To the Editor.
—The study by Dr Rosenthal and colleagues' presents the results of risk-adjusted mortality models. Their case inclusion rules for skilled nursing facility (SNF) patients are a potentially serious source of bias in comparisons of outcomes by their 3 hospital types. Of their original 89 851 cases, 10316 were admitted from SNFs, but 8421 of these original cases discharged to a SNF were omitted from subsequent analyses. One would expect that nearly all patients admitted from a SNF who survived their acute hospital stay would have been discharged to a SNF. Thus, the elimination of all cases discharged to a SNF virtually ensures biases in the hospital mortality data for SNF admissions. If patients died during their index hospital admission, they were included, but if they survived this hospital admission and were discharged to a SNF, they were excluded. The least adverse impact on mortality outcomes from this