Author + information
- Received December 7, 2016
- Revision received February 21, 2017
- Accepted February 21, 2017
- Published online June 14, 2017.
- Andreas P. Kalogeropoulos, MD, MPH, PhDa,∗ (, )
- Ayman Samman-Tahhan, MDa,
- Jeffrey S. Hedley, MD, MSca,
- Andrew A. McCue, MDa,
- Jonathan B. Bjork, MDa,
- David W. Markham, MD, MSca,
- Kunal N. Bhatt, MDa,
- Vasiliki V. Georgiopoulou, MD, MPH, PhDa,
- Andrew L. Smith, MDa and
- Javed Butler, MD, MPH, MBAb
- aDepartment of Medicine, Emory University, Atlanta, Georgia
- bDepartment of Medicine, Stony Brook University, Stony Brook, New York
- ↵∗Address for correspondence:
Dr. Andreas P. Kalogeropoulos, Emory Clinical Cardiovascular Research Institute, 1462 Clifton Road NE, Suite 535B, Atlanta, Georgia 30322.
Objectives This study sought to estimate the rate of progression to Stage D heart failure (HF) among outpatients with Stage C HF and to identify risk factors for progression.
Background The pool of patients who may be candidates for advanced HF therapies is growing.
Methods We estimated 3-year progression to clinically determined Stage D HF and competing mortality among 964 outpatients with Stage C heart failure with reduced ejection fraction (HFrEF), where ejection fraction is ≤40%.
Results The mean age of patients was 62 ± 15 years; 35% were women; 47% were white; 46% were black, and 7% were of other races; median baseline ejection fraction was 28% (25th to 75th percentile: 20% to 35%); and 47% had ischemic heart disease. After 3.0 years (25th to 75th percentile: 1.7 to 3.2 years), 112 patients progressed to Stage D (3-year incidence: 12.2%; 95% confidence interval [CI]: 10.2% to 14.6 annualized: 4.5%; 95% CI: 3.8% to 5.5%), and 116 patients died before progression (3-year competing mortality: 12.9%; annualized: 4.7%; 95% CI: 3.9% to 5.6%). By 3 years, 25.1% of patients (95% CI: 22.2% to 28.1%) had either progressed to Stage D or died (annualized: 9.2%; 95% CI: 8.1% to 10.5%]). Annualized progression rates were higher in black versus white patients (6.3% vs. 2.7%, respectively; p < 0.001), nonischemic versus ischemic patients (6.1% vs. 2.9%, respectively; p < 0.001), and in New York Heart Association functional class III to IV versus I to II patients (7.5% vs. 1.9%, respectively; p < 0.001) but were similar for men and women (4.7% vs. 4.2%, respectively; p = 0.53). Lower ejection fraction and blood pressure, renal and hepatic dysfunction, and chronic lung disease rates were additional predictors of progression. Predictors of competing mortality were different from those of disease progression.
Conclusions Among patients with Stage C HFrEF receiving care in a referral center, 4.5% progressed to Stage D HF each year, with earlier progression among black and nonischemic patients. These findings have implications for healthcare planning and resource allocation for these patients.
- advanced heart failure
- disease progression
- heart failure
- heart failure with reduced ejection fraction
Over time, a proportion of patients with ambulatory (Stage C) heart failure (HF) will develop advanced symptoms, usually defined as New York Heart Association (NYHA) functional class IIIB to IV, refractory to optimal guideline-recommended therapy, therefore reaching Stage D (also referred to as “advanced”, “refractory”, or “end-stage”) HF (1). Advanced HF symptoms are not exclusively encountered among patients with HF and reduced ejection fraction (HFrEF), who account for approximately 50% of HF cases (1). However, patients with Stage D HFrEF are a special population because these patients are candidates for advanced HF therapies, including orthotopic heart transplantation (OHT) and long-term mechanical circulatory support (MCS) (2).
Professional societies have long identified the gaps in the epidemiology of Stage D HF (3,4). However, it was the introduction of left ventricular assist devices as destination therapy that has generated intense focus on the pool of candidates for MCS because of the health care system and public health implications (5). Currently, the estimates of the pool of potential left ventricular assist device candidates are rough projections (1), partially because estimates of Stage D HF prevalence have varied widely (6,7) and partially because the candidate population is a moving target (8). One important knowledge gap is the lack of data for rates and risk factors for progression to Stage D HF among patients with stable Stage C HFrEF patients. This information would be important for referral centers in order to counsel patients and families and communicate expectations appropriately and plan for adequate resources. Estimating the future pool of potential candidates for advanced HF therapies would be important from a public health perspective also.
In this work, we estimated the rate of progression from Stage C to Stage D HF in patients receiving outpatient HF care in an academic center over 3 years of follow-up and evaluated clinical risk factors that are associated with progression to Stage D HF. In addition, we investigated clinical risk factors that are associated with competing mortality in these patients.
We reviewed the records of consecutive adult (age ≥18 years) outpatients who received care associated with International Classification of Diseases-9th Revision-Clinical Modification (ICD–9-CM) codes 402.X1, 404.X1, 404.X3, and 428.XX (9) between January 1, 2012, and March 31, 2012, in Emory Healthcare by cardiologists, including HF specialists. The inception timeframe was selected to allow for ≥3 years of follow-up. Figure 1 summarizes the cohort selection process. We verified HF diagnosis based on documentation of symptoms, signs, and guideline-based treatment. We defined HFrEF as HF with current left ventricular ejection fraction (LVEF) of ≤40% (4). We excluded patients with: 1) specific cardiomyopathies (e.g., characterized as hypertrophic, stress-induced, infiltrative, restrictive, chemotherapy-induced); 2) complex congenital heart defects; 3) primary valvular disease; and 4) primary right-sided disease (e.g., right ventricular cardiomyopathy, Class I pulmonary arterial hypertension). The final cohort consisted of 964 patients. The Emory Institutional Review Board approved the study.
Baseline characteristics and conditions
Baseline demographic, anthropometric, and vital sign data were collected through the Clinical Data Warehouse from the index clinic visit record. For laboratory values and medications, we extracted the last information available up to the index visit date. The presence of baseline chronic conditions was adjudicated and categorized according to the latest (2014) algorithm proposed by Chronic Conditions Data Warehouse (10), which is used to analyze Medicare data.
Definition of stage D heart failure
In the absence of a widely acceptable definition of Stage D HF, progression to Stage D HF was based on clinical assessment from treating physicians and independent verification of patient status by 2 investigators. A patient was considered to have progressed to Stage D HF if the patient assessment concluded that the patient was in “advanced” or “Stage D” HF and/or the patient was scheduled for evaluation for advanced HF therapies, including OHT, long-term MCS (either as bridge-to-transplant or destination therapy), or home inotrope therapy as palliative care. Need for long-term inotropes as bridge therapy to OHT or MCS or decision was also considered a corroborating indicator of progression to Stage D HF. Two investigators independently reviewed each patient record for patient status and time of progression to Stage D HF, and a third investigator with experience in advanced HF resolved discordant assessments.
Additional outcomes data
Competing mortality and post-Stage D HF outcomes (death, OHT, MCS, and palliative care) data were collected through April 30, 2015, thus providing for a minimum of 3 years of follow-up for each living patient who continued to receive care at Emory Healthcare. For patients who were alive during the last Emory Healthcare encounter but did not continue to receive care in our system, the last encounter was considered the last date of follow-up for analysis purposes.
The primary endpoint was progression to Stage D HF in a time-to-event analysis accounting for the competing risk of death. The secondary endpoints were: 1) the combined endpoint of either death or progression to Stage D HF (time to event); and 2) competing mortality. The composite endpoint was selected to reflect progression rates from stable Stage C HF to unstable status.
We powered our study to demonstrate progression rates to Stage D HF within prespecified confidence intervals. Based on an interim analysis, the estimated 3-year rate of progression to Stage D HF was 13.0%. We therefore targeted a final sample size of 930 patients with 10% margin for loss to follow-up. This sample size would provide 80% power at the 2-sided α value of 0.05 to reject the hypothesis that the lower 95% confidence interval (CI) of 3-year progression to Stage D HF would cross the 10% boundary. Because data extraction and adjudication was done in monthly increments, the final sample size slightly exceeded the required sample size.
To identify risk factors for progression to Stage D HF (primary endpoint), taking into account competing mortality, we used Fine and Gray competing risks proportional hazards models (11), which are appropriate here because of the high rate of the competing event (12). We used the same approach for competing mortality. For the secondary endpoint of progression to Stage D HF or death, we used Cox proportional hazards models. We also report mortality as a noncompeting event (regardless of progression to Stage D or advanced HF therapies). There was no indication of nonproportionality in any model, as evaluated by interaction terms with time. The Breslow method was used to decide ties.
We considered the following covariates as potential predictors of progression to Stage D HF based on association with outcomes, clinical interpretation, and previous work: age, sex, race (white, black, other), body mass index, NYHA functional class, last reported LVEF, smoking status, presence of coronary artery disease and other vascular disease, comorbid conditions (e.g., hypertension, diabetes mellitus, depression, chronic lung disease and sleep apnea, chronic kidney disease, atrial fibrillation), blood pressure and heart rate, hematology, renal function, albumin, serum sodium and potassium, total bilirubin and cholesterol, and concomitant therapy for HF. We used stepwise backward regression to reach the final model for each outcome, with a p value of <0.10 as the threshold to retain a variable. To avoid overfitting, the Bayesian information criterion, which penalizes for unwarranted model complexity, was used for guidance. Restricted cubic splines and fractional polynomials were used to evaluate for nonlinear effects (13). To facilitate clinical interpretation, we used multivariate step functions to categorize continuous variables (14). However, we provide results for both continuous and dichotomized variable models. In addition, we converted the final regression model for HF progression to a simple score for practical application, using previously described methods (15). Additional details for the derivation and application of the score are provided in the Online Appendix. We used Stata version 14.1 software (StataCorp LP, College Station, Texas) for all analyses.
Baseline characteristics are presented in Table 1. The mean age was 62 ± 15 years; 35% were women; 47% were white, and 46% were black; median (25th to 75th percentile) LVEF was 28% (20% to 35%); 47% had ischemic heart disease; and 49% were in NYHA functional class I to II, whereas 51% were in NYHA functional class III to IV. Beta-blockers were used in 90% of patients, angiotensin-modulating agents in 72%, and aldosterone antagonists in 28% of patients; 48% had an implantable cardioverter-defibrillator; and 20% were receiving cardiac resynchronization therapy.
Progression to stage D heart failure
The median follow-up was 3.0 years (25th to 75th percentile: 1.7 to 3.2 years). More than 90% of patients had at least 1 year of follow-up. During that time, 112 patients were deemed to have progressed to Stage D, and 116 died before progressing to Stage D HF. The 3-year incidence of progression to Stage D HF (primary endpoint) after accounting for competing mortality was 12.2% (95% CI: 10.2% to 14.6%) with an annualized rate of 4.5% (95% CI: 3.8% to 5.5%) (Figure 2A). Unadjusted rates of progression to Stage D were higher among blacks than among whites (6.3%/year vs. 2.7%/year, respectively; p < 0.001), higher in those with HF of nonischemic than ischemic causes (6.1%/year vs. 2.9%/year, respectively; p < 0.001), and higher in those with NYHA functional class III to IV than in those with functional class I to II symptoms (7.5%/year vs. 1.9%/year, respectively; p < 0.001). There were no differences between men and women (4.7%/year vs. 4.2%/year, respectively; p = 0.53).
Composite of progression to stage D HF or death
The rate of the composite of progression to stage D HF or death by 3 years (secondary endpoint) was 25.1% (95% CI: 22.2% to 28.1%) with an annualized rate of 9.2% (95% CI: 8.1% to 10.5%) (Figure 2B). The rates of progression to stage D HF or death were similar between blacks and whites (10.2%/year vs. 8.5%/year, respectively; p = 0.19) and between nonischemic and ischemic patients (10.2%/year vs. 8.2%/year, respectively; p = 0.14). Also, there were no differences between men and women (9.1%/year vs. 9.3%/year, respectively; p = 0.89). However, the composite endpoint rates were significantly higher in patients with baseline NYHA functional class III to IV than in those with functional class I to II symptoms (13.2%/year vs. 5.2%/year, respectively; p < 0.001).
Predictors of progression to stage D HF
In models including the clinical characteristics described in Table 1, nonwhite race (mostly black), worse NYHA functional class, nonischemic cause, chronic lung disease requiring bronchodilators, lower LVEF and systolic blood pressure, and higher blood urea nitrogen and total bilirubin levels, but not sex, were independent predictors of earlier progression to Stage D HF (Table 2). Results were similar in models with continuous variables (Online Tables 1 and 2). To facilitate clinical application of this model for projections of disease progression within 3 years, we created a HF progression risk score (Box 1) based on the coefficient of the covariates described in Table 2. The score ranges from 0 to 10.5 (Table 3) and can be converted into a 3-year risk of disease progression by using the probabilities presented in Figure 3. In the subset of 749 patients (77.7% of cohort) with available B-type natriuretic peptide (BNP) data, BNP did not independently predict progression when added to the variables described in Table 2 (adjusted subhazard ratio per 100 pg/ml: 1.01; 95% CI: 0.99 to 1.03; p = 0.22). However, these findings should be viewed as suggestive, as these subsets may be subject to selection bias.
Predictors of progression to stage D HF or death
In adjusted models (Table 3), older age, worse NYHA functional class, nonischemic cause, chronic lung disease requiring bronchodilators, lower LVEF, systolic blood pressure, serum albumin, and higher serum blood urea nitrogen, creatinine, and total bilirubin levels were independent predictors of earlier progression to Stage D HF or death; use of angiotensin-modulating agents and beta-blockers was associated with lower composite event rates. Results were similar in models with continuous variables (Online Tables 3 and 4). In the subset with available BNP data, BNP did not independently predict the composite endpoint in models including the clinical predictors in Table 3 (adjusted hazard ratio per 100 pg/ml: 1.01; 95% CI: 0.99 to 1.02; p = 0.51).
The 3-year competing mortality rate (i.e., death before progression to stage D) was 12.9% (95% CI: 10.8% to 15.4%), with an annualized rate of 4.7% (95% CI: 3.9% to 5.6%) (Figure 2C). Among major subgroups, unadjusted competing mortality was significantly higher in patients with baseline NYHA functional class III to IV than in those with functional class I to II symptoms (5.7%/year vs. 3.4%/year, respectively; p = 0.007) and higher in whites than in blacks (5.8%/year vs. 3.9%/year, respectively; p = 0.039). There were no differences in mortality between men and women (4.4%/year vs. 5.2%/year, respectively; p = 0.41) or between nonischemic and ischemic patients (4.1%/year vs. 5.4%/year, respectively; p = 0.13). When evaluated as a noncompeting event (i.e., counting all deaths before or after progression and regardless of advanced HF therapies), 3-year mortality was 17.9% (95% CI: 15.4% to 20.7%), with an annualized rate of 6.7% (95% CI: 5.7% to 7.8%).
In adjusted models (Table 4), older age, worse NYHA functional class, history of hypertension, presence of atrial fibrillation, serum sodium, and creatinine levels were independent predictors of competing mortality; use of angiotensin-modulating agents and beta-blockers was associated with lower mortality. Sex, race, and ischemic cause did not independently predict competing mortality. Results were similar in continuous variable models (Online Tables 5 and 6). BNP was not an independent predictor of competing mortality when added to the variables described in Table 4 (adjusted subhazard ratio per 100 pg/ml: 1.00; 95% CI: 0.98 to 1.03; p = 0.67).
Outcomes among incident stage D HF patients
Among the 112 patients who progressed to Stage D HF, 7 received OHT, 21 long-term MCS, and 42 received palliative inotrope therapy at home as final disposition by the end of the 3-year period. In addition, 26 patients died before reaching a final disposition for Stage D HF, and 16 patients were still under evaluation for advanced therapies by the end of the 3-year period.
Several studies have reported rates and risk factors of mortality, alone or in combination with hospitalizations, in ambulatory patients with Stage C HF (16,17). Instead, we provide data for progression to Stage D HF among these patients. We observed that 12.2% of outpatients with Stage C HF receiving care in an academic center progressed to Stage D after 3 years, whereas the combined death or progression rate was 25.1%. We have identified several clinical risk factors for earlier progression: black race, nonischemic cause, lower baseline LVEF and systolic blood pressure, chronic obstructive pulmonary disease, and evidence of renal or hepatic dysfunction. These risk factors were different from those that predicted competing mortality in our cohort.
In our cohort, with a median 62 years of age, 1 of 8 outpatients with Stage C HF progressed to Stage D after 3 years, whereas another 1 of 8 died, despite receiving care in an academic center with a dedicated HF program offering OHT and MCS services. Our data restate the grim reality that HF is a disease with severe prognosis, despite significant progress over the past 20 years (18). Half the patients die before progressing to Stage D, underscoring the importance of early optimization of available medical and implantable device therapy (19).
Stage D HF is a resource-intensive condition requiring specialized interventions. Our data suggest that, with conservative estimates, more than 100,000 patients with HFrEF will progress to Stage D HF in the United States annually (1). Although our estimates need confirmation, these data are important for the scientific community and the healthcare system. With current rates of OHT and long-term MCS (20,21), our data imply that most patients with Stage D HF will eventually receive palliative care regardless of workup and planning for advanced HF therapies.
Many patients with HF remain stable over long periods whereas others deteriorate rapidly despite therapy optimization. Our study does provide a set of risk factors associated with earlier disease progression. Although these factors overlap with “generic” indicators of poor outcomes in HF, competing mortality analysis suggests that progression and mortality are driven by different risk factors. Knowing what factors impact risk for progression would be helpful in communicating expectations with patients and families to make informed decisions. This is important when considering that most patients with Stage D HF will eventually receive palliative care.
Our data suggest that blacks progress to Stage D HF earlier. Evidence suggests that HF is more common in blacks and often diagnosed at a younger age (22). The exact mechanisms that promote development of clinical disease and lead to disease progression in blacks are not clear. Blacks are more susceptible to sodium retention and vascular injury and less responsive to angiotensin-converting enzyme inhibitors and beta-blockers (23,24). Socioeconomic factors may play a role, as disparities in access to care have been described (25). Therefore, the observed propensity of blacks to progress to Stage D HF needs further investigation.
Patients with nonischemic cardiomyopathy encountered higher rates of progression to Stage D HF. This finding may reflect referral bias. However, it may also reflect the progressive nature of the primary cardiomyopathic process in combination with lower prevalence of cardiovascular disease and younger age and, thus, lower competing mortality. However, the small difference in mortality alone could not adequately explain the difference in rates of progression. Of note, we excluded patients with special causes of cardiomyopathy, most of which have worse prognosis than idiopathic dilated cardiomyopathy (26). Nonischemic HFrEF has a better response to a number of therapies and a better prognosis than ischemic HF (27). Although worse outcomes and poorer response to treatment in ischemic patients in some clinical trials may have been an effect of patient selection (e.g., patients suitable for revascularization might have been excluded), our findings highlight the importance of differentiating between risk for disease progression versus risk for mortality from a clinical and health care system planning perspective.
In our study, worse HF status at inception, as expressed by worse functional class, LVEF, and lower blood pressure, was associated with progression to Stage D, which comes as no surprise. Many patients with HF have renal impairment. Interestingly, blood urea nitrogen concentration was closely associated with HF progression in our analysis, whereas creatinine concentration predicted mortality. Elevated blood urea nitrogen level probably reflects the effects of poor intrarenal hemodynamics and suboptimal volume status and is therefore an indicator of HF status (28). In contrast, creatinine is a closer marker of limited organ reserve. Similarly, elevated bilirubin, a marker of hepatic dysfunction, was a predictor of disease progression. Interestingly, chronic lung disease requiring bronchodilators predicted progression, confirming that combined cardiopulmonary impairment amplifies disease progression in HF. Taking these findings together, it becomes obvious that the number of extracardiac systems involved is an indicator of disease progression risk.
First, this is a single-center study in an academic setting. Therefore, patient characteristics, designation of Stage D HF, and rates of disease progression reflect our practice and are subject to referral bias. Therefore, results may not be generalizable. On the other hand, our patient population should represent the HFrEF population that is likely to receive advanced HF therapies. The distribution of demographics, HF characteristics, and comorbidities in our study was similar to that of a recent pivotal trial in HFrEF patients, although our population was more symptomatic (29). Our racial distribution reflects metropolitan Atlanta demographics, so the proportion of blacks is higher than those in nationwide data. Because blacks are at higher risk of HF progression, this may have overestimated HF progression rates. Second, our definition of Stage D HF was based on physician assessment and might not have captured Stage D patients who were not yet candidates for advanced HF therapies. Despite the need for operational definitions, disease progression is a continuum, thus any threshold is inherently arbitrary. Objective assessment of exercise capacity would be valuable for Stage D HF diagnosis. However, in practice, exercise testing is considered only in the context of evaluation for OHT or MCS and the latter only in non–inotrope-dependent patients (i.e., the minority of current MCS recipients). Third, modes of death were not available in our study. Because we did not have approval to contact families and obtain death certificates and information on mode of death, adjudication of causes would only be possible for in-hospital deaths. In the Framingham Heart Study (30), 70% of community patients with HFrEF died from cardiovascular causes; of these, two-thirds died from pump failure and one-third from sudden cardiac death. Similar findings have been reported by others (31). Fourth, individual-level socioeconomic data, which may influence disease progression, were not available. Finally, the cohort inception was cross-sectional, that is, we did not capture patients at HF onset. Therefore, we did not assess the impact of factors like age at onset, duration of HF, or initial response to pharmacotherapy.
In this study from an academic center, 1 of 8 outpatients with Stage C HFrEF receiving care in cardiology (including HF) clinics progressed to Stage D within 3 years and another 1 of 8 patients died within the same period; in all, 1 of 4 patients either died or progressed to Stage D within 3 years. We also observed earlier progression among black and nonischemic patients. Factors associated with progression to Stage D HF differed from factors associated with competing mortality. Although our estimates need confirmation in larger, preferably multicenter studies, our findings highlight the need for intensified research efforts to improve outcomes for patients with HF and have implications for healthcare system planning.
COMPETENCY IN MEDICAL KNOWLEDGE: Despite substantial therapeutic advances in the past 20 years, systolic heart failure remains a condition with high morbidity and mortality, with 1 of 4 stable outpatients progressing to Stage D heart failure or dying within 3 years of initial assessment. A number of risk factors place the stable patient with systolic heart failure at higher risk for disease progression. Providers may want to treat patients with multiple progression risk factors more aggressively and potentially engage earlier in discussions about pathways of care.
TRANSLATIONAL OUTLOOK 1: Rates of progression to Stage D heart failure vary within a stable systolic heart failure population, and a set of clinical risk factors can identify those at higher risk for earlier progression. Addition of echocardiographic and circulating markers may refine prognosis.
TRANSLATIONAL OUTLOOK 2: Multicenter studies with prospective inception of newly identified systolic heart failure patients would be able to project more accurately the pool of patients who may become candidates for advanced heart failure therapies and provide additional insights into risk factors for progression.
For statistical methods and supplemental tables, please see the online version of this article.
Dr. Kalogeropoulos has received support from the National Institutes of Health, the American Heart Association, the Centers for Disease Control and Prevention, the Atlanta Clinical and Translational Science Institute, and Critical Diagnostics. Dr. Butler has received support from the National Institutes of Health and European Union; and is a consultant for Amgen, Bayer, Cardiocell, Novartis, Boehringer Ingelheim, Trevena, Relypsa, Z Pharma, Pharmain, Merck, and Gilead. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- B-type natriuretic peptide
- heart failure
- heart failure with reduced ejection fraction
- International Classification of Diseases-9th Revision-Clinical Modification
- left ventricular ejection fraction
- mechanical circulatory support
- New York Heart Association
- orthotopic heart transplantation
- subhazard ratio
- Received December 7, 2016.
- Revision received February 21, 2017.
- Accepted February 21, 2017.
- 2017 American College of Cardiology Foundation
- Miller L.W.,
- Guglin M.
- Mancini D.,
- Colombo P.C.
- Yancy C.W.,
- Jessup M.,
- Bozkurt B.,
- et al.
- Kirklin J.K.
- Zannad F.,
- Briancon S.,
- Juilliere Y.,
- et al.
- Ammar K.A.,
- Jacobsen S.J.,
- Mahoney D.W.,
- et al.
- Estep J.D.,
- Starling R.C.,
- Horstmanshof D.A.,
- et al.
- ↵Centers for Medicare and Medicaid Services. Medicare demonstration projects. Available at: https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Downloads/HIQR-ICD9-to-ICD10-Tables.pdf. Accessed February 17, 2017.
- ↵Centers for Mediare and Medicare Services. Chronic conditions data warehouse. Available at: https://www.ccwdata.org/web/guest/condition-categories. Accessed February 17, 2017.
- Royston P.,
- Sauerbrei W.
- Butler J.,
- Kalogeropoulos A.,
- Georgiopoulou V.,
- et al.
- Vazquez R.,
- Bayes-Genis A.,
- Cygankiewicz I.,
- et al.
- Hummel S.L.,
- Ghalib H.H.,
- Ratz D.,
- Koelling T.M.
- Loh J.C.,
- Creaser J.,
- Rourke D.A.,
- et al.
- Packer M.
- Sharma A.,
- Colvin-Adams M.,
- Yancy C.W.
- Johnson J.A.
- Levy W.C.,
- Mozaffarian D.,
- Linker D.T.,
- et al.
- Schrier R.W.
- Lee D.S.,
- Gona P.,
- Albano I.,
- et al.