FUTURE BLOOD PRESSURE PREDICTION ARTIFICIAL INTELLIGENCE PROGRAM: THE DR. ANSWER MEDICAL SOFTWARE

Abstract

Objective: Current guidelines recommend home blood pressure (HBP) monitoring in patients diagnosis with hypertension and those with white coat hypertension or masked hypertension. Delicate classification of these patients are important because the intensification of antihypertensive drug therapy for patients with white coat uncontrolled hypertension or white coat hypertension may not be needed because cardiovascular risk seems not to be greater. Therefore, its detection is important to avoid overtreatment. Meanwhile patients with masked hypertension or masked uncontrolled hypertension and uncontrolled hypertension need further intensification of antihypertensive treatment. Dr. Answer project consists of 21 artificial intelligence (AI) medical software that cover eight major diseases including hypertension. The main object of Dr. Answer hypertension project is to predict the patients next visit office BP according to the home BP data and provide effective and precise treatment plan. Design and method: 2,057 patients and 156,117 pure HBP data were collected in the Dr. Answer hypertension project. Patients were enrolled in 1 tertiary center (Chonnam National University Hospital) and other 21 non-tertiary centers from January 2022 to September 2022. 1,447 patients and 153,712 were excluded due to insufficient HBP data and follow up loss. Finally 610 patients and 2,405 HBP data were evaluated in the AI program model. Results: Patients were 66.7 ± 14.4 years and proportion of women was 24.1%. and patients enrolled in tertiary center was 73.4%. Mean HBP 8weeks of systolic blood pressure and 4 weeks of diastolic blood pressure data were evaluated in the AI program model. Recurrent Neural Network, Long Short-term Memory, Attention algorism were selected in AI model program. Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) and Accuracy data were used as the AI model parameter. MAE of systolic BP was 3.96 and MAE of diastolic BP was 2.1. Meanwhile, RMSE of systolic BP was 5.81 and RSME of diastolic BP was 2.88. Accuracy data showed 74.1% and 91.4% respectively. Conclusions: Dr. Answer hypertension AI medical software showed promising results for predicting future blood pressure data in hypertension patients. Further advancement in the AI program model would be needed.