Reparameterizing Time to Adjust for Longitudinal Effects

• Activity Number: 18
• Type: Topic Contributed
• Date/Time: Sunday, July 31, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Teaching of Statistics in the Health Sciences
• Author(s): Louis Fogg*+ and Louis Fogg
• Companies: Rush University and Rush University
• Keywords: longitudinal data ; nursing research ; length of stay

Abstract

It is important how long a patient actually stays in the hospital. This is referred to as a patient's length of stay (LOS). But varying LOS creates problems for a researcher interested in looking at longitudinal events associated with hospital stay. The most popular way to address this issue is with survival analysis. But in this case, that didn't work. We were interested in finding out if mothers in the hospital tended to forego breastfeeding before they were discharged. The solution was to perform a Back to the Future (BTTF) analysis. This involves reparameterizing days in hospital so that the last day is always 1 and then one counts backwards from there. The analysis has two steps. The first is graphing the data so that the BTTF structure is clear. The second is to use the BTTF parameterization to generate new parameter estimates. Both steps were used with the breastfeeding data, with the results showing a clear tendency for mothers to cut back on breastfeeding in the two-week period before their baby was discharged from the hospital. The technique has implications for discharge planning, and estimating how care might change for patients near the end of their lives.