Cox regression models for intermediate events, with discharge from hospital as an example

Epidemiology. 1993 Mar;4(2):120-7. doi: 10.1097/00001648-199303000-00007.

Abstract

In studies of mortality or morbidity of hospitalized patients, discharge from hospital is an intermediate event between hospital experience and disease outcomes, as disease onset may occur after release from hospital. This study explored the role that discharge might have in risk for surgical infections after hernia repair operations, where follow-up continued for 1 month after operation, and 50% of infections occurred at home. Possible direct and interactive effects were evaluated in the presence of two major methodologic difficulties: waiting-time bias, because patients became candidates for home infections only after leaving the hospital, and selective discharge bias, because discharge carried much prognostic information. It was possible, using Cox models, to correct for the waiting-time bias, but the strong protective effect of termination of hospitalization on the risk for infection remained difficult to interpret. The strengths and limitations of various Cox models in dealing with these issues are discussed.

Publication types

  • Multicenter Study

MeSH terms

  • Adult
  • Cohort Studies
  • Female
  • Follow-Up Studies
  • Hernia / epidemiology
  • Herniorrhaphy
  • Hospitalization
  • Humans
  • Israel / epidemiology
  • Male
  • Middle Aged
  • Patient Discharge / statistics & numerical data*
  • Proportional Hazards Models
  • Regression Analysis*
  • Risk Factors
  • Surgical Wound Infection / epidemiology