We demonstrate the potential of machine learning to gather, track, and analyze symptoms experienced by cancer patients during chemotherapy. Although our initial model requires further optimization to improve the performance, further model building may yield machine learning methods suitable to be deployed in routine clinical care, quality improvement, and research applications.

RationaleMultiple clinical trials support the effectiveness of cardiac resynchronization therapy (CRT); however, optimal patient selection remains challenging due to substantial treatment heterogeneity among patients who meet the clinical practice guidelines.ObjectiveTo apply machine learning to create an algorithm that predicts CRT outcome using electronic health record (EHR) data avaible before the procedure.Methods and resultsWe applied machine learning and natural language processing to the EHR of 990 patients who received CRT at two academic hospitals between 2004–2015. The primary outcome was reduced CRT benefit, defined as <0% improvement in left ventricular ejection fraction (LVEF) 6–18 months post-procedure or death by 18 months. Data regarding demographics, laboratory values, medications, clinical characteristics, and past health services utilization were extracted from the EHR available before the CRT procedure. Bigrams (i.e., two-word sequences) were also extracted from the clinical notes using natural language processing. Patients accrued on average 75 clinical notes (SD, 29) before the procedure including data not captured anywhere else in the EHR. A machine learning model was built using 80% of the patient sample (training and validation dataset), and tested on a held-out 20% patient sample (test dataset). Among 990 patients receiving CRT the mean age was 71.6 (SD, 11.8), 78.1% were male, 87.2% non-Hispanic white, and the mean baseline LVEF was 24.8% (SD, 7.69). Out of 990 patients, 403 (40.7%) were identified as having a reduced benefit from the CRT device (<0% LVEF improvement in 25.2%, death by 18 months in 15.6%). The final model identified 26% of these patients at a positive predictive value of 79% (model performance: Fβ (β = 0.1): 77%; recall 0.26; precision 0.79; accuracy 0.65).ConclusionsA machine learning model that leveraged readily available EHR data and clinical notes identified a subset of CRT patients who may not benefit from CRT before the procedure.

No abstract

7 Background: Natural Language Processing (NLP) presents a novel method of extracting text-embedded information from the electronic health record (EHR) to improve routine assessment of palliative quality metrics such as timely advance care planning (ACP), palliative care provision (PC), and hospice referral. Methods: We identified cancer patients (ICD-9-CM codes 140-209) who received a gastrostomy tube (ICD-9-CM 43.11, 43.19, 44.32; CPT code 49440) from Jan 1, 2012, to Mar 31, 2016 at an academic medical center. We used NLP to identify palliative indication for gastrostomy tube placement by labeling clinical notes from the EHR containing the key word “venting” near the time of the procedure. Documentation of ACP, PC, and hospice referral was identified by NLP using a validated key term library. The sensitivity and specificity of the NLP method was determined by comparing outcome identification to manual chart abstraction performed by two clinicians. All NLP code was written in the open-source programming language Python. Results: NLP was performed for 75,626 documents. Among 305 cancer patients who underwent gastrostomy, 75 (24.6%) were classified by NLP as having a palliative indication for the procedure compared to 72 patients (23.6%) classified by human coders. Manual chart abstraction took > 2,600 times longer than NLP (28 hrs vs. 38 seconds). NLP identified the correct patients with high precision (0.92) and recall (0.96). ACP was documented during the index admission for 89.3% of patients. PC was documented for 85.7% and hospice referral was documented for 64.3% of these patients with advanced cancer during the index hospitalization. NLP identified ACP, PC and hospice referral with high precision (0.88-1.0) and recall (0.92-1.0) compared to human coders. Median survival was 37 days following gastrostomy tube procedure. Conclusions: NLP can greatly speed the assessment of established palliative quality metrics with an accuracy approaching that of human coders. These methods offer opportunities for facilitate quality improvement in palliative care for patients with advanced cancer.