To the Editor We would like to draw your attention to 2 interesting statistical points in regards to the Original Investigation “Comparison of Hospital Mortality and Readmission Rates for Medicare Patients Treated by Male vs Female Physicians” by Tsugawa et al1 in a recent issue of JAMA Internal Medicine.
First, the main outcome measures—patient 30-day mortality and readmission rates—are frequently used measures of quality of care. However, when 30-day mortality is treated as a binary variable (as the authors have done), the implication is that all patients were at equal risk of dying during the 30-day after admission. That is unlikely to be accurate and may violate the basic assumptions of logistic regression, namely equal risk of outcome (here mortality) among those exposed for the entire risk period.2 Perhaps a time-to-event analyses to evaluate the actual differences in risk of mortality for patients treated by male and female physicians may provide a more accurate picture.3
Chowdhury R, Payal AR. Limitations Concerning the Association of Physician Sex and Patient Outcomes. JAMA Intern Med. 2017;177(7):1057-1058. doi:10.1001/jamainternmed.2017.2139