Investigating Medical Patient Records and Administrative Data in Case Identification and Treatment (IMPACT)BackgroundThe purpose of this study is to identify the existence and extent of biases associated with HMO full electronic and claims-type encounter data when they are used to characterize patterns of care and to analyze the relationship between treatment and outcomes of breast and cervical cancers. Electronic data assessed will include inpatient and outpatient visits, pharmacy, cancer registry, pathology, radiology, laboratory, and electronic physician notes. This project is a joint venture between four members of the HMO Cancer Research Network: Fallon/Meyers Primary Care Institute (Fallon) on the East Coast, Kaiser Permanente - Northern California (KPNC) on the West Coast, HealthPartners Research Foundation (HPRF), and Henry Ford Health System (HFHS) in the Midwest. Study AimsThe goal of this project is to determine the completeness and accuracy of HMO comprehensive electronic data systems to identify breast and cervical cancer patients, treatment received, and delivery system factors associated with differences in outcomes. Electronic data assessed will include inpatient and outpatient visits, pharmacy, cancer registry, pathology, radiology, laboratory, and electronic physician notes. We will also take advantage of the opportunity to assess the usefulness of claims-type encounter data alone for tracking patterns of care and outcomes. The specific aims of this project are to:
MethodsThe study will compare information in the electronic data sources to information abstracted from medical records, and assess differences in completeness and accuracy of diagnostic, treatment, and outcomes by patient characteristics, among HMOs, and by source of data. Medical records of randomly sampled cancer patients from all four HMOs, stratified by the age of patient and clinical stage of cancer, will be audited to assess the accuracy and completeness of full electronic and claims-type encounter data on diagnosis, cancer stage, treatment, comorbidities, and outcome. Breast Cancer: A total of 925 women with breast cancer will be randomly selected for the chart audit. From these cases, 500 patients will be sampled from HMOs with cancer registries and the other 425 from sites without cancer registries. They will be selected within categories of age and stage of cancer from the breast cancer study population, assembled through claims-type encounter and full HMO electronic data. Cervical Cancer: Similarly, 995 women with cervical cancer will be randomly selected for the chart audit. Among these cases, 350 patients will be obtained from an HMO with cancer registries and the other 645 from HMOs without cancer registries. SignificanceThis study will provide important information about the feasibility of using computerized claims data as well as other computerized resources for the study of cancer treatment and outcomes. The primary strength of the study is that it uses routinely and efficiently collected population-based electronic data to identify cancer patients, treatments, and outcomes. If these databases are shown to be valid, they provide promising resources for cancer studies with extensive information on treatment, follow-up, and outcome far exceeding those available from traditional cancer registries. Project StatusThis project was funded as an investigator-initiated R01, and is closely affiliated with the CRN. The principal investigator is Feifei Wei, PhD, from HealthPartners Research Foundation. Data collection is ongoing. |
Health Services & Outcomes Research:
|
Home |
About CRN |
Collaborating with CRN |
Project Portfolio
Scientific & Data Resources | Dissemination | Related Links |
Maintained by the
Applied Research Program,
Division of Cancer Control & Population Sciences. Accessibility Feedback | Privacy Policy |