We evaluated the effect of a disease management (DM) program on adherence with recommended laboratory tests, health outcomes, and health care expenditures for patients with type 2 diabetes. The study was a natural experiment in a primary care setting in which the intervention was available to 1 group and then compared to the experience of a matched control group. Univariate analysis and difference in differences analysis were used to test for any significant differences between the 2 groups following a 12-month intervention period. A payer perspective was used to estimate the health care cost consequences based on hospital and physician utilization weighted by Medicare prices. The results were nonsignificant at the .10 level, except for compliance with recommended tests, which showed significant results in the univariate analysis. The intervention increased compliance with testing for HbA1c, microalbuminuria, and lipids, and decreased HbA1c value and the percent of patients with HbA1c >or=9.5%. The point estimates showed small reductions in health care cost; only reductions in costs for office visits were significant at the .10 level. We concluded that while there were signs of improvement in adherence to testing, the low effectiveness may be attributed to existing diabetes management activities in this primary care setting, high compliance rates for testing at the beginning of the study, and a steep learning curve for this complex, information-technology-based DM system. The study raises questions about the incremental gains from complex systems approaches to DM and illustrates a rigorous method to assess DM programs under "real-world" conditions, with control for possible selection bias.