The goal of all health care systems is to provide the best possible care using available resources while decreasing inappropriate use of services. Often, a small proportion of patients with complex medical conditions contribute to the majority of health care utilization and spending.1 This is most apparent in the use of emergency and urgent care (ED) services, which are of high visibility and high cost to the health care system.2
Individuals experiencing homelessness are disproportionately high users of health care services in general and ED care in particular as they face barriers to accessing primary and specialty care.3–5 Although US military Veterans eligible to receive care in Veterans Health Administration (VHA) medical facilities face little or no barriers to accessing care in VHA from an insurance perspective, those Veterans experiencing homelessness are also reported to be high utilizers of ED services.6 The extent of contemporary literature on approaches to decrease ED use in general or decrease inappropriate use of the ED by Veterans is modest.6–13 Thus, there are significant gaps in both our knowledge of the utilization of the ED by homeless Veterans and the effects of strategies to reduce excessive or inappropriate use.
Since 2010, VHA has been implementing the patient-centered medical home model of care, called the Patient-aligned Care Team (PACT), with the goal of restructuring primary care to provide team-based care that is timely, comprehensive, coordinated, and patient centered.14–16 PACT allows patients to have a more active role in their health care and is associated with increased quality improvement, patient satisfaction, decreased ED visits, and decreased costs due to fewer hospital visits and readmissions.17–19 In 2011, the PACT model was expanded to include Veterans experiencing homelessness (H-PACT).7 The objectives were to engage homeless Veterans with tailored approaches to care that prevents complications, accommodates access, establishes longitudinal relationships that encourages behavior change, and intercedes with targeted efforts to prevent recidivism once housed.20,21
A recent description of the H-PACT model provides a preliminary description of a gross change in ED usage and hospitalizations in those enrolled in H-PACT7; however, no study has systematically evaluated the impact of enrollment of homeless Veterans in H-PACT programs and treatment engagement on utilization of services preenrollment and postenrollment with a focus on ED care or other services such as primary care, mental health, substance abuse services, and specialty care. Therefore, in this study, we compared utilization of services among a group of (1) homeless Veterans with H-PACT assignment (enrolled), with (2) a group of homeless Veterans with no H-PACT assignment at those centers with H-PACT (nonenrolled) and with (3) a group of homeless Veterans at medical centers that had not yet implemented H-PACT (usual care). The goal of this study was to address the hypothesis that enrollment in any primary care engagement represented by H-PACT for these Veterans would reduce excessive or inappropriate ED usage among homeless Veterans and identify who among the homeless are amenable to personalized care-delivery interventions.
Setting and Study Population
A group of homeless Veterans enrolled in H-PACT (enrolled) was identified from 20 VHA medical centers that implemented the H-PACT program between January 2012 and June 2013. These centers provide emergency, outpatient, and inpatient services to allow for assessment of the utilization of the full spectrum of health care services by Veterans. For this study, VHA sites that did not offer ED services were excluded to allow for comparability between comparator groups. Veterans experiencing homelessness were identified using a combination of administrative codes indicating homelessness (ICD-9-CM code V60.0, lack of housing)22,23 and enrollment in the H-PACT program using administrative databases. To be included in the analysis, a Veteran was required to have had at least 2 visits with their VHA medical center in the 6 months before enrollment in H-PACT and at least 1 visit in the 6 months after enrollment. Veterans were selected for enrollment in H-PACT teams based on a review of patient files by team members and the decision was made locally. The general principles for enrollment to H-PACT included a history of homelessness and utilization of VHA resources.7,24
The main comparison group of homeless Veterans was identified from patients seen at 1 of 12 VA integrated medical centers that had not yet implemented the H-PACT program (usual care). Around the country, VHA has nearly 152 medical centers and they are classified according to their size and complexity as 1A, 1B, 1C, 2, and 3. The designation is based not only on size but on the complexity of services offered. The usual care group were selected such that the centers all had equal complexity level classifications to centers that had implemented H-PACT and had an ED attached to the medical center. Homeless Veterans at these locations were identified by having been assigned a V60.0 ICD-9-CM code at least twice between January 1, 2012 and December 31, 2012. These Veterans were verified to not have had an assignment with a PACT team at their site. This comparator group was chosen to assess effect of any primary care engagement for homeless Veterans (in this study, H-PACT) versus no primary care engagement (usual care).
Homeless Veterans who were seen at VHA facilities but who were not enrolled in H-PACT or any other PACT team available at those facilities were identified as a second comparison group (nonenrolled). These Veterans were identified by having been assigned a V60.0 ICD-9-CM code at least twice between January 1, 2012 and December 31, 2012 and never having any evidence of enrollment in H-PACT or any other primary care team assignment (PACT) during the observation period. This comparator group was chosen to assess possible spillover of H-PACT engagement on those nonenrolled at the same sites.
Measures and Data Analysis
VHA administrative data pertaining to Veteran demographics (sex, age, evidence of having been in combat as noted by the combat flag in administrative data, marital status, and race), outpatient visits in various categories, and associated ICD-9 diagnoses stored in the VHA corporate data warehouse were accessed through VA Informatics Computing Infrastructure, a secure research portal.25 The Charlson comorbidity index (CCI) is a validated measure of the number and severity of coexisting diagnoses. We calculated the CCI using ICD-9-CM diagnostic codes related to inpatient and outpatient visits during the 12-month observation period for Veterans in all 3 groups using established algorithms.26 Differences in the distribution of these variables among the 3 groups were analyzed for statistical significance. Outpatient visits were grouped into various categories using sets of administrative 3-digit codes (called stop codes) used to identify clinics in VHA facilities.23,27,28 Visits were reduced to 1 per category per day. ED visits (primary stop code 130) were combined with urgent care visits (primary stop code 131) to constitute the ED encounters.
The primary outcome was the mean number of monthly ED and other health care visits per Veteran for H-PACT enrollees in the 6 months before and after enrollment in H-PACT. For Veterans in the nonenrolled and usual care groups, the mean number of monthly ED visits were analyzed during the same 12-month period as the H-PACT program, but split into two 6-month periods (first: January 1–June 30 and second: July1–December 31, 2012). In addition, a subset of high ED utilizers was identified among each of the 3 groups as those with ≥2 ED visits in a 6-month period, along with those who did not use the ED in the baseline period (0 ED group) or those who used the ED once (1 ED group). Identifying these different ED groups helped us to control for regression to the mean. Differences in the mean monthly ED visits between the 2 time periods and pairs of groups were analyzed using t tests and a difference-in-differences (DD) approach.29–31 Differences in other outpatient visits in various categories, including primary care, were analyzed using t tests. All P-values were 2-sided and defined to be significant at P<0.05. Analyses were conducted using Stata statistical software and R Statistical Package.
We fit Poisson DD regression models to the number of days in which a patient was seen in either an emergency department or urgent care clinic (ED). The key predictors were pretime/posttime period (each of 6 mo), if the patient was enrolled in H-PACT, and ED usage group at baseline. Interactions were included in the DD model between the key terms. We also included demographic variables to adjust our estimates: age (grouped into 10 y blocks), religion (grouped into Christian, other, and no preference), marital status, race, and CCI.
Our main DD model compared enrolled versus usual care groups before and after the H-PACT intervention. Using the DD approach allowed us to compare the average change in ED visits for our control group (usual care) with the average change in ED visits for our treatment group (enrolled). The difference of these differences (the DD estimator) represents the average effect of primary care engagement (in this study, H-PACT as they received no other primary care engagement) on our study population. The rationale and strong justification for the DD method is the ability to account for contemporaneous effects (changes over time that affect both the enrolled and usual care groups) and time-invariant differences (fixed effects) between the enrolled and usual care groups.
For checks on this main DD model, we also used the DD approach to compare the enrolled and nonenrolled groups as well as the nonenrolled and usual care groups. For the DD model with enrolled versus nonenrolled, the DD estimator would be biased if there was a spillover effect (ie, Veterans in the nonenrolled group were impacted by H-PACT being at the same site). Thus, we compared the DD estimator for the enrolled versus usual care model with the DD estimator for the enrolled versus nonenrolled model to see how similar they were. In addition, we ran a DD model with nonenrolled versus usual care groups. By comparing this DD estimator with the other DD estimators, it helped us understand if there was possibly a spillover effect or if we had not accounted for time-varying differences between the nonenrolled versus usual care sites.
The secondary outcome was to determine the appropriateness of ED visits before and after enrolling in H-PACT. The rationale for the choice of the secondary outcome is: (1) published literature on the topic of appropriate use of the ED (avoidable and/or preventable visits versus truly emergent ones) as an outcome metric for health care systems and (2) a second outcome to support the case management and patient education efforts of H-PACT. This outcome was addressed by categorizing the appropriateness of ED visits by using the New York University algorithm.32,33 The primary ICD-9 code for each ED visit was classified first as either nonemergent or emergent. The emergent ED visits were then categorized to 3 mutually exclusive categories: (1) ED care needed, not preventable/avoidable, (2) ED care needed preventable/avoidable, or (3) primary care treatable.34 The change in various categories of ED visits within each group (enrolled, nonenrolled, and usual care) were analyzed for statistical significance in the preenrollment and postenrollment periods for H-PACT enrollees by computing P-values with an ordinal logistic regression model comparing time periods within comparator groups. For the nonenrolled and usual care site Veterans, the change was measured from the first to the second 6 months of the 2012 calendar year.
A total of 3981 homeless Veterans were enrolled in H-PACT at 20 VHA sites during the study period (Table 1). The comparator groups consisted of 24,363 homeless Veterans at the same 20 VHA sites who were not enrolled in H-PACT or any other PACT team (nonenrolled) and 23,542 homeless Veterans at 12 usual care sites who were not enrolled in any PACT team (usual care). Overall, H-PACT enrollees were older white males who were not married and were less likely to have been in combat; these Veterans had a higher CCI as compared with the other groups and were, in general, noted to have a higher baseline use of emergency and urgent care, homeless care, primary care, mental health, and substance abuse services.
As shown in Table 2, the DD analyses demonstrated that when Veterans in H-PACT were compared with usual care, there was a significant DD in ED visits (mean count per Veteran); this was most pronounced for the 2+ ED usage group (−4.43). There was a decrease in ED visits when the H-PACT group was compared with the nonenrolled; however, the magnitude of decrease was minimal even for the high ED usage group (−0.29). There was a decrease in ED visits noted when the nonenrolled group was compared with the usual care group, with the magnitude approaching that of the H-PACT group for the high ED usage group (−3.93).
As prominent DD were noted for the high ED utilizer group, further analyses were restricted to this group. As shown in Table 3, when the visits in various service categories were analyzed for the high ED utilizers (≥2 ED visits in a 6-mo period), there was a 51% decrease in ED visits noted for H-PACT enrollees between the preenrollment and postenrollment period (0.12 vs. 0.059 mean visits per Veteran per month, P<0.001). These visits were unchanged for the H-PACT nonenrolled and usual care site Veterans between the first and second observation periods. There were significant increases in homeless care visits and significant decreases in mental health and social work visits among H-PACT Veterans. Other than an increase in homeless care visits, all other visits were unchanged in the usual care group.
For all H-PACT enrollees (Table 4), there was a significant increase in the proportion of ED visits classified as emergent ED care needed: not preventable/avoidable in the 6 months post-enrollment period. There were no changes noted in the visits classified as emergent ED care needed: preventable/avoidable or emergent/primary care treatable. Surprisingly, there was an increase noted in non-emergent visits. For both the nonenrolled and usual care group, there was an increase noted in emergent/primary care treatable visits in the second 6 months of observation. There was an increase noted in the ED visits classified as emergent ED care needed: not preventable/avoidable for the usual care group in the second 6 months of observation. For the high ED utilizers (Table 5), for the H-PACT enrollees, the only significant changes were an increase in the proportion of nonemergent visits. The high utilizers at the usual care sites also demonstrated an increase in the proportion of emergent ED care needed, not preventable/avoidable encounters, even though they too had a decrease in overall ED encounters.
Patient-centered medical home models of care have been shown to be effective in decreasing use of high-cost services, such as emergency care, and to improve overall health for Veterans in general.14,19 A chronic care model tailored to homeless Veterans (a precursor to H-PACT) demonstrated reductions in unnecessary ED use and improved chronic disease management.21 A preliminary descriptive analysis showed a decrease in ED use for those homeless Veterans enrolled in H-PACT.7 We extended these studies through a systematic analysis of a large group of homeless Veterans enrolled in H-PACT and by comparing their ED care to 2 comparator groups. H-PACT enrollees as an overall group had significantly reduced ED care visits in the 6 months after enrollment with the maximum decrease noted in the high ED utilizer group. DD analyses demonstrated a significant decrease in ED visits for the H-PACT group (that represents any primary care engagement) to the usual care group (no primary care engagement).
At the population level, though homeless Veterans are heterogenous with regard to their pathways to homelessness and comorbidities, it seems that as a group, all H-PACT enrollees are benefiting in the short term (6 mo postenrollment) with regard to their accessing and receiving appropriate care. This extended significantly to the subset who are high ED utilizers; thus this group may derive the most benefit from treatment engagement through H-PACT.
It is interesting to note that there was a decrease in ED visits in the nonenrolled group as compared with the H-PACT group, though the decrease was minimal (−0.29 DD, Table 2). This may demonstrate a spillover effect of H-PACT at those sites to those Veterans who were not enrolled in H-PACT. Because of this spillover effect, when the nonenrolled were compared with the usual care group, there was a significant DD noted for ED visits in the nonenrolled.
With regard to appropriateness of ED encounters, even with a decrease in ED visits, the H-PACT enrollees as a group did increase their proportion of ED care needed/not preventable encounters. However, there was an increase noted in the proportion of nonemergent encounters, too, indicating a possibility of unmet needs from VHA outpatient services. Homeless Veterans at the usual care sites also demonstrated an increase in the proportion of ED care needed/not preventable visits along with an increase in proportion of emergent/primary care treatable encounters, indicating a possibility of increased overall demand for ED services among homeless Veterans, a phenomenon that has been described in VHA.6,11
The large number of H-PACT enrollees represent, to the best of our knowledge, the largest trial of patient engagement in homeless health care and a strength of our study. Similarly, the large number of homeless Veterans with no known treatment engagement in PACT teams in the comparator groups may also be considered a strength of our study. The availability of longitudinal administrative data to assess outpatient service use among a group of homeless Veterans from 32 VHA medical centers nationwide allowed us to examine trends in usage among H-PACT enrollees and 2 comparator groups.
We acknowledge several limitations. Although there were guidelines provided to the H-PACT sites, the selection of homeless Veterans into the H-PACT program was not randomized and was based on decisions by staff at the local VHA medical center. This likely contributed to some variation in selection criteria across sites. However, when examined at the VHA level, the profile of the typical H-PACT enrollee seemed to be similar, as noted in Table 1. We were not able to influence randomization in the selection of homeless Veterans to H-PACT; however, we were able to obtain multiple baseline measurements of emergency and urgent care use before the intervention for the H-PACT enrollees before the intervention.
It has been noted that homeless Veterans are among those of the highest users of the ED in VHA10,11 and thus, for the high ED utilizers among all 3 groups, there may have been a natural decline in emergency and urgent care use over time. The decrease in ED visits for the overall H-PACT group and for the usual care group as a whole may be considered a regression to the mean. The DD approach29 provided unbiased treatment effect estimates when, in the absence of treatment, the average outcomes for the treated and control groups would have followed parallel trends over time.35 Furthermore, if ED usage was trending down before the treatment as would be posited by the regression to the mean phenomenon, the DD approach allows us to use a comparison group that is experiencing the same trend but is not exposed to the treatment (in this case, H-PACT engagement).31 The decrease in ED visits was noted only for the high ED utilizers in the H-PACT group (Table 3) and the DD estimator was significant for H-PACT as compared with usual care (Table 2). The magnitude of the effect was highest for the high ED utilizers in the H-PACT enrolled group versus usual care.
The H-PACT program, which has now moved from pilot testing to nationwide roll-out among VHA medical centers, offers a comprehensive treatment engagement strategy that seems to be of benefit to homeless Veterans enrolled.7 Even with modest tangible outcomes in emergency and urgent care use, it is likely that these Veterans are well served by the treatment engagement and navigation through the health care system in accessing much-needed other services. Our results demonstrate that the intended impact of H-PACT engagement may be most evident for H-PACT enrollees who are high ED utilizers. Long-term outcomes such as trends in ED usage over time, an increase in primary care usage in lieu of a decrease in ED use, and overall improvement in health due to chronic conditions merit further study.
The authors acknowledge in-kind funding and support from the National Center on Homelessness Among Veterans for J.D.F. and T.P.O. The authors wish to thank our research team members Yiwen Yao and Ying Suo for assistance in data management, Deborah Hofmann for project management, and Warren Pettey for editorial assistance. These contributors did not receive compensation for their work.
1. Powers BW, Chaguturu SK. ACOs and high-cost patients. N Engl J Med. 2016;374:203–205.
2. Althaus F, Paroz S, Hugli O, et al. Effectiveness of interventions targeting frequent users of emergency departments: a systematic review. Ann Emerg Med. 2011;58:41.e42–52.e42.
3. Locker TE, Baston S, Mason SM, et al. Defining frequent use of an urban emergency department. Emerg Med J. 2007;24:398–401.
4. Han B, Wells BL. Inappropriate emergency department visits and use of the Health Care for the Homeless Program services by Homeless adults in the northeastern United States. J Public Health Manag Pract. 2003;9:530–537.
5. D’Amore J, Hung O, Chiang W, et al. The epidemiology of the homeless population and its impact on an urban emergency department. Acad Emerg Med. 2001;8:1051–1055.
6. Tsai J, Doran KM, Rosenheck RA. When health insurance is not a factor: national comparison of homeless and nonhomeless US veterans
who use Veterans
Affairs Emergency Departments. Am J Public Health. 2013;103(suppl 2):S225–S231.
7. O’Toole TP, Johnson EE, Aiello R, et al. Tailoring care to vulnerable populations by incorporating social determinants of health: the veterans
health administration’s “Homeless Patient Aligned Care Team” program. Prev Chronic Dis. 2016;13:150567.
8. Montgomery AE, Byrne TH. Services utilization among recently homeless veterans
: a gender-based comparison. Mil Med. 2014;179:236–239.
9. LePage JP, Bradshaw LD, Cipher DJ, et al. The effects of homelessness
’ health care service use: an evaluation of independence from comorbidities. Public Health. 2014;128:985–992.
10. Gabrielian S, Yuan AH, Andersen RM, et al. VA health service utilization for homeless and low-income Veterans
: a spotlight on the VA Supportive Housing (VASH) program in greater Los Angeles. Med Care. 2014;52:454–461.
11. Tsai J, Rosenheck RA. Risk factors for ED use among homeless veterans
. Am J Emerg Med. 2013;31:855–858.
12. O’Toole TP, Bourgault C, Johnson EE, et al. New to care: demands on a health system when homeless veterans
are enrolled in a medical home model. Am J Public Health. 2013;103(suppl 2):S374–S379.
13. Hastings SN, Smith VA, Weinberger M, et al. Emergency department visits in Veterans
Affairs medical facilities. Am J Manag Care. 2011;17(spec no.):e215–e223.
14. Nelson KM, Helfrich C, Sun H, et al. Implementation of the patient-centered medical home in the Veterans
Health Administration: associations with patient satisfaction, quality of care, staff burnout, and hospital and emergency department use. JAMA Intern Med. 2014;174:1350–1358.
15. Yano EM, Bair MJ, Carrasquillo O, et al. Patient aligned care teams (PACT): VA’s journey to implement patient-centered medical homes. J Gen Intern Med. 2014;29(suppl 2):S547–S549.
16. Rosland AM, Nelson K, Sun H, et al. The patient-centered medical home in the Veterans
Health Administration. Am J Manag Care. 2013;19:e263–e272.
17. Randall I, Mohr DC, Maynard C. VHA patient-centered medical home associated with lower rate of hospitalizations and specialty care among veterans
with posttraumatic stress disorder. J Healthc Qual. 2017;39:168–176.
18. Hebert PL, Liu CF, Wong ES, et al. Patient-centered medical home initiative produced modest economic results for Veterans
Health Administration, 2010-12. Health Aff (Millwood). 2014;33:980–987.
19. Chaiyachati KH, Gordon K, Long T, et al. Continuity in a VA patient-centered medical home reduces emergency department visits. PLoS ONE. 2014;9:e96356.
20. O’Toole TP, Pirraglia PA, Dosa D, et al. Building care systems to improve access for high-risk and vulnerable veteran populations. J Gen Intern Med. 2011;26(suppl 2):683–688.
21. O’Toole TP, Buckel L, Bourgault C, et al. Applying the chronic care model to homeless veterans
: effect of a population approach to primary care on utilization and clinical outcomes. Am J Public Health. 2010;100:2493–2499.
22. US Department of Veterans
Affairs Office of Inspector GeneralVA Office of Inspector General. Homeless Incidence and Risk Factors for Becoming Homeless in Veterans
. Washington, DC: VA Office of Inspector General; 2012. Available at: https://http://www.va.gov
23. Peterson R, Gundlapalli AV, Metraux S, et al. Identifying homelessness
using VA administrative data: opportunities to expand detection criteria. PLoS ONE. 2015;10:e0132664.
24. O’Toole TP, Johnson EE, Borgia ML, et al. Tailoring outreach efforts to increase primary care use among homeless veterans
: results of a randomized controlled trial. J Gen Intern Med. 2015;30:886–898.
26. Charlson ME, Charlson RE, Peterson JC, et al. The Charlson comorbidity index is adapted to predict costs of chronic disease in primary care patients. J Clin Epidemiol. 2008;61:1234–1240.
27. Mohanty AF, Helmer DA, Muthukutty A, et al. Fibromyalgia syndrome care of Iraq- and Afghanistan-deployed Veterans
Health Administration. J Rehabil Res Dev. 2016;53:45–58.
28. Mohanty AF, Muthukutty A, Carter ME, et al. Chronic multisymptom illness among female Veterans
deployed to Iraq and Afghanistan. Med Care. 2015;53(suppl 1):S143–S148.
29. Angrist JD, Pischke JS eds. Parallel worlds: fixed effects, differences-in-differences, and panel data. Mostly Harmless Econometrics. Princeton: Princeton University Press; 2009:227–243.
30. Chumbler NR, Vogel WB, Garel M, et al. Health services utilization of a care coordination/home-telehealth program for veterans
with diabetes: a matched-cohort study. J Ambul Care Manage. 2005;28:230–240.
31. Dimick JB, Ryan AM. Methods for evaluating changes in health care policy: the difference-in-differences approach. JAMA. 2014;312:2401–2402.
32. Gandhi SO, Sabik L. Emergency department visit classification using the NYU algorithm. Am J Manag Care. 2014;20:315–320.
33. Ballard DW, Price M, Fung V, et al. Validation of an algorithm for categorizing the severity of hospital emergency department visits. Med Care. 2010;48:58–63.
35. O’Neill S, Kreif N, Grieve R, et al. Estimating causal effects: considering three alternatives to difference-in-differences estimation. Health Serv Outcomes Res Methodol. 2016;16:1–21.
Keywords:Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.
homelessness; Veterans; emergency care