Disease management is believed to hold promise for improving health in populations with chronic conditions like asthma, but little evidence exists about its cost-effectiveness. This study's objective was to determine whether computer-based prediction models could potentially improve the cost-effectiveness of asthma disease management. In this retrospective cohort design, computerized hospital, emergency department, and pharmacy data were used to identify children aged 0-14 with asthma-related utilization in a regional health maintenance organization. Demographic, medication and health services utilization variables from a 6-month baseline period were used to predict hospitalization during the following 12 months.

Among the 16,520 children in the study cohort, 1.8% were hospitalized during the follow-up year. In the proportional hazards model, having filled an oral steroid prescription (RR 1.9, 95% CI 1.3-2.8) or having been hospitalized(RR 1.7, 95% CI 1.1-2.7) during the prior 6 months, and not having an identified personal physician (RR 1.6, 95% CI 1.1-2.3) were associated with future hospitalization. A cutoff drawn at the 92nd percentile of predicted risk based on the model's coeffecients would have a sensitivity of 36% and a specificity of 92% to identify patients who were subsequently hospitalized.

The cost-effectiveness analysis used a hypothetical intervention, an asthma education program with a cost of $170 and effectiveness of 50% to prevent future hospitalization. If no prediction model were used, the projected net cost per hospitalization prevented would be $17,000. If a prediction model with a cutoff drawn at the 92nd percentile were used to select patients for intervention, the cost per hospitalization prevented would be $1,600. If the cutoff were drawn at the 97th percentile, cost savings were projected.

We conclude that prediction models which use computerized utilization data will be crucial for pediatric asthma disease management to be cost-effective. Interventions should be directed toward the highest-risk patients for disease management to have reasonable cost-effectiveness.