RETAIN is an interpretable predictive model for healthcare applications. Given patient EMR records, it can predict what diagnoses the patient will receive at each visit such as will he be diagnosed with heart failure in the future? will he get readmitted? RETAIN, uses attention based Recurrent Neural Network (RNN) . The key idea of RETAIN is to improve the prediction accuracy through a sophisticated attention generation process, while keeping the representation learning part simple for interpretation, making the entire algorithm accurate and interpretable. RETAIN trains two RNN in a reverse time order to efficiently generate the appropriate attention variables. The model explainability is thus ingrained into the algorithm.
Input variables : Featuers like Patient ID, Visit dates, Diagnosis, Medication, Procedure codes, Patient demographics
Output Variables : Prediction of Disease
| 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 |
| 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 : arxiv.org
Additional links : arxiv.org | youtu.be
| Model Category | : | Public |
| Date Published | : | February, 2017 |
| Healthcare Domain | : |
Payer
Provider |
| Code | : | github.com |
| Health Risk Management |
| Disease Detection |