Automated identification of pneumonia in chest radiograph reports in critically ill patients.

BACKGROUND: Prior studies demonstrate the suitability of natural language processing (NLP) for identifying pneumonia in chest radiograph (CXR) reports, however, few evaluate this approach in intensive care unit (ICU) patients. METHODS: From a total of 194,615 ICU reports, we empirically developed a lexicon to categorize pneumonia-relevant terms and uncertainty profiles. We encoded lexicon items into unique queries within an NLP software application and designed an algorithm to assign automated interpretations ('positive', 'possible', or 'negative') based on each report's query profile. We evaluated algorithm performance in a sample of 2,466 CXR reports interpreted by physician consensus and in two ICU patient subgroups including those admitted for pneumonia and for rheumatologic/endocrine diagnoses. RESULTS: Most reports were deemed 'negative' (51.8%) by physician consensus. Many were 'possible' (41.7%); only 6.5% were 'positive' for pneumonia. The lexicon included 105 terms and uncertainty profiles that were encoded into 31 NLP queries. Queries identified 534,322 'hits' in the full sample, with 2.7 ± 2.6 'hits' per report. An algorithm, comprised of twenty rules and probability steps, assigned interpretations to reports based on query profiles. In the validation set, the algorithm had 92.7% sensitivity, 91.1% specificity, 93.3% positive predictive value, and 90.3% negative predictive value for differentiating 'negative' from 'positive'/'possible' reports. In the ICU subgroups, the algorithm also demonstrated good performance, misclassifying few reports (5.8%). CONCLUSIONS: Many CXR reports in ICU patients demonstrate frank uncertainty regarding a pneumonia diagnosis. This electronic tool demonstrates promise for assigning automated interpretations to CXR reports by leveraging both terms and uncertainty profiles.

Risk adjusting community-acquired pneumonia hospital outcomes using automated databases.

OBJECTIVE: To describe the development and assessment of the Abbreviated Fine Severity Score (AFSS), a simplified version of the Pneumonia Severity Index (PSI) suitable for providing risk-adjusted reports to clinicians caring for patients hospitalized with community-acquired pneumonia. STUDY DESIGN: Retrospective cohort study. METHODS: We defined the AFSS based on data available in administrative and laboratory databases. We downloaded and linked these hospitalization and laboratory data from 2 cohorts (11,030 patients and 6147 patients) hospitalized with community-acquired pneumonia in all Kaiser Permanente Medical Care Program hospitals in northern California. We then assessed the relationship between the AFSS and mortality, length of stay, intensive care unit admission, and the use of assisted ventilation. Using logistic regression analysis, we assessed the performance of the AFSS and determined the area under the receiver operating characteristic curve (c statistic). Using a combination of manual and electronic medical record review, we compared the AFSS with the full PSI in 2 subsets of patients in northern California and Denver, Colorado, whose medical records were manually reviewed. RESULTS: The AFSS compares favorably with the PSI with respect to predicting mortality. It has good discrimination with respect to inhospital (c = 0.74) and 30-day (c = 0.75) mortality. It also correlates strongly with the PSI (r = 0.87 and r = 0.93 in the 2 medical record review subsets). CONCLUSIONS: The AFSS can be used to provide clinically relevant risk-adjusted outcomes reports to clinicians in an integrated healthcare delivery system. It is possible to apply risk-adjustment methods from research settings to operational ones.