Patients with cancer that are at risk of short-term mortality could be identified using machine learning algorithms. It is unclear how different machine learning algorithms relate and whether they can help clinicians to have timely discussion regarding treatment and end-of-life preferences. In the study, short-term mortality risk for cancer patients can be predicted accurately using structured EHR data. The data was collected for patients receiving care at medical oncology clinics at University of Pennsylvania Health System (UPHS) who were listed in Clarity, a reporting database that includes EHRs for patients, including data on demographic profiles, comorbidities, and test results from laboratories. Logistic regression model was implemented using stepwise variable selection with backward elimination. For random forest and gradient boosting, hyperparameters were obtained using a gid search and 5-fold CV on training cohort. All three algorithm perform very well for predicting 180-day mortality from the date of encounter at an oncology practice. Clinicians believe that when gradient boosting algorithm was applied in real time, most patients who were identified as at high risk were appropriate for a timely conversation about treatment and end-of-life preferences.
Input variables : Demographic characteristics, comorbidities, and laboratory results from EHR
Output Variables : 180 day mortality prediction
| Statistical | : | Somers D | Accuracy | Precision and Recall | Confusion Matrix | F1 Score | Roc and Auc | Prevalence | Detection Rate | Balanced Accuracy | Cohen's Kappa | Concordance | Gini Coefficent | KS Statistic | Youden's J Index |
| Business | : | Population at High Risk of Disease | Risk by Geography | Risk by Demographics | Risk by Clinical Parameters |
| Infrastructure | : | Log Bytes | Logging/User/IAMPolicy | Logging/User/VPN | CPU Utilization | Memory Usage | Error Count | Prediction Count | Prediction Latencies | Private Endpoint Prediction Latencies | Private Endpoint Response Count |
Visit Model : github.com
Additional links : jamanetwork.com
| Model Category | : | Public |
| Date Published | : | January, 2020 |
| Healthcare Domain | : |
Payer
Provider |
| Code | : | github.com |
| Health Risk Management |
| Risk Progression |