This study aimed to examine the usefulness of AI in predicting patient prognosis..
The subjects for the prediction model were 3,819 patients with cerebrovascular or musculoskeletal diseases who were admitted to a recovery-phase rehabilitation ward between 2016 and 2022. The subjects for the validation of the prediction model were 746 patients with cerebrovascular or musculoskeletal diseases who were admitted to a recovery-phase rehabilitation ward of a different facility in 2022. The AI was trained on age, period from disease onset to hospitalization, disease category, scores for each item of the Functional Independence Measure (FIM), and cognitive FIM score, and a prediction model was created using the motor FIM score at the time of discharge as the objective variable. “Prediction One” (Sony Network Communications Corporation, Japan) was used as the AI predictive analysis tool. To validate the prediction model, the median error, median error rate, mean error, root mean square error (RMSE), and coefficient of determination were calculated.
The results showed a median error, a median error rate, an error mean, an RMSE, and a coefficient of determination of 5.5 points, 7.7%, 10.0 points, 15.3 points, and 0.66, respectively.
The results were considered non-inferior to those of previous studies on disease-limited FIM prediction. In addition, we were able to exclude the influence of facility characteristics because we used data from a different facility for verification. AI is a major “machine learning” technology that finds regularities and features in large datasets. The use of AI technology has made it possible to predict the prognosis of a combination of cerebrovascular and musculoskeletal diseases. AI-based forecasting models are highly useful because of their versatility. In the future, we believe that it will be necessary to re-examine the items employed in the forecasting model and verify them in the context of an increased number of facilities.
With the practical application of AI, physiotherapists who do not have a high level of expertise in each disease can predict the prognosis. We believe that the ability to predict prognosis, even for physiotherapists with limited experience, will lead to assurance of the quality of physiotherapy and is significant from the perspective of physiotherapy management. In addition, we expect to improve the accuracy of the predictions by comparing the results from clinical practice with clear indicators, which is also meaningful in terms of education.
Functional independence measure
Prognostic prediction
