September 1, 2004
Volume 10, Issue 9 Abstract

Development and Application of a Computer Algorithm to Identify Epilepsy Cases in Managed Care Organizations

Nelson LS, Gunter, MJ, Robinson S


Maggie J. Gunter, PhD, Lovelace Clinic Foundation, 2309 Renard Place SE, Suite 103, Albuquerque, NM 87106. E-mail: maggie@lcfresearch.org.

Objective: To develop and apply computer algorithms to an administrative dataset to identify the prevalence and incidence of epilepsy and epilepsy-related mortality within a managed care organization.

Methods: The study consisted of exploratory and confirmatory phases, followed by application of the algorithm to estimate prevalence, incidence, and mortality. In the exploratory phase, potential epilepsy patients were identified based on epilepsy-related codes in administrative data; a random sample of charts were reviewed to confirm epilepsy cases. An algorithm was developed utilizing combinations of epilepsy-related diagnoses, procedures, and medications based on chart review results. In the confirmatory phase, the algorithm derived in the exploratory phase was then applied to a new dataset from the same MCO; a second confirmatory chart review was conducted. Further algorithm refinement was accomplished by applying logistic regression models to the combined chart review data from both phases. In the application phase, the final models were applied to 1-, 3-, and 5-year datasets to identify prevalent and incident cases that were then linked to a statewide death registry to derive mortality estimates.

Results: The best model used diagnoses and anti-epileptic drugs as predictors, had a positive predictive value of 84% (sensitivity, 82%, specificity 94%), and correctly classified 90% of the cases. Prevalence rates of 7—10/1000 across the 1-, 3-, and 5-year datasets, depending on age, gender, and ethnicity, were obtained. Annualized incidence for members continuously enrolled for 3 years was 47/100 000 and 71/100 000 for members continuously enrolled for 5 years. Crude mortality rates were 2-2.5 times higher for epilepsy patients identified with the best model compared to controls.

Conclusion: The algorithm developed in this project can be used to monitor trends in incidence, prevalence, and mortality to inform decisions critical to improving the healthcare and overall quality of life for epilepsy patients.

Impact of Differential IRB Processes in a Multisite Study

Greene SM, Geiger AM, for the Cancer Research Network PROTECTS Project


Sarah Greene, MPH, Group Health Cooperative-Center for Health Studies, 1730 Minor Ave, Suite 1600, Seattle, WA 98105.E-mail: greene.sm@ghc.org.

Background: In multisite studies, including the Cancer Research Network (CRN), every participating site is obliged to obtain IRB approval to conduct the project. Variability in IRB review considerations, coupled with complex regulations that may also vary by institution, result in multiple, iterative reviews that can potentially compromise the consistency of study methodology. Reconciling this process, and meeting all IRBs' requirements is time- and resource-intensive, jeopardizing established study timelines and budget.

Objective: To profile IRB approval process among 6 CRN sites participating in a study assessing long-term outcomes and satisfaction with the decision to have a prophylactic mastectomy.

Methods: Specific aims of the original study were to assess long-term outcomes and satisfaction with the decision to have a prophylactic mastectomy, among women who had undergone this procedure between 1980-1988. Data were collected via a 7-page cross-sectional survey. Materials reviewed by the IRB included the questionnaire, introductory letter, reminder letter, and script used for telephone reminders to participants to complete the survey. At 5 sites, an external survey firm handled data collection. Data on IRB processes included total time from initial submission to final approval, number of iterations of IRB review for each site, nature of review, and any unique stipulations.

Results: Multiple review cycles were necessary at all sites. Although 3 months were built into the original study timeline to accommodate an iterative review process, completing IRB review at all sites ultimately took twice that long. In this period, investigators tackled Health Insurance Portability and Accountability Act regulations, requests from some IRBs to add a physician consent process, and a requirement by 1 IRB that the site notify prospective participants that the outside survey firm would be contacting them. Often, 1 IRB's stipulations necessitated repeated resubmissions of materials at other sites.

Institutional Review Board review is an essential aspect of research, but one that presents substantial challenges to multisite studies. Scientists, funders, and policy-makers should be cognizant of these challenges and consider developing strategies to facilitate IRB review while maintaining scientific rigor and adequate human subjects protections.

Discussion: In this minimal risk survey of women's outcomes and attitudes following prophylactic mastectomy, with no physical or behavioral intervention, numerous human subjects complexities were encountered, largely due to the multisite nature of the study. This experience underscores the need to examine approaches that collaborative projects and IRBs could take to implement more efficient and standardized processes.