Risk-adjusted adverse outcomes for elective colorectal surgery vary significantly across regions in the United States, and, therefore, regionally based Medicare payments could disadvantage some hospitals.
The findings, presented at the annual Central Surgical Association meeting, suggest that alternative payment models (APMs) should consider regional benchmarks as a variable when evaluating quality and pricing of episodes of care to level the playing field among hospitals, the study authors said.
To evaluate risk-adjusted outcomes in elective colorectal resections, lead study author Donald E. Fry, MD, FACS, of Northwestern University in Chicago and his colleagues used the 2012-2014 Medicare Limited Data Set to develop prediction models for inpatient deaths (IpDs), prolonged postoperative length-of-stay (prLOS) outliers, 90-day postdischarge deaths without readmission (PD90), and 90-day postdischarge readmissions (RA90) after nonassociated readmission events were removed.
All hospitals with a minimum of 20 evaluable colorectal resection cases from the master data set regardless of coding quality were identified for comparative outcomes. For hospital analysis, the total number of patients with one or more adverse outcomes (AOs) was tabulated. The total predicted AOs were then set equal to the number of observed events for each hospital by multiplication of the hospital-specific predicted value by the ratio of observed-to-predicted events for the entire final hospital population of patients.
Hospitals were then sorted by the nine Census Bureau regions: region 1 (New England), region 2 (Middle Atlantic), region 3 (South Atlantic), region 4 (East South Central), region 5 (West South Central), region 6 (East North Central), region 7 (West North Central), region 8 (Mountain), and region 9 (Pacific).
Within each region, total patients, total observed AOs, and total predicted AOs were derived from the prediction models. Region-specific standard deviations (SDs) were computed and overall region z scores and risk-adjusted AO rates were calculated for comparison.
A total of 1,497 hospitals had 86,624 patients for the comparative analysis of hospitals with 20 or more qualifying cases. Hospitals averaged 57.9 cases with a median of 43 for the study period. Among the AOs, there were 947 IpD (1.1%), 7,268 prLOS (8.4%), 762 PD90 (0.9%), and 14,552 RA90 (16.8%) patients. An additional 1,130 patients died during or following readmission within the 90-day postdischarge period for total postoperative deaths including inpatient and 90 days following discharge of 2,839 (3.3%). Total patients with one or more AOs were 21,064 (24.3%).
Among the hospitals, 49 (3.3%) had z scores of –2.0 or less. These best-performing hospitals had a median z score of –2.24 and a median risk-adjusted AO rate of 10.8%. A total of 159 hospitals (10.6%) had z scores that were greater than –2.0 but less than or equal to –1.0. These hospitals had a median risk-adjusted AO rate of 15.1%. There were 66 hospitals (4.4%) with z scores greater than +2.0. These suboptimal-performing hospitals had a z score of +2.39 and a median risk-adjusted AO rate of 38.8%. A total of 209 hospitals (14.0%) had z scores greater than or equal to +1.0 but less than +2.0. They had a median risk-adjusted AO rate of 32.5%.
Findings showed a nearly 5-SD difference between the Pacific region and the New England region. In addition, the results showed a 2.5% absolute risk–adjusted adverse outcome rate difference between the top- and the lowest-performing regions.
What findings could mean for bundled payments
The findings raise concerns about a lopsided playing field for hospitals when it comes to bundled payments, according to the study authors.
“What that means is if you have more readmissions and more complications and your historical profile is being used to pay for care going forward, regions of the country with poorer outcomes would get higher prices than those areas with better outcomes,” Dr. Fry, executive vice president, clinical outcomes management, MPA Healthcare Solutions, Chicago, said in an interview.
The Centers for Medicare & Medicaid Services has used performance of the Census Bureau region as a major factor in defining target price at the beginning of the Comprehensive Care for Joint Replacement bundled payment program. Regional performance will become the exclusive basis for the target price as the program matures into successive years, the study authors noted. A colorectal surgery bundle has not yet been proposed by Medicare, but because it’s a common operation with relatively high adverse outcomes rates, it is expected to be included in a future bundled payment strategy, according to the investigators.
Dr. Fry and his coauthors concluded that hospitals and surgeons may find meeting the target price of a bundled payment to improve margins or avoid losses more difficult for any inpatient operation if they are in a best-performing region.
A 2.5% adverse outcome rate difference between regions may not seem like much, but the variance could mean a wide disparity in payments, Dr. Fry said.
“We have done previous research with colon operations and identified that a readmission after an elective colon operation costs about an additional $30,000,” he said. “If cases being done in poorer-performing areas of the country have two or three more readmissions per 100 patients, then it means those areas are going to be paid on average $1,000 more per case than would be the circumstance for those areas where outcomes are better.”
By these parameters, the CMS would basically be rewarding care that is suboptimal in the regions with high adverse outcome rates, he said.
“APMs are going to evolve in health care,” Dr. Fry said. “I feel that regional and local outcomes, as illustrated in this article, are different, and that a national standard for expected outcomes needs to be the benchmark. The national benchmark becomes a method to stimulate hospitals and surgeons to know what the results of their care really are and how they compare nationally.”
Perspectives on the study
The study showed that regional and other types of outcome variation need to be considered when using quality measures as a gate to incentive payments under alternative payment models, said Win Whitcomb, MD , a practicing hospitalist and chief medical officer for Remedy Partners, an episodic care company.
However, it remains unclear whether adjustment for quality measures should be based on regions, and, if so, how those regions should be broken down, said Dr. Whitcomb, who is cofounder and past president of the Society of Hospital Medicine. Another question is whether adjustments should take into consideration the characteristics of hospitals, he said, for example, the general demographics of patients who visit academic medical centers, compared with the demographics of community hospitals. Dr. Whitcomb also noted that programs such as the Hospital Readmissions Reduction Program has faced criticism for not factoring in the disproportionately high share of low socioeconomic status patients at some hospitals.
“The paper raises the important issue of comparing apples to apples when quality measures are used to determine payment under alternative payment models,” Dr. Whitcomb said in an interview. “But a number of questions remain about how risk should be adjusted and how benchmarking should occur between hospitals.”
Henry Pitt, MD , FACS, chief quality officer for Temple University Health System in Philadelphia and professor of surgery at Temple University said the study was interesting and raises some important questions, but he had some issues with the methodology. The methodology overweights readmissions and also does not consider the socioeconomic status of patients at the hospitals studied, he said.
“The overall average adverse outcome in the study was 24.3%, but 16.8% of that 24% is due to readmissions,” he said in an interview. “Length of stay and readmissions are correlated – people who tend to have long lengths of stay tend to be readmitted. If you add those two rates together in the study, almost all of the outcome rate is the length of stay and readmissions. You could argue [that] mortality [should] be more heavily weighted than length of stay and readmissions.”
In regard to adjusting risk by region, there is good research that utilization of procedures varies dramatically by region and that some of this variation may have more to do with overutilization of procedures, Dr. Pitt added.
“I would think that risk adjusting for type of operation, diagnosis, and socioeconomic status –across the country – would be more appropriate than risk adjusting by region where there may be major differences in patient selection and indication for operation,” he said.
For example, there is currently an international debate about the management of diverticulitis, including whether and when to operate as well as what procedure(s) to perform.
“It may be in one region, there’s a very low threshold to operate, whereas in another region, there’s a high threshold to operate,” he said. “And the operations that are done may be very different in one region than another.”
Dr. Fry and his colleagues are planning future research in this area, and he said he hopes their studies will impact how the CMS rolls out its bundled payment programs in the future.
“What we’re trying to stimulate is for payment models to be nationally indexed and not regionally indexed,” he said. “CMS is doing that now with [Medicare Severity–Diagnosis Related Groups]. They are paying a price for a total joint replacement, a price for a colon resection, a price for a heart operation – and they do make adjustments based on the local wage and price index, but the core payment is linked to a national payment model, and that’s what we would like to see happen with the bundled payment initiative.”
On Twitter @legal_med