Developing an Algorithm to Identify History of Cancer Using the Electronic Medical Record
cancer, tumor registry
Background/Aims: In studies of cancer incidence, it is often desirable to exclude individuals with any prior cancer. While the virtual tumor registry (VTR) can be a valuable tool in identifying cancer cases, these data may be incomplete. In particular, cancers diagnosed prior to health plan enrollment, or diagnosed out of network, may not be captured in the VTR. Further, data in the VTR may lag several months behind, so recently diagnosed cases may be missed. As part of a new study, we needed to identify cancer-free adults in the current Kaiser Permanente Colorado member population. To accomplish this, we developed an algorithm using our electronic medical record (EMR) data to identify individuals who have never had cancer.
Methods: Kaiser Permanente Colorado members between the ages of 40–75 years who were continuously enrolled during 2013 were included. We used administrative and EMR data, including the VTR, chemotherapy files, and inpatient and outpatient claims, to identify those with a high likelihood of prior cancer.
Results: A total of 203,888 members met our initial inclusion criteria. The algorithm flagged 31,885 (15.6%) people as having a history of cancer. Of these, 10,494 (33%) cancer cases were identified by the VTR, 9,125 (29%) were identified with only a diagnosis code of cancer, and 7,812 (25%) had only a receipt of chemotherapy. Having at least 2 visits to oncology accounted for 1,324 (4%) cases. The remaining 3,130 (10%) cases were a combination of the aforementioned categories. Most cases, 29,126 (91%), were identified as having cancers prior to 2013. Of those with prior cancer, 61% were women. Those flagged as having cancer were significantly older, with mean age 61 (9.2 standard deviation) compared to 56 (9.4 standard deviation) for those without cancer (P
Discussion: It is important to consider data sources outside the VTR to capture all health plan members with a history of cancer. Casting a wide net will help ensure that anyone with a history of cancer, whether diagnosed recently or prior to joining the health plan, will be identified.
Clarke CL, Feigelson HS, Wrenn MG. Developing an Algorithm to Identify History of Cancer Using the Electronic Medical Record. J Patient Cent Res Rev 2015;2:87. http://dx.doi.org/10.17294/2330-0698.1077