Dynamic Rolling Severity Score for ICU Patient Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems in intensive care units (ICUs) lack a continuous severity of illness score that can accurately predict patient deterioration in real-time, often resulting in false positives, missed detections, and a lack of memory in scoring systems, which hampers early intervention and patient safety.
Innovation Solution
A predictive analytics system that continuously monitors high-frequency vital signs data, incorporates past information, and sends alerts to clinicians when a patient's severity of illness score crosses predefined thresholds, using an exponential weighted moving average to update scores and account for previous interventions like mechanical ventilation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional early warning scores are used to identify at-risk patients, then some patients can be detected, but the system produces a high number of false positives leading to alarm fatigue and many deteriorating patients go undetected
Solution Approach 1:
The system transitions from static early warning scores to a dynamic rolling seventy calculation that continuously updates patient risk assessment by incorporating the most recent vital sign data while weighting recent measurements more heavily than historical data. This dynamic approach adapts to changing patient conditions in real-time, improving detection accuracy while reducing false alarms through continuous re-evaluation rather than fixed threshold comparisons
Solution Approach 2:
The system changes the parameter weighting scheme by applying exponential decay to historical data points, where more recent measurements receive higher weights and older measurements receive progressively lower weights. This parameter transformation allows the system to prioritize current patient status while retaining contextual information from the past seventy hours, thereby improving reliability without generating excessive false positives
2Device complexity
If static scores at individual timepoints are used, then the calculation is simple, but the system lacks memory and cannot incorporate past scores in the calculation of present scores
Solution Approach 1:
The system implements continuous monitoring and calculation of the rolling seventy metric by continuously ingesting new vital sign data and updating the risk score in near-real-time. This continuous action ensures that the system maintains an ongoing assessment of patient status incorporating both historical context and current measurements, eliminating the discontinuous nature of static scores while managing computational complexity through efficient data processing
Solution Approach 2:
The system dynamically balances the inclusion of historical data with computational efficiency by applying exponential weighting that naturally diminishes the influence of older data points. This dynamic weighting scheme allows the system to incorporate a seventy-hour history of patient data without requiring equal computational resources for all historical points, thus maintaining simplicity while preserving memory
3Measurement precision
If manual entry of information is required for severity of illness scores, then data accuracy may be improved, but the score generation is considerably delayed
Solution Approach 1:
The system implements self-service by automatically extracting vital sign data and clinical measurements from existing electronic medical record systems and bedside monitors. The rolling seventy calculation is automatically performed by the system using predefined algorithms, eliminating the need for manual data entry while maintaining measurement precision through direct integration with clinical data sources. This automation enables near-real-time score generation without sacrificing data accuracy
Data Source
AI summary
A system and method for creating and presenting a rolling severity of illness acuity score are comprised of two phases. The first phase analyzes historical physiologic and clinical data from a hospital's EMR system and discovers data patterns that presage a deleterious outcome. These data patterns are stored in a first database. Once these patterns have been revealed, the second phase involves embedding these patterns into a hospital's EMR, collecting data on new patients in near-real time, matching up this data against the patterns uncovered in Phase One, and producing a score based on how many of these deleterious patterns a patient is exhibiting. This score is updated continuously, weighting the current value higher than previous scores. If this rolling score surpasses a specific value, an alert is generated. The alert, in one example a color-coded signal, is then sent to a mobile device or tablet of the clinicians caring for the patient.