A Data-Driven Legal Analytics Framework for Healthcare Litigation Outcome Prediction

Abstract

A worldwide public healthcare concern that has been gaining more notice is healthcare legal litigation. For the results of medical litigation involving physicians, there are no predictive algorithms that employ traditional logistical approaches and machine learning (ML). To create a prediction system for medical litigation outcomes, lower their frequency, and support the safe and proper growth of the healthcare field, factors driving medical litigation among inpatients are investigated employing ML methods. Subjects with the identical healthcare grade and admission year as the litigant set were chosen at random in a 1:1 proportion to serve as the control set, and every litigation case submitted to the Government Hospital Arbitration Council between 2020 and 2023 was chosen. Single-factor assessment (a=0.06) of the 14 potentially contentious features was conducted using the SPSS program, and variables with statistically significant variations were chosen. At a 7:3 proportion, the statistics were split into training and testing collections. After choosing 5 ML algorithms, a disagreement prediction system was created using Python tools. Decision graph evaluation (DGE) was employed to assess the systems' medical practicality, while the algorithms' suitability and accuracy were described using the AUC, specificity, sensitivity, accuracy, precision, mean precision (MP), and F1 value. After extracting 1190 patients from the control and litigation sets, 11 key variables were chosen. The random forest approach had greater AUC, accuracy, ML, sensitivity, and F1-value than various approaches, and the DGE graph showed that it had significant medicinal advantages. The key influencing factors are discharge kind, admission costs, and the inpatient unit. RF is anticipated to be encouraged offline in litigation forecasting due to its strong medical potential and litigation prediction capabilities.

Ähnliche Arbeiten