PURPOSE:Timely recognition of acute lung injury (ALI) is essential for appropriate therapy and enrollment in clinical trials. However, ALI is often not recognized in a timely manner. We have previously developed an electronic alert system for identification of ALI in the medical intensive care unit (ICU). This study aimed to assess its accuracy for identifying ALI in two tertiary care hospitals.
METHODS:The “ICU data mart”, a Microsoft SQL-based integrative database with data abstracted within one hour entry into the electronic medical records, was used to detect ALI cases according to standard criteria. The sensitivity and specificity of the electronic alert were validated in different type of ICUs, using prospective assessment by trained study coordinator, blinded to the ALI electronic alert, as the gold standard. Physician notes were queried for the presence of ALI.
RESULTS:From 3795 consecutive patients admitted to the ICUs, a retrospective screening identified 316 patients with ALI (8.3%). The ALI electronic alert had a sensitivity of 99.1% and specificity 90.0%, with a positive predictive value of 46.0% and a negative predictive value of 100%. The results according to the ICU type are presented in the table. Only 81 (25.6%) cases were recognized in the physicians’ notes.
CONCLUSION:An automatic screening based on electronic medical records accurately identifies patients with ALI. Positive predictive value varies based on the ICU type.
CLINICAL IMPLICATIONS:Smart alerts based on electronic medical records can be used for timely identification of patients with specific ICU syndromes.
DISCLOSURE:Vitaly Herasevich, No Financial Disclosure Information; No Product/Research Disclosure Information