Dissertation Title:  "Evaluating the Effects of Two Successive Policies Regarding Obstetrical Lengths of Stay"

This dissertation evaluates effects of two policies affecting a large health maintenance organization (HMO) population: an "early discharge" program featuring one hospital overnight plus a nurse home visit after uncomplicated vaginal delivery; and, subsequent Massachusetts law guaranteeing 48 hours of hospital coverage. Retrospective data on 20,366 normal vaginal births between October 1990 and March 1998 spanned both interventions. Outcome measures were aggregated into quarterly intervals, and changes in outcomes coincident with the policies were estimated using interrupted time series analysis.

In the first of three papers, we examine breastfeeding initiation and continuation. As these key outcomes are not readily available in administrative datasets, we developed a text-search algorithm to determine feeding practice during infants' first 90 days from automated medical records. We found that despite large, expected shifts in the percentage of mother-infant pairs having shorter stays, there were no changes due to either policy in the gradual upward trend in breastfeeding initiation or the constant rate of continuation among initiators.

The second paper describes post-discharge newborn service utilization, plus financial impacts on the HMO. Program-related home visits contributed to an increase of about 260% in the proportion of newborns receiving any clinical evaluation on Day 3 or 4. The corresponding relative drop after the law was 18%  less substantial, because home visits were guaranteed for stays <48 hours. Many experts consider follow-up on Day 3 or 4 more crucial than discharge timing. Routine health center visits increased with program implementation, but there were no significant changes in adverse events such as emergency room visits or rehospitalizations. Neither policy altered total HMO payments for hospitalizations and home care by more than $100 per delivery. The impact on HMO expenditures was surprisingly small, because of hospital price changes in apparent response to the HMO program.

Our final paper describes increases in various measures of newborn jaundice and feeding difficulties at the start of the HMO program. We conclude that the elevated measures are not attributable to earlier discharge, because they continued during the post-mandate period. Ascertainment bias best explains these findings, and is likely due to increases in follow-up care.