Type and combination of health data influences policy insights

Data from electronic health records (EHRs) are increasingly used for scientific research and health policy. However, not all healthcare data are equally suitable or available for identifying patients or determining key moments in their care trajectory. This is shown by a study conducted by Nivel, Tilburg University, Jeroen Bosch Hospital, the Dutch National Health Care Institute (Zorginstituut Nederland), and Equalis. The researchers investigated which types of EHR data are most suitable for identifying patients with colon cancer and determining their date of diagnosis. The results show that combining different types of data leads to more accurate results, but also requires additional effort and careful consideration of privacy.

Different types of healthcare data

EHRs contain different types of information. Reimbursement-oriented data are primarily recorded for healthcare reimbursement, while care-oriented data are intended to support the diagnostic and treatment process. In addition, EHRs contain both structured data, such as diagnosis codes, and unstructured data, such as clinicians' notes. For research and policymaking, it is not always clear which data are most appropriate or how much effort is required to make them suitable for analysis. 

Careful choices are needed when using healthcare data

The results show that reimbursement-oriented data alone have limited value for accurately identifying patients. Adding clinical information from EHRs improves the quality of patient identification, but requires more time, increases costs, and raises additional privacy considerations. Manual validation also showed that some patients identified through automated selection had been incorrectly included. A targeted combination of data sources can therefore improve the accuracy of patient identification without requiring access to all available EHR data.

Research findings based on routinely collected healthcare data are used to inform important policy decisions, such as how healthcare is organised and which treatments are reimbursed. That is why it is essential to make well-informed choices about which data sources to use. A decision framework could help researchers and policymakers strike the right balance between accuracy, the effort required, and privacy

Melissa van Essen

PhD candidate at TRANZO Tilburg University

Want to know more?

Read the full results of the study in the publication in Health Policy and Technology.

About the study

This case study analysed data from patients treated for colon cancer at Jeroen Bosch Hospital in 2019. Different combinations of reimbursement-oriented and care-oriented EHR data were evaluated and subsequently manually validated to assess the accuracy of the various patient selection strategies.