Case:

It’s 5pm on an average shift when you encounter a 58 year-old male with a history of CAD, CHF & HTN who complains of 4 days’ progressive dyspnea on exertion and lower extremity edema. He has been taking his furosemide as prescribed, but admits to dietary indiscretions over the recent holiday weekend.

Vital signs on presentation are: T 36.8, HR 78, BP 165/98, RR 18 & SpO2 94%(RA)

On examination he is in no distress, has no visible JVD, cardiac murmurs or gallop. He is breathing comfortably but has fine crackles in the bilateral bases. There are B-lines in bilateral lung bases on pulmonary ultrasound.

EKG shows no ischemic changes. CXR shows mild pulmonary edema. Laboratory results are significant for normal renal function, an undetectable troponin, a BNP of 300, and CO2 34.

You administer a dose of IV furosemide, and wonder if home with outpatient follow-up is a reasonable disposition, so you perform a literature search to investigate this question.

Background:

There are nearly 1 million US ED visits for acute heart failure annually, with more than 80% of patients requiring hospital admission (1). These admissions represent a significant burden on the health care system and are expected to rise due to the aging population (2).

Despite these concerns, there has been relatively little research into effective strategies to reduce admission rates from the ED.  Discharge is challenging as previous studies have previously reported increased risk for readmission and death (3).

There have been a small number of both retrospective studies and large database studies that have attempted to identify risk factors for poor outcomes, but there has not been a widely accepted clinical decision tool developed to aid physicians in identifying low risk patients for discharge (4). Two recent studies attempted to risk stratify these heart failure patients and create a prediction tool to aid in discharge.

Reference 1: Identification of Emergency Department Patients With Acute Heart Failure at Low Risk for 30-Day Adverse Events: The STRATIFY Decision Tool (5)

A. Evaluation of Study Design

  1. What is the Study design?: Prospective observational cohort study
    1. What is the Population? 2 University EDs and 2 community EDs in Nashville and Cincinnati
      1. Who was Excluded?
        1. Patients not willing to take part in the study
        2. <18 yrs of age
        3. Should be noted that low BNP <100 was not used to exclude pts
      2. What was the Intervention?
        1. No specific intervention; study created to validate a clinical decision rule
      3. What was the Comparison?
        1. None
      4. What were the Outcomes?
        1. Adverse events over a 5 day and 30 day period
      5. Will the rule serve a purpose if it is valid? Does the rule make clinical sense?
        1. Yes, if external validity is shown. This is due to the large health care cost and pressure put on providers regarding AHF
      6. Is the outcome of interest clearly defined and clinically important?
        1. Yes the outcomes are clearly defined. This is clinically important in order to help ease the burden of AHF cases in regards to health care cost
      7. Are the predictors clearly defined, sensible, reliable and reproducible as determined by different clinicians?
        1. Predictors were clearly defined; ie elevated troponin and BUN. Should be noted that early adverse events were shown to be related to ACS. Later adverse events were related with death
      8. How did the authors construct the list of potential predictors?
        1. They attempted to study variables that were easily attainable in the ED.
      9. Were those who assessed predictors blinded to outcomes?
        1. Those assessing patients were blinded to inpatient medical records, however those that were admitted to the hospital were still followed as inpatients
      10. Was the study site described?
        1. Yes, 2 academic sites and 2 community sites. 1 in Nashville and3 in Cincinnati
      11. Was the data collected prospectively?
        1. Yes

B. Results

1.     What were the results? Does the rule have predictive power?

a.     In patients with risk thresholds of 1%, 3%, 5% and 10%, found that 0%, 1.4%, 13%, and 49.5% of pts were considered low risk

b.     7% of patients experienced a 5 day event and 12% of patients experienced a 30 day event

i.     88% of patients experienced no adverse events

c.     negative predictive values are 100%, 96%, and 93% for identifying true low-risk patients at 3%, 5%, and 10% risk of death or subsequent adverse events.

d.     Elevated troponin and BUN were found to be significant predictors of adverse events

2.     Was there an adequate sample size, including an adequate number of positive outcomes?

a.     1033 subjects followed that were believed to be presenting with AHF.

C. Validity

1.     Was the rule/tool validated in a separate population?

a.     No. Study only found internal validity and did not include separate populations

2.     Was the rule consistent and valid when applied?

a.     study did not prospectively apply the rule

D.  Has an impact analysis been performed? No

 

 

Reference 2: Prospective and Explicit Clinical Validation of the Ottawa Heart Failure Risk Scale, With and Without Use of Quantitative NT-proBNP (6)

 

  1.  Evaluation of Study Design

    1. What is the Study design?
      1. Prospective observational cohort study
    2. What is the population? 1100 patients with “acute heart failure” at six tertiary hospital EDs
      1. >50 years old
      2.  Chief complaint of SOB, due to acute heart failure (European Society of Cardiology – underlying abnormality and appropriate symptoms and signs of fluid retention)
    3. Who was excluded?
      1. Those who did not meet acute acute heart failure definition (many excluded because their SOB lasted >7 days, no clear heart failure on CXR)
      2. too ill to consider discharge (<85% SpO2 resting on baseline O2, HR >120, SBP <85, confusion/disorientation/dementia, MI, STEMI, terminal illness weeks, nursing home/chronic care faility, previously enrolled, chronic hemodialysis)
      3. What was the intervention?
        1. Application of Ottawa Heart Failure Risk Score (OHFRS)
      4. What was the comparison?
        1. Clinical practice
      5. What were the outcomes? Primary: SAE, Secondary: OHFRS- accuracy (physician vs criterion interpretation by steering committee), acceptability, potential impact)
        1. 30-day all cause mortality from ED visit
        2. 14-day admission to ICU/stepdown, new NIV or ETT, MI, major procedure, return to ED for related problem requiring admission
    4. Will the rule serve a purpose if this is valid? Does the rule make clinical sense?
      1. Yes, no clear guideline for which patients needs admission
      2.  Potential for substantial resource saving if discharge rather than admission
    5. Is the outcome of interest clearly defined and clinically important?
      1. Combined endpoint with some clinically relevant components (death) and some with questionable relevance (readmission)
    6. Are the predictors clearly defined, sensible, reliable and reproducible as determined by different clinicians?
      1. Poorly reproducible by different clinicians? agreement 59% for category, 95% category +/- 1
    7. How did the authors construct the list of potential predictors?
      1. Previous  OHFRS with and without NT-proBNP (in initial OHFRS, multivariate analysis with stepwise logistic regression analysis to determine independent predictors of SAEs)
    8. Were those who assessed predictors blinded to outcomes?
      1. those who assessed outcomes were blinded to predictors (319)
      2. Steering committee who made criterion interpretation blinded to outcomes (319)
      3. Did having clinicians perform the risk scale change what they would have done by normal judgement? Potential confounder?
    9. Was the study site described? Yes, canadian teruary care EDs
    10. Was the data collected prospectively? Yes
  2.  Results

    1. What were the results? Does the rule have predictive power?
      1. Compared to current actual practice, OHFRS score threshold >1 would improve sensitivity for SAEs with tradeoff of more admissions; threshold >2 would have similar sensitivity with similar admissions
      2. NT-proBNP values seem to improve sensitivity
      3. Without NT-proBNP,   +LR at threshold >1 was 1.22, -LR 0.33, +LR at >2 was 1.61, -LR 0.52
      4. with NT-proBNP, +LR at >1 was 0.11, -LR 0.31, +LR at >2 was 1.34, -LR 0.50
      5. LR not overly convincing as -LR is greater than 0.1
    2. What were the statistical tests used and analysis appropriate?
      1.   Using criterion interpretation, calculated SAE rate, admission rate, sensitivity, and specificity at thresholds defined with cutoff at each OHFRS score. Interestingly, paper does note that OHFRS score was originally designed to present a calculator of risk rather than cut point to guide admissions, though this does become part of the interpretation in this paper.
      2.  Notably, much higher admission rate and SAE rate in 2017 paper than 2013 paper where rule was developed, which might be due to clinicians using the OHFRS in their clinical decision making.
    3. Was there an adequate sample size, including an adequate number of positive outcomes?
      1. Power calculation for goal sample size are not explicitly mentioned
      2. NT-proBNP sample size limited by whether test was performed.

C.    Validity

a.     Was the rule/tool validated in a separate population? Yes

b.     Was the rule consistent and valid when applied? Not consistent between providers using; also, providers were advised not to make their decisions based on rule

D.    Has an impact analysis been performed?

a.     Did clinicians use the rule? Advised not to use rule as only decision maker.

b.     Did the rule work in practice? Unclear what uptake would be; seems to have issues with reproducibility between physicians

 

Discussion:

Patients with a history of heart failure have a baseline risk of adverse events (death, ACS, intubation, dialysis) that would be unacceptable in a healthy patient. These patients are also complex, and the decision of whether they are safe for discharge cannot be easily pared down to a few simple clinical factors. The adequacy of follow up, and ability of patients to reliably administer home medications also must be considered when making the decision to discharge. It is exceptionally difficulty to factor these multi-factorial issues into a safe and accurate clinical prediction score.

Both of these Clinical Decision Tools attempt to distill this complex disease into a few variables. The Stratify Decision Tool represents a US population from Cincinnati and Nashville, but has not been externally validated. Its tool is also complex and requires the use of a logarithmic scale to calculate a score, which is onerous to use while on shift. When used, this tool does not identify a large enough population that is safe for discharge. The 3% risk of adverse event group only rules out 1.4% of studied patients. If you apply the 5% risk of adverse event group, this would rule out 13.4% of patients, However, this cut off does not safely exclude adverse events, with a negative LR of 0.34.

The Ottawa Heart Failure Risk Scale has the benefit of being externally validated. Its score is also much simpler to apply. However, in the external validation it was shown to be hard to implement, with only 59% of physicians accurately calculating the correct score for their patients.  The study also excluded a significant proportion of patients such as patients with > 7 days of symptoms and patients from nursing homes. These populations represent a significant proportion of our patients and exclusion limits the overall applicability of the tool. When looking at outcomes, these included patient centered outcomes such as death, but also included readmission, which has questionable significance. In this validation study, the score has similar test characteristics to the STRATIFY Tool, with a cut off of >1 having a negative likelihood ration of 0.31 and would allow for 22.4% of patients to be discharged. If we assume that some of these adverse events are preventable with admission, then neither of these clinical decision rules effectively rule out the risk of adverse events.

One of the limitations of both papers is determining what adverse events could have been prevented. Due to the baseline heath of these patients, it is unknown if admitting these patients can prevent or even change their risk of adverse events. The superior study to determine if these events are preventable is a large RCT. Until an RCT is performed, we must assume that some of these events are preventable, and neither tool safely rules out the risk of an adverse event.

Both tools provide a framework for providers to use when risk stratifying these patients and might have a use in shared decision making among the patients and providers when determining the need for admission. However, neither tool can safely rule out the risk of an adverse event at the above cut offs. Future trials should focus on whether admission changes the risk of adverse events as well as cost effectiveness.

Bottom Line:

Acute Heart failure is a complex presentation with a myriad of variables that factor into a clinician’s disposition decisions. Many of these, such as a patient’s ability to take their home medications and availability of follow up, cannot be captured in a decision rule. Both of these decision rules make a valiant effort to distill many of the clinical variables into a usable algorithm and have created an acceptable risk prediction score. Use of this risk prediction score may be used to aid in shared decision-making, but at this time, both are not accurate enough to determine disposition decisions in isolation.

References

  1. Storrow AB, Jenkins CA, Self WH, Alexander PT, Barrett TW, Han JH, et al. The burden of acute heart failure on U.S. emergency departments. JACC Heart Fail. 2014;2(3):269-77.
  2. Benjamin EJ, Blaha MJ, Chiuve SE, Cushman M, Das SR, Deo R, et al. Heart Disease and Stroke Statistics—2017 Update: A Report From the American Heart Association. Circulation. 2017.
  3. McCausland JB, Machi MS, Yealy DM. Emergency physicians’ risk attitudes in acute decompensated heart failure patients. Acad Emerg Med. 2010;17(1):108-10.
  4. Auble TE, Hsieh M, Gardner W, Cooper GF, Stone RA, McCausland JB, et al. A prediction rule to identify low-risk patients with heart failure. Acad Emerg Med. 2005;12(6):514-21.
  5. Collins SP, Jenkins CA, Harrell FE, Jr., Liu D, Miller KF, Lindsell CJ, et al. Identification of Emergency Department Patients With Acute Heart Failure at Low Risk for 30-Day Adverse Events: The STRATIFY Decision Tool. JACC Heart Fail. 2015;3(10):737-47.
  6. Stiell IG, Perry JJ, Clement CM, Brison RJ, Rowe BH, Aaron SD, et al. Prospective and Explicit Clinical Validation of the Ottawa Heart Failure Risk Scale, With and Without Use of Quantitative NT-proBNP. Academic Emergency Medicine. 2017;24(3):316-27.

 

Co-Authors: Vivian Lam, MD, Matthew Macias, MD

Faculty Reviewer: Brendan Byrne, MD

(Visited 54 times, 1 visits today)
mm

Alex Beyer, MD

EM Resident Class of 2019

More Posts

Follow Me:
Twitter

Leave a Reply

Your email address will not be published. Required fields are marked *