Assessing Prognosis Using Risk Scores
I. Risk stratification in Heart Failure: What every physician needs to know.
Myriad aspects of a patient's medical history impact his or her prognosis after a diagnosis of heart failure. In the general population, reasonably accurate predictions of life expectancy can be made based solely on a patient's age and gender.
In the heart failure population, risk prediction is more challenging, with the severity of the heart failure having much more impact on expected survival than simple patient status as younger versus older. Factors that affect patient survival include the patient's demographics and history (particularly, the etiology of the patient's heart failure), functional status (as determined by New York Heart Association classification or exercise testing), results of imaging studies (including ejection fraction or severity of valvular disease as measured by echocardiography), laboratory markers (e.g., serum hemoglobin, serum sodium, renal function), and biomarkers (e.g., B-type natriuretic peptide, troponin, ST2, gal-3).
Multivariate risk scores are generated by combining a selection of these individual markers, generally using a logistic regression or Cox proportional hazard model. Overall, they provide a powerful means of predicting the likelihood of outcomes in heart failure patients, making them quite valuable in selecting patients for advanced therapies and in helping patients and their families plan for the future. This chapter will focus on a selection of the more commonly used risk models and their clinical applications.
Univariate risk predictors
Many of the individual risk factors that predict patient outcomes are intuitive. Reduced ejection fraction, in particular <45%, is associated with progressively increased mortality. In outpatient populations, around a 40% increase in mortality has been seen with each 10% reduction in ejection fraction.
Ejection fraction has not, however, been as predictive of outcomes in hospitalized cohorts. Each one unit increase in New York Heart Association class is associated with a doubling of mortality; an NYHA IV patient has eight times the mortality of an NYHA I patient. The etiology of a patient's heart failure can be a powerful predictor, with infiltrative etiologies of heart failure auguring as much as a 400% increase in risk of mortality as compared to idiopathic cardiomyopathies.
Abnormalities of many biomarkers predict worsened outcomes. Peak consumption of oxygen with maximal exertion (peak VO2) has been used as a criterion for selection of heart transplantation candidates since the demonstration that patients with a value >14 ml/kg/min could be expected to have as good of a survival rate with medical management as with cardiac transplantation.
More recent studies have suggested a nearly 30% increase in mortality for each 1 unit decrease in peak VO2in patients receiving beta-blockers and argued for a lower value of peak VO2(in the range of 10 ml/kg/min) to qualify for transplant listing. High-sensitivity troponins have shown predictive power, as have newer markers such as ST2 or gal-3.
The most widely validated lab marker is B-type natriuretic peptide (BNP), with a doubling of mortality risk seen for normal versus abnormal range BNP. BNP also correlates with other important clinical outcomes, such as increased risk of hospitalization with progressively more abnormal values.
Hyponatremia or hypernatremia, along with polycythemia or anemia, are predictors of worsened outcomes, with each 1 g/dl reduction in hemoglobin associated with a 20% increase in multivariate adjusted risk of death. Some clinical data points show a reverse epidemiology, with lower body mass index (BMI) and cholesterol levels serving as favorable prognostic markers in the general population, but as unfavorable prognostic markers in the heart failure population.
Each 1 kg/m2increase in BMI portends a 5% increase in risk of developing heart failure, while for a patient who already has heart failure each 5 kg/m2 increase in BMI suggests a 10% reduction in mortality.
Multivariate Risk Predictors
I. The Seattle Heart Failure Model (SHFM)
The SHFM was developed as a predictor of life expectancy and 1-, 2-, and 5- year mortality in heart failure patients. It was derived in the Prospective Randomized Amlodipine Survival Evaluation (PRAISE1) trial of amlodipine and has since been validated in tens of thousands of patients across dozens of studies.
SeattleHeartFailureModel.org features an easy to use web calculator. There is also a downloadable application that is available for Windows, Macintosh, and several portable platforms.
The risk score was developed via Cox proportional hazards modeling. It is scored as a continuous variable that can be converted to a percentage chance of survival at various time points. The most recent version includes a differential benefit for patients with implantable cardiac defibrillators based on the observation that very high-risk patients do not show improved mortality after implantable cardioverter-defibrillator (ICD) placement (>25% annual mortality). A screenshot of the Windows version of the web calculator is shown in
Seattle Heart Failure Model Calculator
Assessing prognosis can be quite challenging. Some patients are clearly floridly decompensated, barely ambulatory, and in need of inpatient management for acute exacerbation of their heart failure.
Others have minimal functional decline or evidence of laboratory or hemodynamic dysfunction and have a much better prognosis. Multiple studies have shown that neither patients nor providers are particularly good at providing accurate assessments of expected mortality.
The value of the SHFM is in allowing more accurate risk stratification than allowed by holistic clinician assessment. Decision points include 1-year expected mortality of >30% in an NYHA Class IV heart failure patient with an ejection fraction <25% as a generally accepted criteria for ventricular assist device placement (some downward revision of the required level of annual mortality is underway as survival with a ventricular assist device increases), while 10% to 20% or greater annual mortality is generally considered to be at the beginning of the appropriate range for consideration of heart transplantation.
An annual mortality of >20% is the level at which ICD placement appears to no longer confer a mortality benefit in heart failure patients as the predominant mode of death is pump failure rather than sudden cardiac death. The value of the SHFM in lower risk patients is both in providing more knowledge for patients and providers about likely prognosis, and in allowing a more clear understanding of the survival changes that can occur with starting new heart failure medications or placing life-saving devices, such as implantable cardiac defibrillators or cardiac resynchronization therapy.
II. The Heart Failure Survival Score (HFSS)
Peak consumption of oxygen with maximal exertion has long been used as a criterion for listing for heart transplantation. The HFSS was developed to integrate additional clinical information and provide more accurate risk stratification. It divides patients into high-, intermediate-, and low-risk strata.
The HFSS is calculated using seven clinical variables. The score is calculated as 0.0464 * (left ventricular ejection fraction in %) + 0.0255 * (mean blood pressure in millimeters of mercury) + 0.0546 * (peak consumption of oxygen with maximal exertion in ml/kg/min) + 0.0470 * (serum sodium in mmol/L) - 0.0216 * (resting heart rate in beats per minute). 0.6931 is subtracted from the score if the patient has ischemic cardiomyopathy, and 0.6083 is subtracted from the score if the patient has intraventricular conduction delay.
Scores less than 7.2 are classified as high risk; 8.1 or above are low risk; values in between are categorized as intermediate risk. In the late 1990s validation cohort, survival was 88%, 60%, and 35% in the low-risk, intermediate-risk, and high-risk groups. More recent studies have suggested 89%, 72%, and 60% 1-year survival rates for the respective risk groups.
Use of the HFSS is limited by the requirement for peak consumption of oxygen with maximal exertion to calculate the score. As such, its use has primarily been in patients presenting to transplantation/specialized heart failure clinics.
Patients in the medium- and high-risk category are appropriate for consideration of listing for heart transplantation or placement of ventricular assist device, while low-risk patients are appropriately deferred for consideration of these advanced heart failure therapies. There is a version available for purchase for Android and iPhones at http://www.mediquations.com
III. The Enhanced Feedback for Effective Cardiac Treatment (EFFECT) model
Approximately 4,000 patients hospitalized with acute decompensation of heart failure in Ontario, Canada, were used to derive and validate a model for predicting 30-day and 1-year all-cause mortality for hospitalized inpatients. It has been prospectively validated in other cohorts.
The EFFECT model can be found at http://www.ccort.ca/Research/CHFRiskModel.aspx. The model looks at a patient's age, respiratory rate, systolic blood pressure, hemoglobin, blood urea nitrogen, sodium, and presence of five comorbid conditions (history of stroke, dementia, chronic obstructive pulmonary disease, hepatic cirrhosis, and cancer).
The multiple logistic regression derived score is an integer, with risk falling into strata that range from a low-risk strata of <60 points, corresponding to 0.4% 30-day and 7.8% 1-year mortality, to a high-risk strata of >150 points, corresponding to 59% 30-day and 78.8% 1-year mortality.
Clinicians have the same pitfalls in predicting mortality in hospitalized patients as in ambulatory patients. Identifying a high-risk cohort of patients allows more quantitative decision-making on discharge from the emergency department versus admission to the floor versus admission to the intensive care unit. Quantitative survival projections can help to facilitate discussions about end-of-life care versus transplant and left ventricular assist device placement in the highest-risk patients.
IV. The Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE) risk model and discharge score
Four hundred twenty-three inpatients receiving Swan-Ganz catheterization for management of their advanced heart failure decompensation were studied in North America. Clinical information from the end of their hospital stays was used to derive a discharge model for 6-month risk of rehospitalization and mortality.
The ESCAPE risk score was derived with a logistic regression model. Points are summed up based on the presence of each of the following risk factors:
Age > 70 yr...........................................1
BUN 41-90 mg/dl...................................1
BUN >90 .............................................2
6-min walk <300 feet..............................1
Sodium <130 mEq/L..............................1
Received CPR or mechanical ventilation..2
Discharge daily diuretic dose >240 mg....1
Not on beta-blocker at discharge.............1
Discharge BNP 501-1,300 pg/mmol.........1
Discharge BNP >1,300 pg/mmol.............4
A risk score of 0 corresponds to 6-month mortality of 7%; 1 to 2 corresponds to 10% to 17%, 6-month mortality; 3 to 4 corresponds to 26% to 45%, 6-month mortality; 5 and higher (approximately 5% of the patients in the study) corresponds to 75% to 100%, 6-month mortality.
Patients in the hospital and recently discharged from the hospital are the highest-risk heart failure patients. Recent discharge from the hospital is a strong predictor of patients who will present to the hospital for readmission. A discharge risk score allows better identification of the highest risk patients and potential targeting for closer follow-up and more aggressive monitoring in the postdischarge period.
V. Other risk scores
Many other risk scores have been reported in the literature. A sampling of other risk scores is presented below.
A. The Acute Decompensated Heart Failure National Registry (ADHERE) risk score
The ADHERE risk score was developed and validated in a registry of 65,275 patients from several hundred medical centers across the United States. A simple tree model (http://jama.ama-assn.org/content/293/5/572/F1.expansion.html) requiring only the variables of blood urea nitrogen, creatinine, and systolic blood pressure, along with a logistic regression derived formula with greater accuracy for risk stratification, were developed to predict in-hospital mortality for this inpatient population. The log odds of mortality are calculated as follows:
0.0212 * (blood urea nitrogen) - 0.0192 * (systolic blood pressure) - 0.0131 * (heart rate) - 0.0288 * (age) - 4.72
B. Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) risk profiles
These qualitative descriptions of patient risk profiles have been suggested as strata to categorize patients being evaluated for left ventricular assist device placement. The profiles of risk are as follows:
1. Life-threatening hypotension and failure of organ perfusion
2. Declining end-organ function and volume status on intravenous vasoactive medications
3. Dependence on intravenous vasoactive medications or temporary circulatory support
4. Fatigue at rest or with minimal exertion despite aggressive oral pharmacotherapy
5. Fatigue with anything more than the simplest activities of daily living
6. Fatigue with a few minutes of any meaningful activity
7. Advanced New York Heart Association class III symptoms
C. The Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF) risk score
Patients (48,612) from the OPTIMIZE-HF inpatient registry were enrolled and used to develop a nomogram (http://www.optimize-hf.org/art/OPT-Mortality.pdf) for prediction of in-hospital mortality. Variables include age, heart rate, systolic blood pressure, serum sodium, serum creatinine, primary cause of patient admission, and presence of left ventricular systolic dysfunction.
The model has been validated with patient outcomes from the AHDERE registry. A small subset of the patients from the full registry was used to derive a different nomogram to predict 60- to 90-day postdischarge mortality.
D. The Irbesartan in Heart Failure with Preserved Ejection Fraction (I-PRESERVE) risk scores
The I-PRESERVE investigators looked at 4,128 patients with preserved ejection fraction and used Cox modeling to develop several scoring systems for prediction of cardiovascular hospitalization, all-cause mortality, and heart failure death or hospitalization.
The models developed to predict the different outcomes did not include the same variables, but the most powerful predictors included log N-terminal pro-B-type natriuretic peptide, age, diabetes mellitus, and history of hospitalization for the composite end-point of all-cause mortality or cardiovascular hospitalization. For all-cause mortality specifically, left ventricular ejection fraction proved a more powerful predictor than previous heart failure hospitalization.
E. The candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity (CHARM) program risk models
The CHARM program evaluated 7,599 patients with Cox regression modeling to develop risk scores to predict all-cause mortality and a composite endpoint of cardiovascular death and heart failure hospitalization. There were 21 predictor variables included in each model, with the most powerful variables being older age, diabetes mellitus, and reduced left ventricular ejection fraction. Interestingly, increased mortality was not observed until patients reached age 60 or had a left ventricular ejection fraction below 45%.
What's the Evidence?
Levy, WC, Lee, KL, Hellkamp, AS. "Maximizing survival benefit with primary prevention implantable cardioverter-defibrillator therapy in a heart failure population". Circulation. vol. 120. 2009. pp. 835-42.
Ketchum, ES, Moorman, AJ, Fishbein, DP. "Predictive value of the Seattle Heart Failure Model in patients undergoing left ventricular assist device placement". J Heart Lung Transplant. vol. 29. 2010. pp. 1021-5.
Aaronson, KD, Schwartz, JS, Chen, TM, Wong, KL, Goin, JE, Mancini, DM. "Development and prospective validation of a clinical index to predict survival in ambulatory patients referred for cardiac transplant evaluation". Circulation. vol. 95. 1997. pp. 2660-7.
Lee, DS, Austin, PC, Rouleau, JL, Liu, PP, Naimark, D, Tu, JV. "Predicting mortality among patients hospitalized for heart failure: derivation and validation of a clinical model". JAMA. vol. 290. 2003. pp. 2581-7.
Fonarow, GC, Adams, KF, Abraham, WT, Yancy, CW, Boscardin, WJ. "Risk stratification for in-hospital mortality in acutely decompensated heart failure: classification and regression tree analysis". JAMA. vol. 293. 2005. pp. 572-80.
O'Connor, CM, Hasselblad, V, Mehta, RH. "Triage after hospitalization with advanced heart failure: the ESCAPE (Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness) risk model and discharge score". J Am Coll Cardiol. vol. 55. 2010. pp. 872-8.
Stevenson, LW, Pagani, FD, Young, JB. "INTERMACS profiles of advanced heart failure: the current picture". J Heart Lung Transplant. vol. 28. 2009. pp. 535-41.
Abraham, WT, Fonarow, GC, Albert, NM. "Predictors of in-hospital mortality in patients hospitalized for heart failure: insights from the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF)". J Am Coll Cardiol. vol. 52. 2008. pp. 347-56.
Komajda, M, Carson, PE, Hetzel, S. "Factors associated with outcome in heart failure with preserved ejection fraction: findings from the Irbesartan in Heart Failure With Preserved Ejection Fraction Study (I-PRESERVE)". Circ Heart Fail. vol. 4. 2011. pp. 27-35.
Pocock, SJ, Wang, D, Pfeffer, MA. "Predictors of mortality and morbidity in patients with chronic heart failure". Eur Heart J. vol. 27. 2006. pp. 65-75.
Mancini, D, Lietz, K. "Selection of cardiac transplantation candidates in 2010". Circulation. vol. 122. 2010. pp. 173-83.
Hunt, SA, Abraham, WT, Chin, MH. "2009 focused update incorporated into the ACC/AHA 2005 Guidelines for the Diagnosis and Management of Heart Failure in Adults: a report of the American College of Cardiology Foundation/American Heart Association task force on practice guidelines: developed in collaboration with the International Society for Heart and Lung Transplantation". Circulation. vol. 119. 2009. pp. e391-479.
Ketchum, ES, Levy, WC. "Establishing prognosis in heart failure: a multimarker approach". Progress in cardiovascular diseases.. vol. 54. 2011. pp. 86-96.
Copyright © 2017, 2013 Decision Support in Medicine, LLC. All rights reserved.
No sponsor or advertiser has participated in, approved or paid for the content provided by Decision Support in Medicine LLC. The Licensed Content is the property of and copyrighted by DSM.
Sign Up for Free e-newsletters
Regimen and Drug Listings
GET FULL LISTINGS OF TREATMENT Regimens and Drug INFORMATION
|Head and Neck Cancer||Regimens||Drugs|
|Renal Cell Carcinoma||Regimens||Drugs|
Cancer Therapy Advisor Articles
- Moving Targets: Off-Label Prescribing of Targeted Therapies
- Conditional 10-Year Breast Cancer Recurrence Rates Decline Regardless of Subtype
- New Algorithm Improves Outcome Prediction for Diffuse Large B-Cell Lymphoma
- Immunotherapy Combo in Head and Neck Cancer Shows Activity in Phase 2 Trial
- Trials, Collaboration Might Help Reduce Survival Disparities Between Community, Academic Cancer Centers
- Nutraceuticals/Supplements and Cancer Prevention: All Hype?
- Encorafenib Plus Binimetinib Offers a New Option for MEK+BRAF Inhibition
- Reclassification of Variants of Uncertain Significance: A Q&A With Theodora Ross MD, PhD
- POT1 Revealed as a Sarcoma Predisposition Gene
- Cabozantinib Shows Promising Activity in Osteosarcoma and Ewing Sarcoma
- Registration Trial to Look at HRAS-Directed Tipifarnib in HNSCC
- 1 in 5 Patients Readmitted Within 30 Days of Reconstruction Surgery for Head and Neck Cancer
- Treatment-Related Hyperprogression: Concerns Continue to Mount
- Activation-Induced Cytidine Deaminase Transcripts Linked To More Aggressive CLL
- Next-Generation Sequencing Helped Identify Pathways for Transformation of WM to DLBCL