Predicting 180-day mortality for women with ovarian cancer using machine learning and patient-reported outcome data

Abstract

Contrary to national guidelines, women with ovarian cancer often receive treatment at the end of life, potentially due to the difficulty in accurately estimating prognosis. We trained machine learning algorithms to guide prognosis by predicting 180-day mortality for women with ovarian cancer using patient-reported outcomes (PRO) data. We collected data from a single academic cancer institution in the United States. Women completed biopsychosocial PRO measures every 90 days. We randomly partitioned our dataset into training and testing samples. We used synthetic minority oversampling to reduce class imbalance in the training dataset. We fitted training data to six machine learning algorithms and combined their classifications on the testing dataset into an unweighted voting ensemble. We assessed each algorithm's accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC) using testing data. We recruited 245 patients who completed 1319 PRO assessments. The final voting ensemble produced state-of-the-art results on the task of predicting 180-day mortality for ovarian cancer paitents (Accuracy = 0.79, Sensitivity = 0.71, Specificity = 0.80, AUROC = 0.76). The algorithm correctly identified 25 of the 35 women in the testing dataset who died within 180 days of assessment. Machine learning algorithms trained using PRO data offer encouraging performance in predicting whether a woman with ovarian cancer will die within 180 days. This model could be used to drive data-driven end-of-life care and address current shortcomings in care delivery. Our model demonstrates the potential of biopsychosocial PROM information to make substantial contributions to oncology prediction modeling. This model could inform clinical decision-making Future research is needed to validate these findings in a larger, more diverse sample.

Using Machine Learning to Predict Unplanned Hospital Utilization and Chemotherapy Management From Patient-Reported Outcome Measures

“…Inclusion of three different PROMs allowed understanding of their individual predictive value. This study addressed the inconsistency in preprocessing methods for class imbalance in existing studies 18,33,[35][36][37] and highlighted differences in results generated from these three techniques. The variety of performance metrics reported allowed between-studies comparison 15 and in-depth understanding of models.…”

“…In previous studies, it was performed on the entire data set 18,35 or training set only. 36,37 ML can also be trained on original data and evaluated using multiple performance metrics. 33 Since there is no consistency in data preprocessing methods, the model performances in these scenarios were compared to discover bias in the results.…”

“…Six ML models, namely, logistic regression (LR), decision tree (DT), adaptive boosting (AB), random forest (RF), support vector machines (SVMs), and neural network (NN), were selected on the basis of their inclusions in previous research. 18,37 Hyperparameter tuning was performed on training sets through grid search with five-fold cross-validation. 37 The models were applied using Python sklearn library.…”

“…18,37 Hyperparameter tuning was performed on training sets through grid search with five-fold cross-validation. 37 The models were applied using Python sklearn library.…”

Early prediction of ovarian cancer risk based on real world data

“…A recent publication illustrates this concept in OC, although the number of samples in the database used, the Surveillance, Epidemiology, and End Results (SEER) dataset, is reduced, which casts doubts on the reproducibility of the results when applied to a broader population [12]. Similar problems due to a low training dataset occur in other works [13,14].…”