Skip to main content
NCBI home page
Lippincott Open Access logoLink to Lippincott Open Access
. 2026 Jan 15;64(4):207–215. doi: 10.1097/MLR.0000000000002287

Risk for Acute Care Utilization in Housing-Insecure Adults

A National Study of a Hot Spotter Program in the US Veterans Health Administration

Daniel M Blonigen *,†,✉, Kathryn S Macia *, Ivan Raikov ‡, Jean Yoon §, Jillian Weber ∥
PMCID: PMC12970552  PMID: 41536037

Abstract

Background:

To identify patients at the highest risk for acute care utilization, health care systems have developed “hot spotter” programs. Homelessness is a robust social determinant of acute care utilization.

Objectives:

To describe the prevalence, patterns, and correlates of meeting criteria for a hot spotter program among housing-insecure adults in the US Veterans Health Administration (VHA).

Research Design:

Among veterans on the VHA Homeless Registry in Fiscal Years 2018–2022 (N=1,469,893), we identified those who met criteria for a Hot Spotter Report [ie, ≥1 hospital admissions and/or ≥2 emergency department (ED) visits in at least one quarter], described their patterns of acute care use, and examined differences in patient characteristics and outpatient service use between those who met report criteria in multiple quarters (vs. one).

Results:

Thirty percent (N=446,974) met report criteria in at least one quarter; most (56%) met report criteria in ≥2 quarters. Diagnoses of depression (58%) and/or a substance use disorder (51%) were common; however, the rate of hospitalization in an acute medical setting during the cohort period was twice that of being hospitalized in an acute mental health setting (50% vs. 25%). Being on the Hot Spotter Report in multiple quarters (vs. one) was associated with more chronic conditions (M=5.08 vs. 3.29), higher rates of suicidality (23.7% vs. 11.7%), and higher likelihood of all types of outpatient care (P<0.0001).

Conclusions:

Given rates of chronic medical conditions and medical hospitalizations, it may behoove hot spotter programs to increase care coordination with medical respite programs to support patients in the postacute phase.

Key Words: homelessness, veterans, acute care, hot spotter programs

Acute care service use [hospitalizations, emergency department (ED) visits] accounts for the vast majority of costs in a health care system.1 In any care system, typically 5%–10% of the highest-need patients contribute half or more of the total cost of acute care.2 These patients are characterized by a significant disease burden and a multitude of social needs that present challenges to patients’ capacity for self-management and to effectively access ambulatory services.2,3

Housing instability is one of the strongest social determinants of acute care utilization.4–6 Homeless persons use acute care services far more frequently than their housed counterparts.7 This pattern is also true of homeless veterans, who represent a disproportionate share of all homeless persons in the US.8 Studies of acute care utilization in integrated care systems such as the Veterans Health Administration (VHA) are advantageous as it ensures that such service use is not driven by a lack of health insurance and less likely to be driven by fragmented care.9 National cohort studies have consistently shown homelessness to be a robust predictor of frequency of acute care use in veterans.5,6,9–11 Further, among users of VHA acute care, homeless veterans (vs. housed veterans) have higher rates of substance use and psychiatric disorders and more use of all types of ambulatory care.5,11,12

To reduce costs and improve the quality of care, health care systems have developed “hot spotter” programs that use real-time data on acute care service use to identify high-need patients and assist with their care coordination.13,14 VHA developed a Hot Spotter Report to identify veterans on their Homeless Registry who had ≥1 hospital admissions and/or ≥2 ED visits in the past quarter.15 These cutoffs were based on analyses of the costs of care–ie, VHA Homeless Program Office found that homeless veterans who met or exceeded this utilization threshold accounted for ∼70% of all acute care service use in this population and comprised the top 10% of VHA patients in terms of costs of care.15 Hot Spotter Reports serve as an early intervention tool to identify those who are most at risk for super-utilization of acute care; reports are available on a dashboard on the Veterans Support Service Center. For staff who chose to access this information, the report is intended to prompt them to develop care plans tailored to factors that may be driving acute care use in these patients and support their primary care engagement and care coordination. Given that VHA is the largest provider of health care for homeless adults in the US,16 research on veterans identified by these reports can provide insights into at-risk homeless adults in other health care systems.

Previously, Szymkowiak et al3 identified 16,912 homeless veterans who met criteria for the Hot Spotter Report in at least one quarter between July 1, 2014, and December 31, 2015; latent class analysis of these veterans’ acute care utilization patterns identified 7 classes, which varied considerably in the extent to which acute care use persisted over time. For example, 43% of the total sample met Hot Spotter report criteria in only one quarter and one-third of the sample met this criteria for ≥3 quarters. Analyses to determine what patient characteristics distinguished these groups were largely limited to sociodemographics—eg, those meeting the report criteria in ≥3 quarters were disproportionately older, White, non-Hispanic, unmarried, and had a rural residence.

To inform national efforts aimed at addressing the health care needs of homeless persons,17–19 we identified all veterans who met criteria for VHA’s Homeless Hot Spotter Report in fiscal years (FY) 2018–2022. This paper expands on the prior work of Szymkowiak et al3 by exploring whether those who met the report criteria in one quarter (vs. multiple quarters) differ on a wider array of patient characteristics than was previously examined and by testing for differences in use of outpatient services. Our specific objectives were to identify the (1) prevalence of meeting criteria for the Hot Spotter Report among all veterans on VHA’s Homeless Registry, (2) patterns of acute care service use in this cohort (eg, frequency and type of hospitalizations, ED visits), and (3) patient characteristics and service utilization patterns of those who met the report criteria in one quarter (vs. multiple quarters).

METHODS

Design

This retrospective-cohort design included all patients on VHA’s Homeless Registry in FYs 2018–2022 (October 1, 2018–September 30, 2022; N=1,469,893). The criteria for being on this registry is receipt of any VHA homeless services within the past 2 years; this includes those unhoused as well as those unstably housed or at risk of housing loss. Therefore, we refer to those on the Homeless Registry as “housing-insecure.” We defined our analytic cohort as those on the Registry who met criteria for the Hot Spotter Report, specifically having ≥1 hospital admissions and/or ≥2 ED visits in at least 1 of the 16 quarters during the cohort period; for the sake of parsimony, we refer to these individuals as “Hot Spotters.” Data on use of homeless services, hospitalizations, and ED visits were obtained from VHA’s Corporate Data Warehouse (CDW) and organized per patient, per quarter. CDW contains data on hospitalizations at VA facilities but not non-VA facilities. We measured patient visits and encounters across different care settings (outpatient, inpatient), relationships with providers, and ICD-10 diagnostic codes related to all encounters. Clinic stop codes were used to classify the various types of outpatient visits, including ED visits, across the entire 4-year cohort period (Supplement, Supplemental Digital Content 1, http://links.lww.com/MLR/D129 for diagnostic codes and for Bedsection labels and stop codes used to identify hospitalizations and outpatient visits, respectively). This study was reviewed and approved by the local institutional review board, with a waiver of informed consent for analysis of the electronic health record data.

Hot Spotter Status

Dichotomous variables of “hot spotter status” were measured per patient, per quarter using the criteria for the Hot Spotter Report (≥1 hospital admissions and/or ≥2 ED visits during the quarter). Hospitalizations were defined through Bedsections from the CDW Inpatient domain and were further organized by the admitting specialty to determine the type of hospitalization: medical-acute (eg, medical-surgical settings, ICU), medical-extended care (eg, short-stay or long-stay rehabilitation; short-stay or long-stay skilled nursing), mental health acute (eg, acute inpatient psychiatry), mental health-extended care (eg, substance use/mental health residential treatment) (eTable 1 in the Supplement, Supplemental Digital Content 1, http://links.lww.com/MLR/D129). ED visits were identified using stop codes 130 and 131 in the CDW Outpatient domain.

Patient Characteristics and Outpatient Service Use

Sociodemographic and mortality data were obtained from the CDW Patient domain; age was calculated from the start of the cohort period (October 1, 2018); sex, race, ethnicity, and marital status were determined from patients’ first encounter during the cohort period. Rurality at the time of the first encounter was determined from VHA’s Planning Systems Support Group Geocoded file, which uses USDA definitions.20 Homelessness (yes/no) was based on Tsai et al21 operational definition using VHA records and calculated per patient, per quarter. Chronic conditions (30 from the Chronic Condition Warehouse)22 and mental health conditions (from VHA’s Program Evaluation Resource Center)23 were based on ICD-10 diagnostic codes and coded as “yes” if present in any inpatient or outpatient encounter during the entire 4-year cohort period (eTable 2 in the Supplement, Supplemental Digital Content 1, http://links.lww.com/MLR/D129). Outpatient service use was determined from the primary stop code for encounters and categorized by mental health, housing, veterans justice programs, primary care, peer support, whole health/complementary integrative health (CIH, eg, acupuncture, chiropractic), employment services, general and specialty medicine, surgical, rehabilitation, home-based care, and extended care (eg, geriatric services) (eTable 3 in the Supplement, Supplemental Digital Content 1, http://links.lww.com/MLR/D129). Use of each of these services was calculated per patient, per quarter.

Statistical Analyses

Descriptive statistics for Hot Spotter status (number of quarters) and acute care service use were calculated for the entire cohort. Patients were categorized by Hot Spotter status (1 quarter vs. ≥2 quarters on the Hot Spotter Report) and compared on all patient variables using t tests for continuous variables and χ2 tests for categorical variables. For outpatient services, χ2 tests were used to compare the Hot Spotter groups in terms of the percentage of patients in each group who had any use of that service during the 4-year cohort period. Standardized Mean Differences (SMDs) and 99% CIs between the Hot Spotter groups were calculated for each variable by adapting the code from the stddiff package in R.24 Specifically, this R package includes a function that allows SMDs to be calculated for binary variables and does not require an assumption of normality for those variables (ie, stddiff.binary). All analyses were conducted in R version 4.4.1.25

RESULTS

Prevalence of Acute Care Utilization and Hot Spotter Status in the VHA Homeless Registry

Out of the 1,469,893 patients in the VHA Homeless Registry, 446,974 (30.4%) met Hot Spotter status in at least one quarter during the 4-year study period (referred to henceforth as the “analytic cohort”) (Table 1; Fig. 1) In any given quarter, 4.8% of Homeless Registry patients, on average, met Hot Spotter status (range=3.7%–5.5%). Figure 1 provides the frequency distribution of the number of quarters that patients in the analytic cohort met Hot Spotter status. On average, those in the analytic cohort met Hot Spotter status in 2.51 quarters (SD=2.15); 56% of the analytic cohort met Hot Spotter status in ≥2 quarters. Table 1 provides descriptives for types of acute care service use in the analytic cohort. Approximately 71% of the analytic cohort was hospitalized at least once in the 4-year period, averaging 2.22 admissions in this timeframe (SD=3.54). Fifty percent of the analytic cohort was hospitalized at least once in an acute medical setting and 25% were hospitalized in an acute mental health setting. Over 90% of the analytic cohort visited the ED over the 4-year period, with an average of 7.12 visits (SD=9.67). The prevalence of hot spotter status and acute care service use in the pre-COVID (October 1, 2018–March 30, 2020) and post-COVID periods (April 1, 2020–September 30, 2022) was comparable (eTable 4 in the Supplement, Supplemental Digital Content 1, http://links.lww.com/MLR/D129).

TABLE 1.

Descriptives of Acute Care Service Utilization in the Analytic Cohort (“Hot Spotters”)

Variable No. (%) with any Mean (SD) of total number Median (IQR) Min, max
Hot Spotter status* (no. quarters) 446,974 (100) 2.51 (2.15) 2 (1–3) 1, 16
Hospitalizations
 All-cause 317,071 (70.9) 2.22 (3.54) 1 (0–3) 0, 137
 Hospitalization type
  Medical (acute) 223,232 (49.9) 1.29 (2.39) 0 (0–2) 0, 100
  Medical (long-term) 23,239 (5.2) 0.07 (0.32) 0 0, 13
  Mental health (acute) 110,657 (24.8) 0.65 (2.12) 0 0, 125
  Mental health (long-term) 58,646 (13.1) 0.21 (0.68) 0 0, 26
  Other 2010 (0.4) 0.01 (0.08) 0 0, 6
Emergency department visits
 All-cause 411,574 (92.1) 7.12 (9.67) 5 (2–9) 0, 732
Open in a new tab
*

Hot Spotter status=≥1 hospitalizations and/or ≥2 Emergency Department visits in a quarter of the fiscal year.

FIGURE 1.

FIGURE 1

Open in a new tab

Frequency distribution of the number of quarters that patients in the analytic cohort met Hot Spotter Status.

Patient Characteristics Between Hot Spotters and Non-Hot Spotters

To understand the analytic cohort in the context of the larger VHA Homeless Registry, the patient characteristics of those who did (vs. did not) meet hot spotter status during the cohort period were compared (Table 2) . On most demographic variables, the SMD between the non-hot spotters and hot spotters was small in magnitude. An exception to this was race (SMD=0.29); 35.9% of hot spotters (vs. 28.1% of non-hot spotters) were Black/African American. On average, hot spotters (vs. non-hot spotters) also had more chronic conditions (4.29 vs. 1.37; SMD=1.07), and higher rates of all mental health diagnoses, including suicidality/self-harm.

TABLE 2.

Patient Characteristics of Hot Spotters and Non-Hot Spotters on the VHA Homeless Registry

Non-hot spotters (N=1,022,919) Hot spotters (N=446,974) Difference tests
Variables n or M (% or SD) n or M (% or SD) t or χ2 P SMDa (99% CI)
Age group <0.0001 0.12 (0.11– 0.12)
 18–34 144,408 (14.1) 56,590 (12.7)
 35–44 142,794 (14) 56,879 (12.7)
 45–44 160,132 (15.7) 70,572 (15.8)
 55–64 279,567 (27.4) 142,029 (31.8)
 ≥65 293,944 (28.8) 120,810 (27.0)
 Missing 2074 (0.2) 94 (0)
Age (M, SD) 54.75 (15.82) 54.83 (14.52) 0.003 0.01 (0.00– 0.01)
Sex <0.0001 0.03 (0.02– 0.03)
 Male 914,991 (89.5) 395,978 (88.6)
 Female 107,844 (10.5) 50,996 (11.4)
 Missing 84 (0.0) 1 (0.0)
Race <0.0001 0.29 (0.28– 0.29)
 White 585,125 (57.2) 243,444 (54.5)
 Black/ African American 287,151 (28.1) 160,520 (35.9)
 Asian 9769 (1) 3431 (0.8)
 Native Hawaiian/Pacific Islander 10,897 (1.1) 4212 (0.9)
 American Indian/Alaskan Native 15,930 (1.6) 5968 (1.3)
 Declined to answer 42,846 (4.2) 21,014 (4.7)
 Missing 71,201 (7) 8385 (1.9)
Ethnicity <0.0001 0.23 (0.22– 0.23)
 Hispanic 73,639 (7.2) 33,647 (7.5)
 Non-Hispanic 861,746 (84.2) 393,435 (88)
 Declined to answer 25,873 (2.5) 12,326 (2.8)
 Missing 61,661 (6) 7566 (1.7)
Rurality <0.0001 0.12 (0.12– 0.12)
 Rural 272,585 (26.6) 96,184 (21.5)
 Urban 750,334 (73.4) 350,790 (78.5)
Marital status <0.0001 0.18 (0.17– 0.18)
 Married 271,965 (26.6) 108,202 (24.2)
 Single (never married) 260,643 (25.5) 113,597 (25.5)
 Separated/divorced 423,673 (41.4) 204,423 (45.7)
 Widowed 40,319 (3.9) 18,171 (4.1)
 Missing 26,319 (2.6) 2581 (0.6)
Housing status
 No. quarters homeless (M, SD) 1.02 (0.28) 1.06 (0.47) <0.0001 0.10 (0.10– 0.11)
Chronic conditions
 None 521,201 (51) 28,609 (6.4)
 No. chronic conditions (M, SD) 1.37 (2) 4.29 (3.29) <0.0001 1.07 (1.07– 1.08)
Mental health diagnoses
 Depression 284,320 (27.8) 261,259 (58.5) <0.0001 0.65 (0.65– 0.66)
 Post-traumatic stress disorder 218,534 (21.4) 184,318 (41.2) <0.0001 0.44 (0.43– 0.44)
 Serious mental illness 70,719 (6.9) 98,441 (22) <0.0001 0.44 (0.44– 0.44)
 Substance use disorders 193,349 (18.9) 227,396 (50.9) <0.0001 0.71 (0.71– 0.72)
 Personality disorders 20,503 (2) 40,884 (9.1) <0.0001 0.32 (0.31– 0.32)
 Anxiety disorders 192,760 (18.8) 175,969 (39.4) <0.0001 0.46 (0.46– 0.47)
 Suicidality/self-harm 20,860 (2) 82,316 (18.4) <0.0001 0.56 (0.56– 0.57)
Mortality
 Deaths during the study period 80,638 (7.9) 58,018 (13) <0.0001 0.17 (0.16– 0.17)
Open in a new tab
a

SMD indicates standardized mean difference. These were calculated using an R package that includes a function that allows SMDs to be calculated for binary variables and does not require an assumption of normality for those variables (ie, stddiff.binary).

Patient Characteristics and Hot Spotter Status

At the outset of the cohort period, patients in the analytic cohort were 54.83 years old, on average (SD=14.52); 27.0% of the sample was age 65 years or older(Table 3) Most patients were male (88.6%), white (54.5%), non-Hispanic (88.0%), lived in an urban setting (78.5%), and were unmarried (75.2%). SMDs between the one-time and repeat Hot Spotter subgroups on these sociodemographics were generally small (range=0.03–0.15). Across the analytic cohort, patients had an average of 4.29 chronic conditions (SD=3.29), most commonly mood disorders (63.8%), hyperlipidemia (54.3%), and diabetes (31.1%) (eTable 5 in supplement, Supplemental Digital Content 1, http://links.lww.com/MLR/D129). The mean number of chronic conditions was greater for repeat versus one-time Hot Spotters (5.08 vs. 3.29; SMD=0.57). Over half of the analytic cohort had diagnoses of depression (58.5%) and substance use disorder (50.9%). Across nearly all conditions and mental health diagnoses, prevalence rates were significantly higher for repeat (vs. one-time) Hot Spotters, with SMDs largest for anemia (0.40), hypertension (0.34), and pneumonia (0.30) (eTable 5 in Supplement, Supplemental Digital Content 1, http://links.lww.com/MLR/D129). Repeat Hot Spotters were more likely than one-time Hot Spotters to receive a diagnostic code for suicidality/self-harm during the cohort period (23.7% vs. 11.7%; SMD=0.32). SMDs comparing the patient characteristics between those who met hot spotter status in 2 versus ≥3 quarters are provided in eTable 6 of the Supplement, Supplemental Digital Content 1, http://links.lww.com/MLR/D129. Among all patients in the analytic cohort, 58,018 (13%) died at some point during the study period.

TABLE 3.

Patient Characteristics (Sociodemographics, Chronic Conditions, and Mental Health Diagnoses) By Hot Spotter Status

Hot spotter status
All patients in analytic cohort (N=446,974) 1 quarter (N=198,376; 44%) ≥2 quarters (N=248,598; 56%) Difference tests between hot spotter groups (1 vs. ≥2 quarters)
Variables n or M (% or SD) n or M (% or SD) n or M (% or SD) t or χ2 P SMDb (99% CI)
Age group <0.0001 0.15 (0.14–0.16)
 18–34 56,590 (12.7) 28,819 (14.5) 27,771 (11.2) a
 35–44 56,879 (12.7) 28,228 (14.2) 28,651 (11.5) a
 45–44 70,572 (15.8) 32,178 (16.2) 38,394 (15.4) a
 55–64 142,029 (31.8) 59,242 (29.9) 82,787 (33.3) a
 ≥65 120,810 (27.0) 49,875 (25.1) 70,935 (28.5) a
 Missing 94 (0) 34 (0) 60 (0)
Age (M, SD) 54.83 (14.52) 55.77 (14.19) 53.65 (14.84) <0.0001 0.15 (0.14– 0.15)
Sex <0.0001 0.03 (0.02– 0.04)
 Male 395,978 (88.6) 174,593 (88) 221,385 (89.1)
 Female 50,996 (11.4) 23,783 (12) 27,213 (10.9)
 Missing 0 0 0
Race <0.0001 0.09 (0.08– 0.09)
 White 243,444 (54.5) 109,896 (55.4) 133,548 (53.7) a
 Black/African American 160,520 (35.9) 67,783 (34.2) 92,737 (37.3) a
 Asian 3431 (0.8) 1798 (0.9) 1633 (0.7) a
 Native Hawaiian/Pacific Islander 4212 (0.9) 1994 (1) 2218 (0.9) a
 American Indian/Alaskan Native 5968 (1.3) 2819 (1.4) 3149 (1.3) a
 Declined to answer 12,326 (4.7) 9616 (4.8) 11,398 (4.6) a
 Missing 8385 (1.9) 4470 (2.3) 3915 (1.6) a
Ethnicity <0.0001 0.05 (0.04– 0.05)
 Hispanic 33,647 (7.5) 15,573 (7.9) 18,074 (7.3) a
 Non-Hispanic 393,435 (88.0) 173,519 (87.5) 219,916 (88.5) a
 Declined to answer 12,326 (2.8) 5370 (2.7) 6956 (2.8)
 Missing 7566 (1.7) 3914 (2) 3652 (1.5) a
Rurality <0.0001 0.11 (0.10– 0.12)
 Rural 96,184 (21.5) 47,667 (24) 48,517 (19.5)
 Urban 350,790 (78.5) 150,709 (76) 200,081 (80.5)
Marital status <0.0001 0.09 (0.08– 0.09)
 Married 108,202 (24.2) 51,330 (25.9) 56,872 (22.9) a
 Single (never married) 113,597 (25.5) 50,651 (25.5) 62,946 (25.3)
 Separated/divorced 204,423 (45.7) 87,840 (44.3) 116,583 (46.9) a
 Widowed 18,171 (4.1) 7194 (3.6) 10,977 (4.4) a
 Missing 2581 (0.6) 1361 (0.7) 1220 (0.5) a
Housing status
 No. quarters homeless (M, SD) 1.06 (0.47) 1.03 (0.37) 1.07 (0.53) <0.0001 0.08 (0.08– 0.09)
Chronic conditions
 None 28,609 (6.4) 19,765 (10) 8844 (3.6) <0.0001 0.57 (0.56– 0.58)
 No. chronic conditions (M, SD) 4.29 (3.29) 3.29 (2.67) 5.08 (3.52) <0.0001 0.57 (0.56– 0.58)
Mental health diagnoses
 Depression 261,259 (58.5) 105,566 (53.2) 155,693 (62.6) <0.0001 0.19 (0.18– 0.20)
 Post-traumatic stress disorder 184,318 (41.2) 76,922 (38.8) 107,396 (43.2) <0.0001 0.09 (0.08– 0.10)
 Serious mental illness 98,441 (22) 34,511 (17.4) 63,930 (25.7) <0.0001 0.20 (0.20– 0.21)
 Substance use disorders 227,396 (50.9) 87,981 (44.4) 139,415 (56.1) <0.0001 0.24 (0.23– 0.24)
 Personality disorders 40,884 (9.1) 11,870 (6) 29,014 (11.7) <0.0001 0.20 (0.19– 0.21)
 Anxiety disorders 175,969 (39.4) 70,311 (35.4) 105,658 (42.5) <0.0001 0.15 (0.14– 0.15)
 Suicidality/self-harm 82,316 (18.4) 23,282 (11.7) 59,034 (23.7) <0.0001 0.32 (0.31– 0.33)
Mortality
 Deaths during the study period 58,018 (13) 23,342 (11.8) 34,676 (13.9) <0.0001 0.07 (0.06– 0.07)
Open in a new tab
a

Categories whose proportions significantly differed between Hot Spotter groups at Bonferroni-adjusted P<0.01.

b

SMD indicates standardized mean difference. These were calculated using an R package that includes a function that allows SMDs to be calculated for binary variables and does not require an assumption of normality for those variables (ie, stddiff.binary).

Outpatient Services and Hot Spotter Status

Most patients in the analytic cohort attended primary care (97.2%), general and specialty medicine (88.7%), surgical care (83.6%), whole health/CIH (76.3%), rehabilitation (74.3%), and mental health (74.2%) outpatient services at least once during the 4-year period (Table 4) Level of engagement for these services was greatest for primary care and mental health (median of 17 and 16 total visits, respectively, across the cohort period). Among mental health services, general outpatient mental health was most commonly utilized (71.1%). More than one-third of patients used a housing service, including nearly 1 in 4 who accessed permanent supportive housing services through the US Department of Housing and Urban Development–Veterans Affairs Supportive Housing (HUD-VASH) program. Outpatient services that were utilized by a minority of patients during the cohort period were peer support (20.8%), employment (17.4%), home-based care (15.4%), veterans justice programs (12.3%), and extended care (10.5%). Compared with one-time Hot Spotters, repeat Hot Spotters were more likely to have utilized all types of outpatient services, with SMDs largest for rehabilitation (0.42), whole health/CIH (0.42), general and medical specialty (0.32), surgical (0.31), and substance use disorder treatment (0.31).

TABLE 4.

Outpatient Service Utilization by Hot Spotter Status

All patients in the analytic cohort (n=446,974) Hot spotter status 1 quarter (n=198,376; 44%) Hot spotter status ≥2 quarters (n=248,598; 56%)
Service use Service use Service use Difference tests of any service use between hot spotter groups (1 vs. ≥2 quarters)
Service type No. (%) of patients with any use Median (IQR) in patients with any use No. (%) of patients with any use Median (IQR) in patients with any use No. (%) of patients with any use Median (IQR) in patients with any use χ2 * P SMD† (99% CI)
Mental health (any) 331,628 (74.2) 16 (48) 135,643 (68.4) 12 (29) 195,985 (78.8) 21 (69) <0.0001 0.24 (0.23– 0.25)
 Geropsychology 8998 (2.0) 2.01 2 (5) 2246 (1.1) 2 (5) 6752 (2.7) 2 (5) <0.0001 0.12 (0.11– 0.12)
 MHICM‡ 15,065 (3.4) 14 (88) 3581 (1.8) 17 (94) 11,484 (4.6) 13 (86) <0.0001 0.16 (0.15– 0.17)
 PRRC§ 31,437 (7.0) 3 (17) 7504 (3.8) 3 (19) 23,933 (9.6) 3 (16) <0.0001 0.24 (0.23– 0.24)
 PTSD∥ 53,285 (11.9) 4 (12) 19,863 (10.0) 5 (11) 33,422 (13.5) 4 (12) <0.0001 0.11 (0.10– 0.11)
 Residential rehabilitation 73,014 (16.3) 53 (113) 20,412 (10.3) 38 (84) 52,602 (21.2) 60 (129) <0.0001 0.30 (0.29– 0.31)
 Substance use disorder 113,907 (25.5) 12 (47) 36,077 (18.2) 9 (35) 77,830 (31.3) 15 (52) <0.0001 0.31 (0.30– 0.32)
 General outpatient 317,956 (71.1) 11 (23) 129,008 (65.0) 8 (16) 188,948 (76.0) 13 (30) <0.0001 0.24 (0.23– 0.25)
Housing (any) 164,132 (36.7) 6 (20) 60,357 (30.4) 5 (17) 103,775 (41.7) 7 (22) <0.0001 0.24 (0.23– 0.25)
 HUD-VASH 100,368 (22.46) 10 (26) 36,065 (18.2) 9 (24) 64,303 (25.9) 11 (28) <0.0001 0.19 (0.18– 0.19)
 Grant per diem 45,799 (10.25) 3 (5) 13,976 (7.0) 3 (5) 31,823 (12.8) 3 (6) <0.0001 0.19 (0.19– 0.20)
 HCHV# 122,725 (27.5) 2 (5) 42,788 (21.6) 2 (3) 79,937 (32.2) 3 (6) <0.0001 0.24 (0.23– 0.25)
Veterans justice programs 54,905 (12.3) 2 (6) 21,067 (10.6) 2 (6) 33,838 (13.6) 2 (6) <0.0001 0.09 (0.08– 0.10)
Primary care 434,484 (97.2) 17 (20) 189,457 (95.5) 14 (16) 245,027 (98.6) 20 (21) <0.0001 0.18 (0.17– 0.19)
Peer support 92,890 (20.8) 4 (11) 28,593 (14.4) 3 (8) 64,297 (25.9) 5 (12) <0.0001 0.29 (0.28– 0.30)
Whole health / CIH** 340,839 (76.3) 7 (14) 131,858 (66.5) 4 (9) 208,981 (84.1) 9 (18) <0.0001 0.42 (0.41– 0.42)
Employment services 77,645 (17.4) 4 (11) 24,754 (12.5) 3 (8) 52,891 (21.3) 5 (12) <0.0001 0.24 (0.23– 0.24)
 Voc Rehab 46,331 (10.4) 3 (7) 13,616 (6.9) 2 (5) 32,715 (13.2) 3 (8) <0.0001 0.21 (0.20– 0.22)
 CWT†† 50,561 (11.3) 4 (12) 15,020 (7.6) 4 (10) 35,541 (14.3) 5 (12) <0.0001 0.22 (0.21– 0.22)
 HVCES‡‡ 16,951 (3.8) 2 (2) 5938 (3.0) 2 (2) 11,013 (4.4) 2 (2) <0.0001 0.08 (0.07– 0.08)
General and specialty medicine 396,271 (88.7) 9 (17) 164,506 (82.9) 6 (10) 231,765 (93.2) 12 (22) <0.0001 0.32 (0.31– 0.33)
Surgical 373,601 (83.6) 8 (15) 153,082 (77.2) 6 (11) 220,519 (88.7) 10 (17) <0.0001 0.31 (0.30– 0.32)
Rehabilitation 332,156 (74.3) 7 (15) 127,410 (64.2) 5 (9) 204,746 (82.4) 9 (21) <0.0001 0.42 (0.41– 0.43)
Home-based care 68,960 (15.4) 5 (13) 21,410 (10.8) 5 (12) 47,550 (19.1) 5 (14) <0.0001 0.24 (0.23– 0.24)
Extended care 47,010 (10.5) 4 (10) 13,782 (7) 3 (9) 33,228 (13.4) 4 (10) <0.0001 0.21 (0.21– 0.22)
Open in a new tab
*

χ2 was used for all variables because we were comparing proportions (eg, the percentage of patients with any use of that outpatient service type).

†

SMD indicates standardized mean difference. These were calculated using an R package that includes a function that allows SMDs to be calculated for binary variables and does not require an assumption of normality for those variables (ie, stddiff.binary).

‡

MHICM indicates Mental Health Intensive Case Management.

§

PRRC indicates Psychosocial Rehabilitation and Recovery Center.

∥

PTSD indicates Post-traumatic Stress Disorder.

HUD-VASH indicates Housing and Urban Development-Veterans Affairs Supported Housing.

#

HCHV indicates Health Care for Homeless Veterans.

**

CIH indicates Complementary and Integrative Health.

††

CWT indicates Compensated Work Therapy.

‡‡

HVCES indicates Homeless Veteran Community Employment Services.

DISCUSSION

Since the inception of the Camden Health Care Coalition, interest in hot spotter programs to identify patients at risk for acute care utilization has grown substantially.26–28 Although there is limited evidence for the efficacy of hot spotter programs to reduce acute care service use,14,29 there may be potential for these initiatives to improve the quality of care for high-need patients, particularly in times of critical need. Through this study, we sought to inform these efforts by outlining the prevalence, patterns, and correlates of hot spotter status for housing-insecure patients in VHA, the largest provider of health care services for homeless adults in the US.16 Several of the findings have clinical and policy implications for this program and similar programs that target high-need homeless adults.

First, despite the high prevalence of mental health conditions in this cohort, medical hospitalizations were twice as common as mental health hospitalizations. This stands in contrast to the fact that substance and mental health-related conditions are typically the most common presenting diagnoses of homeless veterans seeking acute care.12 Similar to prior research in homeless adults, the prevalence of substance use and mental health conditions in the current cohort were high.30 Nevertheless, the pattern of hospitalization type highlights the scope of medical care management needed by this population. Medical respite care may play a valuable role in the postacute phase of medical hospitalizations. Also known as “recuperative care” and resembling short-term skilled nursing facilities, these programs provide 24/7 shelter and medical care to homeless persons who no longer meet criteria for hospitalization but are too ill be discharged to locations that are unsheltered or lacking nursing services.31 Partnerships between health care systems and nonprofits to create medical respite programs can provide a key service for homeless hot spotter patients and potentially reduce days of hospitalization and risk of readmission.32 Respite care programs may also be an ideal context to orient and train patients on telehealth protocols that use data from digital devices to facilitate monitoring of health status for common medical and mental health conditions in this population (eg, glucose levels for diabetes; self-reported mood symptoms; and suicidal ideation).33,34 The National Institute for Medical Respite Care estimates that there are 152 respite programs in the US but gaps remain in their availability in many areas due to limited funding and lack of community engagement.31

Second, consistent with nationwide trends as well as the high prevalence of chronic conditions and medical hospitalizations in the current sample,35 patients on VHA’s Homeless Hot Spotter Report are aging. Since the prior study publication of this program in 2016,3 the proportion of homeless veterans aged 65 years or older has doubled (13.8% vs. 27.0%). Further, individuals without stable housing have rates of chronic and geriatric conditions similar to adults in the general population who are 10–20 years older.36 With homelessness expected to triple in those aged 65 years or older over the next decade,37 it may behoove hot spotter programs to increase coordination and linkage to geriatric programs and/or tailor their services to the needs of older adults.38,39 For example, mobile medical units and similar outreach services may be critical for increasing access to care for an aging, and increasingly immobile, homeless population.40

Third, in general, hot spotter programs aim to increase patients access and engagement in ambulatory services to minimize their use of acute care. Notably, the present sample had high levels of acute care use despite being highly engaged in primary care and mental health care. This pattern, which has also been observed in other studies,5 may be attributable to the level of acuity of the patient population, difficulties managing their chronic conditions, and the general accessibility of EDs and other acute care services.5,41 By contrast, employment services, peer support, and justice programs were used by a minority of patients. Nonetheless, these services target social needs that are likely to be common in homeless patients identified by hot spotter programs (eg, unemployment, social isolation, and legal needs).42 More systematic screening of social needs and incorporation of this information into hot spotter initiatives may facilitate tailored linkage to services to directly address these needs. In VHA, this may involve administration of the Assessing Circumstances and Offering Resources for Needs screener of social risk and needs to all patients on the Hot Spotter Report.43

Fourth, though all patients may have been housing-insecure, most were not homeless per se during the cohort period. For some patients, this may reflect placement in permanent supportive housing, which was utilized by nearly 1 in 4 patients. The benefits of such programs are often driven by the case management services that accompany them and are likely to be most critical for repeat hot spotters.44,45 For other patients, housing may have been a time-limited need and acute care utilization may not be directly driven or exacerbated by homelessness per se. In either case, there was a significant degree of heterogeneity of housing status in the present cohort, which suggests value in increasing adoption of hot spotter initiatives for these patients outside of housing services. For example, though some VHA facilities have specialized primary care teams for homeless patients, the majority of these patients are on general primary care teams.46 Thus, implementation of the homeless hot spotter initiative in primary care more generally in VHA may be recommended.

Limitations and Future Directions

Several study limitations must be acknowledged. First, the findings describe a specific hot spotter initiative in VHA—a health care system in which insurance status is not a barrier to care access. Consequently, the findings may not fully generalize to hot spotter programs in other health care systems. This notwithstanding, some generalizability of the findings are warranted given that VHA is the largest provider of health care for homeless adults in the US.16 Second, legislation over the past decade has increased the use of both VHA-purchased and non-VHA care for veterans.47,48 However, acute care service use outside of VHA facilities is not captured on the Hot Spotter Report; therefore, our analysis may be underestimating total acute care use in the cohort. Future research should describe the prevalence and pattern of community-based acute and nonacute service use of Hot Spotter veterans. Third, the cohort timeframe coincided with the onset of the COVID-19 pandemic, which may have impacted the frequency of both acute and nonacute service use. However, prior research has demonstrated VHA’s ability to sustain primary care and mental health care engagement for patients during the early stages of the pandemic,49 including those with housing instability.50 Fourth, the present analysis was descriptive and did not test how service use engagement may have impacted change in Hot Spotter status over time; such analyses are a key direction for future research.

Among housing-insecure patients in VHA, nearly 1 in 3 met criteria for a Hot Spotter Report in at least one quarter from FY18–22, with most meeting criteria in ≥2 quarters. Though the majority of patients in the cohort had a substance use and/or mental health condition, the proportion of patients aged 65 years or older, the number of chronic medical conditions, and the rate of medical hospitalizations in the cohort highlight the complex medical care management needs of this population that may need to be prioritized in interventions offered to Hot Spotter patients. Future work should endeavor to identify patterns and types of care utilization that lower the risk of acute care utilization among patients in hot spotter programs and identify facilitators and barriers to the adoption and sustainment of these programs in health care systems.

Supplementary Material

mlr-64-207-s001.docx (62.6KB, docx)
Open in a new tab

Footnotes

This work was supported by a VA Research Career Scientist Award for Dr Blonigen (IK6HX003763) and intramural grant funding from the VA National Center for Homelessness Among Veterans (NCHAV). The views expressed here are the authors’ and do not necessarily represent those of NCHAV or the US Department of Veterans Affairs.

The authors declare no conflict of interest.

Supplemental Digital Content is available for this article. Direct URL citations are provided in the HTML and PDF versions of this article on the journal's website, www.lww-medicalcare.com.

Contributor Information

Daniel M. Blonigen, Email: daniel.blonigen@va.gov.

Kathryn S. Macia, Email: Kathryn.Macia2@va.gov.

Ivan Raikov, Email: iraikov@stanford.edu.

Jean Yoon, Email: Jean.Yoon@va.gov.

Jillian Weber, Email: Jillian.Weber@va.gov.

REFERENCES

Articles from Medical Care are provided here courtesy of Wolters Kluwer Health

RESOURCES