The purpose of this dissertation was to examine the determinants of unplanned emergency department (ED) use by home care clients, the profile of older ED patients, their transitions from the ED, as well as the determinants of post discharge outcomes among older ED patients. The goal of this dissertation was to create theoretically driven, evidence-based, and practical risk identification methods for home care and the ED.
First, a multi-year, census-level cohort study was conducted on home care clients in two Canadian provinces (N=617,035). Census-level data from RAI-HC assessments were linked to census-level ED records. A needs-based decision tree model – the ED Model – informed by the Andersen Behavioural Model, was created using decision tree analyses. The final model was validated on a separate data partition and compared to the ERA Index and the CARS. Multilevel analyses were conducted to test regional variation in model performance. Disease stratified analyses were also conducted to test model generalizability across common disease classes. Regression analyses determined the effect of predisposing and enabling factors within ED Model strata.
Second, a multi-site, multi-province prospective cohort study was conducted, termed the Management of Older Persons in Emergency Departments (MOPED) Study, using a clinically representative sample of 2,101 older ED patients. The interRAI ED-CA was used to assess older ED patients, and a 90-day disposition was collected. The profile of older ED patents was examined. Best-subset regression analyses identified person-level determinants of acute inpatient admission. Two needs-based decision tree models – the ALC/LTC and ED Revisit Models – were created using decision tree analyses to determine the risk of ALC designation or LTC placement, and unplanned repeat ED visits, respectively. Both models were validated on separate data partitions. Multilevel analyses were conducted to test site-level variation in the models’ performance.
Overall, 41.2% of home care clients had at least one unplanned emergency department visit within 6 months of an assessment. Previous ED use, cardio-respiratory symptoms, cardiac conditions, and mood symptoms featured heavily in the ED Model. The ED Model provided moderate risk differentiation and clinical utility. It achieved an area under the curve of 0.62 (95% CI: 0.61-0.62) and showed clear differentiation in Kaplan-Meier plots using validation data. Multi-level analyses showed no regional variation. The ED Model significantly outperformed the similar tools specific to primary care with respect to overall accuracy and perceived clinical utility. Predisposing and enabling characteristics provided little added differentiation beyond evaluated need.
The majority of older ED patients were dependent on others for basic tasks of daily living, and many had fragile informal care or lived alone. Triage acuity generally did not differentiate chronic geriatric disabilities and conditions. Previous ED or hospital use was associated with chronic geriatric disabilities and conditions as well as informal caregiver distress. The Admission Model found that multiple factors were associated with admission to inpatient acute care, including: acuity, instability, changes in ADL function, cognition, nutrition, and anhedonia. Overall, 20.7% of older ED patients admitted to acute care were designated ALC or discharged to LTC; whereas 39.5% of older ED patients discharged home had one or more repeat ED visits within 90 days. Cognitive, functional, and informal care indicators were predictive of ALC/LTC; whereas functional status and symptoms were predictive of repeat ED use. The ALC/LTC and ED Revisit Models provided good risk differentiation, achieving AUC’s of 0.74 (95% CI: 0.69-0.79) and 0.69 (95% CI: 0.63-0.74), respectively. The ALC/LTC and ED Revisit Models showed clear differentiation in Kaplan-Meier plots. Multi-level analyses showed no site-level variation in each models’ performance.
This dissertation produced tangible and empirically-based risk assessment models for clinical practice in home care and the ED. The models developed in this dissertation can support the targeting of preventative services as well as better communication strategies between the ED and community supportive care, primary care, and inpatient acute care. Key questions related to the prevention of the risk pathways identified in each risk assessment model remain unanswered, and should be a focus of future research.||en