When it comes to recommend medication combination for patients with complex health conditions, existing deep learning approaches either do not customize based on health history of patient, or do not consider existing drug-drug interaction (DDI) that might lead to adverse effect. GAMENet is an end to end model which fills this gap. It is based on graph convolutional networks (GCN) and Memory augmented neural networks (MANN). Patient's health history and drug-drug interaction knowledge are used to provide safe and personalized recommendation of medication combination. It is tested on real world MIMIC-III dataset for effectiveness and safety by comparing with several state-of-the-art methods. GAMENet outperformed all baselines in effectiveness measures and achieved 3.60% DDI rate reduction from existing EHR data.
Input variables : Patient EHR (health history and visit details), drug-drug interaction graph
Output Variables : Recommended medication
| 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 | : | Drug Cost per Visit |
| 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 : arxiv.org
| Model Category | : | Public |
| Date Published | : | November, 2018 |
| Healthcare Domain | : | Life Sciences |
| Code | : | github.com |
| Clinical Trials |
| Drug Utilization |