Defining emergency department episodes by severity and intensity: A 15-year study of Medicare beneficiaries

Publication Type:
Journal Article
Citation:
BMC Health Services Research, 2010, 10
Issue Date:
2010
Full metadata record
Background. Episodes of Emergency Department (ED) service use among older adults previously have not been constructed, or evaluated as multi-dimensional phenomena. In this study, we constructed episodes of ED service use among a cohort of older adults over a 15-year observation period, measured the episodes by severity and intensity, and compared these measures in predicting subsequent hospitalization. Methods. We conducted a secondary analysis of the prospective cohort study entitled the Survey on Assets and Health Dynamics among the Oldest Old (AHEAD). Baseline (1993) data on 5,511 self-respondents 70 years old were linked to their Medicare claims for 1991-2005. Claims then were organized into episodes of ED care according to Medicare guidelines. The severity of ED episodes was measured with a modified-NYU algorithm using ICD9-CM diagnoses, and the intensity of the episodes was measured using CPT codes. Measures were evaluated against subsequent hospitalization to estimate comparative predictive validity. Results. Over 15 years, three-fourths (4,171) of the 5,511 AHEAD participants had at least 1 ED episode, with a mean of 4.5 episodes. Cross-classification indicated the modified-NYU severity measure and the CPT-based intensity measure captured different aspects of ED episodes (kappa = 0.18). While both measures were significant independent predictors of hospital admission from ED episodes, the CPT measure had substantially higher predictive validity than the modified-NYU measure (AORs 5.70 vs. 3.31; p < .001). Conclusions. We demonstrated an innovative approach for how claims data can be used to construct episodes of ED care among a sample of older adults. We also determined that the modified-NYU measure of severity and the CPT measure of intensity tap different aspects of ED episodes, and that both measures were predictive of subsequent hospitalization. © 2010 Kaskie et al; licensee BioMed Central Ltd.
Please use this identifier to cite or link to this item: