Author + information
- Received February 8, 2012
- Revision received June 27, 2012
- Accepted July 2, 2012
- Published online October 9, 2012.
- Chohreh Partovian, MD, PhD⁎,†,
- Scott R. Gleim, PhD⁎,
- Purav S. Mody, MD⁎,†,
- Shu-Xia Li, PhD†,
- Haiyan Wang, MD, MS†,
- Kelly M. Strait, MS†,
- Larry A. Allen, MD, MHS‡,
- Tara Lagu, MD, MPH§,
- Sharon-Lise T. Normand, PhD∥ and
- Harlan M. Krumholz, MD, SM⁎,†,¶,#,⁎ ()
- ↵⁎Reprints requests and correspondence:
Dr. Harlan Krumholz, Internal Medicine, Yale University School of Medicine, 1 Church Street, Suite 200, New Haven, Connecticut 06510
Objectives This study sought to determine hospital variation in the use of positive inotropic agents in patients with heart failure.
Background Clinical guidelines recommend targeted use of positive inotropic agents in highly selected patients, but data are limited and the recommendations are not specific.
Methods We analyzed data from 376 hospitals including 189,948 hospitalizations for heart failure from 2009 through 2010. We used hierarchical logistic regression models to estimate hospital-level risk-standardized rates of inotrope use and risk-standardized in-hospital mortality rates.
Results The risk-standardized rates of inotrope use ranged across hospitals from 0.9% to 44.6% (median: 6.3%, interquartile range: 4.3% to 9.2%). We identified various hospital patterns based on the type of agents: dobutamine-predominant (29% of hospitals), dopamine-predominant (25%), milrinone-predominant (1%), mixed dobutamine and dopamine pattern (32%), and mixed pattern including all 3 agents (13%). When studying the factors associated with interhospital variation, the best model performance was with the hierarchical generalized linear models that adjusted for patient case mix and an individual hospital effect (receiver operating characteristic curves from 0.77 to 0.88). The intraclass correlation coefficients of the hierarchical generalized linear models (0.113 for any inotrope) indicated that a noteworthy proportion of the observed variation was related to an individual institutional effect. Hospital rates or patterns of use were not associated with differences in length of stay or risk-standardized mortality rates.
Conclusions We found marked differences in the use of inotropic agents for heart failure patients among a diverse group of hospitals. This variability, occurring in the context of little clinical evidence, indicates an urgent need to define the appropriate use of these medications.
Heart failure is a leading cause of hospital admission, accounting for almost 1 million hospitalizations in the United States annually (1). In the absence of major advances in the treatment of this condition, early mortality has declined only modestly over the past 2 decades (2). Outcome measures have revealed that hospitals vary in their 30-day risk-standardized mortality rates (RSMR), indicating that hospital-level differences in treatment patterns may affect patient outcomes (3,4).
Positive inotropic agents are used in the treatment of the highest-risk patients hospitalized with heart failure. Dopamine and dobutamine have been available for decades and were approved by the Food and Drug Administration before the mandate to evaluate the benefits and risks of new drugs in large trials. A third positive inotropic agent, milrinone, was approved in 1988 for the treatment of acute decompensated heart failure based on its hemodynamic effects, rather than on clinical endpoints. Data on the comparative effectiveness of these agents on the outcomes of patients with heart failure are lacking (5). The only large clinical trial of milrinone compared its effect with placebo in hospitalized patients without end-organ hypoperfusion and found an increased risk of adverse events (6). Other positive inotropic agents that were tested in trials, such as amrinone and vesnarinone, were shown to increase mortality (7,8).
The most recent guidelines for the diagnosis and management of heart failure from the American College of Cardiology/American Heart Association recommend limited use of these agents, stating that “intravenous inotropic drugs such as dopamine, dobutamine or milrinone might be reasonable for those patients presenting with documented severe systolic dysfunction, low blood pressure and evidence of low cardiac output, with or without congestion, to maintain systemic perfusion and preserve end-organ performance” (9). The guidelines explicitly state that intravenous positive inotropic agents are not recommended for hospitalized patients with heart failure who do not have evidence of decreased organ perfusion. The clinical practice guidelines of the Heart Failure Society of America (10) and the European Society of Cardiology (11) mirror the American College of Cardiology/American Heart Association recommendation. The recommendations are based on expert opinion.
Little information is available about how use of positive inotropic agents varies among hospitals. Scarce evidence and relatively weak guideline recommendations indicate the potential for marked variation. Accordingly, we investigated treatment patterns of inotrope use among patients hospitalized for heart failure in a large network of hospitals in the United States. We also report the relationship between inotrope use and in-hospital RSMR and length of stay, including comparisons of hospitals with high and low use patterns.
We conducted a cross-sectional study using data from Perspective, a voluntary, fee-supported database developed by Premier, Inc. (Charlotte, North Carolina), for measuring quality and healthcare use. As of 2010, Perspective contained data from more than 500 hospitals in the United States, including more than 130 million cumulative hospital discharges. Inpatient discharges represent approximately 20% of all acute care inpatient hospitalizations nationwide. In addition to the information available in the standard hospital discharge file, Perspective contains a date-stamped log of all billed items at the patient level, including medications and laboratory, diagnostic, and therapeutic services. For this study, patient data were deidentified in accordance with the Health Insurance Portability and Accountability Act, and hospitals were identified by a random identifier assigned by Premier. The Yale University Human Investigation Committee reviewed the protocol for this study and determined that it is not considered to be Human Subjects Research as defined by the Office of Human Research Protections.
Patients and hospitals
Our analysis included the first episode of hospitalization per patient between January 1, 2009, and December 31, 2010, that had a principal diagnosis of heart failure (International Classification of Diseases, Ninth Revision, Clinical Modification codes 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, 428.0, 428.1, 428.20, 428.21, 428.22, 428.23, 428.30, 428.31, 428.32, 428.33, 428.40, 428.41, 428.42, 428.43, and 428.9) or a principal diagnosis of respiratory failure (International Classification of Diseases- Ninth Revision, Clinical Modification code 518.81) combined with a secondary diagnosis of congestive heart failure (428.0). We excluded patients younger than 18 years or those whose physicians were pediatricians, because our focus was not on congenital disease. We excluded hospitalizations with a duration of 1 day, as well as transfers to or from another acute care facility because we could not accurately assess treatment with inotropic therapy during an entire hospitalization.
In addition to patient age, sex, race or ethnicity, and insurance status, we used software (version 3.4, 3.5, and 3.6 for federal fiscal years 2009, 2010, and 2011, respectively) provided by the Healthcare Costs and Utilization Project of the Agency for Healthcare Research and Quality to classify comorbidities from the standard hospital discharge file based on methods described by Elixhauser and Steiner (12). This tool provides a Diagnosis Related Group screen of International Classification of Diseases-Ninth Revision-Clinical Modification secondary diagnoses.
For each hospital, Perspective contains information, collected from the American Hospital Association database, on bed count, teaching status, geographic location (by census division), and whether it serves an urban or rural population. Participating hospitals were geographically diverse, with a composition similar to that of acute care hospitals nationwide. They were predominantly small to midsized nonteaching facilities that serve a largely urban population.
Hospital use and clinical outcomes
We assessed the use of dopamine, dobutamine, and milrinone, which are the 3 positive inotropic agents that were noted in the American College of Cardiology/American Heart Association Heart Failure Guidelines. We also evaluated the hospitals' use of a number of diagnostic or therapeutic procedures, including pulmonary artery catheterization, ventricular assist device, heart transplantation, mechanical ventilation, and implantable cardioverter defibrillator, with and without cardiac resynchronization therapy. We also assessed median length of stay per hospital and in-hospital RSMR.
We constructed summary statistics using frequencies and proportions for categorical data and means, and medians and interquartile ranges (IQR) for continuous variables.
To determine the patient and hospital characteristics that were associated with the use of inotropic agents, we constructed 4 logistic regression models: 1 for overall inotropic use and 1 for each of the 3 inotropic agents. Patient characteristics including age group, sex, and comorbidities were considered as candidate covariates. We selected the variables for the final model using a stepwise algorithm. After controlling for selected patient characteristics, we fit logistic regression models to evaluate further the effects of hospital characteristics (hospital size, heart failure volume, urban or rural setting, geographic location by census division, and teaching status). We report odds ratios (ORs) and corresponding 95% confidence intervals (CIs) for each significant factor.
We used hierarchical generalized linear models (HGLM) to calculate risk-standardized use rates for use of any inotropic agent and for use of each inotropic agent (4,13). We selected patient characteristics used as covariates for risk adjustment by stepwise algorithm using logistic regression models. We performed additional scaling to ensure that the unadjusted and adjusted rates were comparable such that the slopes of the weighted linear regression of the unadjusted rate and the adjusted rate were equal to 1. We also used hierarchical logistic regression to estimate the in-hospital RSMR, adjusting for patient characteristics including age, gender, and all comorbidities as well as a hospital individual effect as a random effect. We used a modified version of a previously published 30-day mortality model with data elements restricted to those available during the index admission (4). We used weighted linear regression models to assess the relationship between hospital inotrope use and RSMR.
To assess further the contribution of institutional effect to the variation in use of positive inotropic agents, we compared the receiver operating characteristic (ROC) curves of the logistic regression models that adjust for only patient case mix with the ROC curves of the HGLM models that also take into account institutional factors. We calculated the intraclass correlation coefficient for the HGLM models as described elsewhere (14).
Next, we categorized hospitals as either a predominant user for 1 of the 3 agents or as a mixed user based on their pattern of inotrope use. For these classifications, we included only hospitals with at least 15 hospitalizations over 2 years with the use of any positive inotropic agent. We calculated the total use by adding the number of hospitalizations using dobutamine, dopamine, or milrinone. Because patients can receive more than 1 agent during a single hospitalization, the total use may exceed the number of hospitalizations. We then calculated the percentage of hospitalizations receiving each inotropic agent by hospital, with the total use as the denominator. If a hospital had at least 55% of hospitalizations receiving any single agent (55% for dobutamine and dopamine, 50% for milrinone), it was deemed a predominant user for this agent. The cutoffs were chosen empirically based on hospital distribution of percent use. If none of the agents exceeded 55% of total use and total dobutamine and dopamine use was 80% or more, the hospital was categorized as a dobutamine and dopamine mixed user. The rest of the hospitals were characterized as dobutamine, dopamine, and milrinone mixed users. We used Kruskal-Wallis and chi-square tests to assess the association between use pattern and different hospital characteristics.
We conducted analyses with SAS software version 9.2 (SAS Institute, Inc., Cary, North Carolina), estimated the hierarchical logistic models using the GLIMMIX macro in SAS, and created the figures with R (version 2.11.1) (15).
Hospital and patient characteristics
We identified 189,948 hospitalizations from 376 hospitals that met our enrollment criteria. Of these hospitals, 53% had more than 250 beds, 73% were nonteaching, and 78% were located in urban settings. The median volume of patients with heart failure per hospital over the 2 years was 394 (IQR: 161 to 770, range: 1 to 2,076).
We assessed patient characteristics by hospital. The median patient age was 76 years (IQR by hospital: 64 to 84 years), the median percent of women was 52.7% (IQR: 49.1% to 55.7%), and the median percent of white patients was 78.4% (IQR: 49.5% to 92.3%). The most common comorbidities included hypertension (median among hospitals: 70.0%), coronary atherosclerosis (55.1%), cardiac dysrhythmias (47.9%), disorders of lipid metabolism (43.5%), renal failure (37.9%), and diabetes without complications (33.7%). Most admissions (median among hospitals: 65.3%, IQR: 52.8% to 75.7%) were through the emergency department. Medicare was the most common form of health insurance, accounting for approximately two thirds (median among hospitals: 65.7%) of patients. The most frequent procedures performed were renal dialysis and mechanical ventilation, with median use among hospitals of 5.3% (IQR: 2.4% to 7.6%) and 5.4% (IQR: 3.6% to 7.4%), respectively. Cardiac procedures were rare, with median use among hospitals of 1.2% (IQR: 0.0% to 4.4%) for automatic implantable cardioverter defibrillators with or without cardiac resynchronization therapy and 0.2% (IQR: 0.0% to 0.8%) for pulmonary artery catheterization.
Hospital use of positive inotropic agents
Of all hospitalizations, 13,676 (7.2%) included a treatment with a positive inotropic agent. Among hospitals, the unadjusted treatment rate ranged from a minimum of 0% to a maximum of 38.0%. The hospital risk-standardized treatment rate ranged from a minimum of 0.9% to a maximum of 44.6% (IQR: 4.3% to 9.2%, median: 6.3%) (Fig. 1A). Dobutamine was most common (range: 0.4% to 47.4%, IQR: 2.2% to 6.8%, median: 3.7%), followed by dopamine (range: 0.6% to 16.3%, IQR: 2.5% to 4.7%, median: 3.3%). Milrinone use was much less on average and highly variable (range: 0.02% to 67.5%, IQR: 0.5% to 2.0%, median: 0.78%) Fig. 1B).
The pattern of agents used varied widely across hospitals. Of the 225 hospitals with at least 15 hospitalizations involving inotrope treatment, 65 (29%) were dobutamine predominant, 56 (25%) were dopamine predominant, 3 (1%) were milrinone predominant, 71 (32%) were dobutamine and dopamine mixed pattern, and 30 (13%) were dobutamine, dopamine, and milrinone mixed (Table 1). After adjustment for patient case mix, the likelihood of treatment with a positive inotropic agent varied by hospital use patterns (p < 0.0001 for the association of use pattern with percent use). There was also a significant association between hospital pattern of inotrope use and hospital size (p < 0.02). We did not find a significant relationship between pattern of inotrope use and teaching status, heart failure volume, median length of stay, or hospital percent of heart transplant or implantation of ventricular assist devices. Figure 2 illustrates various hospital patterns based on overall percentage of inotrope use as well as mix of agents used.
There were 151 hospitals with fewer than 15 cases of inotrope use over 2 years that were not included in the pattern classifications. Their risk standardized median percent of inotrope use was 4.7% (IQR: 3.8% to 6.2%). These hospitals were small (median number of beds: 121, IQR: 71 to 216), had a low volume of patients with heart failure (median: 155, IQR: 87 to 275), and were mainly nonteaching (85.4%).
Patient and hospital characteristics associated with the use of positive inotropic agents
We assessed the association between patient characteristics and inotrope use, including overall use and use of individual inotropic agent (Online Appendix). Patients with cardiac arrest and ventricular fibrillation had the highest likelihood of receiving inotropes, both for combined inotrope use and for use of each individual agent. The likelihood of receiving inotropes also was higher in younger patients, most notably for milrinone. Other comorbidities associated with higher likelihood of receiving inotropic treatment included acute myocardial infarction, valvular disease, fluid and electrolyte disorders, coagulopathy, cardiac dysrhythmias, coronary atherosclerosis and other heart disease, renal failure, and aortic and peripheral arterial embolism or thrombosis. Female patients had a lower likelihood of being treated with positive inotropic agents. The ROC curves of these logistic regression models ranged from 0.69 for dobutamine to 0.75 for dopamine and milrinone.
After adjusting for patient characteristics, we assessed the association between inotrope use and hospital characteristics by adding the following hospital characteristics to the model: size, heart failure volume, urban versus rural setting, teaching versus nonteaching status, and geographic location by census division (Table 2). These logistic regression models showed the likelihood of receiving dopamine (OR: 1.13, 95% CI: 1.07 to 1.21) or milrinone (OR: 1.23, 95% CI: 1.10 to 1.36) to be higher in teaching hospitals and the odds of receiving dobutamine (OR: 1.31, 95% CI: 1.20 to 1.44) or milrinone (OR: 1.65, 95% CI: 1.33 to 2.05) to be higher in urban hospitals. Hospitals with the highest odds of using milrinone had a lower volume of patients with heart failure (between 26 and 200 hospitalizations over 2 years). The odds of using positive inotropic agents were highest in the east south central region (OR: 1.52, 95% CI: 1.41 to 1.65) and lowest in New England (OR: 0.43, 95% CI: 0.37 to 0.50). The odds of being treated with any positive inotropic agent were highest in hospitals having between 251 and 400 beds.
The addition of hospital characteristics to the logistic regression models was associated with improvement in the performance of all 5 models (Table 3). The ROC curves ranged from 0.71 for dobutamine to 0.76 for dopamine and 0.77 for milrinone. However, the best performance was obtained with the HGLM models that adjusted for patient case mix and an individual hospital effect as random effects. The ROC curves for HGLM ranged from 0.77 to 0.88. Furthermore, the intraclass correlation coefficients of the HGLM indicated that a noteworthy proportion of the variance in inotrope use could be explained by the individual institutional effect after accounting for differences in patient case mix. Individual institutional effect could explain 34% of variability in milrinone use, 19% of variability in dobutamine use, and 10% of variability in dopamine use.
Use of positive inotropic agents and overall hospital mortality
When all patients were included, the median of the unadjusted in-hospital mortality rates was 4.4% (IQR: 3.2% to 5.7%). The median of the in-hospital RSMR was 4.7% (IQR: 3.9% to 5.5%). There was no significant relationship between RSMR and hospital percent of inotrope use or hospital pattern of use (Table 1 and 4) (Fig. 2). When we stratified hospitals by the crude percent of inotrope use weighted for heart failure volume, we found no difference in RSMR between hospitals in the top and bottom tenth percentiles. When we stratified hospitals by use of heart transplant or ventricular assist device, we found no association between RSMR and percent of inotrope use.
In this large observational study, we found marked differences in the patterns of use of positive inotropic agents among a diverse group of hospitals in the United States. Variations in rates and types of medication reflected differences in hospitals as well as patient characteristics. We did not find an association between patterns of use and in-hospital RSMR or length of stay.
Despite the potential harm associated with positive inotropic agents (16,17) and the lack of strong endorsement by clinical practice guidelines, they are commonly used. Guidelines state that inotropes should be confined to carefully selected patients with low blood pressure and reduced cardiac output who can have blood pressure and heart rhythm monitored closely (9). Registries suggest that this group represents approximately 3% of all patients hospitalized with heart failure (18). Our study and others suggest that many more (7% to 12%) patients are being treated with these agents (19,20). In the Acute Decompensated Heart Failure National Registry, the mean systolic blood pressure for patients treated with dobutamine was 124.0 ± 29.3 mm Hg and 121.3 ± 27.4 mm Hg for those treated with milrinone. Of the 6,198 patients (9% of the total cohort) who were treated with these agents, only 507 (8%) had a systolic blood pressure of <90 mm Hg (19). The Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness reported that the use of vasoactive therapy among its participants was not significantly influenced by blood pressure or cardiac index (21).
The pattern of agents used also was quite variable between hospitals. Overall rates of use were related to the mix of agents, with the highest percentage of overall inotropic use found in milrinone-predominant hospitals. In the absence of evidence about the comparative effectiveness of these drugs in patients with heart failure, the likelihood of a patient being treated with a specific agent seems dependent mostly on the institution to which the patient is admitted. Despite concerns about the safety of inotrope use, we failed to find differences in mortality or length of stay. Nevertheless, given that hospitalizations for heart failure are common and that inotropic agents have potential for harm in at least some patient populations, the observed variation in patterns of practice highlights the urgent need for greater evidence to guide these care decisions.
There are several limitations to consider. First, hospitals in the Premier network may not be a representative sample of all hospitals in the United States. However, preliminary comparisons between patient and hospital characteristics for the hospitals that submit data to Premier and those of the probability sample of hospitals and patients selected for the National Hospital Discharge Survey suggest that the patient populations are similar with regard to age, gender, length of stay, mortality, primary discharge diagnosis, and primary procedure groups. In addition, the patients included in our study had very similar characteristics to those of the heart failure patients described in registries such as the ADHERE (Acute Decompensated Heart Failure National Registry) or Get With the Guidelines-Heart Failure (18,21). Second, this database does not include clinical data such as left ventricular ejection fraction, vital signs (e.g., heart rate and blood pressure), or laboratory test results (serum creatinine) that are important determinants of inotrope use and may contribute to improving the risk adjustment for patient case mix across hospitals. However, the differences we observed are larger than would be expected based on differences in case mix. Despite the lack of these clinical and biological data, the performance of the models (predictive ability) showed that inotrope use at the hospital level can be modeled adequately when accounting for both patient case mix and institutional clustering effects. Third, we included only the first admission, rather than all hospitalizations, per patient. This was because analyses showed that percentage of inotrope use was higher in patients with multiple hospitalizations. Therefore, including all hospitalizations would have overestimated the relative importance of institution-related factors (versus patient-related factors) in explaining the variation in inotrope use. Indeed, when all hospitalizations per patient were included, the ICC were slightly higher (0.131 for any inotrope use, 0.216 for dobutamine, 0.102 for dopamine, and 0.372 for milrinone).
Our analyses demonstrate that a noteworthy proportion of the variation observed in inotrope use was related to an individual institutional effect. This finding is in agreement with the Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness, in which the most important predictor of use was the study site (hospital) to which the patient was admitted, despite the inclusion of more patient-level clinical and biological data in the multivariate analysis (21).
The marked differences that we observed in the rates and patterns of inotrope use in the treatment of patients hospitalized with heart failure in the United States are attributed in part to unmeasured institutional factors, making the likelihood and type of treatment with an inotropic agent for any given patient highly dependent on the hospital to which the patient is admitted. This study heralds an urgent need for further investigation to define the proper role of inotropic agents in the treatment of patients with decompensated heart failure.
For supplemental tables, please see the online version of this article.
Hospital Patterns of Use of Positive Inotropic Agents in Patients with Heart Failure
This work was supported by grant DF10-301 from the Patrick and Catherine Weldon Donaghue Medical Research Foundation, West Hartford, Connecticut, and by grant UL1 RR024139-06S1 from the National Center for Advancing Translational Sciences, Bethesda, Maryland. Dr. Krumholz is supported by grant U01 HL105270-02 (Center for Cardiovascular Outcomes Research at Yale University), and Dr. Allen is supported by grant K23 HL105896-01A1, both from the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Maryland. Dr. Gleim is supported by grant 11 POST-77100000 from the American Heart Association Founders Affiliate in Dallas, Texas. Dr. Krumholz is the recipient of a research grant from Medtronic, Inc., through Yale University and is chair of a cardiac scientific advisory board for UnitedHealth. Dr. Allen is a consultant for Amgen, Inc. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- hierarchical generalized linear model
- interquartile range
- odds ratio
- receiver operating characteristic
- risk-standardized mortality rate
- Received February 8, 2012.
- Revision received June 27, 2012.
- Accepted July 2, 2012.
- American College of Cardiology Foundation
- Roger V.L.,
- Go A.S.,
- Lloyd-Jones D.M.,
- et al.
- Bernheim S.M.,
- Grady J.N.,
- Lin Z.,
- et al.
- Krumholz H.M.,
- Wang Y.,
- Mattera J.A.,
- et al.
- Hunt S.A.,
- Abraham W.T.,
- Chin M.H.,
- et al.
- Dickstein K.,
- Cohen-Solal A.,
- Filippatos G.,
- et al.
- Team R.R.
- Adams K.F.,
- DeMarco T.,
- Berkowitz R.,
- et al.
- Adams K.F. Jr.,
- Fonarow G.C.,
- Emerman C.L.,
- et al.
- Abraham W.T.,
- Adams K.F.,
- Fonarow G.C.,
- et al.
- Yancy C.W.,
- Lopatin M.,
- Stevenson L.W.,
- et al.