BACKGROUND: Cardiac risk assessment may not be routinely performed. Electronic health records (EHRs) offer the potential to automate risk estimation. We compared EHR-based assessment with manual chart review to determine the accuracy of automated cardiac risk estimation and determination of candidates for antiplatelet or lipid-lowering interventions. METHODS: We performed an observational retrospective study of 23,111 adults aged 20 to 79 years, seen in a large urban primary care group practice. Automated assessments classified patients into 4 cardiac risk groups or as unclassifiable and determined candidates for antiplatelet or lipid-lowering interventions based on current guidelines. A blinded physician manually reviewed 100 patients from each risk group and the unclassifiable group. We determined the agreement between full review and automated assessments for cardiac risk estimation and identification of which patients were candidates for interventions. RESULTS: By automated methods, 9.2% of the population were candidates for lipid-lowering interventions, and 8.0% were candidates for antiplatelet medication. Agreement between automated risk classification and manual review was high (kappa = 0.91; 95% confidence interval [CI], 0.88-0.93). Automated methods accurately identified candidates for antiplatelet therapy [sensitivity, 0.81 (95% CI, 0.73-0.89); specificity, 0.98 (95% CI, 0.96-0.99); positive predictive value, 0.86 (95% CI, 0.78-0.94); and negative predictive value, 0.98 (95% CI, 0.97-0.99)] and lipid lowering [sensitivity, 0.92 (95% CI, 0.87-0.96); specificity, 0.98 (95% CI, 0.97-0.99); positive predictive value, 0.94 (95% CI, 0.89-0.99); and negative predictive value, 0.99 (95% CI, 0.98-> or =0.99)]. CONCLUSIONS: EHR data can be used to automatically perform cardiovascular risk stratification and identify patients in need of risk-lowering interventions. This could improve detection of high-risk patients whom physicians would otherwise be unaware.
BACKGROUND: Cardiac risk assessment may not be routinely performed. Electronic health records (EHRs) offer the potential to automate risk estimation. We compared EHR-based assessment with manual chart review to determine the accuracy of automated cardiac risk estimation and determination of candidates for antiplatelet or lipid-lowering interventions. METHODS: We performed an observational retrospective study of 23,111 adults aged 20 to 79 years, seen in a large urban primary care group practice. Automated assessments classified patients into 4 cardiac risk groups or as unclassifiable and determined candidates for antiplatelet or lipid-lowering interventions based on current guidelines. A blinded physician manually reviewed 100 patients from each risk group and the unclassifiable group. We determined the agreement between full review and automated assessments for cardiac risk estimation and identification of which patients were candidates for interventions. RESULTS: By automated methods, 9.2% of the population were candidates for lipid-lowering interventions, and 8.0% were candidates for antiplatelet medication. Agreement between automated risk classification and manual review was high (kappa = 0.91; 95% confidence interval [CI], 0.88-0.93). Automated methods accurately identified candidates for antiplatelet therapy [sensitivity, 0.81 (95% CI, 0.73-0.89); specificity, 0.98 (95% CI, 0.96-0.99); positive predictive value, 0.86 (95% CI, 0.78-0.94); and negative predictive value, 0.98 (95% CI, 0.97-0.99)] and lipid lowering [sensitivity, 0.92 (95% CI, 0.87-0.96); specificity, 0.98 (95% CI, 0.97-0.99); positive predictive value, 0.94 (95% CI, 0.89-0.99); and negative predictive value, 0.99 (95% CI, 0.98-> or =0.99)]. CONCLUSIONS: EHR data can be used to automatically perform cardiovascular risk stratification and identify patients in need of risk-lowering interventions. This could improve detection of high-risk patients whom physicians would otherwise be unaware.
Authors: Beverly B Green; Melissa L Anderson; Andrea J Cook; Sheryl Catz; Paul A Fishman; Jennifer B McClure; Robert Reid Journal: Am J Prev Med Date: 2012-04 Impact factor: 5.043
Authors: John Heintzman; Steffani R Bailey; Megan J Hoopes; Thuy Le; Rachel Gold; Jean P O'Malley; Stuart Cowburn; Miguel Marino; Alex Krist; Jennifer E DeVoe Journal: J Am Med Inform Assoc Date: 2014-02-07 Impact factor: 4.497
Authors: Stephen D Persell; Donald M Lloyd-Jones; Elisha M Friesema; Andrew J Cooper; David W Baker Journal: J Gen Intern Med Date: 2012-11-11 Impact factor: 5.128
Authors: James Oujiri; Abdul Hakeem; Quinn Pack; Robert Holland; David Meyers; Christopher Hildebrand; Alan Bridges; Mary A Roach; Bennett Vogelman Journal: BMJ Qual Saf Date: 2011-02 Impact factor: 7.035