Pediatric carve outs: The use of disease-specific conditions as risk adjusters in capitated payment systems

John S. Andrews, Gerard F. Anderson, Carol Han, John M. Neff

Research output: Contribution to journalArticlepeer-review

22 Scopus citations


Objective: To address the issue of adverse selection in capitated payment systems by developing a list of disease-specific pediatric conditions (ie, 'carve outs') to be considered for separate reimbursement. Design: A descriptive study using a large Medicaid database. Intervention: With the use of fiscal year 1993 Washington State Medicaid cost data for 302 240 pediatric patients, a list of disease-specific carve outs was developed to meet the following criteria: high cost, low variability in cost, and association with a large proportion of medical spending. Results: Six-hundred seventy-three patients (0.2%) in the database had annual costs totaling $25 000 or more. Ten percent of the cases accounted for approximately two thirds of spending, while the least expensive 70% of cases made up only 15% of the expenditures. Prematurity and complications of prematurity, neoplasms, congenital heart disease, organ transplantations, congenital anomalies, and respiratory problems were general categories of disease that met our criteria for a carve out. The association of a major surgical procedure with a diagnosis increased the predictive accuracy for high cost. Conclusion: Lists of disease-specific carve outs such as this one can be used by Medicaid plans and other insurers who are concerned about risk selection to identify conditions for separate reimbursement in capitated payment systems.

Original languageEnglish (US)
Pages (from-to)236-242
Number of pages7
JournalArchives of Pediatrics and Adolescent Medicine
Issue number3
StatePublished - 1997

ASJC Scopus subject areas

  • Pediatrics, Perinatology, and Child Health


Dive into the research topics of 'Pediatric carve outs: The use of disease-specific conditions as risk adjusters in capitated payment systems'. Together they form a unique fingerprint.

Cite this