EHRs can help detect patients with a history of cancer

Electronic health records can help identify patients with a history of cancer, according to research published this month in eGEMs (Generating Evidence and Methods to Improve Patient Outcomes).

Tumor registries are the "gold standard" in identifying people with incident cancer. However, they don't necessarily capture those with prevalent cancer or a history of it, making the registry and any research based on it, less accurate. They're also not necessarily up to date.

The researchers, from the Kaiser Permanente Institute for Health Research, created an algorithm to identify members with a history of cancer. They included all Kaiser Permanente Colorado members ages 40 to 75 enrolled in 2013 (201,787 members), and used administrative and EHR data, including Kaiser Colorado's tumor registry, their chemotherapy files and inpatient and outpatient claims to create an algorithm, which they tweaked after validating it using chart review.

The final algorithm achieved a sensitivity of 100 percent and specificity of 84.6 percent for identifying cancer. The algorithm identified 10.7 percent of all plan members with a history of cancer. If the researchers had relied on just the tumor registry, 47 percent of those with a history of cancer would have been missed. 

The researchers suggested that the algorithm could be useful to identify cancer cases, especially for organizations that don't have a tumor registry or a less complete one. The algorithm is also helpful for excluding people with cancer from research analysis, and useful to be able to identify a cancer free study population to study--for instance, heart disease in patients without cancer, or the incidence of cancer over time.

To learn more:
- read the study (.pdf)