This retrospective study evaluated the clinical performance of Vital-PICASO, an artificial intelligence-based biological signal analysis software designed to predict the risk of hypotension within 24 hours using vital-sign data from adult general ward inpatients. Electronic medical record data from patients aged 19 years or older who were admitted to a general ward at Seoul National University Hospital were retrospectively reviewed. Hypotension was defined as systolic blood pressure below 90 mmHg. Eligible data were classified as hypotension-positive or hypotension-negative according to predefined reference-standard criteria. Vital-PICASO was applied to the selected vital-sign datasets, and the software-generated hypotension risk scores were compared with the reference-standard classifications. Predictive performance was evaluated separately using data that included oxygen saturation and data that excluded oxygen saturation. Because the study used previously collected medical records, there was no direct participant contact and the software results were not used to guide patient care.
This study was a retrospective, single-center, superiority-confirmatory clinical performance study conducted at Seoul National University Hospital. Data Source and Study Population Electronic medical records were reviewed for patients aged 19 years or older who had been admitted to a general ward for at least 12 hours between January 1, 2023, and May 31, 2025. Clinical data previously used in the development of Vital-PICASO were excluded. The collected variables included systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate, body temperature, oxygen saturation when available, and the corresponding measurement times. Additional variables included the occurrence, time, and location of hypotension; month and year of birth; sex; height and weight when available; do-not-resuscitate status; pregnancy status; date of admission to the general ward; clinical department; and the vital-sign data collection period. Data Selection Potentially eligible hypotension-positive and hypotension-negative data were identified from the electronic medical record system. To reduce selection bias, candidate data were numbered and randomly selected by an independent statistician using R or SAS software. Hypotension-positive data were defined as data from patients who experienced hypotension, defined as systolic blood pressure below 90 mmHg, in a general ward. Vital-sign data from up to 72 hours before the first eligible hypotension event were collected. If hypotension occurred within 24 hours after admission or transfer to the general ward, all available vital-sign data recorded before the event were collected, provided that at least 12 hours of data were available. Hypotension-negative data were obtained from patients who did not experience hypotension during the selected general ward admission. Negative data were randomly selected from the same wards and from time intervals comparable with those of the positive cases. Vital-sign data from up to 72 hours after general ward admission or around the selected reference time point were collected. Only one eligible data interval was selected per patient. The final study dataset consisted of 648 data cases, including 324 hypotension-positive cases and 324 hypotension-negative cases. Two analysis datasets were evaluated: 168 cases for the analysis including oxygen saturation, consisting of 84 positive and 84 negative cases, and 480 cases for the analysis excluding oxygen saturation, consisting of 240 positive and 240 negative cases. Data cases could be used in both analysis datasets. Reference Standard Establishment The initial hypotension-positive or hypotension-negative classification was determined according to the predefined systolic blood pressure criterion. To improve the reliability of the study dataset, a clinician with at least 1 year of clinical experience independently reviewed the data collection reference time point and relevant patient information while blinded to the initial positive or negative classification. The clinician's final classification was used as the reference standard. Application of the Investigational Device The medical device operator received vital-sign datasets identified only by screening numbers and was blinded to the reference-standard classification. Each dataset was analyzed using Vital-PICASO to estimate the risk of hypotension within 24 hours. The peak risk score and corresponding time within the analysis interval were recorded. Statistical Analysis The software-generated risk scores were compared with the reference-standard classifications. The primary performance measures were the area under the receiver operating characteristic curve for: 1. the model including oxygen saturation; and 2. the model excluding oxygen saturation. The AUROC for each model was compared with the prespecified comparator performance value of 0.7847. The performance criterion for each primary outcome was met if the lower bound of the 95% confidence interval for the difference between the Vital-PICASO AUROC and the comparator AUROC was greater than zero. The overall study was considered successful if both primary outcomes met their respective performance criteria. Secondary performance measures included comparison with the predictive performance of the National Early Warning Score and Modified Early Warning Score, area under the precision-recall curve, precision, recall, F1 score, threshold-specific performance, performance by prediction-time interval, and cumulative-time performance. Because the study used previously collected electronic medical record data, there was no direct participant contact, additional examination, treatment, or change in clinical care.
Study Type
OBSERVATIONAL
Enrollment
648
Seoul National University Hospital
Seoul, South Korea
Area Under the Receiver Operating Characteristic Curve for Hypotension Prediction With SpO2
The area under the receiver operating characteristic curve (AUROC) was calculated by comparing the hypotension risk scores generated by Vital-PICASO with the reference-standard classification of hypotension-positive or hypotension-negative. The analysis included systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate, body temperature, and oxygen saturation as input variables. Hypotension was defined as systolic blood pressure below 90 mmHg. AUROC ranges from 0 to 1, with a higher value indicating better discrimination. The prespecified success criterion was met if the lower bound of the 95% confidence interval for the difference between the Vital-PICASO AUROC and the comparator AUROC of 0.7847 was greater than 0.
Time frame: During the 24-hour period before the hypotension event or corresponding reference time point
Area Under the Receiver Operating Characteristic Curve for Hypotension Prediction Without SpO2
The area under the receiver operating characteristic curve (AUROC) was calculated by comparing the hypotension risk scores generated by Vital-PICASO with the reference-standard classification of hypotension-positive or hypotension-negative. The analysis included systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate, and body temperature as input variables, without oxygen saturation. Hypotension was defined as systolic blood pressure below 90 mmHg. AUROC ranges from 0 to 1, with a higher value indicating better discrimination. The prespecified success criterion was met if the lower bound of the 95% confidence interval for the difference between the Vital-PICASO AUROC and the comparator AUROC of 0.7847 was greater than 0.
Time frame: During the 24-hour period before the hypotension event or corresponding reference time point
Area Under the Precision-Recall Curve for Hypotension Prediction With SpO2
The area under the precision-recall curve (AUPRC) was calculated using the Vital-PICASO risk scores and the reference-standard hypotension classifications when oxygen saturation was included as an input variable. AUPRC summarizes the relationship between precision and recall across possible classification thresholds. A higher value indicates better predictive performance.
Time frame: During the 24-hour period before the hypotension event or corresponding reference time point
Area Under the Precision-Recall Curve for Hypotension Prediction Without SpO2
The area under the precision-recall curve (AUPRC) was calculated using the Vital-PICASO risk scores and the reference-standard hypotension classifications when oxygen saturation was excluded from the input variables. AUPRC summarizes the relationship between precision and recall across possible classification thresholds. A higher value indicates better predictive performance.
Time frame: During the 24-hour period before the hypotension event or corresponding reference time point
Precision, Recall, and F1 Score Across Prespecified Risk-Score Thresholds
Precision, recall, and F1 score were calculated at prespecified Vital-PICASO risk-score thresholds. The analyses were performed separately for the model including oxygen saturation and the model excluding oxygen saturation. Precision represents the proportion of predicted positive cases that were reference-standard positive. Recall represents the proportion of reference-standard positive cases that were correctly predicted as positive. The F1 score is the harmonic mean of precision and recall. Each measure ranges from 0 to 1, with a higher value indicating better classification performance.
Time frame: During the 24-hour period before the hypotension event or corresponding reference time point
Predictive Performance by Prediction-Time Interval
Vital-PICASO predictive performance was evaluated separately within prespecified time intervals before the hypotension event or corresponding reference time point. AUROC, AUPRC, and F1 score were calculated for the following intervals: 0 to 1 hour; 1 to 6 hours; 6 to 12 hours; and 12 to 24 hours. The analyses were conducted separately for the model including oxygen saturation and the model excluding oxygen saturation.
Time frame: At 0 to 1, 1 to 6, 6 to 12, and 12 to 24 hours before the hypotension event or corresponding reference time point
Predictive Performance by Cumulative Prediction Window
Vital-PICASO predictive performance was evaluated over cumulative time windows before the hypotension event or corresponding reference time point. AUROC, AUPRC, and F1 score were calculated for the following cumulative prediction windows: 0 to 1 hour; 0 to 6 hours; 0 to 12 hours; and 0 to 24 hours. The analyses were conducted separately for the model including oxygen saturation and the model excluding oxygen saturation.
Time frame: Up to 24 hours before the hypotension event or corresponding reference time point
Difference Between Vital-PICASO AUROC and the Prespecified NEWS and MEWS Comparator AUROC
The AUROC of Vital-PICASO was compared with the prespecified comparator AUROC of 0.7847 derived from the weighted average predictive performance of the National Early Warning Score, Modified Early Warning Score, and relevant published hypotension-prediction studies. The difference between the Vital-PICASO AUROC and the comparator AUROC was calculated with a 95% confidence interval. A positive difference favored Vital-PICASO. The comparison was conducted separately for the model including oxygen saturation and the model excluding oxygen saturation.
Time frame: During the 24-hour period before the hypotension event or corresponding reference time point
Sensitivity, Specificity, Positive Predictive Value, and Negative Predictive Value at the Selected Risk-Score Threshold
Classification performance at the selected Vital-PICASO risk-score threshold was evaluated using sensitivity, specificity, positive predictive value, and negative predictive value. Sensitivity was defined as the proportion of reference-standard positive cases correctly classified as positive. Specificity was defined as the proportion of reference-standard negative cases correctly classified as negative. Positive predictive value was defined as the proportion of predicted positive cases that were reference-standard positive. Negative predictive value was defined as the proportion of predicted negative cases that were reference-standard negative. The analyses were conducted separately for the model including oxygen saturation and the model excluding oxygen saturation.
Time frame: During the 24-hour period before the hypotension event or corresponding reference time point
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.