Risk factors and screening instruments to predict adverse outcomes for undifferentiated older emergency department patients: a systematic review and meta-analysis

Abstract

Objectives: A significant proportion of geriatric patients experience suboptimal outcomes following episodes of emergency department (ED) care. Risk stratification screening instruments exist to distinguish vulnerable subsets, but their prognostic accuracy varies. This systematic review quantifies the prognostic accuracy of individual risk factors and ED-validated screening instruments to distinguish patients more or less likely to experience short-term adverse outcomes like unanticipated ED returns, hospital readmissions, functional decline, or death.

Methods: A medical librarian and two emergency physicians conducted a medical literature search of PubMed, EMBASE, SCOPUS, CENTRAL, and ClinicalTrials.gov using numerous combinations of search terms, including emergency medical services, risk stratification, geriatric, and multiple related MeSH terms in hundreds of combinations. Two authors hand-searched relevant specialty society research abstracts. Two physicians independently reviewed all abstracts and used the revised Quality Assessment of Diagnostic Accuracy Studies instrument to assess individual study quality. When two or more qualitatively similar studies were identified, meta-analysis was conducted using Meta-DiSc software. Primary outcomes were sensitivity, specificity, positive likelihood ratio (LR+), and negative likelihood ratio (LR-) for predictors of adverse outcomes at 1 to 12 months after the ED encounters. A hypothetical test-treatment threshold analysis was constructed based on the meta-analytic summary estimate of prognostic accuracy for one outcome.

Results: A total of 7,940 unique citations were identified yielding 34 studies for inclusion in this systematic review. Studies were significantly heterogeneous in terms of country, outcomes assessed, and the timing of post-ED outcome assessments. All studies occurred in ED settings and none used published clinical decision rule derivation methodology. Individual risk factors assessed included dementia, delirium, age, dependency, malnutrition, pressure sore risk, and self-rated health. None of these risk factors significantly increased the risk of adverse outcome (LR+ range = 0.78 to 2.84). The absence of dependency reduces the risk of 1-year mortality (LR- = 0.27) and nursing home placement (LR- = 0.27). Five constructs of frailty were evaluated, but none increased or decreased the risk of adverse outcome. Three instruments were evaluated in the meta-analysis: Identification of Seniors at Risk, Triage Risk Screening Tool, and Variables Indicative of Placement Risk. None of these instruments significantly increased (LR+ range for various outcomes = 0.98 to 1.40) or decreased (LR- range = 0.53 to 1.11) the risk of adverse outcomes. The test threshold for 3-month functional decline based on the most accurate instrument was 42%, and the treatment threshold was 61%.

Conclusions: Risk stratification of geriatric adults following ED care is limited by the lack of pragmatic, accurate, and reliable instruments. Although absence of dependency reduces the risk of 1-year mortality, no individual risk factor, frailty construct, or risk assessment instrument accurately predicts risk of adverse outcomes in older ED patients. Existing instruments designed to risk stratify older ED patients do not accurately distinguish high- or low-risk subsets. Clinicians, educators, and policy-makers should not use these instruments as valid predictors of post-ED adverse outcomes. Future research to derive and validate feasible ED instruments to distinguish vulnerable elders should employ published decision instrument methods and examine the contributions of alternative variables, such as health literacy and dementia, which often remain clinically occult.