Development and validation of a natural language processing algorithm using electronic health record data to identify patients with breast cancer with low social support.

• DOI: 10.1200/op.2024.20.10_suppl.421
• Publication Date: 2024-09-30T13:33:33Z
• Authors: Candyce Kroenke; Rhonda Aoki; Lauren Mammini; David Cronkite; Stacey Alexeeff; Salene M. W. Jones; Lawrence H. Kushi; Shaila Strayhorn; Jessica Mogk; David Mosen; David Carrell

421 Background: Social support is important to the management of breast cancer treatment. Our team has developed data from electronic health record (EHR) into structured ‘concept groups’ that will form basis for development EHRsupport, a computable, EHR-based measure social support. We report evaluation these concept groups against chart review as part our validation. Methods: built natural language processing (NLP) algorithm on clinical notes in 7,989 women diagnosed January 2006 September 2021 with invasive cancer. identified and 10 unstructured data: 1) living situation, 2) marital/partner status, 3) parenthood 4) visit (accompanied patient ≥1 visit, attended alone at visit), 5) friends/other support, 6) explicit positive or negative mentions 7) mention deceased person, 8) transportation issues, 9) relationship conflict stress, 10) isolation. validated charts 100 patients randomly drawn broader population (nonoverlapping training set) also around time (-1 +3 months) diagnosis. Results: Concept group availability ranged 1.3% isolation 98.3% situation. Specificity predictive values were moderate high all groups. Sensitivity (Table) lower low availability. Conclusions: Data have been available since advent Epic NLP-based accurately captured within EHR systematically collected supporting tool can be used identify risk EHRsupport groups, (n=7,989), vs. review. Availability (%) PPV NPV Living situation (e.g., not) 98.3 92 61 Partner/spouse 92.0 85 81 Parenthood status 88.8 95 80 93 83 Visit 82.9 77 96 59 Positive *46.9 75 70 Negative 33 99 67 Friend/other 46.8 79 Deceased person 39.9 86 97 90 Transportation issues 14.6 22 Relationship conflict/stress 7.7 25 29 94 *Availability