Regression analysis of health care utilization by respiratory symptoms. (A) Model outcome: all cause health care utilization. (B) SCD-related acute health care utilization. (C) VOE health care utilization. (D) ACS acute care utilization. Estimates are exponentiated coefficients (eg, RR, due to negative binomial regression). Final model includes M3 + fixed effects for year and site (state) of care. M1, unadjusted model (respiratory phenotypes as only covariate); M2, M1 + demographic variables (age, sex, and race); M3, M2 + SDH proxy variables (home median income quartile and home urbanization). RR, relative risk; VOE, vaso-occlusive event.