Clinical notes provide valuable information about patients as they describe symptoms, reasons for diagnosis and patient history. Analysis of such text can be useful in various medical predictions. However, being high-dimensional and unstructured, clinical notes are underused as compared to the structured data like demographics and lab results as it is difficult for clinical machine learning models to utilize these texts. ClinicalBERT applies the BERT model to clinical corpora in order to utilize clinical notes and predict hospital readmission. Huang et. al. have developed a flexible framework to represent clinical texts by applying transformers to clinical notes, evaluate these clinical texts representations and fine tune the network to predict readmissions. The model predicts the probability of 30 – day readmission at different timepoints of admission. This model outperforms many baselines on predicting 30-day readmission.
Input variables : Clinical notes
Output Variables : Predicted probability of readmission
| 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 | : | Readmission Rate | Avg Hospital LOS |
| 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, 2020 |
| Healthcare Domain | : | Provider |
| Code | : | github.com |
| Utilization Management |
| Readmission |