Abstract
Evaluating the performance of a health promotion program to lower the risk of cardiovascular disease requires comparison with a suitably selected reference. Typically, the reference includes easily accessible participants who declined or were not enrolled in the intervention program. Unfortunately, the selected reference often fails to be comparable for many of the risk factor attributes that might determine the success or failure of the intervention. In response, we developed a propensity score that results in a reference sample with features that are similar to those enrolled in the health promotion program.
Participants who received repeat physical examinations in Japan from 2008 to 2009 were identified as candidates for enrollment in an intensive health promotion program to lower the risk of cardiovascular disease (N = 33,009). Those who attended at least one session of the health promotion program were selected as the intervention group (N = 1,114). The remainder were selected as reference group I (N = 31,895). Among the latter, those with a high propensity for enrollment into the intervention group based on a logistic regression model were selected as reference group II (N = 3,008). A group III reference was similarly defined based on a linear regression propensity score (N = 2,992). Characteristics of the intervention and reference groups were compared.
In the intervention group, subjects were younger and less likely to smoke cigarettes or eat breakfast than reference group I (P<0.001). Declines in body weight, waist circumference, and blood pressure between the repeated examinations from 2008 to 2009 were significantly greater in the intervention versus reference group I. Other than age, differences between the intervention and reference groups II and III failed to persist.
In conclusion, identifying a reference group based on a propensity score results in a group of individuals with characteristics similar to those enrolled in a health promotion program. Comparison of the reference and intervention groups could result in improved assessments of the performance of health promotion programs.
