Abstract
Background/Aims In an era of increasing openness in health plans, use of electronic medical records (EMR) for detailed chart abstraction for all enrollees poses challenges. At Group Health Cooperative (GH), EMR data are most complete for enrollees in the Group Practice Division (GPD). However, all GH enrollees (from both the GPD and the Contracted Network Division (CND) have EMRs. CND providers do not have access to patients’ GH EMR. Information for CND enrollees is only entered when the patient or provider interacts with the GPD, such as a patient receiving GPD specialty care while receiving CND primary care. In the interest of achieving a larger sample size for an ongoing GH-based case-control study (FOCUS), we evaluated the feasibility of including CND enrollees by determining the proportion with sufficient EMR information for fracture adjudication.
Methods The FOCUS study is adjudicating and recruiting incident osteoporotic fracture cases in women across the menopausal transition using the GPD. We used identical automated sampling methods to identify, via ICD-9 codes, a pool of potential adjudicable fracture cases among CND enrollees.
Results For the GPD, from a denominator of 58,328 age-eligible women, 2,310 (4%) were identified with fractures January 2008–March 2011. For the CND, among 38,041 age-eligible women, 1,040 (3%) were identified with fractures. Following automated exclusions, 944 CND potential cases remained. Initial review of these 944 EMRs found that 685 charts had no data; 54 had data unrelated to fracture; 145 had sufficient fracture-related data to adjudicate; and 60 had some fracture-related data. Preliminary chart adjudication of these last two groups found that 88% (127/145) of the “complete” EMR group and 10% (6/60) of the “partially complete” group had qualifying fractures.
Discussion About 23% of GH network enrollees with automated fracture codes had evaluable EMR data. Preliminary adjudication has identified 65% (133/205) as additional potential FOCUS cases to contact for interviews. It appears that, even for a study requiring sufficient chart information to adjudicate a health outcome, the CND can contribute cases for study inclusion. We also have developed network control selection methods. Still to be ascertained at the survey step are continued eligibility and study participation.
