Toward Predicting 30-Day Readmission Among Oncology Patients: Identifying Timely and Actionable Risk Factors.

PURPOSE: Predicting 30-day readmission risk is paramount to improving the quality of patient care. In this study, we compare sets of patient-, provider-, and community-level variables that are available at two different points of a patient's inpatient encounter (first 48 hours and the full encounter) to train readmission prediction models and identify possible targets for appropriate interventions that can potentially reduce avoidable readmissions. METHODS: Using electronic health record data from a retrospective cohort of 2,460 oncology patients and a comprehensive machine learning analysis pipeline, we trained and tested models predicting 30-day readmission on the basis of data available within the first 48 hours of admission and from the entire hospital encounter. RESULTS: Leveraging all features, the light gradient boosting model produced higher, but comparable performance (area under receiver operating characteristic curve [AUROC]: 0.711) with the Epic model (AUROC: 0.697). Given features in the first 48 hours, the random forest model produces higher AUROC (0.684) than the Epic model (AUROC: 0.676). Both models flagged patients with a similar distribution of race and sex; however, our light gradient boosting and random forest models were more inclusive, flagging more patients among younger age groups. The Epic models were more sensitive to identifying patients with an average lower zip income. Our 48-hour models were powered by novel features at various levels: patient (weight change over 365 days, depression symptoms, laboratory values, and cancer type), hospital (winter discharge and hospital admission type), and community (zip income and marital status of partner). CONCLUSION: We developed and validated models comparable with the existing Epic 30-day readmission models with several novel actionable insights that could create service interventions deployed by the case management or discharge planning teams that may decrease readmission rates over time.

Anticancer Therapy at the End of Life: Lessons From a Community Cancer Institute.

INTRODUCTION: Studies have shown aggressive cancer care at the end of life is associated with decreased quality of life, decreased median survival, and increased cost of care. This study describes the patients most likely to receive systemic anticancer therapy at the end of life in a community cancer institute. MATERIALS AND METHODS: We performed a retrospective cohort study of 201 patients who received systemic anticancer therapy in our institution and died between July 2016 and April 2017. Data collected included primary malignancy, hospice enrollment, healthcare utilization, Oncology Care Model (OCM) enrollment, and clinical assessments at last office visit prior to a treatment decision before death. We defined our outcome variable as the receipt of anticancer treatment in the last 14 days of a patient's life. We evaluated 20 clinical exposure variables with respect to the outcome classes. Risk ratios along with their associated confidence intervals and P values were calculated. Significance was determined using the Benjamini-Hochberg procedure to account for multiple testing. RESULTS: Of the 201 patients who died of cancer, 36 (17%) received anticancer therapy within the last 14 days of life. Several risk factors were significantly positively associated with receiving anticancer therapy at the end of life including hospitalization within 30 days of end of life, number of hospitalizations per patient (≥2), death in hospital, enrollment in OCM, and a diagnosis of hematologic malignancy. CONCLUSION: Our findings demonstrate those enrolled in the OCM and those with hematologic malignancies have a higher risk of receiving anticancer therapy in the last 14 days of life. These observations highlight the need for better identifying the needs of high-risk patients and providing good quality care throughout the disease trajectory to better align end-of-life care with patients' wishes.

Prevalence and Characterization of Yoga Mentions in the Electronic Health Record.

BACKGROUND: There is a growing patient population using yoga as a therapeutic intervention, but little is known about how yoga interfaces with health care in clinical settings. PURPOSE: To characterize how yoga is documented at a large academic medical center and to systematically identify clinician-derived therapeutic use cases of yoga. METHODS: We designed a retrospective observational study using a yoga cohort (n = 30,976) and a demographically matched control cohort (n = 92,919) from the electronic health records at Penn Medicine between 2006 and 2016. We modeled the distribution of yoga notes among patients, clinicians, and clinical service departments, built a multinomial Naïve Bayes classifier to separate the notes by context-dependent use of the word yoga, and modeled associations between clinician recommendations to use yoga and 754 diagnostic codes with Fisher's exact test, setting an false discovery rate (FDR)-adjusted P-value ≤ .05 (ie, q-value) as the significance threshold. RESULTS: Yoga mentions in the electronic health record have increased 10.4-fold during the 10-year study period, with 2.6% of patients having at least 1 mention of yoga in their notes. In total, 30,976 patients, 2398 clinicians, and 41 clinical service departments were affiliated with yoga notes. The majority of yoga notes are in primary care. Nine diagnoses met the significance criteria for having an association with clinician recommendations to use yoga including Parkinson's disease (Odds ratio [OR], 6.3 [3.7 to 11.4]; q-value < 0.001), anxiety (OR, 5.8 [3.8 to 9.0]; q-value < 0.001), and backache (OR, 3.8 [2.4 to 6.3]; q-value = 0.001). CONCLUSIONS: There is a widespread and growing trend to include yoga as part of the clinical record. In practice, clinicians are recommending yoga as a nonpharmacological intervention for a subset of common chronic diseases.

Using Machine Learning on Home Health Care Assessments to Predict Fall Risk.

Falls are the leading cause of injuries among older adults, particularly in the more vulnerable home health care (HHC) population. Existing standardized fall risk assessments often require supplemental data collection and tend to have low specificity. We applied a random forest algorithm on readily available HHC data from the mandated Outcomes and Assessment Information Set (OASIS) with over 100 items from 59,006 HHC patients to identify factors that predict and quantify fall risks. Our ultimate goal is to build clinical decision support for fall prevention. Our model achieves higher precision and balanced accuracy than the commonly used multifactorial Missouri Alliance for Home Care fall risk assessment. This is the first known attempt to determine fall risk factors from the extensive OASIS data from a large sample. Our quantitative prediction of fall risks can aid clinical discussions of risk factors and prevention strategies for lowering fall incidence.