Applying a model to predict neutropenia risk in patients with cancer using electronic data [presentation]
- View All
Background/Aims: Granulocyte-colony stimulating factors (G-CSFs) are indicated to decrease the incidence of infections associated with chemotherapy-induced neutropenia (a severe reduction in white blood cells) in cancer patients receiving myelosuppressive chemotherapy. Severe neutropenia or febrile neutropenia occurs in 25-40% of treatment-naïve patients and are associated with increased infection rates, treatment delays, dose reductions, hospitalizations, deaths, and increased costs. In addition to the chemotherapy treatment regimen, patient-specific factors influence neutropenic risk. Current guideline recommendations do not include a mechanism to quantify patient risk when determining need to G-CSF support. Studies report over and under use of G-CSF among those at low and high risk for neutropenia. A neutropenia risk model developed by Lyman, et al. showed prior chemotherapy, relative dose intensity (RDI), abnormal liver and renal function, and low white cell count are major risk factors. We are testing the feasibility of using readily available clinical and administrative data from the HealthPartners electronic medical record and Oncology Treatment Module to apply the Lyman model to data from adult cancer patients receiving first cycle chemotherapy. Methods: We included breast, colorectal, lymphoid, lung, and ovarian cancer patients to compare abstracted chart data to electronically extracted data. We will adapt the Lyman model to predict neutropenia risk using our database. We will also extract electronic data from our site’s Virtual Data Warehouse (VDW) for comparison with sites participating in a separate CRN-funded pilot project to evaluate the feasibility of using VDW data to predict neutropenic risk. Results: We identified and abstracted data on 235 patients who received one of 60 treatment regimens. The median age at diagnosis was 61 (range 30-86) and 32% were male. Of those, 92 (39%) received G-CSF. The most commonly administered chemotherapy included cyclophosphamide, etoposide, 5-fluorouracil and doxorubicin. Manually abstracted data will be used to validate the electronically extracted data. Data validation, including inter-rater reliability and RDI, is ongoing. Discussion: G-CSF use among chemotherapy patients can be identified in electronic oncology treatment records. Validation of key variables to determine the feasibility of using electronic data for risk determination is ongoing.
occurs within event