Date of Award
10-16-2023
Degree Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Nursing
First Advisor
Patricia Friend
Abstract
Growing evidence supports the benefits of serious illness communication including goals of care (GOC) discussions and documentation in the electronic health record (EHR). Patients who discuss end-of-life (EOL) care with their clinicians are more likely to have positive outcomes including better-reported quality of life, less distress, and a higher likelihood of receiving care consistent with their preferences. Limited research suggests only a small fraction of ovarian cancer patients have such discussions with their clinicians. Using a novel natural language processing (NLP) methodology, this retrospective and descriptive study explores EHRs for patients with ovarian cancer to characterize documentation of GOC. Using concept unique identifiers (CUIs) as the primary data organizer and means for semantic analysis, a rules-based NLP algorithm was built, refined, and validated that uncovered GOC documentation from the EHR with good accuracy and discrimination. GOC documentation was characterized including evaluation for possible disparities. Elements of GOC documentation were identified for 67.3% of the overwhelmingly Non-Hispanic, White patient sample. Eleven distinct disciplines were identified as clinician authors of GOC-positive notes. Missed opportunities were identified to offer the support of palliative care, and to improve the quality of patients’ EOL experience. While the study investigated for possible disparities based on variables of age, race, ethnicity, and insurance class, the only statistically significant finding was that more GOC-positive notes were identified for Non-White patients compared to Whites (p < .003). This may represent discordance between the health care team’s recommendations and the preferences and GOC expressed by non-White patients. Use of NLP shows promise for future study, interventions, and clinical practice to improve care and nudge closer to delivering goal concordant care for patients dealing with ovarian cancer.
Recommended Citation
Joehl, Hillarie Elizabeth, "Using Natural Language Processing to Evaluate Electronic Health Records of Patients with Ovarian Cancer for Documentation of Goals of Care" (2023). Dissertations. 4066.
https://ecommons.luc.edu/luc_diss/4066
