NEJM: Prevalence Data on Diastolic Heart Failure Flawed

To the non-medical types, I apologize for this post, but I have to rant a bit to my colleagues. For you to keep up, here's a definition:

Diastolic heart failure occurs in people with a normally contracting heart muscle that has difficulty relaxing to fill with blood. Because it cannot fill with blood, the amount of blood ejected with each heartbeat is compromised. Diastolic heart failure is different from the more commonly understood "systolic" heart failure caused by a weakened heart muscle).

In today's New England Journal of Medicine, an interesting article entitled "Trends in Prevelence and Outcomes of Heart Failure With Preserved Ejection Fraction" by Owan, et al. describes what they perceive as an increasing prevalence of "heart failure in patients with preserved ejection fraction" (also known as "diastolic heart failure") in Olmstead County, Minnesota between 1987 and 2001. Wonderful graphs are depicted showing a rising prevalence of this problem.

But how was the diagnosis of heart failure made for this study? Not by physical exam or chart review, but rather by searching the Medicare database for "DRG 127" - the "Diagnosis Related Group" code for congestive heart failure, or ICD-9 code 428 (Congestive heart failure).

This study is another failed attempt to make meaningful population prevelance data from these Medicare codes, and amazingly, neither the authors nor editors considered the healthcare and reimbursement climate in which those codes were generated. They forgot the most SERIOUS source of bias in their study: money.

Bruce Psaty, MD, PhD et al in a commentary from the 1 Jul 1999 issue of the Am J of Cardiology, entitled "The Potential Costs of Upcoding for Heart Failure in the United States" stated:

"Although the incidence of validated episodes of heart failure did not differ significantly between 1981 and 1991 in Olmsted County, Minnesota, the number of initial hospitalizations coded for heart failure among persons ≥65 years of age in the United States (US) increased from 631,306 in 1986 to 803,506 in 1993. Potential explanations include the aging population, the improved survival after myocardial infarction, and the burden of risk factors (e.g., hypertension and diabetes)."

But they also offer another rationale which I think is compelling:

"The US Health Care Financing Administration reimburses hospitals for Medicare patients on the basis of diagnosis-related groups (DRGs). Hospitalizations for heart failure (DRG = 127) received a reimbursement estimated to be $7,057.98 in 1997. For other primary diagnoses, the presence of a complication increased the level of reimbursement, and heart failure qualified as a complication. The additional discharge diagnosis of heart failure for a Medicare patient in 1997 increased the reimbursement for an acute myocardial infarction by $3,345.06, for pneumonia by $2,872.70, and for a gastrointestinal hemorrhage by $3,071.40. For these 3 common conditions, a secondary diagnosis of heart failure increased the reimbursement to hospitals by an average of $3,096.39."

While this certainly does not diminish the importance of physicians appreciating that heart failure can occur in people with normal ejection fractions, I am skeptical that the prevalence of this disorder increases over time as the study suggests. To reinforce this finding, Psaty and colleagues noted:

"Among 485 subjects with a primary or secondary Medicare discharge diagnosis of heart failure, we were able to validate only 303 (positive predictive value 62.5%). For the other 182 patients, we were not able to find even modest levels of supporting evidence for the diagnosis of heart failure. The proportion of false-positive diagnoses was 37.5% (95% confidence interval [CI] 33.2% to 41.8%). In contrast, the proportion of false-negative diagnoses was low (10.1% of 832). The false-positive rate seen in the CHS is consistent with the findings of other studies, and it is much higher than the false-positive rate for diagnoses such as myocardial infarction, which tend to be ≤15%.

"Using the false-positive rate from the CHS, we estimate that in 1993, 301,315 of the 803,506 hospitalizations for heart failure may reflect incorrect diagnoses (95% CI 266,764 to 335,866). If the average excess cost of an upcoded diagnosis is $3,096.39, then US hospitals may have received excess reimbursements from Medicare of as much as $933 million a year (95% CI, $826 to $1,040 million). The high proportion of false-positive compared with false-negative diagnoses favors reimbursement to hospitals. If the CHS sample is representative, the estimated costs of upcoding, even after adjustment for the revenues lost by the false-negative diagnoses, would be still be high—about $502 million."

I think we have to rethink this paper with this incentive in mind.

"A wise man should have money in his head, but not in his heart."
Jonathan Swift
Irish essayist, novelist, & satirist (1667 - 1745)

--Wes

Quote reference link.