Comparing different definitions of chronic conditions in a national data set

Objectives. - To compare the application of 4 different definitions of chronic conditions and disabilities in children using a large national database. Methods. - The study used data on children 0-17 years of age from the 1994 and 1995 National Health Interview Survey (NHIS) Core Interview and Disability Supplement. We selected items to create algorithms that operationalized 4 extant definitions and compared the proportions of children identified by each. For each definition we conducted bivariate and logistic regression analyses of age and sociodemographic and health-related characteristics potentially related to presence of disability. We then compared the results across definitions to determine if they were equally likely to identify children across a full range of sociodemographic variables, ages, and health-related characteristics. We also examined the level of agreement among definitions and determined the degree to which they identified the same or different children. Results. - Using a single data set, the 4 operational definitions identified somewhat different proportions of children as having chronic conditions or disability overall (the range was 13.7%-17%). However, the ways in which the definitions identified children along a spectrum of variables did not differ substantially across the definitions. The percent agreement in identification between any two individual definitions was ≥95%. In general, a higher proportion of children were identified among males, whites, those in older age groups, and among those whose parents had lower levels of education and those who were living below the poverty level. Although the same set of variables identified children with conditions who fit all 4 definitions compared with children who met the criteria for 1 to 3 definitions, higher rather than lower parent education was associated with agreement across all definitions. Conclusions. - Little has been known about the comparability of different definitions for identifying children with chronic conditions, special health care needs, and disability. Our findings suggest substantial overlap in the numbers and characteristics of the children that several different conceptual definitions identify. It remains unclear whether subgroups with higher or lower proportions of children with chronic disorders represent a substantive finding or whether they result from methodological problems shared by the ways in which the definitions are operationalized.