Modified CARG score using data from the electronic health record to predict chemotherapy toxicity in older adults.

Abstract

12038 Background: Older adults starting chemotherapy are at greater risk of toxicity compared with younger patients. Additional tools are needed to aid in management decisions for this population. The Cancer and Aging Research Group (CARG) chemotherapy toxicity calculator is one such tool, which stratifies older adults into high, intermediate, or low risk for chemotherapy toxicity (Hurria, A., Mohile, S., Tew, W. P., & et al. (2016). Validation of a prediction tool for chemotherapy toxicity in older adults with cancer. Journal of Clinical Oncology, 34(20), 2366–2371. https://doi.org/10.1200/jco.2015.65.4327). This tool relies on a face-to-face encounter to ask questions, e.g. “How is your hearing?” or “Can you take your own medicines?” in addition to lab values such as hemoglobin and creatinine clearance. We modified the CARG chemo toxicity calculator to include data points which could be pulled from the electronic health record (EHR) without the need for a face-to-face encounter to assess for an association with emergency department (ED) visits and hospital admissions. Methods: Retrospective data analysis was conducted using the EHR of patients over age 65 diagnosed with a solid tumor from 1/1/2019 to 12/31/2020 who started chemotherapy. A modified CARG score was calculated using age, cancer type, number of drugs, hemoglobin, creatinine clearance, and falls within the past 6 months. The remaining items needed to calculate the complete CARG score were excluded since they were not accessible in the EHR. We assessed ED visits leading to admission, ED visits leading to discharge, direct admissions, and the total of all 3 visit types for all patients. Results: A modified CARG score was calculated for 763 patients. Multiple models were evaluated and negative binomial distribution was found to be the best fitted for our data. For every one unit increase in our calculated score the number of ED visits which lead to hospital admission increased by 6% (p-value = 0.0156). Additionally, there was a 5% increase in combined ED visits, ED visits leading to admission, and direct admissions for every one unit increase in risk score (p-value = 0.0063). ED visits that did not lead to admission were found not to have an association with the risk score (p-value= 0.1263). (Table) Conclusions: A modified CARG score using data obtained from patients’ EHR had a statistically significant association with increased ER visits that resulted in hospital admission and with the total of ED visits leading to admission, ED visits leading to discharge, direct admissions. Using these outcomes as a surrogate for toxicity, we deduce that a simple tool could be used to predict chemotherapy toxicity in older adults. [Table: see text]