Hot Off the Press: Prospective and Explicit Clinical Validation of the Ottawa Heart Failure Risk Scale, With and Without Use of Quantitative NT‐proBNP

Congestive heart failure admissions are common and expensive, and high percentages of patients are readmitted within 3 to 6 months.1-5 Canadian admission rates for heart failure are much lower than in America.6 There is a lack of evidence to guide physicians in disposition decisions—who can go home and who requires admissions? Many risk stratification scales discuss the risk of mortality, but do not include broader adverse events that may occur.7-16 The Ottawa Heart Failure Risk Score (OHFRS) was derived to predict significant adverse events (SAEs) at 14 days.17 The current study aims to validate the tool in a real-time clinical setting. This prospective cohort study included patients ≥ 50 years old with dyspnea of <7 days’ duration, due to acute heart failure.18 Patients too ill to be discharged were excluded. Treating physicians assessed the OHFRS approximately 2 to 8 hours after ED presentation. The primary outcome measured was SAE within 14 days, with SAEs including death from any cause within 30 days, admission to a monitored unit, any positive pressure ventilation, myocardial infarction, major cardiac procedure, new dialysis, or subsequent hospital admission if the patient was initially discharged from the ED. At an admission threshold of ≥1, the OHFRS would increase sensitivity from 71.8% to 91.8% for SAEs, but also increase admission rates. A threshold of ≥2 had a similar sensitivity, but decreased admissions (57.2% vs. 48.3%.) Addition of NT-proBNP levels did not substantially change the results. This was a multicenter, prospective, ED-based study, with explicitly specified predictor variables and outcomes. Some limitations were identified. First of all, clinicians were not blinded to the OHFRS score, and in fact the treating clinician calculated the score. Although they were instructed not to use the score when making decisions, this could result in incorporation bias. Patients were not enrolled consecutively, and patients presenting at night could be higher risk than those during the day. At lower scores (<1), the OHFRS increases sensitivity for SAEs, and as a result, admissions increase. However, it is unclear whether admitting patients would prevent any of the SAEs from occurring. The use of a composite outcome can combine outcomes that are clinically dissimilar, making interpretation more difficult. For example, readmission to hospital (without any other complication) was included as an adverse event in this study, but is debatable whether this really represents an important patient-oriented outcome. Follow-up was stated to be complete, yet relied primarily on hospital records with phone follow-up as needed; no information is given as to the proportion which needed phone follow-up. A total of 1,100 patients were enrolled. Forty-three percent were discharged from the ED and 57% were admitted at the index visit. Mortality was 3.7% and SAEs occurred in 15.5% of patients (19.4% in admitted patients, 10.2% in discharged patients). Using an OHFRS cutoff of ≥1, when compared to actual care, sensitivity for SAEs increased from 71.8% to 91.8%, with admissions increasing from 57.6% to 77.6%. Increasing the cutoff to ≥2 resulted in similar sensitivity (71.8% vs. 71.2%) but decreased admissions (57.2% vs. 48.3%.) With the use of NT-proBNP, a score of ≥1 was 95.8% sensitive, compared to ED physician decisions at 69.8% sensitivity. The admission rate would have increased from 60.8% to 88.0%. Twelve percent of physicians rated themselves as being uncomfortable or very uncomfortable using the OHFRS. The OHFRS is a clinical decision “tool” (as opposed to a “rule”) and as such can be use at different cutoffs to help guide shared decision making. At values ≥ 1, sensitivity is prioritized and admissions would be increased. Increasing the cutoff to ≥2 would result in fewer admissions while preserving sensitivity (compared to actual practice). The underlying question remains as to whether reducing (or increasing) admissions will result in fewer adverse events. Ideally, we would like to see a prospective, randomized implementation trial to understand the true benefits of harms of using this scale in practice. I would use it as a way of supporting my decision in the EMR, the same way I use the HEART score when I think a patient can be safely discharged: “The patient has low-risk chest pain, with a HEART score of 2, and is comfortable being discharged to follow up with primary care.” In addition to efforts to externally validate the Ottawa CHF “tool,” I envision two research priorities that I've not seen suggested in the article or by other bloggers. First, there seems to be a natural fit for this risk stratification instrument to identify ED patients most amenable and likely to benefit to community paramedicine interventions for CHF (http://pmid.us/26169927). Second, the hierarchy of diagnostic evidence holds that further validation attempts demonstrate patient-centric benefits in applying the Ottawa instrument. In other words, do patients at higher risk for adverse outcomes actually benefit from admission or does admission just mean that the adverse outcomes occur in the hospital rather than at home? I'm also hoping to see a head-to-head prospective comparison of this new instrument with others like the new Lee et al. instrument that the authors discuss.” Thanks for the great feedback. Creating new decision tools is a multi-step process and we have only completed 2 steps, derivation and validation. Next step will be implementation to see if patients are really better off when the OHFRS is incorporated into practice. Finally there is the KT phase of spreading the word. Where we are now is a scale that has been validated but not yet tested. Clinicians are free to use it now because there are no other clinically validated tools out there for AHF in the ED. We know that U.S. practice is hugely different than that in Canada with the majority of AHF patients being admitted. In Canada we send many patients home but that decision is often based on gestalt. Now we have provided a bit evidence to help understand the medical risk. Other factors obviously include the opportunity to optimize outpatient therapy, early follow-up, and strong home support. Stay tuned for more AHF evidence from Canada. The SAEs are really a composite endpoint—this is understandable in terms of creating a predictor model/Clinical Decision Rule—however patient may care less about re-admission than about death, ventilation or PCI. A sensitivity analysis for just death, ICU admission or revascularisation would provide insight in the robustness of the rule. Twitter Poll Paper-in-a-pic from Kirsty Challen, @EMOttawa The OHFRS can be used to risk stratify acute heart failure patients at the bedside. Higher scores correlate to a higher risk of significant adverse events. A score of <1 denotes a patient at low risk of SAEs, while scores of ≥1 correlate with higher risk. Adding the results of a NT-proBNP increases sensitivity even further, but also results in increased admissions.