | −.01* | −.01, −.00 |
| Native Hawaiian or other Pacific Islander | .01* | .01, .01 | −.01* | −.01, −.01 |
| Multiple races | .00 | −.01, .01 | −.01* | −.01, −.00 |
| Non-Hispanic ethnicity (reference: Hispanic) | .01* | .01, .01 | .00* | .00, .00 |
| Marital status (reference: never married) | ||||
| Married | .00 | −.00, .01 | .00 | −.00, .01 |
| Previously married | .00 | −.00, .01 | .00* | .00, .00 |
| Not homeless (reference: homeless) | .02* | .02, .02 | .02* | .02, .02 |
| Access to health care systems | ||||
| Distance to VHA primary or secondary care ≤40 miles (reference: >40 miles) | −.05* | −.06, −.05 | −.03* | −.03, −.03 |
| Rurality (reference: urban) | ||||
| Rural | −.03* | −.03, −.03 | −.01* | −.01, −.01 |
| Highly rural | −.06* | −.06, −.06 | −.01* | −.01, −.01 |
| VHA priority group assignment 1–6 high priority service (reference: lower priority; above annual income threshold) | .01* | .01, .02 | .00* | .00, .00 |
| Health insurance (reference: no health insurance) | −.05* | −.07, −.03 | −.00 | −.02, .01 |
| County characteristic | ||||
| VA medical center or community-based outpatient center (reference: none) | −.00 | −.02, .02 | .01 | −.01, .02 |
| Ratio of population to behavioral health providers | −.00 | −.02, .02 | −.00 | −.01, .01 |
| Median income | −.00 | −.03, .03 | −.00 | −.01, .01 |
| Unemployment rate | −.00 | −.06, .05 | −.00 | −.01, .01 |
| Percentage of households with income below federal poverty level | −.00 | −.04, .04 | −.00 | −.02, .01 |
| Fiscal year (reference: FY2016) | ||||
| 2017 | .01 | −.01, .02 | .01 | −.01, .02 |
| 2018 | .01 | −.01, .02 | .02 | −.00, .03 |
| 2019 | −.02* | −.03, −.00 | .03* | .01, .04 |
| Intercept | −.08 | −.11, −.05 | .14 | .13, .16 |
TABLE 4. Regression analyses of variables as predictors of behavioral health outpatient visits and inpatient stays in community care purchased by the Veterans Health Administration, fiscal years 2016–2019
Veterans with serious mental illness, PTSD, personality disorders, anxiety disorders, and diagnoses that were non–behavioral health specific had a significantly lower probability than veterans with other behavioral health disorders of receiving outpatient behavioral health care in the community versus VHA (Table 4). Additionally, veterans who were male, from racial minority groups (versus White), who lived nearer to VHA primary care or in rural or highly rural areas (versus urban) had significantly lower probabilities of using community care, compared with VHA, for outpatient behavioral health care. Veterans who were non-Hispanic (versus Hispanic), previously married (versus never married), not homeless, with high VHA priority status, or receiving services in FY2019 (versus FY2016) had significantly higher probabilities of using community care, compared with VHA, for outpatient behavioral health care.
Discussion
Although some prior studies have examined use of behavioral health care by veterans in VHA and Medicare (3) or Medicaid (7), this study uniquely examined behavioral health care delivered or purchased by VHA following the Veterans Choice Act of 2014, which vastly increased community care. Additionally, this study included approximately 3.5 million VHA users of both inpatient and outpatient behavioral health care, whereas the VHA-Medicare study was limited to about 15,000 VHA primary care users and the VHA-Medicaid study was limited to about 7,000 nonelderly veterans with behavioral health conditions.
Thus, this study provided a broad view of behavioral health care utilization in VHA-delivered and VHA-purchased community care for the entire VHA-enrolled population while adjusting in predictive models for non-VHA coverage through public or private health insurance. In line with our first hypothesis, we found that a substantial portion of inpatient behavioral health care was provided in the community (20% of inpatient stays), with a smaller portion of outpatient behavioral health care provided in the community (3% of visits). Our second hypothesis—that more severe behavioral health conditions would be treated in VHA than in the community—was supported for inpatient care but not for outpatient care, for which a clear pattern did not emerge. This finding likely resulted from a larger portion of outpatient behavioral health visits in VHA not having a specific behavioral health diagnosis, compared with community care visits (21% versus 4%; see online supplement), which could be related to different coding practices in VHA versus the community, where more diagnoses tend to be documented (20). As we anticipated, a larger portion of behavioral health care was provided in VHA by clinicians with more years of clinical training, compared with clinicians in the community.
We also found that group therapy accounted for a larger proportion of services provided by VHA, compared with the community. This difference was likely attributable to VHA’s focus on peer support for treatment, its need to meet access standards, and its relative absence of fee-for-service incentives to provide individual care, compared with community settings. We also found higher rates of methadone prescribing in the community than in VHA, which is not surprising because VHA runs very few opioid treatment programs authorized to prescribe methadone (21, 22); furthermore, veterans can more easily meet the requirement for multiple methadone maintenance visits each week through the closer proximity of community providers. In summary, these results demonstrate that although a small portion of VHA enrollees use outpatient behavioral health care in the community, there is demand for psychotherapy and alcohol and drug services that will require good coordination between VHA and the community. Additionally, because one-fifth of inpatient behavioral health stays occurred in the community, it is critical that information about these admissions flow back to VHA so that appropriate follow-up care occurs.
Differences in utilization patterns for VHA-delivered and VHA-purchased behavioral health care reflect veterans’ choice of behavioral health care and the supply of this care. An eligible veteran can receive care from any Community Care Network (CCN) provider willing to accept Medicare payment rates and not suspended by CMS. VHA itself cannot engage in selective contracting, although it could encourage the third-party administrators who oversee the CCN to do so.
A limitation of this study is that it is possible that VHA providers and community providers coded care differently. Whether coding practices would attenuate or exacerbate the differences we found is not immediately known.
We expect that broader eligibility criteria in the MISSION Act will increase the number of veterans seeking VHA-purchased behavioral health care. Thus, differences in military cultural competence and training in evidence-based treatment need to be addressed (8, 9). In response to the MISSION Act (Sections 123, 131, and 133), trainings are available through VHA Training Finder Real-Time Affiliate Integrated Network MISSION Act Curriculum on topics such as PTSD, opioids, and military cultural awareness (23). Additionally, VHA clinicians and staff from VHA’s Office of Mental Health and Suicide Prevention created a Community Provider Toolkit with sections on asking about military experience, working with veteran populations, supporting veteran mental health and wellness, and navigating veteran benefits and services (24). VHA’s National Center for PTSD and Rocky Mountain Mental Illness Research Education and Clinical Center also offer, respectively, free consultation on PTSD and suicide risk management for community providers working with veterans (25, 26). Opportunities thus exist for community care providers to leverage VHA resources on veteran-specific behavioral health needs.
Conclusions
As demand for behavioral health care outpaces supply, VHA may need to leverage community care to meet behavioral health needs and optimize veteran care. This is the first study to illustrate how much behavioral health care is VHA-delivered and VHA-purchased—a first step in assessing VHA enrollees’ demand for and providers’ supply of behavioral health care. Understanding differences in VHA-delivered and VHA-purchased community care provides an opportunity to help veterans receive high-quality behavioral health care. Future studies should assess the consequences of dual use of VHA-delivered and VHA-purchased behavioral health care, along with the health outcomes, quality, timeliness, and cost of care in the community, to ensure that it meets or exceeds VHA standards.
1. : Innovations in community care programs, policies, and research. Med Care 2021; 59:S229–S231Crossref, Medline, Google Scholar
2. : Prevalence, comorbidity, and prognosis of mental health among US veterans. Am J Public Health 2015; 105:2564–2569Crossref, Medline, Google Scholar
3. : Use of outpatient care in Veterans Health Administration and Medicare among veterans receiving primary care in community-based and hospital outpatient clinics. Health Serv Res 2010; 45:1268–1286Crossref, Medline, Google Scholar
4. : Veterans Health Administration and Medicare outpatient health care utilization by older rural and urban New England veterans. J Rural Health 2005; 21:167–171Crossref, Medline, Google Scholar
5. : Dual use of VA and non-VA services among primary care patients with depression. J Gen Intern Med 2009; 24:305–311Crossref, Medline, Google Scholar
6. : Does consideration of Medicare use affect VA evaluations of treatment for new episodes of depression? Adm Policy Ment Health 2008; 35:468–476Crossref, Medline, Google Scholar
7. : The impact of Medicaid enrollment on Veterans Health Administration enrollees’ behavioral health services use. Health Serv Res 2018; 53(suppl 3):5238–5259Crossref, Medline, Google Scholar
8. : Psychotherapy practices for veterans with PTSD among community-based providers in Texas. Psychol Serv 2018; 15:442–452 Crossref, Medline, Google Scholar
9. : Ready to Serve: Community-Based Provider Capacity to Deliver Culturally Competent, Quality Mental Health Care to Veterans and Their Families. Santa Monica, CA, RAND, 2014. http://www.rand.org/content/dam/rand/pubs/research_reports/RR800/RR806/RAND_RR806.pdf Google Scholar
10. : Addressing patients’ veteran status: primary care providers’ knowledge, comfort, and educational needs. Fam Med 2018; 50:455–459Crossref, Medline, Google Scholar
11. : Readiness of non–Veteran Health Administration advanced practice registered nurses to care for those who have served: a multimethod descriptive study. Policy Polit Nurs Pract 2020; 21:82–94Crossref, Medline, Google Scholar
12. : Health care utilization patterns among high-cost VA patients with mental health conditions. Psychiatr Serv 2015; 66:952–958Link, Google Scholar
13. Clinical Classifications Software (CCS) for ICD-10-PCS (Beta Version). Rockville, MD, Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project, 2019. https://www.hcup-us.ahrq.gov/toolssoftware/ccs10/ccs10.jsp Google Scholar
14. : Development and Structure of BETOS 2.0 With Illustrative Data. Washington, DC, Urban Institute, 2020. https://www.urban.org/research/publication/development-and-structure-betos-20-illustrative-data Google Scholar
15. BETOS 2.0 Classification Code Assignments 2019. Washington, DC, Urban Institute, 2019. https://datacatalog.urban.org/dataset/betos-20-classification-code-assignments-2019 Google Scholar
16. : A combined comorbidity score predicted mortality in elderly patients better than existing scores. J Clin Epidemiol 2011; 64:749–759Crossref, Medline, Google Scholar
17. : Evaluation of the Department of Veterans Affairs Mental Health Services. Washington, DC, National Academies Press, 2018. https://doi.org/10.17226/24915 Google Scholar
18. : A Unified Approach to Measuring the Effect Size Between Two Groups Using SAS. Presented at the
19. : Statistical Power Analysis for the Behavioral Sciences. Cambridge, MA, Academic Press, 2013 Crossref, Google Scholar
20. : Comorbidity assessment is uneven across Veterans Health Administration and Medicare for the same patient: implications for risk adjustment. Med Care 2020; 58:717–721Crossref, Medline, Google Scholar
21. Substance Use Disorder (SUD) Program: Locations. Washington, DC, Department of Veterans Affairs. https://www.va.gov/directory/guide/sud.asp. Accessed Jul 27, 2021 Google Scholar
22. : Characteristics of veterans receiving buprenorphine vs methadone for opioid use disorder nationally in the Veterans Health Administration. Drug Alcohol Depend 2016; 160:82–89Crossref, Medline, Google Scholar
23. Welcome to VHA TRAIN. Washington, DC, Veterans Health Administration, Employee Education System. https://train.missionact.org/main/welcome Google Scholar
24. Community Provider Toolkit. Washington, DC, Department of Veterans Affairs, 2021 Google Scholar
25. PTSD Consultation Program. Washington, DC, Department of Veterans Affairs, National Center for PTSD. https://www.ptsd.va.gov/professional/consult/index.asp. Accessed Jan 9, 2021 Google Scholar
26. Suicide Risk Management Consultation Program (SRM): Supporting Providers Who Serve Veterans. Aurora, CO, and Salt Lake City, Rocky Mountain Mental Illness Research Education and Clinical Center for Veteran Suicide Prevention, 2022. https://www.mirecc.va.gov/visn19/consult/ Google Scholar
