Abstract
Background: It is often desirable to include enrollment and disenrollment (or, conversely, retention) information in grant applications. Overall disenrollment has been evaluated as part of Virtual Data Warehouse quality assurance work, but disenrollment by member characteristics has been incompletely characterized. This information is important in longitudinal outcomes research.
Aim: To calculate member disenrollment across HMORN sites by age, gender, and selected disease states.
Methods: Baseline membership cohorts were identified for 8 consecutive years (2000–2007) at 9 HMORN sites. Members were included in a baseline cohort year if they were enrolled for at least 305 days in that year. Disenrollment, defined as having fewer than 305 days of membership in a year, was assessed at 1, 3, and 5 years for each cohort. Members who died before the end of each assessment period were censored. Comorbidity was assigned based on ICD9 diagnosis code(s) for diabetes, asthma, heart failure, hypertension, hyperlipidemia or depression over a three-year period ending with the end of the cohort baseline year. SAS code was prepared at one site for distribution to other sites to run against VDW tables.
Results: While overall disenrollment varied across the participating sites, within-site disenrollment changed little from year to year. In general, 1-year disenrollment rates varied from 7% to > 22%; 5-year disenrollment rates ranged from < 30% to > 60%. Using one site as an example, peak 1-year disenrollment was seen among the 20–29 year old age group; lowest rates were seen among those age 60 years and older. Similar patterns were seen at 5 year follow-up, although differences between highest and lowest rates were smaller. Disenrollment rates were slightly higher among males. Disenrollment among members with any one of the measured diseases was lower than among members without that condition.
Conclusions: Disenrollment (and retention) rates vary by age and disease state and less so by gender. While overall disenrollment is useful for many purposes, rates by member characteristics can be essential while planning studies of specific populations.
- Received May 27, 2010.
- Accepted May 27, 2010.
