Derivation and validation of the automated search algorithms to identify cognitive impairment and dementia in electronic health records

详细信息    查看全文

• 作者：Sakusic Amra ; MD^a ; ^b ; ^c ; ^{Sakusic.Amra@mayo.edu} ; John C O'Horo ; MD ; MPH^a ; ^d ; ^e ; Tarun D Singh ; MBBS^d ; ^f ; Gregory A Wilson ; RRT^d ; Rahul Kashyap ; MBBS^d ; ^g ; Ronald Petersen ; MD ; PhD^f ; ^h ; ⁱ ; Rosebud O Roberts^f ; ^h ; ⁱ ; John D Fryer^j ; Alejandro A Rabinstein ; MD^d ; ^f ; Ognjen Gajic ; MD ; MSc^a ; ^d
• 关键词：Dementia ; Cognitive impairment ; Cognitive decline ; Critical illness ; Electronic search strategy ; Electronic medical records
• 刊名：Journal of Critical Care
• 出版年：2017
• 出版时间：February 2017
• 年：2017
• 卷：37
• 期：Complete
• 页码：202-205
• 全文大小：484 K
• 卷排序：37

文摘

Long-term cognitive impairment is a common and important problem in survivors of critical illness. We developed electronic search algorithms to identify cognitive impairment and dementia from the electronic medical records (EMRs) that provide opportunity for big data analysis.Materials and methodsEligible patients met 2 criteria. First, they had a formal cognitive evaluation by The Mayo Clinic Study of Aging. Second, they were hospitalized in intensive care unit at our institution between 2006 and 2014. The “criterion standard” for diagnosis was formal cognitive evaluation supplemented by input from an expert neurologist. Using all available EMR data, we developed and improved our algorithms in the derivation cohort and validated them in the independent validation cohort.ResultsOf 993 participants who underwent formal cognitive testing and were hospitalized in intensive care unit, we selected 151 participants at random to form the derivation and validation cohorts. The automated electronic search algorithm for cognitive impairment was 94.3% sensitive and 93.0% specific. The search algorithms for dementia achieved respective sensitivity and specificity of 97% and 99%. EMR search algorithms significantly outperformed International Classification of Diseases codes.ConclusionsAutomated EMR data extractions for cognitive impairment and dementia are reliable and accurate and can serve as acceptable and efficient alternatives to time-consuming manual data review.