Editor’s Note: Health Affairs has recently published two studies looking at the association between hospital costs and quality. The first, by Ashish Jha and coauthors, appeared in our May-June issue, and the second by Laura Yasaitis, Amitabh Chandra, and coauthors, was published online.
Variations in spending and intensity of care, and the effects of these variations on quality and outcomes, have become a major focus of the current health reform debate. Therefore, the Health Affairs Blog asked Jha and Chandra to each describe the major findings of the two studies, and then to explain whether they viewed the studies as consistent with each other or as contradictory. Chandra’s response is below and Jha’s response is here.
The two researchers touch on differences between the methods and conclusions of the two studies. However, both researchers appear to essentially agree that, in Jha’s words, “unlike most sectors of the U.S. economy, where we usually have to make a cost-quality trade-off, no such sacrifice is likely to be necessary in the health care sector. There is ‘plenty of fat’ that can be removed to identify the high-quality, low-cost institutions.”
In a study comparing hospital-level spending and quality, we found no evidence that hospitals with higher spending provide better care. In fact, in some cases hospitals that spend more provide worse care.
In a climate of growing concerns about how much our country spends on health care, it is increasingly important to know what we’re getting for our money. Previous research has demonstrated a negative relationship between quality and spending at the regional level. However, findings at the level of the Hospital Referral Region (HRR) — a geographic designation devised for the Dartmouth Atlas of Health Care, reflecting hospitals’ typical service areas — do not lend themselves to actionable policy. Our study is one of the first to examine the relationship between hospital-level quality and spending.
What we did
We examined whether hospitals that spent more Medicare dollars on their patients in the last two years of life performed better on quality indicators. Our quality measures came from the publicly available Hospital Compare data set for the years 2004-2007. We examined ten measures that collectively encompass care delivered for acute myocardial infarction (AMI), congestive heart failure (CHF), and pneumonia.
We wanted a spending measure that captured the intensity of care that hospitals provide. We used end-of-life Parts A and B Medicare spending for chronically ill beneficiaries. Because all these beneficiaries died within two years, they were comparably sick, and we further adjusted for comorbidities to reduce the influence of illness in affecting spending differences. (And it’s not just the last few months of life that drive these measures; spending in the twenty-four to six months prior to death, excluding the last six months, was very highly correlated with spending in the last six months of life.) We determined how much of that spending was due to specific quantity inputs by the hospital. By using only the amount of spending predicted by what the hospital actually did, as opposed to what the hospital was reimbursed, we sought to reduce the influence that graduate medical education (GME), disproportionate-share hospital (DSH), and geographic price adjustments may have had in affecting spending differences between hospitals.
What we found
When we compared spending to performance on quality indicators, we found absolutely no evidence of a positive relationship; in fact, there was a significant negative trend for spending and quality on AMI and pneumonia. We obtained similar results when we limited our investigation to academic medical centers.
When we examined the performance of hospitals within specific regional areas (Exhibit 5 in our article shows hospitals within New York and Los Angeles), there again was no correlation within regions. It’s clear that there is a wide range of both spending and quality, and that their distribution is far from predictable. In fact, even within these regions, many hospitals were able to achieve exemplary quality with lower spending.
From these data, it seems possible that higher end-of-life spending can result from more chaotic, unorganized care and that this same care setting also fails to deliver the correct, appropriate care when needed. Care systems that haven’t figured out how to coordinate care at the end of life may also be falling behind in their performance on basic, proven medical interventions in acute settings.
How our findings compare with those of Jha and colleagues
Our findings suggest that it may be possible to provide better care at a lower cost to the health care system. It is of interest to compare our results to a study published in this journal last month by Ashish Jha and colleagues, using the same data on quality. They argued that low-cost hospitals were more likely to perform worse on the same quality indicators. How can two different studies, looking at the same quality data, come to different conclusions?
First, as the study by Jha and colleagues noted, we measured costs in different ways. They used the accounting costs as reported by hospitals to Medicare to measure how much it actually cost the hospital to pay for the patient’s admission. By contrast, we used what Medicare pays for a given procedure. Suppose, for example, that a hospital with a strong, um, financial orientation were to churn patients through, finding the most profitable diagnosis-related groups (DRGs) and admitting patients as often as possible to maximize Medicare billings. They’d look costly by our measure (because on a per capita basis, their patients would be costing Medicare real money), but they could be cheaper by Jha and colleagues’ measure, because on a per admission basis, the costs actually incurred by the hospital could be quite modest. Now clearly, this is all quite speculative, but the point is that we are trying to measure actual inputs, while Jha and colleagues are more concerned about financially motivated hospitals undercutting services (like fewer nurses) in order to increase profits. Both quite important margins, but measuring quite different things.
Second, it’s important to note that the findings from both studies are actually quite similar — there is little or no difference in average quality across spending groups, for either study. Jha and colleagues considered six quality measures: three for measures of process (i.e., a summary score for whether the hospital provided effective care for their AMI patients), and three for risk-adjusted outcomes, such as risk-adjusted thirty-day mortality rates following AMI. None of the outcomes measures was significantly different between the highest and lowest quartiles. For the process measure, only one, for CHF, showed a modest clinical significance, with the summary process measure equal to 77 percent for the lowest-cost hospitals and 82 percent for the highest-cost hospitals. The fifth measure, for pneumonia, is insignificant. In the sixth category, AMI process, the difference is clinically negligible — 89 percent for the lowest-cost and 91 for the highest-cost quartile — with no apparent significance between quartiles 1 and 4 (although the authors did find a significant difference using a trend significance test).
Thus, our interpretation of Jha and colleagues’ study is that, as for our study, there little or no association between spending and quality. Whether the observed correlation tips in one direction or the other may be less important than the fact that spending and quality are nearly independent of one another! Both studies support the view that (1) spending more doesn’t seem to do much of anything, at least with our current quality measures, and (2) there is tremendous potential to save money and save lives by moving the vast majority of U.S. hospitals closer to the efficiency “frontier” — those hospitals able to provide high-quality care at lower costs.