Towards proactive palliative care in oncology: developing an explainable EHR-based machine learning model for mortality risk prediction.

BACKGROUND: Ex-ante identification of the last year in life facilitates a proactive palliative approach. Machine learning models trained on electronic health records (EHR) demonstrate promising performance in cancer prognostication. However, gaps in literature include incomplete reporting of model performance, inadequate alignment of model formulation with implementation use-case, and insufficient explainability hindering trust and adoption in clinical settings. Hence, we aim to develop an explainable machine learning EHR-based model that prompts palliative care processes by predicting for 365-day mortality risk among patients with advanced cancer within an outpatient setting. METHODS: Our cohort consisted of 5,926 adults diagnosed with Stage 3 or 4 solid organ cancer between July 1, 2017, and June 30, 2020 and receiving ambulatory cancer care within a tertiary center. The classification problem was modelled using Extreme Gradient Boosting (XGBoost) and aligned to our envisioned use-case: "Given a prediction point that corresponds to an outpatient cancer encounter, predict for mortality within 365-days from prediction point, using EHR data up to 365-days prior." The model was trained with 75% of the dataset (n = 39,416 outpatient encounters) and validated on a 25% hold-out dataset (n = 13,122 outpatient encounters). To explain model outputs, we used Shapley Additive Explanations (SHAP) values. Clinical characteristics, laboratory tests and treatment data were used to train the model. Performance was evaluated using area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC), while model calibration was assessed using the Brier score. RESULTS: In total, 17,149 of the 52,538 prediction points (32.6%) had a mortality event within the 365-day prediction window. The model demonstrated an AUROC of 0.861 (95% CI 0.856-0.867) and AUPRC of 0.771. The Brier score was 0.147, indicating slight overestimations of mortality risk. Explanatory diagrams utilizing SHAP values allowed visualization of feature impacts on predictions at both the global and individual levels. CONCLUSION: Our machine learning model demonstrated good discrimination and precision-recall in predicting 365-day mortality risk among individuals with advanced cancer. It has the potential to provide personalized mortality predictions and facilitate earlier integration of palliative care.

Maximum expected survival rate model for public access defibrillator placement.

AIM: Mathematical optimization of automated external defibrillator (AED) placement has demonstrated potential to improve survival of out-of-hospital cardiac arrest (OHCA). Existing models mostly aim to improve accessibility based on coverage radius and do not account for detailed impact of delayed defibrillation on survival. We aimed to predict OHCA survival based on time to defibrillation and developed an AED placement model to directly maximize the expected survival rate. METHODS: We stratified OHCAs occurring in Singapore (2010-2017) based on time to defibrillation and developed a regression model to predict the Utstein survival rate. We then developed a novel AED placement model, the maximum expected survival rate (MESR) model. We compared the performance of MESR with a maximum coverage model developed for Canada that was shown to be generalizable to other settings (Denmark). The survival gain of MESR was assessed through 10-fold cross-validation for placement of 20 to 1000 new AEDs in Singapore. Statistical analysis was performed using χ2 and McNemar's tests. RESULTS: During the study period, 15,345 OHCAs occurred. The power-law approximation with R2 of 91.33% performed best among investigated models. It predicted a survival of 54.9% with defibrillation within the first two minutes after collapse that was reduced by more than 60% without defibrillation within the first 4 minutes. MESR outperformed the maximum coverage model with P-value < 0.05 (<0.0001 in 22 of 30 experiments). CONCLUSION: We developed a novel AED placement model based on the impact of time to defibrillation on OHCA outcomes. Mathematical optimization can improve OHCA survival.