In May 2013, the Centers for Medicare and Medicaid Services (CMS) released CMS Medicare Provider Analysis and Review (MEDPAR) inpatient data that contain discharge information for 100% of Medicare fee-for-service beneficiaries using hospital inpatient services. This data shows what more than 3,200 hospitals in the United States were being paid for the most frequently performed 100 inpatient procedures. The variations were extraordinary. Some hospitals in the New York State were being paid 40 times as much as the world-famous Mayo clinic for some treatments.
This kind of variation is understandably a huge cause of concern at a time when health care costs are widely seen to be spinning out of control. Our research suggests, however, that the data contains a silver lining: The bulk of excess costs to CMS and inpatients for all the procedures — a total of $5.3 billion above the average across all hospital by procedure — are highly concentrated in just a small number of hospitals.
When we applied the techniques of Six Sigma analysis to the CMS data, we found that just 32 hospitals — less than 1% of the hospitals in the data — accounted for about 25% of the excess accepted charges. (Hospitals determine what they will charge, or bill, for items and services, and CMS then decides how much of that amount is appropriate and will be paid.) A handful of hospitals in New York State accounted for nearly half of them. Add some hospitals in Baltimore, Maryland, and some in the cities of San Francisco, Stanford (Palo Alto), and Los Angeles in California, and the figure goes to nearly 80%. If the excess is that highly concentrated, it is likely that significant efficiency gains can be achieved with relatively little effort.
The discipline of Six Sigma measures the variability of a process attribute, say, costs or prices, in terms of number of standard deviations away from the average. Indeed, the term Six Sigma refers to the three standard deviations (or “sigmas”) below and three above the average for the bell-shaped normal distribution. In this context, the definition of a ‘defective’ cost to patients and CMS is being a statistical outlier: that is, a procedure at a hospital is cost defective if the portion of the charge the CMS and patients paid is more than three standard deviations higher than the average across all hospitals for that procedure.
The CMS data provides summary information on nearly 7 million records of the top 100 most frequently performed procedures performed in 2011. Of these 100 procedures, the top two types alone account for more than a tenth of all procedures performed: major joint replacement (6.1%) and septicemia (4.6%). With such high frequencies, one would expect a more-or-less standard cost to CMS and patients across hospitals. However, for major joint replacement (procedure 470) the cheapest hospital received $9,000 per treatment on average while the most expensive received $39,000; for septicemia (procedure 871), total accepted charges ranged from $7,500 per treatment on average in the lowest-cost hospital to about $44,000 in the highest-cost one.
By carrying out this analysis on the full dataset, we found that fewer than 2% out of the nearly 7 million procedures — 123,243 to be precise — were ‘defects’ in the sense that the average total payment received per procedure in a particular hospital was a statistical outlier across all hospitals for the respective procedure. This means the amount paid to the hospital for the procedure was more than three standard deviations higher than the average payment for that procedure across all hospitals. Note that the standard deviation is already quite high as reflected in the wide variation in costs to CMS and patients; so to be an outlier is quite something.
(A simplifying assumption necessary for using the CMS data is that each hospital charges the same amount for any particular procedure for each time it carried out this procedure was carried out in this hospital in 2011. Getting rid of this assumption by dealing with all procedures at a given hospital individually would not change the overall results, though.)
The exhibit below shows the 50 hospitals responsible for 80% of the total 123, 243 cost-defective procedures across all hospitals in the United States in the CMS data. As a result, these hospitals received extra revenues.
The exhibit below lists the 32 hospitals (less than 1% of all the hospitals in the data) responsible for more than a quarter of the $5.3 billion extra paid for cost-defective interventions and the excess amount for those procedures that each received.
Owing to the skewed distribution, a majority of the hospitals in the country have below-average accepted charges. Moreover the providers with the most egregious costs to CMS and patients are concentrated in such locations as Brooklyn (nearly 15% of all defects), Bronx (about 9%) and New York City (slightly less than 9%). Providers in New York State alone are responsible for half of all the cost-defective procedures followed by California (about 18%) and Maryland (about 11%).
The conclusion we draw from this is that before fretting about fixing the health care system in general, the U.S. government and the hospitals themselves should consider engaging in a Six-Sigma-type analysis every year for continuous improvement. They can identify the tiny minority of hospitals — or areas within a hospital — that are enormously expensive across many procedures they perform and need better understanding of what is behind their extraordinarily high accepted charges. It may well turn out that many of these excess charges levied by these few hospitals can be significantly reduced simply through better management.
From the Editors of Harvard Business Review and the New England Journal of Medicine