Brain trauma care targets analysis using a high-rate recording and computing network

Introduction

We analyze information on brain-injured patients' monitoring and care provided by a powerful information system.

Methods

We analyzed 543 hours on 11 patients, limited to 72 hours per patient when available: mean arterial pressure (MAP), intracranial pressure (ICP) and cerebral perfusion pressure (CPP) values were plotted against guideline thresholds, respectively 90 mmHg, 20 mmHg and 60 mmHg. The data were sampled every 2 seconds. Extraction was performed using a 3 teraflops supercomputer. We developed a method to detect periods of abnormal values.

Results

The calculated CPP and monitored CPP differed despite a good correlation (r = 0.91, P < 0.0001). Fifty-seven percent, 40% and 27% of the recorded MAP, ICP and CPP values reached thresholds. The time distributions of abnormal CPP, ICP and MAP values are detailed in Table 1: 51.7% of the MAP periods, 48.8% of the ICP periods, 51.8% of the calculated CPP were short episodes (<30 min). Mortality was associated with CPP < 60 (OR = 4.13 – logistic regression model, P < 0.0001) and inversely associated with MAP drops and IC hypertension episodes (OR 0.58 and 0.45, respectively). The mean time spent in each episode was higher in the NS group (76 ± 6 vs 48 ± 5 min). Caregivers' actions are perceptible on a CPP distribution chart.

Table 1 Detected abnormal episodes

Conclusion

Monitoring artifacts should be better identified when monitoring-based targets are used to guide therapy. Computer-based data analysis shows evidence of frequent episodes requiring therapeutic actions according to published guidelines, assuming that multimodal monitoring is not limited to the three studied parameters. Caregivers need new tools for data management to provide a better quality of care.