OBJECTIVE: Performance of computerized adverse drug event (ADE) monitoring of electronic health records through a prospective ADE Monitor and ICD9-coded clinical text review operating independently and simultaneously on the same patient population for a 10-year period are compared. Requirements are compiled for clinical decision support in pharmacy systems to enhance ADE detection. METHODS: A large tertiary care facility in Utah, with a history of quality improvement using its advanced hospital information system, was leveraged in this study. ICD9-based review of clinical charts (ICD9 System) was compared quantitatively and qualitatively to computer-assisted pharmacist-verified ADEs (ADE Monitor). The capture-recapture statistical method was applied to the data to determine an estimated prevalence of ADEs. RESULTS: A total estimated ADE prevalence of 5.53% (13,420/242,599) was calculated, with the ICD9 system identifying 2,604 or 19.4%, and the ADE monitor 3,386 or 25.2% of all estimated ADEs. Both methods commonly identified 4.9% of all estimated ADEs and matched 62.0% of the time, each having its strength in detecting a slightly different domain of ADEs. 70% of the ADE documentation in the clinical notes was found in the discharge summaries. CONCLUSION: Coupled with spontaneous reporting, computerized methods account for approximately half of all ADEs that can currently be detected. To enhance ADE monitoring and patient safety in a hospitalized setting, pharmacy information systems should incorporate prospective structuring and coding of the text in clinical charts and using that data alongside computer-generated alerts of laboratory results and drug orders. Natural language processing can aid computerized detection by automating the coding, in real-time, of physician text from clinical charts so that decision support rules can be created and applied. New detection strategies and enhancements to existing systems should be researched to enhance the detection of ADEs since approximately half are not currently detected.
OBJECTIVE: Performance of computerized adverse drug event (ADE) monitoring of electronic health records through a prospective ADE Monitor and ICD9-coded clinical text review operating independently and simultaneously on the same patient population for a 10-year period are compared. Requirements are compiled for clinical decision support in pharmacy systems to enhance ADE detection. METHODS: A large tertiary care facility in Utah, with a history of quality improvement using its advanced hospital information system, was leveraged in this study. ICD9-based review of clinical charts (ICD9 System) was compared quantitatively and qualitatively to computer-assisted pharmacist-verified ADEs (ADE Monitor). The capture-recapture statistical method was applied to the data to determine an estimated prevalence of ADEs. RESULTS: A total estimated ADE prevalence of 5.53% (13,420/242,599) was calculated, with the ICD9 system identifying 2,604 or 19.4%, and the ADE monitor 3,386 or 25.2% of all estimated ADEs. Both methods commonly identified 4.9% of all estimated ADEs and matched 62.0% of the time, each having its strength in detecting a slightly different domain of ADEs. 70% of the ADE documentation in the clinical notes was found in the discharge summaries. CONCLUSION: Coupled with spontaneous reporting, computerized methods account for approximately half of all ADEs that can currently be detected. To enhance ADE monitoring and patient safety in a hospitalized setting, pharmacy information systems should incorporate prospective structuring and coding of the text in clinical charts and using that data alongside computer-generated alerts of laboratory results and drug orders. Natural language processing can aid computerized detection by automating the coding, in real-time, of physician text from clinical charts so that decision support rules can be created and applied. New detection strategies and enhancements to existing systems should be researched to enhance the detection of ADEs since approximately half are not currently detected.
Entities:
Keywords:
Clinical decision support; data collection; error management and prevention; pharmacy information systems; system improvement
Authors: E P McCarthy; L I Iezzoni; R B Davis; R H Palmer; M Cahalane; M B Hamel; K Mukamal; R S Phillips; D T Davies Journal: Med Care Date: 2000-08 Impact factor: 2.983