An application of survival analysis to population health management program evaluation

Evaluating the financial impact of long-term lifestyle management interventions is difficult because of the variability across the program members' enrollment and participation levels. This variability is often referred to as "censored" health claims data. Survival analysis is designed to overcome censored data of various kinds (e.g., disenrollment, loss to follow-up, missing data).

A quasi-experimental program versus matched reference group study design was used. Participants (N = 142) in smoking cessation or weight management programs in 1997 were retrospectively matched to nonparticipants (N = 142).

The study location was a regional hospital in Tennessee.

Hospital employees with health plan benefits served as subjects.

Program interventions included health risk appraisals, screenings, and telephonic health coaching for weight management and smoking cessation.

The major dependent measure was paid claims information.

Program participants experienced more health care claims but significantly lower cost than nonparticipants. Nonparticipants averaged $40.77 more per claim than participants. Participants were less likely than nonparticipants to experience a high-cost or outlier claim event. For example, by month 72 nonparticipants had twice the probability of experiencing an outlier claim as participants.

Effective lifestyle health management may increase low-cost health claims activity, because many patients improve at self-care. As demonstrated here, preventive initiatives result in lower cumulative costs and decrease the risk of high-cost or outlier claim events.