The 1980S and 1990s have seen a rapid diffusion of neonatal intensive care units (NICUs) into community hospitals, and an associated shift of high-risk cases from regional NICUs to these community hospitals. Most of these newer NICUs are smaller, intermediate level NICUs that do not provide a full range of care for critically ill infants. While previous studies have found lower mortality for births that occur at hospitals with regional (level III) NICUs that can provide the full range of intensive care services for infants, much remains to be learned about these differences. Specifically, none of these studies have examined how the relative performance of different levels of NICUs is changing over time and most of the previous studies did not control for clinical risks. This time period has also seen rapid growth in health maintenance organizations (HMOs) and other forms of managed care, and increasing financial pressures on hospitals. But there are no studies that examine how hospital competition, financial pressures on hospitals, or managed care effect NICU growth, care for high-risk newborns, or neonatal outcomes. In previous research we have begun to develop models that control much more fully for clinical risks using a single year of data (Phibbs, 1996). The most striking findings were: 1) births in hospitals with intermediate and expanded intermediate level NICUs, regardless of patient volume, had mortality that was no different than births at hospitals without NICUs; and 2) patient insured by HMOs and Medicaid had higher mortality. But, this study was based on data from only one year (1990), from a single state (California), and it provides no information on how these differences are changing over time. How these effects are changing (or not changing) over time has very important policy implications. If these differences persist, or are growing larger, then policy interventions may be necessary since the number of potentially preventable deaths would be very large. Conversely, if these differences are steadily diminishing over time, then disruptive policy interventions may be unnecessary. This proposal will build on the analyses we have already done using the California linkage of the birth certificates with the discharge abstracts for the mothers and infants, extending the linked data to the years 1983-1998. It will also include data from other states (Florida, New York, Washing, and Wisconsin) for varying time periods. This large panel of data will allow a careful examination of how the diffusion of NICUs and the growth of managed care have effected neonatal mortality. We will use logistic regression models, including models that account for the time series nature of the data to examine the association between levels of NICU care, market forces, and neonatal mortality, controlling for the wide range of risk factors that are available in the linked data.
