Association for Public Policy Analysis & Management

The U.S. is engaged in a number of payment reform activities that are designed to reduce health care costs. The most ambitious of these programs is the Maryland All-Payer Model which prospectively sets global budgets for hospitals in the state. The program, launched in 2014, was designed to incentivize hospitals to replace hospital services with lower-cost population health investments. While it is hoped that hospitals will substitute unnecessary high-cost care for lower-cost alternatives that produce equal or superior health outcomes, the program may inadvertently reduce services in ways that decrease health. This paper examines program effects on neonatal intensive care utilization and infant mortality. Neonatal Intensive Care Units (NICU) are expensive, but highly effective for infants that need them. However, previous research suggests that not all infants treated in the NICU require it.

We obtained NICU admission status, birthweight, infant mortality, and other individual characteristics from the restricted-use 2011-2017 Natality and Linked Birth-Death Files. These data were merged with state-by-year characteristics from the American Community Survey and the Area Health Resources Files. Outcomes were NICU admission and the infant mortality rate. Effects were estimated using a difference-in-differences model that included state fixed effects, year fixed effects, and other individual and state-by-year controls. The post period started in 2015, the first full year of program implementation. Control states included 20 states and DC that collected NICU admission on the birth certificate since 2011 and adopted the ACA Medicaid Expansion as Maryland did. Statistical inference was obtained from a bootstrap method that accounted for serial correlation in the presence of a single treated cluster.

Effects were estimated for all infants and for subgroups defined by birthweight. Most very-low birthweight (<1,500g) infants are likely to need NICU services. However, fewer low (1,500-2,499g) and normal birthweight (2,500-3,999g) infants are likely to have clear indications for NICU care and NICU utilization for such infants may be more sensitive to financial incentives.

NICU admission rates increased at the same rate in Maryland and control states prior to program implementation, suggesting that assumptions of the study design were met. Difference-in-differences results suggested that the program led to a 17.1% decline in NICU admissions. Effects were large and statistically significant for the low and normal birthweight groups (12.1% and 23.4% decline, respectively) while small and not significant for the very-low and high birthweight groups. There was no significant change in the infant mortality rate. Results were robust to excluding early and late Medicaid expansion states, to the set of covariates, to measuring time on a fiscal year basis or using 2014 as the cut-off year. We also come to similar conclusions using a synthetic control model that relies on a different set of assumptions.

Our results suggest that Maryland’s global budget program decreased NICU admissions and that program effects were concentrated among infants that had less indication of clinical need. Results appear to have been achieved without adverse health outcomes. Our results complement a growing literature that evaluates Maryland’s program, which has so far come to mixed conclusions.