P049: A novel administrative database solution for capturing ED patient co-morbidity - the derived Charlson Comorbidity Index

Abstract

Introduction: ED patient comorbidity is difficult to ascertain for research. Traditional surrogates such as triage acuity, admission rate, and age have been used to approximate patient complexity. Differences between EDs for the management of similar conditions are nevertheless difficult to reconcile. The Charlson Comorbidity Index (CCI) contains 19 categories and is a validated predictor of the ten-year mortality for a patient who may have a range of comorbid conditions. CCI is based on the International Classification of Diseases (ICD) diagnosis codes found in administrative data such as the Discharge Abstract Database (DAD). The DAD collects this, and other inpatient information, for all Canadian hospitals. We sought to develop a linkage between the regional ED database and the regional inpatient DAD in order to derive a CCI score for each ED patient as a surrogate of comorbidity. Methods: We used regional data from Vancouver Coastal Health (VCH) over a 2.5 year period from April 2013 - September 2015. An algorithm was created to identify CCI conditions in the regional DAD. Whenever a patient visited the ED a query was made to the DAD going back for 5 years to acquire CCI relevant diagnoses and enter these diagnoses as well as the CCI weighting into the ED database. Patient DAD records from VCH were utilized no matter in which ED a patient presented. No information from admissions outside the region was available. Results: There were 931,596 regional ED visits made by 446,579 unique patients in a total of 11 EDs (6 urban and 5 rural). In total there were 127,233 patients with a CCI score (13.7% of total visits). The average CCI was 0.40 (SD 1.31) with a range of 0.12 at the urban urgent care centre to 0.52 at the urban tertiary care centre. More isolated rural EDs tended to have higher percentages of patients with CCI scores than community urban EDs. Higher acuity, age, and ambulance arrival, ED death, all correlated to higher CCI scores. The most common CCI conditions were “diabetes with complications” (10/11 EDs) and was present in 35,816 (3.8%) visits and “cancer” (10/11 EDs) present in 34,624 (3.7%) ahead of COPD (26,451 visits) and CHF (25,233 visits). Conclusion: Use of the CCI is a novel way to passively capture patient comorbidities without reliance on a data entry technician. Limitations include the inability to link to hospitalization data outside a specific health region.