Baseline Housing Stability Positively Impacts the Effect of an Ambulatory Intensive Care Unit on Decreasing Hospitalization Rates

INTRODUCTION

Over the past decade, organizations have funded Intensive Primary Care (IPC) teams to improve chronic disease management and reduce healthcare costs for patients at risk of high acute care utilization. Ambulatory Intensive Care Units (A-ICUs) are a type of IPC, providing high-touch, coordinated care through multidisciplinary teams embedded in primary care clinics.¹ While some studies show promise that IPC interventions improve outcomes for high-need, high-cost (HNHC) patients, randomized controlled trials have shown limited effectiveness of IPCs in reducing acute care utilization, at least in the short term.^2–5

HNHC patients face a disproportionate burden of adverse social determinants of health (e.g., housing instability, substance use, poverty).⁶ Housing instability is linked to poor chronic disease management, high hospital readmission rates, and early mortality. Recent IPC trials have enrolled participants with unstable housing, but have not addressed the impact of housing instability on IPC effectiveness.^2–4 This study examines how baseline housing stability affects the outcomes of an A-ICU intervention in a healthcare for the homeless setting.

METHODS

This was an exploratory secondary analysis of the SUMMIT A-ICU (“SUMMIT”) randomized trial at a Federally Qualified Health Center (FQHC) in Portland, OR.⁴ SUMMIT provided medically and socially complex patients with coordinated, multidisciplinary care, and consisted of a physician, nurse, care coordinators, social workers, pharmacist, and team manager. Patients were eligible if they were 18 years or older, had one or more hospitalizations in the past 6 months, and had either two chronic medical conditions or one chronic medical condition and a mental health or substance use diagnosis. Participants were randomized to either SUMMIT or usual FQHC care.

We conducted baseline demographic surveys and follow-up surveys at 6-months post-randomization. Participants identified their current housing situation, which was dichotomized as “Stable” or “Unstable” based on literature and input from clinical staff. The primary outcome was change in hospitalization rate over 6 months, comparing pre-and post-randomization periods.

We used two sample t-tests and chi-square tests to compare baseline characteristics by housing status. To evaluate whether housing status modified the intervention’s effect, we fit a linear effects model with a three-way interaction between randomization arm, time, and housing status. We reported the mean change in hospitalization rates for each group, stratified by baseline housing status.

RESULTS

A total of 156 patients with complete data were included. Differences in demographic data between stably and unstably housed patients are reported in Table 1.

Table 1.

SUMMIT Trial Participant Characteristics by Baseline Housing Status

	Total	Stable housing	Unstable housing
	N = 156	N = 79	N = 77
Randomization arm
Enhanced usual care	78 (50.0)	43 (54.4)	35 (45.5)
SUMMIT	78 (50.0)	36 (45.6)	42 (54.5)
Female	53 (34.9)	29 (38.2)	24 (31.6)
Age†, mean (SD), years	55.2 (9.6)	56.9 (8.4)	53.3 (10.3)
Social support (ENRICHD score), mean (SD)*	18.9 (6.7)	18.0 (6.6)	19.8 (6.8)
Gross HH income last month	133 (85.8)	67 (84.8)	66 (86.8)
High school educ or less	95 (60.9)	44 (55.7)	51 (66.2)
Presence of opioid use disorder	43 (27.6)	22 (27.8)	21 (27.3)
Self-rated health, mean (SD)	5.3 (2.3)	5.4 (2.2)	5.2 (2.4)
Cognitive impairment† (TICS score < 20)*	62 (40.3)	25 (32.1)	37 (48.7)
Asian	3 (1.9)	0 (0.0)	3 (3.9)
Black/African American	20 (12.8)	9 (11.4)	11 (14.3)
Hispanic/Latino	5 (3.2)	1 (1.3)	4 (5.2)
Native American/Native Alaskan	20 (12.8)	9 (11.4)	11 (14.3)
Native Hawaiian/Pacific Islander	1 (0.6)	0 (0.0)	1 (1.3)
White	120 (76.9)	63 (79.7)	57 (74.0)
Other	5 (3.2)	3 (3.8)	2 (2.6)
Drug abuse screening (DAST-10 score)*
No problems reported	62 (39.7)	37 (46.8)	25 (32.5)
Low/moderate level	55 (35.3)	26 (32.9)	29 (37.7)
Substantial/severe level	39 (25.0)	16 (20.3)	23 (29.9)
Current alcohol problem† (AUDIT-10 score > 7)*	32 (20.6)	11 (14.1)	21 (27.3)
Presence of depression (PHQ-9 score > 9)*	82 (53.2)	39 (49.4)	43 (57.3)
Number of comorbidities, mean (SD)	5.1 (2.1)	5.3 (2.1)	4.9 (2.1)

^*AUDIT-10, 10-item Alcohol Use Disorders Identification Test; DAST-10, 10-item Drug Abuse Screening Test; ENRICHD, Enhancing Recovery in Coronary Heart Disease Patients Social Support Instrument; PHQ-9, 9-item Patient Health Questionnaire; TICS, Telephone Interview Cognitive Status

^†p < 0.05

In the SUMMIT group, stably housed participants experienced an average decrease in hospitalization rates of − 1.12 (95% CI, − 1.79, − 0.44). Unstably housed participants had a non-significant increase in hospitalization rates at 6 months (0.40; 95% CI, − 0.89, 1.69). The three-way interaction term for intervention group, housing status, and time was not significant at the 0.05 level (χ²₁ = 1.73; p = 0.19). However, among stably housed participants, there was a non-significant trend towards fewer hospitalizations in the SUMMIT group compared to the control group (− 0.71; 95% CI, − 1.58, 0.17), while unstably housed participants saw non-significant increases in hospitalizations (0.55; 95% CI, − 0.99, 2.09) (Fig. 1).

Figure 1.

Mean hospitalization rates by baseline housing and randomization arm.

DISCUSSION

Our findings suggest stable housing may be important for the success of IPC interventions like SUMMIT in reducing utilization. We found that stably housed patients in SUMMIT experienced decreases in hospitalization rates, but those with unstable housing did not. Housing instability poses barriers to healthcare engagement, including difficulty accessing appointments and adhering to treatment plans. These barriers may have prevented unstably housed patients from benefiting from SUMMIT activities with the potential to decrease hospitalizations.

These results contrast with those of the VA Homeless Patient Aligned Care Teams (HPACT), which incorporated tailored IPC services for homeless veterans (e.g., housing assistance, meals, transportation).⁷ HPACT decreased utilization rates, suggesting that IPC interventions enrolling unstably housed patients might benefit from incorporating services targeted to the unique priorities of those facing housing instability.

Unstably housed patients may benefit from SUMMIT’s outreach activities through early detection of health deterioration. It is possible that the increased outreach and coordination lead to a paradoxical increase of appropriate hospitalizations for unstably housed patients. While this was an exploratory secondary analysis and was not powered to detect a significant interaction, our findings suggest utility in disaggregating HNHC patients by housing status in future studies to better understand how participant housing status interacts with intervention outcomes.

Funding

This analysis was funded by Portland State University Homelessness Research in Action Collaborative Faculty Grant (Chan), AHRQ K12HS022981 (Chan), and NIDA K23 053390 (Chan). Dr. Edwards was additionally supported by VA HSR CDA 16-152.

Data Availability

The data that support the findings of this study are available from the corresponding author, A.G., upon reasonable request.

Declarations:

Conflict of Interest:

The authors declare that they do not have a conflict of interest.