Background: Socioeconomic status (SES) has been shown to have an association with patient health outcomes, including risk of hospital readmission. However, most SES data involve self-reported or aggregated information rather than objective individual measures. HOUSES provides an objective individual metric based upon real property data collected from county assessors that has been validated in other studies. We propose that lower individual socioeconomic status as measured by HOUSES is associated with increased likelihood of 30-day hospital readmission.
Methods: Subjects were 13,034 adult primary care patients from a Patient-Centered Medical Home practice initially admitted to the hospital between January 1, 2011 and December 31, 2013. Individual SES was measured utilizing the HOUSES-index (an individual housing-based index, derived from the subject’s address information, presented as a standardized z-score) at the time of index hospital admission. HOUSES-index is based on housing information available from public record. This includes the value of housing unit, number of bedrooms, and number of bathrooms. The z-scores derived from the HOUSES-index were divided into quartiles. Primary endpoint was a composite of hospital readmission or death within 30 days of initial hospital admission.
Results: 97% (12,588 of 13,034) of eligible subject addresses were successfully geocoded and matched to the HOUSES database. As noted above, subjects were separated into four quartiles based on their HOUSES-index score. Of the 12,588 patients matched to the HOUSES database and assigned a z-score, there were 3,671 (29.16%) in the first quartile (H1, lowest SES), 3,493 (27.75%) in the second quartile (H2), 3,052 (24.25%) in the third quartile (H3), and 2,372 (18.84%) in the fourth quartile (H4, highest SES). There were 26,279 total hospital admissions. Initial bivariate analysis showed that higher SES was associated with a lower likelihood of 30-day hospital readmission or death. However, when multivariate analysis was performed with other known factors affecting likelihood of hospital readmission (Charlson comorbidity index, total number of hospitalizations, etc.), this association no longer existed. Multivariate odds ratios comparing likelihood of 30-day readmission or death for H1 vs H2 was 1.00, H1 vs H3 was 1.01, and H1 vs H4 was 0.99.
Conclusions: Multivariate analysis, which took into account other known factors that increase likelihood of hospital readmission, showed no association between lower SES as measured by HOUSES and likelihood of hospital readmission or death within 30 days of index hospitalization. Our initial hypothesis, therefore, was disproven in this study. However, lower SES as measured by HOUSES is associated with factors typically thought of as increasing likelihood of hospital readmission, such as total number of hospitalizations and total number of emergency department visits. Additional research in this area is needed, as HOUSES-index may be a clinically useful tool to measure SES from information obtained objectively.