Dealing with endogeneity in non-randomized medical studies: a study of acute kidney injury following cardiopulmonary bypass surgery

Stewart R. Miller, Richard Gretz, Joshua Walker, Mitchell Katona, Christian Jacobsen, Nitin A Das, John Calhoon, Edward Sako

Research output: Contribution to journalArticlepeer-review

Abstract

Many medical studies have used non-randomized sampling, which tends to be the case in research that involves a surgical procedure. Intra-operative procedures and actions conducted by the attending surgical team may be based on pre-operative conditions of the patient, which can introduce endogeneity. Using acute kidney injury (AKI) following cardiopulmonary bypass surgery as a research setting, the present study uses a control function approach to explain why two perfusionist-directed principle components of delivered oxygen (DO2) while the patient is on pump—hemoglobin and cardiac index—need to be treated as endogenous variables rather than exogenous ones. We further show conditions in which the exogenous model understates and overstates the predicted probability of AKI.

Original languageEnglish (US)
JournalHealth Services and Outcomes Research Methodology
DOIs
StateAccepted/In press - 2022

Keywords

  • Acute kidney injury
  • Cardiopulmonary bypass surgery
  • Control function approach
  • Endogeneity

ASJC Scopus subject areas

  • Health Policy
  • Public Health, Environmental and Occupational Health

Fingerprint

Dive into the research topics of 'Dealing with endogeneity in non-randomized medical studies: a study of acute kidney injury following cardiopulmonary bypass surgery'. Together they form a unique fingerprint.

Cite this