OBJECTIVES: To describe how household surveys can be used to assess the effects of the new State Children's Health insurance Program (SCHIP) , review methodologic issues associated with household survey data, and propose solutions for dealing with these issues. PRINCIPAL FINDINGS: To estimate the effect of SCHIP, analysis must explicitly recognize and control for the fact that other factors that could affect the outcomes of interest besides the new program will change over the analysis period. In assessing SCHIP's effect, SCHIP-eligible children must be identified using a detailed simulation model. Analyses that use either a simple eligibility model or only examine children with incomes between 100 and 200 percent of poverty will not accurately identify SCHIP-eligible children. Under these circumstances estimates of the effect of SCHIP will be biased downward. In addition analyses must rely on the same survey in the pre- and post- SCHIP periods to obtain reliable estimates. Moreover, the survey must attempt to obtain data on separate SCHIP programs, and analysts must consider the implications of the possible increasing underreporting of public health insurance coverage. Finally, analysts should be cautious about evaluating SCHIP's success before the program is mature. CONCLUSION: While evaluating SCHIP using household surveys has some challenges, if conducted carefully such analyses will provide important in formation on the effect of the SCHIP program that can not be obtained elsewhere.
OBJECTIVES: To describe how household surveys can be used to assess the effects of the new State Children's Health insurance Program (SCHIP) , review methodologic issues associated with household survey data, and propose solutions for dealing with these issues. PRINCIPAL FINDINGS: To estimate the effect of SCHIP, analysis must explicitly recognize and control for the fact that other factors that could affect the outcomes of interest besides the new program will change over the analysis period. In assessing SCHIP's effect, SCHIP-eligible children must be identified using a detailed simulation model. Analyses that use either a simple eligibility model or only examine children with incomes between 100 and 200 percent of poverty will not accurately identify SCHIP-eligible children. Under these circumstances estimates of the effect of SCHIP will be biased downward. In addition analyses must rely on the same survey in the pre- and post- SCHIP periods to obtain reliable estimates. Moreover, the survey must attempt to obtain data on separate SCHIP programs, and analysts must consider the implications of the possible increasing underreporting of public health insurance coverage. Finally, analysts should be cautious about evaluating SCHIP's success before the program is mature. CONCLUSION: While evaluating SCHIP using household surveys has some challenges, if conducted carefully such analyses will provide important in formation on the effect of the SCHIP program that can not be obtained elsewhere.