Location
Hilton Waikoloa Village, Hawaii
Event Website
https://hicss.hawaii.edu/
Start Date
7-1-2025 12:00 AM
End Date
10-1-2025 12:00 AM
Description
Vital signs are precise indicators of patient deterioration, and early warning scores have been introduced to identify high-risk patients in hospital wards. However, these scores often rely solely on current readings, ignoring trends over time. In this work, the prediction of vital signs—Diastolic Blood Pressure, Systolic Blood Pressure, Heart Rate, Respiratory Rate, Oxygen Saturation, and Temperature—is made using past vitals, demographics, and admission data from the MIMIC-III dataset. Deep learning models were trained for multi-task learning, with LSTM-Attention (LSTM-ATTN) outperforming others. It achieved a mean squared error of 0.0022, surpassing Long Short Term Memory (LSTM), Bidirectional Long Short Term Memory (BiLSTM), Multi-head Attention. This underscores its potential for deployment not only in hospital settings but also in the context of virtual ward management, where real-time prediction of patients’ next vital signs and early detection of deterioration can be invaluable.
Recommended Citation
Jehangir, Basra and Li, Weizi, "Enhancing Early Warning Systems: Predicting Next Vital Signs Using Recurrent Neural Networks and Attention Models" (2025). Hawaii International Conference on System Sciences 2025 (HICSS-58). 3.
https://aisel.aisnet.org/hicss-58/hc/virtual_hospitals/3
Enhancing Early Warning Systems: Predicting Next Vital Signs Using Recurrent Neural Networks and Attention Models
Hilton Waikoloa Village, Hawaii
Vital signs are precise indicators of patient deterioration, and early warning scores have been introduced to identify high-risk patients in hospital wards. However, these scores often rely solely on current readings, ignoring trends over time. In this work, the prediction of vital signs—Diastolic Blood Pressure, Systolic Blood Pressure, Heart Rate, Respiratory Rate, Oxygen Saturation, and Temperature—is made using past vitals, demographics, and admission data from the MIMIC-III dataset. Deep learning models were trained for multi-task learning, with LSTM-Attention (LSTM-ATTN) outperforming others. It achieved a mean squared error of 0.0022, surpassing Long Short Term Memory (LSTM), Bidirectional Long Short Term Memory (BiLSTM), Multi-head Attention. This underscores its potential for deployment not only in hospital settings but also in the context of virtual ward management, where real-time prediction of patients’ next vital signs and early detection of deterioration can be invaluable.
https://aisel.aisnet.org/hicss-58/hc/virtual_hospitals/3