Needle in a Haystack: Natural Language Processing to Identify Serious Illness

Abstract

Background: Alone, administrative data poorly identifies patients with palliative care needs. Objective: To identify patients with uncommon, yet devastating, illnesses using a combination of administrative data and natural language processing (NLP). Design/Setting: Retrospective cohort study using the electronic medical records of a healthcare network totaling over 2500 hospital beds. We sought to identify patient populations with two unique disease processes associated with a poor prognosis: pneumoperitoneum and leptomeningeal metastases from breast cancer. Measurements: Patients with pneumoperitoneum or leptomeningeal metastasis from breast cancer were identified through administrative codes and NLP. Results: Administrative codes alone resulted in identification of 6438 patients with possible pneumoperitoneum and 557 patients with possible leptomeningeal metastasis. Adding NLP to this analysis reduced the number of patients to 869 with pneumoperitoneum and 187 with leptomeningeal metastasis secondary to breast cancer. Administrative codes alone yielded a 13% positive predictive value (PPV) for pneumoperitoneum and 25% PPV for leptomeningeal metastasis. The combination of administrative codes and NLP achieved a PPV of 100%. The entire process was completed within hours. Conclusions: Adding NLP to the use of administrative codes allows for rapid identification of seriously ill patients with otherwise difficult to detect disease processes and eliminates costly, tedious, and time-intensive manual chart review. This method enables studies to evaluate the effectiveness of treatment, including palliative interventions, for unique populations of seriously ill patients who cannot be identified by administrative codes alone.

‘Caveat emptor’: the cautionary tale of endocarditis and the potential pitfalls of clinical coding data—an electronic health records study

“…Finally, it suggests more work is needed to explore novel methods of improving case identification using electronic health records, such as improving data linking between admissions and microbiology results [30], using natural language processing methods [50], machine learning approaches [51]or healthcare process modelling [52], and supporting efforts to share, evaluate and refine these methods [53].…”

Prognostic Indices for Advance Care Planning in Primary Care: A Scoping Review

“…With a growing interest in the use of machine learning and artificial intelligence in medical care, our results can guide researchers who wish to test multiple algorithms simultaneously. 51,52 Our work has implications for practice-based research networks that wish to expand the implementation of ACP in the primary care setting. For example, the Patient-Centered Outcomes Research Institute recently funded 7 studies to encourage the expansion of ACP and palliative care.…”

Natural Language Processing to Assess Palliative Care and End-of-Life Process Measures in Patients With Breast Cancer With Leptomeningeal Disease

“…It also allowed for rapid semiautomated review to ensure that keywords have not been taken out of context. 6 Sections of reports that included the term "leptomeningeal" were highlighted by the program for review by a neurologist (K.B. ), who then was able to assess the context in which the term was used.…”

“…3,4 However, identifying EOL process measures has been a challenge due to a lack of consistent use of associated billing codes and the limitations of manual chart review. 5,6 Within the electronic health record (EHR), the majority of information (>70%) resides in free-text notes, especially that pertaining to EOL care. 7 The development of novel computer-assisted abstraction techniques, including natural language processing (NLP), offers new tools for capturing and assessing the performance of established EOL process measures.…”