The following is a detailed discussion of the data and how they should be interpreted.

Death Rates

In choosing a hospital, you will want to know how successful the facilities you are considering have been in delivering good results for other patients. The most important result is keeping patients alive.

On our ratings tables, we report risk-adjusted death rates. These rates were adjusted in an effort to take into account the fact that some hospitals treat a relatively high percentage of sicker and frailer patients, who would have a relatively high risk of dying at any hospital.

The adjusted death rates are based on analysis of records of hospital stays of Medicare patients 65 or older admitted to hospitals during a three-year period (fiscal years 2012 to 2014) for medical cases and a four-year period (fiscal years 2011 to 2014) for surgical cases. This is the only available uniform, nationwide data file of hospital cases.

The hospitals submit records to Medicare to get reimbursed for services rendered. Medicare adds one fact to the records submitted by the hospitals—whether the patient died within a specified number of days after hospital admission. Medicare gets this information on deaths, even on deaths that occurred after a patient’s discharge from the hospital, by using Social Security records.

Our mortality rate analysis was conducted by Michael Pine and Associates, a Chicago-based firm that is expert in evaluating the clinical quality of hospitals.

The analysis began with the selection of a subset of cases. We selected types of cases that are relatively common and that have substantial death rates that might be affected by the quality of a hospital’s care. The cases included acute myocardial infarction (heart attack), obstructive pulmonary disease, cerebrovascular accident (stroke), and three other types of cases in which patients were treated medically. They also included 12 types of surgical cases, such as coronary artery bypass graft surgery, bowel surgery, and total hip replacement. The cases were selected from the Medicare records based on detailed definitions using standard diagnosis and procedure codes.

On the ratings tables, you will see, for example, a column showing “Adjusted death rate for all selected cases.” You will see that the death rates range from below 11 percent to more than 15 percent. This difference suggests that, among similar patients with serious medical problems or surgical procedures, those going into one of the high-death-rate hospitals have a four-percentage-point higher chance of dying than those going into one of the low-death-rate hospitals.

A four-percentage-point difference in chance of dying in a few days is something that most people will be intently interested in avoiding. To put this figure in perspective, consider a four-percentage-point difference in death rates for a few days of stay in two different hotels—one hotel with a zero-percent death rate and the other with a four-percent death rate among visitors staying there for a few days. Such a difference would certainly be headline news.

It is important to note that we did not look at all cases. The average death rate across all hospitals for all the types of Medicare cases we selected is higher than the death rate would be if we reported on all cases, including low-risk cases, the hospitals treated. So hospitals are not as dangerous overall as the death rates for our selected cases might suggest, but the adjusted death rates we present are useful in comparing the hospitals for a broad group of serious types of cases.

Adjusted death rates were calculated in several steps.

First, for each type of case, we calculated the actual death rate for each hospital. We counted all deaths that occurred within 90 days of admission, even if they occurred after the patient was discharged from the hospital. Checking for this 90-day period eliminates the possibility that a hospital might have relatively low death rates only because it discharges patients to their homes, hospice care, or nursing homes when they are on the verge of death.

We then calculated a “predicted” death rate for each type of case for each hospital. The predicted death rate tells what percent of the hospital’s patients would have died if the hospital were just as successful as the average of all U.S. hospitals in keeping similar patients alive. The patient characteristics that were taken into account in determining whether patients were similar were age, gender, the presence or absence of selected principal and secondary diagnoses, and whether certain surgical procedures were performed. For example, a hospital that had heart attack patients who were mostly over age 85 with secondary diagnoses like congestive heart failure, diabetes, and malnutrition, might have a considerably higher predicted death rate for heart attack cases than a hospital whose heart attack patients were mostly age 65 to 70 and had few other medical problems. (More detail on the methods used by Michael Pine and Associates to calculate predicted death and adverse outcome rates is included in the technical report.

Next, we used each hospital’s predicted death rate and actual death rate along with the national-average death rate to calculate an “adjusted” death rate for the hospital. The simplest way to calculate an adjusted rate is in two steps. First, we can calculate the ratio of the actual rate to the predicted rate. If, for example, Hospital A has an actual death rate of five percent but would be predicted to have a death rate of 10 percent based on how sick and frail its patients are, then the ratio is 0.5 (five percent divided by 10 percent). Second, we can multiply this ratio by the national average death rate to get the adjusted death rate. If the national average death rate were 12 percent, then the adjusted death rate for Hospital A would be six percent (12 percent multiplied by the 0.5 ratio). In fact, we used a more complicated formula (using odds ratios) for calculating adjusted rates, but the result is very nearly the same.

For each hospital for each category of cases, we also checked whether the difference between the actual death rate and the predicted death rate was “statistically significant.” For example, if the actual rate was nine percent and the predicted death rate was 10 percent, what are the chances that the one-percentage-point difference was the result of the hospital’s simply having had unusually good luck with its patients during the four-year period we analyzed?

We know that some patients survive when the average patient in a similar condition who experienced the same treatment would be expected to die, and some die when the average similar patient given the same treatment would be expected to survive. Since the difference in result can’t be explained, we call it good luck or bad luck. And any given hospital might have a string of good luck or bad luck with its patients. But big differences between actual and predicted death rates for large numbers of cases are not likely to be the result of luck alone. For each hospital for each category of cases on the ratings tables, we have used different typefaces for the adjusted death rate to indicate whether the difference between the hospital’s actual death rate and its predicted death rate was “statistically significant”:

• Green type—The hospital did better than predicted, and there’s less than one chance in 40 that this better-than-predicted experience is just the result of good luck.
• Regular type—although the hospital may have done better or worse than predicted, there’s at least one chance in 20 that its better- or worse-than-predicted experience is just the result of good or bad luck.
• Red type—The hospital did worse than predicted, and there’s less than one chance in 40 that this worse-than-predicted experience is just the result of bad luck.

On the “Overall Death & Adverse Outcome Rates” tab of the ratings tables, we show adjusted death rates and statistical significance for two categories of cases: all selected case types (medical and surgical combined) and all selected medical cases.

On the ratings tables under the “Death and Adverse-Outcome Rates for Specific Cases” tab, we present death analysis results on those three categories of cases and also on more specific categories of medical cases, such as heart attack, stroke, and pneumonia. For specific surgical cases, we present both adjusted death rates and adverse outcomes rates. (See below for a discussion of adverse outcomes rates.) These tables present the data differently from how they are presented for the “Overall” tables. Instead of reporting adjusted death rates, we simply indicate whether the hospital’s death rate for each category of cases was statistically significantly better or worse than predicted, given the hospital’s mix of patients. We score the hospitals as follows:

• means the hospital did better than average given its mix of cases and that there’s less than one chance in 40 that this better-than-average experience is just the result of good luck.
• means the hospital did better than average given its mix of cases and that there’s less than one chance in 20 (but more than one chance in 40) that this better-than-average experience is just the result of good luck.
• means the hospital was so close to the average given its mix of cases that there’s more than one chance in 10 that its difference from the average is just a result of good or bad luck.
• means the hospital did worse than average given its mix of cases and that there’s less than one chance in 20 (but more than one chance in 40) that this worse-than-average experience is just the result of bad luck.
• means the hospital did worse than average given its mix of cases and that there’s less than one chance in 40 that this worse-than-average experience is just the result of bad luck.

If a hospital’s number of cases or predicted number of deaths was too small, we reported neither adjusted death rates nor the significance of the difference between actual and predicted death rates. For such hospitals, the tables show “Insufficient data.”

Interpreting the Death Rate Data

Let’s look a little more fully at the strengths and weaknesses of the death rate data. How valuable are they in choosing a hospital for your care?

The data are helpful in predicting a hospital’s outcomes—especially among hospitals with a reasonably large number of cases. We have found over the years that hospitals that have significantly better-than-average adjusted death rates in one period are substantially more likely than other hospitals to have significantly better-than-average death rates in subsequent years.

But, while the data have predictive power, there is much debate about the usefulness of such data for comparing hospitals. The following are some limitations that are important to keep in mind when considering the data on different hospitals—

• From the billing records submitted to Medicare by the hospitals and used for our analysis, one cannot always be sure whether secondary diagnoses existed when the patient entered the hospital or whether they occurred during the hospital stay. Consider heart attack cases. If the patient’s record says the patient had diabetes as a secondary diagnosis, we can be confident that the diabetes was there on admission. In contrast, if the record says the patient had pneumonia, we can’t confidently make an assumption as to whether the patient came into the hospital with the pneumonia or acquired it in the hospital. Without knowing this, we can’t know whether to give the hospital credit for having more difficult cases if it has an unusually large number of heart attack cases with pneumonia. We wouldn’t want to give the hospital such credit in our analysis if the hospital is causing the pneumonia. Fortunately, Medicare since October 2008 (after much urging by Michael Pine and Associates and others) has been requiring hospitals to have a flag for each secondary diagnosis indicating whether it was present upon admission to the hospital. So, in our analysis, we could use that flag in a heart attack case with a secondary diagnosis of pneumonia to decide whether to give the hospital credit for having a patient who was relatively sick on admission as opposed to having had the pneumonia occur in the hospital.
• The presence of these flags has improved the validity of our analysis. But we know that the hospitals do not always apply the flags accurately-and that a hospital may have incentives to code secondary diagnoses as being present on admission since that will make the hospital’s adjusted results look better than they otherwise would. So this aspect of our analysis can be affected by inaccurate data from the hospitals. We are able to spot, and re-code, in the case of some errors-for example, if diabetes were coded as acquired in the hospital, we would re-code it as present on admission. But there are some types of secondary diagnoses for which coding errors would not be obvious and correctable. So there is still some possibility that our results could be affected by coding errors.
• Because of data limitations, various underlying characteristics of patients could not be considered. Suppose, for example, you are looking at a public hospital that caters to low-income, uninsured patients. There’s a good chance that the hospital’s patients might have social problems—such as the absence of emotionally supportive family members—that are not reported in the data available for analysis but that might affect death rates within 90 days of hospital admission.
  
  
• Within any one of the types of cases we looked at, patients may have diseases at different stages of progression, with very different risks of death. Some hospitals’ pneumonia cases, for example, might include a disproportionately large number of cases in which the disease was at an advanced stage by the time the patient was admitted. The data we were working with did not include information on laboratory or X-ray results, which would make it possible to distinguish among patients on the basis of these findings. Undetected differences in patient mix are especially likely when comparing hospitals if one hospital is a regional referral center to which other hospitals send their difficult cases. 
• Cases were followed for only up to 90 days after admission. Problems caused by some hospitals may not result in death until later than that. (Longer follow-up periods, of course, have their own set of problems since more time increases the chances of death from causes unrelated to the hospital stay.)
• Some differences in death rates may result from differences in community practices or in the availability of non-hospital facilities to care for patients. In some communities, for example, patients in final stages of emphysema (obstructive pulmonary disease) may be allowed to die in their homes or in nursing facilities, while in other communities these patients may be admitted to hospitals for their final few days.
• Some of the data for some hospitals may not be accurate. There are, no doubt, many innocent errors when so many records are processed by hospital coding staffs. In addition, it is likely that hospitals follow different coding guidelines in describing diagnoses in the records they report to Medicare in their efforts to get the highest allowable reimbursements for the cost of care.
• Some of the data are incomplete. For example, the billing record that is the source of the data has space for hospitals to list only eight secondary diagnoses in addition to the principal diagnosis. If a patient had nine or more secondary diagnoses, the adjustment process was not able to allow for the secondary diagnoses in excess of eight.
• Time has elapsed since the period to which the data apply. The Medicare records of hospital cases don’t become immediately available to the public or to researchers, and it took time for us to do our analyses.
• The data are for patients 65 or older. It is possible that hospitals that perform well with that age group don’t do so well with younger patients.
• High or low hospital death rates may result from the quality of treatment provided by specific doctors, not from the quality of the hospital’s performance. If a hospital does well because of specific doctors, that may do you no good if you use a different doctor.

To use our data on death rates, first look for categories of cases like yours on the “specific cases” tabs. If you are looking for a hospital to use for major bowel surgery, for example, look for hospitals with adverse outcomes rates that were significantly better than predicted in that category.

If your case doesn’t fit any of the categories on our table, or if you are selecting a hospital in advance of the need (for example, as a consideration in choosing a physician or HMO), look for a hospital with a favorable score in the “all selected cases” category and possibly in several other categories of cases of interest to you.

It is interesting to note that hospitals that had significantly better-than-predicted mortality rates in the “all selected cases” category also tended to get higher ratings than other hospitals from surveyed physicians. As the figure below shows, the hospitals that had significantly better-than-predicted mortality rates were rated “very good” or “excellent” on average for “surgery on an adult in cases where the risk of complications is high” by 47 percent of surveyed physicians. Hospitals that had significantly worse-than-predicted mortality rates got such favorable ratings from substantially fewer surveyed physicians (only 25 percent).

Be sure to discuss the death rate data with your doctor. Ask for any information he or she has that might explain an especially high or low adjusted death rate. In addition, we recommend asking hospitals for their comments on their death rates. Call the hospitals’ public relations departments.

Rates of Adverse Outcomes

The death rate information we have presented focuses only on one bad outcome: death. But there are other bad outcomes. You don’t want to contract an infection in the hospital, have a bad reaction to a drug, fall out of bed, have to be readmitted for additional care, or have any of many other types of complications even if you ultimately survive. Some complications cause permanent disability or disfigurement; others just make your hospital stay longer and more unpleasant. You want neither.

To give you information that might alert you to high complication rates, we took a roundabout approach. We looked for complications only in surgical cases. Our assumption was that for most of the surgeries, timing was discretionary, and patients would not generally be admitted to a hospital for surgery if they currently had an infection or some other medical problem that might be expected to go away if the surgery were simply delayed. Using these cases, Michael Pine and Associates, which did the analysis for us, developed a proxy indicator for complications in cases where death did not occur within 90 days of hospital admission. The proxy indicator is intended to highlight complications regardless of whether they are reported as such in the hospital records. This proxy indicator looks for prolonged hospital lengths of stay. Analyses of medical records have shown that a large proportion of prolonged lengths of stay are associated with important complications.

Here is a simplified explanation of how that analysis was done. The analysis recognized that, for a given category of cases, a given hospital will have varying lengths of stay, even after allowing for differences in patients’ characteristics. But after allowing for differences in patient characteristics, most of this variation will be clustered around the hospital’s average length of stay for that category of cases. Cases in which the length of stay is not within a hospital’s predicted cluster are likely to involve complications. For each category of cases for each hospital, the analysis identified cases that had lengths of stay outside the predicted cluster of lengths of stay. These were deemed to be prolonged lengths of stay. Such prolonged lengths of stay, like deaths, might occur more often in hospitals with especially sick or frail patients, so the analysis calculated a predicted percentage of prolonged lengths of stay for each hospital based on the mix of characteristics of the hospital’s patients.

The predicted percentage of prolonged lengths of stay was then combined with the predicted percentage of deaths to come up with a predicted percentage of “adverse outcomes” for each hospital. At the same time, the actual percentage of prolonged lengths of stay was combined with the actual percentage of deaths to come up with an actual percentage of “adverse outcomes” for each hospital. The predicted adverse outcome rate was compared to the hospital’s actual adverse outcome rate to calculate a ratio that was in turn used to calculate a risk-adjusted adverse outcome rate. This was done by following the same steps, described above, that were used in calculating a risk-adjusted death rate from the actual, predicted, and all-hospital death rates.

The overall risk-adjusted adverse outcome rates for each hospital, which take into account both deaths and prolonged lengths of stay, are shown on the “Overall Death and Adverse Outcome Rates” tab. On the “Death and Adverse Outcome Rates for Specific Cases” tab, we also report for each hospital whether its actual adverse outcomes rate is significantly different from the predicted adverse outcomes rate for several specific types of surgical cases.

It is important to keep in mind that most of the caveats set out above with regard to adjusted death rates also apply to adjusted adverse outcome rates. In addition, while the death rates measure directly something we care about—death—the adverse outcome rates use a proxy—length of stay—as an indicator of the thing we care about—complications.

Surprisingly, we found no substantial relationship between adjusted death rates and adjusted adverse outcome rates (once we eliminated deaths from the adverse outcome rates). While these two rates are measures of different types of outcomes, one might expect that hospitals that are relatively good at preventing complications would also be relatively good at preventing deaths. But similar analyses done by others, also looking at complications and deaths, have had similar findings. When we ranked hospitals, we put weight on death rates rather than on adverse outcome rates, but both types of rates appear on our ratings tables.

Patient Ratings

The ratings from patients shown in our ratings tables come from federally sponsored surveys, using a standardized questionnaire and survey procedure. The survey asked a random sample of recently discharged patients about important aspects of their hospital experience. The survey attempts to collect at least 300 completed surveys for each hospital every year.

On our ratings tables, we have used Green type to highlight the highest scores and we have marked in red type the lowest scores.

When using these data, keep in mind that the mix of patients can differ from one hospital to the next, and these differences in patient mix can affect a hospital’s survey results. The analysis has tried to take into account these differences so that survey results reported are what would be expected for each hospital if all hospitals had a similar mix of patients, but these adjustments might not completely ensure fair comparisons for all hospitals.

Also, the content of several of the questions is to some degree subjective, and you may have different critical standards than those of the surveyed patients.

Interestingly, as the figure below indicates, hospitals that got relatively high ratings from the doctors we surveyed (see below) also tended to get relatively high ratings from surveyed patients.

Hospitals are not required to participate in the survey of patients, although most larger hospitals will get somewhat lower reimbursement rates on their claims for Medicare patients if they do not participate. We believe all hospitals should participate (with the possible exception of very small hospitals that have too few patients to yield a meaningful sample of patient survey responses.)

Leapfrog Group’s Hospital Safety Grade^SM

These grades come from The Leapfrog Group, an organization that tracks and encourages hospitals’ efforts to improve patient safety. Leapfrog calculates its Safety Grade for more than 2,500 hospitals in the U.S.; grades shown on our ratings table were those reported by Leapfrog in November 2018. The grades used in Leapfrog’s program are derived from expert analysis of publicly available data using national evidence-based measures of patient safety. The Leapfrog Hospital Safety Grade program rates hospitals on their overall performance in keeping patients safe from preventable harm and medical errors. For more information, visit hospitalsafetygrade.org.

Medical School Affiliation

On our ratings tables, we show which hospitals had major affiliations with medical schools according to federal government records. An affiliation with a medical school generally means that various doctors and medical students check patients’ records and interview the patients. That can result in your getting a variety of perspectives brought to bear on your case. Also, doctors who are at the cutting edge of research and practice tend to have at least some of their practice at hospitals affiliated with medical schools. This means that the doctors practicing at these hospitals tend to be exposed to the newest developments and to have their ideas challenged by sophisticated colleagues. In addition, hospitals affiliated with medical schools are likely to have advanced diagnostic and treatment equipment.

On the other hand, hospitals affiliated with medical schools are generally larger than average. The care may be relatively impersonal. And it can be annoying to be bothered by a stream of students and trainees, all with an academic—but often not very personal—interest in your case.

The figure above shows the relationship between major medical school affiliation and physician ratings. As you can see, the hospitals with major medical school affiliations were rated much higher than other hospitals by surveyed physicians. And as shown on the figure below, hospitals with major medical school affiliations also tend to have better adjusted mortality rates for all selected cases than other hospitals.

How Many Cases

Many studies have concluded that, in some types of cases, hospitals that handle large volumes of cases have better results than other hospitals—that, in effect, practice makes perfect. Our tables don’t report on hospitals’ volumes for specific types of cases, but you can ask your doctor or the public information staffs of hospitals you are considering how many cases like yours they handle each year. Your doctor may have to help you define your case type in a precise enough way for the information to be meaningful.

At a very general level, we have reported on our ratings tables the total number of Medicare cases each hospital discharged over the time periods for the selected medical and surgical case types we included in our analysis. This gives you a rough indicator of hospital size. Interestingly, as the figure below shows, the larger hospitals (those with more cases in this total case category) had somewhat lower adjusted death rates for “all selected cases” than hospitals with lower total numbers of cases.

 

Physician Ratings

To collect physicians’ ratings of hospitals, we mailed questionnaires to physicians in the seven metro areas where we publish Consumers’ Consumers' Checkbook magazine (the Boston, Chicago, Philadelphia, San Francisco, Seattle, Twin Cities, and Washington, DC, areas). We surveyed virtually all of the actively practicing, office-based physicians in these areas. Our survey was conducted during the spring and summer of 2014.

We gave each physician a list of area hospitals and asked the physician to rate each for “surgery on an adult in cases where the risk of complications is high.” The physicians were asked to use a five-point scale: “excellent,” “very good,” “good,” “fair,” or “poor,” and simply to leave the form blank for hospitals for which they couldn’t answer.

Our ratings tables show for each hospital the percentage, among the physicians who rated it, who said it was either “very good” or “excellent.” The tables also show the number of raters for each hospital. We have not reported data on any hospital rated by fewer than 20 physicians. Some hospitals were rated “very good” or “excellent” by more than 70 percent of the physicians who rated them, while some got such favorable ratings from fewer than 10 percent. We use Green type to denote the most favorable scores and red type to denote the least favorable scores.

We believe these ratings are a useful indicator of hospital quality, but you should keep several caveats in mind—

• In some cases, physicians may not have had a good basis for judging hospitals. A physician who regularly visits a hospital can get a good perspective on how responsive and thorough the nurses and other staff are, how clean and organized the facility is, and other aspects of quality. But most physicians don’t see all parts of a hospital or see how it performs with all types of cases. And many of the ratings were likely based on reputations among the physicians’ peers rather than direct experience.
• Physicians may have had biases—and possibly financial and professional interests based on their hospital affiliations—that influenced their ratings of specific hospitals.
• The types of physicians who responded to our survey might have had different opinions of the hospitals than the types who chose not to respond.
• Even if there was no consistent difference between the types of physicians who responded and the types who did not, we know that for each hospital there is a mix of opinions among physicians, and it is possible that some hospitals got relatively low (or high) scores simply because they had bad (or good) luck in who happened to respond. For example, for one hospital, our survey might have reached a few physicians with the most negative opinions on a day when these physicians had time to respond while physicians who would have given negative ratings of another hospital were too busy to respond on the day the survey reached them. Naturally, such bad or good luck is less likely to be the explanation for one hospital’s being rated higher than another if the number of raters was large or the two hospitals’ scores are very different.

We also asked the physicians to tell us which two hospitals in the area were most desirable and which were least desirable for various types of cases—

• Surgery on an adult in cases where the risk of complications is low
• Medical care for an adult in complex cases
• Surgery on a child in cases where the risk of complications is high
• High-risk delivery of a baby
• Uncomplicated delivery of a baby

For each type of case, we counted the total number of times each hospital was mentioned favorably and compared that number to the total number of times it was mentioned at all (either favorably or unfavorably). The ratings tables show, for each hospital for the types of cases listed, the percentage of mentions that were favorable. For example, if a hospital was mentioned for a given type of case as most desirable by eight doctors and as least desirable by two doctors, that would give it a score of 80 (eight favorable mentions divided by 10 total mentions equals 80 percent). On the ratings tables, we have not reported scores for a type of case for any hospital that was mentioned fewer than six times, and we have marked with an asterisk (*) any score that is based on fewer than 10 mentions.

This most desirable/least desirable approach to scoring doesn’t allow formal assessments of the statistical significance of hospital-to-hospital differences. But hospitals that look good based on the most desirable/least desirable approach tend also to look good when rated on the five-point poor-to-excellent scale.

You can, on a smaller scale, collect the same kind of information that we collected from physicians. You can ask physicians you know for their recommendations, and you can ask your friends to get the thoughts of physicians they know.

How Our Analysis of Hospital Death Rates Compares to the Government’s Analysis

The federal government’s Hospital Compare identifies a small number of hospitals that have significantly better- or worse-than-average risk-adjusted death rates (adjusted for patient age, multiple conditions, and other characteristics) for heart attack, congestive heart failure, pneumonia, stroke, and obstructive pulmonary disease patients.

We support the government’s efforts to measure and identify differences in hospital outcomes. Unfortunately, however, at the time of this writing, Hospital Compare singles out very few hospitals as having significantly better- or worse-than-average death rates. For example, the government identifies fewer than seven percent of hospitals as either better- or worse-than-average for heart failure cases. So the government’s Hospital Compare website’s information on these measures is not of much use to most consumers.

Consumers' Checkbook, on the other hand, has identified more hospitals as performing significantly better- or worse-than-average. For example, for heart failure, we identified about 31 percent of hospitals we evaluated nationwide as having statistically significantly either better- or worse-than-average death rates. We also report on results for more types of cases—for six serious medical conditions and 12 major surgical procedures.

The good news is that hospitals that did well in the government’s analysis for the most part also scored well in our analysis. For instance, among the hospitals the government’s Hospital Compare website identified as having statistically significantly better-than-average risk-adjusted death rates for congestive heart failure patients, 68 percent also show up in Consumers' Checkbook’s analysis as having statistically significantly better-than-average risk-adjusted death rates in heart failure cases. And all of the hospitals with statistically significantly better-than-average risk-adjusted death rates in the government’s analysis for heart failure cases had at least better-than-average death rates for heart failure patients in our analysis, though not all showed up as being statistically significantly better-than-average in our analysis. We found similar results for the four other conditions that both the government and we analyzed.

This similarity in results exists despite the fact that the patients whose outcomes are being measured were not the same in the government’s analysis as in Consumers' Checkbook’s analysis and despite the fact that there were differences in analysis methods and available data details.