Table 2.

Extended Cox regression analysis estimating the hazard associated with homelessness^a for readmission to a large urban hospital in Vancouver, British Columbia, 2016–2020 (n = 3,907)

Follow-up time	Homeless status	Total events	Total PDs	Incidence per PY	Unadjusted HRb (95% CIc)	Adjusted HRd (95% CI)
≤ 30 days	No Yes	340 149	91,713 19,500	1.35 2.79	Reference 2.05 (1.67, 2.53)	Reference 2.04 (1.65, 2.51)
31–90 days	No Yes	237 85	181,213 39,188	0.48 0.79	Reference 1.66 (1.25, 2.21)	Reference 1.65 (1.24, 2.19)
91–365 days	No Yes	504 179	756,492 170,294	0.24 0.38	Reference 1.58 (1.28, 1.95)	Reference 1.56 (1.26, 1.94)
Overall	No Yes	1,081 413	1,029,418 228,982	0.38 0.66

CI Confidence Interval, HR Hazard Ratio, PDs Person-Days, PY Person-Year

^aHomelessness was used as time-varying covariate

^bThis cox model includes homelessness and the interaction terms with time (at 30 & 90 days)

^c95% CIs and both unadjusted and adjusted hazard ratios were estimated using Robust Standard Errors

^dThe multivariable Cox model was controlled for age at index visit (continuous measure), gender (men, women & unknown/other); length of stay at index visit (continuous measure), substance use disorder (no vs. yes) as reason of stay at index visit and severe mental illness (either schizophrenia or bipolar disorder) (no vs. yes) as reason of stay at index visit