Differential associations of homelessness with emergency department visits and hospitalizations by race, ethnicity, and gender

Abstract

Objective: To investigate the differential associations of homelessness with emergency department (ED) visits and hospitalizations by race, ethnicity, and gender.

Data sources: California Medicaid enrollment and claims.

Study design: We identified beneficiaries experiencing homelessness (BEH) and those who did not (NBEH) using diagnosis and place of service codes and residential addresses. Outcomes include four ED visit measures and four hospitalization measures. We compared the use of these services by BEH to NBEH overall and by race, ethnicity, and gender groups in regression models controlling for covariates.

Data collection: We used a sample of Medicaid beneficiaries who met eligibility criteria for a California Medicaid demonstration program in 2017 and 2018 but were not enrolled in the program. We identified 473,069 BEH, and the rest (1,948,422) were considered NBEH. We used the 2018 data for utilization analyses and most covariates. We constructed lagged measures of health conditions using 2017 data.

Principal findings: We found that homelessness was significantly associated with 0.34 more ED visits (p < 0.01) and a higher likelihood of frequent ED visits (2.77 percentage points [pp], p < 0.01), any ED visits due to mental health conditions (0.79 pp, p < 0.01), and any ED visits due to substance use disorders (1.47 pp, p < 0.01). Experiencing homelessness was also significantly associated with 0.03 more hospitalizations (p < 0.01), a higher likelihood of frequent hospitalizations (0.68 pp, p < 0.01) and high frequent hospitalizations (0.28 pp, p < 0.01), and a longer length of stay (0.53 days, p < 0.01). We found a larger association for American Indian and Alaska Native, Black, Native Hawaii or Pacific Islander, and White populations than that for Asian and Hispanic populations. The associations are larger for males than females.

Conclusions: Our findings identified distinct utilization patterns by race, ethnicity, and gender. They indicated the need for developing race, ethnicity, and gender-specific strategies to reduce ED visits and hospitalizations of BEH.

Keywords: emergency department visits; ethnicity; gender; health equity; homelessness; hospitalizations; race.

FIGURE 1

Overview of the two‐stage analytic process to study the associations of homelessness with emergency department visits and hospitalizations. WPC denotes California's Whole Person Care (WPC) Pilot program. MC‐SIMEX denotes the misclassification simulation extrapolation approach. “E‐value” was developed by VanderWeele and Ding (2017).

FIGURE 2

Disparities in emergency department visits and hospitalizations among beneficiaries experiencing homelessness by race, ethnicity, and gender groups. (A) Emergency department visits. (B) Hospitalizations. Shown are unadjusted health care utilization measures for beneficiaries experiencing homelessness (N = 473,069) by race and ethnicity‐gender groups. ED denotes Emergency Department. AIAN denotes American Indian and Alaska Native. NHOPI denotes Native Hawaiian and other Pacific Islander. Tables A2 and A3 describe ED visits and hospitalizations by homelessness status for all race, ethnicity, and gender groups, including those who identified their race/ethnicity as “Other or Multiracial,” and “Unknown.” Race and ethnicity categories are listed in alphabetical order.

FIGURE 3

Differential associations of homelessness with emergency department visits and hospitalizations by race, ethnicity, and gender groups. (A) Emergency Department Visits. (B) Hospitalizations. Shown are marginal effects calculated after regression models with a three‐way interaction term for race and ethnicity, gender, and homelessness indicator (N = 2,421,491). Negative binomial regressions were used for count outcomes, while logistic regressions were used for binary outcomes. ED denotes emergency department. pp denotes percentage points. AIAN denotes American Indian and Alaska Native. NHOPI denotes Native Hawaiian and Pacific Islander. Numbers are available in Tables A4 and A5, including those who identified their race and ethnicity as “Other or Multiracial,” and “Unknown.” Statistical significances on pairwise comparisons of marginal effects after Bonferroni correction are consistent with the nonoverlapping of 95% confidence intervals shown in the above figures. Race and ethnicity categories are listed in alphabetical order. [Color figure can be viewed at wileyonlinelibrary.com]