Finding an algorithm for heart failure


Decompensated congestive heart failure (CHF) leads to more than one million hospital admissions annually in the United States. Patients admitted for a heart failure exacerbation have a broad range of treatment options, yet in-hospital mortality remains at about 4%.

An algorithm to risk-stratify CHF patients could improve resource allocation and possibly survival, according to William T. Abraham, FACP, director of the division of cardiovascular medicine at Ohio State University Medical Center. Dr. Abraham co-led the OPTIMIZE-HF study, a multi-center trial that enrolled 48,612 hospitalized heart failure patients to determine which clinical signs relate to poor outcomes.

William T Abraham, FACP
William T. Abraham, FACP

The study's results (see Table) showed that age, heart rate, systolic blood pressure, serum creatinine, serum sodium and presence or absence of left ventricular systolic dysfunction predicted likelihood of in-hospital mortality. Low systolic blood pressure, high admission creatinine, and older age were the most important predictive factors. [J Am Coll Cardiol. 2008 Jul 29;52(5):347-56] Based on these results, Dr. Abraham and his study colleagues created an algorithm for use in stratifying patients admitted for CHF. Dr. Abraham recently talked to ACP Hospitalist about the study and its implications.

Q: Do you have recommendations on how physicians should modify individual treatment plans based on the risk factors you found?

A: Absolutely. It would make good sense to put the low mortality risk patients in a subacute hospital environment and treat them more conservatively. Many of those patients are likely to respond to intravenous diuretics alone. On the other hand, the higher risk patients should be in an ICU and treated more aggressively with IV vasoactive therapy, diuretics, and if necessary, positive inotropic agents.

A typical scenario is that patients are admitted to the hospital with volume overload, and they are given some diuretic, they begin to excrete fluid, and clinicians have a sense that everything is going in the right direction, until the laboratory results come back showing worsening kidney function, hyponatremia, or some other evidence that things are not going so well. It may take two to three days into the hospitalization for this to become apparent, and now their mortality risk may be increased.

The idea is to understand the patient's risk at the very front end. Apply the nomogram, risk-stratify them, get them into the right bed or setting, and develop the right care plan for them, prospectively.

Q: Your study found that heart failure patients taking ACE inhibitors or beta-blockers at the time of admission have a lower risk of in-hospital mortality. How can hospitalists use this information?

A: One of the messages here is that patients who are already on evidence-based, guideline-recommended therapy for heart failure should continue to receive these medications throughout the hospitalization, if possible. Often when patients are admitted with worsening heart failure, there is a tendency to discontinue those medications. The message is that they should be continued, unless there is some extreme circumstance that requires discontinuation.

Some other work we have published from OPTIMIZE-HF suggests that discharging patients on these life-saving medications improves 60- to 90-day outcome following hospitalization. These drugs form the cornerstone of heart failure treatment, but they are also important around the time of decompensation. If the patient is already receiving them, we should try not to discontinue them, and if the patient is not yet prescribed these medications, once their acute decompensation is treated, an attempt should be made to prescribe these medications prior to discharge.

Q: Do you know of any hospitals that are using the algorithm suggested in your study?

A: I have talked to some colleagues around the country and many have suggested that it's a good tool, but I can't tell you of any hospitals that are using it in a routine or systematic way. It often does take a while for new research to translate into clinical practice. The heart failure world is a few decades behind the acute MI/chest pain world. We have been risk-stratifying patients coming into the ED with chest pain for many years. To date, the approach with heart failure patients has been a onesize- fits-all approach. Ninety percent of patients with worsening heart failure who present to an emergency department are admitted to the hospital, and there are few risk-based algorithms to guide their disposition and their treatment.

Q: Did any of the study's results surprise you?

A: I don't think that they are that surprising. If you ask many clinicians who take care of a lot of these patients what types of things might predict outcome, they would predict things like blood pressure, kidney function and age. The unique contribution of this particular paper is that it puts it into a model that allows a risk prediction. Rather than just having some impression that a patient with bad kidney function or advanced age or low blood pressure is likely to be at higher risk, we can put all of that together, and we can come up with a relatively precise risk assessment.

The way that patients have been assessed previously is based on bedside gestalt, but that is probably an extremely crude way of determining who is at high mortality risk and who is not. This nomogram has tremendous power to allow a better risk assessment. The good news is that it does not require the assessment of anything that is not routinely done in practice at the time of presentation or admission of these patients.

Q: Are you planning any future studies based on these results?

A: We are interested in doing some prospective work in using the risk assessment to make these triage and treatment decisions, and evaluating the impact on outcome. If you make better triage decisions, appreciating the mortality risk, getting patients into the right setting, and moderating the aggressiveness of treatment based on mortality risk, do you improve outcome? That would require a prospective interventional study. It takes quite a bit to organize, so we have not quite gotten there yet.