Interestingly, the kind of interventions that are modeled here include better access to nutritious food, more supervised recreational opportunities for kids, better nutritional labeling and higher tobacco taxes. Note these are all interventions aimed at the entire community, not at individual patients. None of these interventions lead to filing of medical claims.
Critical assumption made for modeling is that these interventions "could reduce uncomplicated diabetes and high blood pressure rates by 5 percent in one to 2 years; heart, stroke, and kidney disease by 5 percent within 5 years, and cancer, arthritis, and COPD by 2.5 percent within 10 to 20 years." The authors state that their modeling is conservative, since only medical savings are counted. However, this depends upon the accuracy of forecasts of program costs and program impact.
Note that the ROI for states is low in the first two years, and of course states are not feeling flush with cash at the moment.
I'm grateful to Jim Pfeffer for pointing this out. Jim scours the quality literature for a weekly post available here.