Using Complications to Evaluate Neonatal Health Care: Controlling for Censoring by Death

• Activity Number: 283
• Type: Invited
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Committee on Applied Statisticians
• Author(s): Dylan S Small*+ and Fan Yang and Jing Cheng and Scott Lorch
• Companies: University of Pennsylvania and University of Pennsylvania and University of California, San Francisco and Children's Hospital of Philadelphia
• Keywords: causal inference ; censoring by death

Abstract

The Institute of Medicine has suggested the use of complication rates as a measure of the quality of care provided by an institution. We consider the use of complication rates as a measure of the quality of care provided by a neonatal intensive care unit (NICU). Two difficulties with using complications are: (1) censoring by death - many complications in neonatal care are not assessed until a newborn is at least 36 weeks old (e.g., bronchopulmonary dysplasia ) and consequently such complications are censored by death if a newborn dies before 36 weeks old; (2) substantial differences on measured and unmeasured patient risk factors between different NICUs. We develop a novel approach to control for censoring by death and unmeasured patient risk factors. Our approach to controlling for censoring by death makes use of multiple possible complications and the fact that there are multiple causal pathways along which a newborn might die, e.g., a child who dies from an intraventricular hemorrhage would not necessarily have later developed bronchopulmonary dysplasia if she had survived. Our approach to controlling for unmeasured patient risk factors makes use of instrumental variables.