Providing high-quality emergency care for older adults is challenging
because older patients tend to have more co-morbidities, higher rates of
serious illnesses, and frequent communication barriers. Use of
screening instruments may allow identification of geriatric patients who are at increased risk for
readmission or other adverse outcomes; the results of the screening could potentially be then used to guide targeted interventions.
Given the nature of the topic, we would like to extend a formal invitation to Academic Life in Emergency Medicine (ALiEM) Journal Club (#ALiEMJC) to weigh in on the discussion!
Given the nature of the topic, we would like to extend a formal invitation to Academic Life in Emergency Medicine (ALiEM) Journal Club (#ALiEMJC) to weigh in on the discussion!
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.
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.
Did you know that the live version of #GeriMedJC runs for an hour and is broadcast to nine different hospitals in four different cities via the Ontario Telemedicine Network?
- Toronto: Mount Sinai Hospital, Sunnybrook Hospital, St. Michael’s Hospital, Baycrest Hospital, North York General Hospital, St. Joseph’s Hospital
- Mississauga: Trillium Hospital
- Kitchener-Waterloo: Grand River Hospital
- Orillia: Soldier's Memorial Hospital
Can't join us live? No worries! Engage in the discussion on Twitter on May 29, 2015 at 08:00 EDT / 12:00 GMT and don't forget to use the hashtag #GeriMedJC.
Several queries -- with my own perspectives/answers following, but would be keen to know the ideas of the #GeriMedJC audience.
ReplyDelete1) Since the meta-analysis by Carpenter et al indicate that none of these instruments (ISAR, TRST, VIP, etc.) or constructs of "frailty" predict any adverse outcome at any post-ED period of time, at any threshold, is it time to de-implement their use? The ISAR in particular is widely used around the world. MY RESPONSE: No, lacking any better alternatives in 2015 screening efforts ought to continue while researchers try to develop a truly accurate prognostic instrument (see below), but "nurse and physician educators need to recognize the limitations of existing instruments as predictors of adverse outcomes following an episode of ED care. Additionally (we) dissuade researchers, payers, and policy-makers from using any of these instruments as a basis for meaningful risk stratification."
2) Why do these instruments fail to predict adverse outcomes and what focus/methods ought future investigators use in an attempt to derive more accurate, yet feasible instruments that can be used at the point of care in the ED? MY RESPONSE (in order of most important to least important): a) test a tiered approach to risk stratification in which the first-level screen assesses age-related vulnerability (as the ISAR and TRST were designed to do) and the second-level screen assesses disease-specific risk of adverse outcome (using validated instruments for CHF, COPD, AMI, pneumonia, etc. prognostic risk); b) assess previously untested constructs of "frailty" in deriving future vulnerability instruments, constructs like formally tested cognitive function, grip strength, health literacy, and socio-economic status/access to care; c) standardize definitions (example "ED returns") and collaborative methodology using large center efforts rather than single-center studies; d) use established methods of deriving/validating clinical decision aids/prognostic models (Stiell et al in Ottawa).