Clinical and Health Affairs
Differences in the Cost of Health Care Provided by Group Practices in Minnesota
By John E. Kralewski, Ph.D., Bryan E. Dowd, Ph.D., and Yi (Wendy) Xu
■ This article reports the findings of a study designed to identify differences in the cost and quality of care provided by medical group practices in Minnesota. Fifty-three practices that provide services to enrollees of employer-based self-insured health plans were included in the study. Costs adjusted for case mix and payment levels were found to vary from $2,400 to nearly $4,700 per member per year. Quality of care had less variance and was not found to be related to cost. The practices that provided high-quality, low-cost care included both relatively small physician-owned practices and large, multi-clinic systems that also owned hospitals.
One proposal for reducing health care costs that is rapidly gaining support is shifting patients to lower-cost providers. It is argued that some physicians and hospitals charge less for services and that substantial savings can be achieved by seeking care from those providers. A counter argument that is gaining currency is that these savings are unduly optimistic because the most important cost issue is the number of units of service used rather than how much providers are paid for those units. This is the basic philosophy behind the medical home and accountable care organization (ACO) movements.1,2
Supporters of medical homes and ACOs argue that they will reduce costs while improving the quality of care through effective coordination and management of services. In order to determine whether or not these models actually do reduce costs, three things need to be understood: the extent to which variation exists among medical group practices in terms of resources used to care for patients; whether quality of care is linked to cost; and the characteristics of the practices that provide low-cost, high-quality care. In this article, we present data related to variation in resource use among 53 medical groups that provide care to individuals enrolled in employer-based self-insured health plans in Minnesota. We also examine the relationships between differences in cost and quality-of-care measures and practice characteristics such as size and ownership.
Methods
Our study is based on data obtained from a firm that manages claims for self-insured health plans in Minnesota. They identified 53 medical groups, each of which provided care for at least 300 individuals enrolled in those plans. Enrollees were assigned to the practice where they received at least 50% of their primary care during 2007 and 2008. We limited the study to those practices that provided care for at least 300 enrollees to assure a stable database for each practice.
The cost of care provided by each practice was calculated by identifying the services, procedures, and prescription drugs used by each enrollee in that practice and assigning the average allowed amount paid for each unit in all of the practices in the study. This captured both insurance and out-of-pocket payments (deductibles and copayments) and removed differences among the practices that were the result of negotiated payment levels. The resulting costs then were risk-adjusted using 3M Clinical Risk Group software to account for differences in illnesses among the enrollees assigned to each practice.
Quality scores were obtained from MN Community Measurement, which collects data reported by medical practices in Minnesota.3 We included in our study six quality measures related to disease prevention, cancer screening, and management of chronic illnesses. In addition, we calculated avoidable hospitalizations and inappropriate emergency department (ED) admissions for each practice.
Avoidable ED visits were calculated using an algorithm developed by Billings, Parikh, and Mijanovich,4 which uses ED diagnoses to assign probability that the ED visit was 1) not emergent; 2) emergent but did not require care in the ED; 3) emergent but could have been avoided with better primary care; or 4) emergent and required emergency care. We calculated an avoidable ED score for each practice based on the frequency of the first two of those measures per 1,000 enrollees.
Avoidable hospital admissions were calculated using an algorithm developed by Bindman, et al.5 Admissions are considered avoidable to the extent that the diagnosis indicates that they resulted from inadequate primary care in the ambulatory care setting.
Findings
The 53 practices included in our study represented a cross-section of practices in Minnesota. They ranged in size from five to more than 1,500 physicians. Forty percent were physician-owned. The remaining practices were owned by hospitals or integrated systems that have multiple clinic sites, extensive specialty services, and at least one hospital. (It is important to note that when a group practice billed as one unit, it was counted as one practice in our study even though it may have several clinic sites.) All of the group practices in our study provide primary care, but 60% also provide at least some specialty services. All but five of the practices were located in the Minneapolis/St. Paul metropolitan area.
Cost
As shown in Figure 1, the risk- adjusted cost of care provided by the 53 practices varies considerably. At the low end was a practice with costs of $2,405 per member per year (PMPY); at the high end was one with costs that were slightly more than $4,700 PMPY. The mean cost for all of the practices was $3,197 PMPY. We found that the physicians in the practice with the highest costs used almost twice as many units of services to care for patients as those in the practice with the lowest costs.
The mean cost for the 10 practices with the highest costs was $4,128 PMPY compared with $2,680 PMPY for the 10 practices with the lowest costs. Consequently, switching enrollees from the 10 most expensive practices to the 10 least expensive ones could conceivably save nearly $1,500 PMPY. We caution that these costs do not represent what specific practices are actually paid by enrollees’ health plans. Rather, these figures represent the risk-adjusted standardized cost of care based on the units of service used to provide that care.
Quality
As previously noted, the quality scores for our analysis are based on data provided by MN Community Measurement for six measures: optimal care for patients with diabetes, asthma, and hypertension, and screening rates for breast, cervical, and colon cancer. These are the measures for which they had the most complete data. The composite score for each practice is a measure of its performance compared with that of all the other practices on each of the measures. Fourteen of the practices had incomplete data and were dropped from this analysis.
As shown in Table 1, the composite quality scores have less variance than the cost data. They range from 2.60 to 3.75 (mean = 3.10) on a scale of one to five with five representing the highest quality rating. However, an examination of the individual quality components included in the composite score shows greater variance across practices. For example, mammography screening rates vary from a low of 65% of women at risk in one practice to a high of 93% in another. Cervical cancer screening varied from 74% to 93% of the women at risk, and blood pressure control ranged from 40% to 80% of patients at risk.
Practices with high scores on one quality measure did not always perform well on the others. Of the 10 practices that had the highest scores for diabetes care, only two were in the top 10 for three of the other measures; two practices were not in the top 10 for any of the other measures (data not shown). Consequently, when these measures are combined, the variance in aggregate quality scores across the practices decreases.
Only 25% of the medical groups in our study had an incidence of inappropriate ED use in 2007 and 2008; but 43% had avoidable hospitalizations. Nine of the practices that had at least one avoidable hospitalization also had at least one inappropriate ED visit, suggesting that these quality measures are related.
We analyzed the relationship between cost and quality of care two ways. First, we compared the cost and composite quality scores for each practice. Second, we calculated the correlation of specific quality scores with cost. Figure 2 (p. 42) displays the relationship between the composite quality scores and the cost of providing care by the practices reporting complete quality data. These data show that quality of care within these practices does not improve as costs increase beyond approximately $3,000 PMPY. The highest quality scores are achieved by practices that have costs in the $3,000 PMPY range, and scores decline as costs go above $3,100 PMPY.
Our next analysis compared the cost and quality scores to each other statistically by calculating Spearman correlation coefficients for each pair. This indicates the degree to which each score is determined by each of the other scores in the practices (Table 2, p. 43). The magnitude of the influence of these variables on each other was obtained by squaring the coefficient. For example, if a measure has a coefficient of 0.30, it explains 9% of the variance in the paired measure. These data further support the finding that higher costs do not result in higher quality. Other than avoidable hospitalization rates, none of the quality measures included in our study were significantly correlated with the cost of care. Moreover, higher quality scores in one measurement area did not translate into higher quality scores in others. Inappropriate ED use and avoidable hospitalizations were weakly correlated probably because both relate to the overall management of patient care by physicians and the use of ED services that often results in the hospitalization of the patient.
These data also support the previous observation that cancer screening varies within practices. Cervical cancer screening was weakly correlated with colorectal cancer screening, but the coefficient was only 0.43 and the correlation with breast cancer screening was even less. The practices in our study were often not consistent in the provision of these cancer prevention screens.
Best Practices
As shown in Figure 2, seven practices had composite quality scores at or above the mean and cost scores that were at or below the mean. Three of these were physician-owned primary care practices, one was a large physician-owned multispecialty practice with multiple clinic sites, two were owned by large integrated delivery systems, and one was owned by a hospital.
Discussion
Our data confirm that there is significant variance among medical group practices in terms of the cost of care provided to individual patients. Patients who receive care from the practices with the highest costs could save more than $1,000 per family member per year if they would shift to practices that are in the middle of the cost distribution and even more if they sought care from those in the lower-cost quartile. Probably the most important finding is that in most cases, these enrollees also would receive higher-quality care by changing to a lower-cost medical group. As to whether the higher-quality, low-cost practices were owned by physicians or part of larger integrated health care systems, the data were mixed. Five of the 10 practices with the lowest costs were owned by physicians, two were owned by hospitals, and three were part of integrated systems. The 10 practices with the highest costs included four that were owned by physicians, three that were owned by hospitals, and three that were part of integrated delivery systems. Clearly, there are factors other than ownership that differentiate high- and low-cost practices. Organizational factors such as electronic health record capacity, staffing, internal practice efficiency, and, most important, the practice culture will be key variables in our next study. Differences in enrollee lifestyles and the degree to which they manage their illnesses also may vary among medical group practices, and that is not accounted for in our analysis. The best practices may be those that provide high-quality care at lower, but not the lowest, costs and have more patients who play an active role in managing their illnesses. MM
John Kralewski and Bryan Dowd are professors in the University of Minnesota School of Public Health’s Division of Health Policy and Management. Yi (Wendy) Xu is a research assistant.
Acknowledgement: We would like to thank Jim Chase, president of MN Community Measurement, for his thoughtful advice regarding quality of care measures. This project was funded in part by a grant from the Robert Wood Johnson HCFO program.
References
1. Fleming C. Health Policy Brief: Accountable Care Organizations. Health Affairs Blog. July 27, 2010. Available at: http://healthaffairs.org/blog/2010/07/28/end-of-life-savings-the-fools-gold-of-reform/. Accessed January 14, 2011.
2. Devers K., Berenson R. Can Accountable Care Organizations Improve the Value of Health Care by Solving the Cost and Quality Quandaries? Policy Briefs/Timely Analysis of Immediate Health Policy Issues. Urban Institute, October 1, 2009. Available at www.urban.org/publications/411975.html. Accessed January 14, 2011.
3. MN Community Measurement. 2010Annual Health Care Quality Report. Available at: http://mncm.org/site/upload/files/HCQRFinal2010.pdf. Accessed January 14, 2011.
4. Billings JN, Parikh N, Mijanovich T. Emergency room use: the New York story. Issue Brief: The Commonwealth Fund. Number 434, November 1, 2000. Available at: www.commonwealthfund.org/Content/Publications/Issue-Briefs/2000/Nov/Emergency-Room-Use--The-New-York-Story.aspx. Accessed January 14, 2011.
5. Bindman AB, Grumbach K, Osmond D, et al. Preventable hospitalizations and access to health care. JAMA. 1995;274(4):305-11.