An expert system for monitor alarm integration

Citation
C. Oberli et al., An expert system for monitor alarm integration, J CLIN M C, 15(1), 1999, pp. 29-35
Citations number
40
Categorie Soggetti
Aneshtesia & Intensive Care
Journal title
JOURNAL OF CLINICAL MONITORING AND COMPUTING
ISSN journal
13871307 → ACNP
Volume
15
Issue
1
Year of publication
1999
Pages
29 - 35
Database
ISI
SICI code
1387-1307(199901)15:1<29:AESFMA>2.0.ZU;2-V
Abstract
Objective. Intensive care and operating room monitors generate data that ar e not fully utilized. False alarms are so frequent that attending personnel tends to disconnect them. We developed an expert system that could select and validate alarms by integration of seven vital signs monitored on-line f rom cardiac surgical patients. Methods. The system uses fuzzy logic and is able to work under incomplete or noisy information conditions. Patient stat us is inferred every 2 seconds from the analysis and integration of the var iables and a uni ed alarm message is displayed on the screen. The proposed structure was implemented on a personal computer for simultaneous automatic surveillance of up to 9 patients. The system was compared with standard mo nitors (Space-Labs (TM) PC2), using their default alarm settings. Twenty pa tients undergoing cardiac surgery were studied, while we ran our system and the standard monitor simultaneously. The number of alarms triggered by eac h system and their accuracy and relevance were compared. Two expert observe rs (one physician, one engineer) ascertained each alarm reported by each sy stem as true or false. Results. Seventy-five percent of the alarms reported by the standard monitors were false, while less than 1% of those reported by the expert system were false. Sensitivity of the standard monitors was 7 9% and sensitivity of the expert system was 92%. Positive predictive value was 31% for the standard monitors and 97% for the expert system. Conclusion s. Integration of information from several sources improved the reliability of alarms and markedly decreased the frequency of false alarms. Fuzzy logi c may become a powerful tool for integration of physiological data.