• Center on Health Equity and Access
  • Clinical
  • Health Care Cost
  • Health Care Delivery
  • Insurance
  • Policy
  • Technology
  • Value-Based Care

Comparing Variation in Medicare and Private Insurance Spending in Texas

The American Journal of Managed CareDecember 2011
Volume 17
Issue 12

We found that, in 2008, variations across Texas in total spending and inpatient utilization are similar in Blue Cross Blue Shield of Texas and Medicare.


A great deal of research has documented the wide variation in Medicare spending across different geographic regions in the United States. However, little research has been done on spending variation in the commercial sector. The objectives of this paper are (1) to compare variations in spending and inpatient utilization in the Blue Cross Blue Shield of Texas (BCBSTX) population and the Medicare population across 32 Texas regions and (2) to investigate if the pattern of widely varying Medicare spending but similar BCBSTX spending found in a previous analysis of El Paso and Hidalgo/McAllen exists across the state.

Study Design:

Retrospective study using 2008 BCBSTX and Medicare data. We used total spending per member/enrollee per month and inpatient admissions per 1000 members/enrollees.


After adjusting BCBSTX and Medicare spending for price and adjusting BCBSTX spending and utilization for age and gender, we computed coefficients of variation, standard deviations from the Texas means, and kernel density estimates for standard deviations from the mean to compare variation in BCBSTX and Medicare spending and inpatient utilization.


Results indicated that variations across Texas in total spending and inpatient utilization are similar in BCBSTX and Medicare both in level and in direction, as the correlations between Medicare and commercial spending and inpatient utilization are positive after excluding the Hidalgo/McAllen regions.


Over the state of Texas, regions of high Medicare spending also tend to be regions of high private insurance spending. McAllen appears to be an outlier for Medicare spending, but not for BCBSTX spending.

(Am J Manag Care. 2011;17(12):e488-e495)

A great deal of research has documented the wide variation in Medicare spending across different geographic regions in the United States, but little is known about spending variation in the commercial sector. Previous research reported widely varying Medicare spending but similar Blue Cross Blue Shield of Texas (BCBSTX) spending in 2 Texas areas, El Paso and McAllen. Using statewide data, we found that:

  • Variation in Medicare and commercial spending is similar across Texas.

  • Regions of high Medicare spending also tend to be regions of high commercial spending.

  • McAllen appears to be an outlier for Medicare spending, but not for BCBSTX spending.

A great deal of research has documented the wide variation in Medicare spending across different geographic regions in the United States. Some experts have cited this variation as evidence of an inefficient healthcare system with estimates reaching 30% for the proportion of US health spending wasted on unnecessary and potentially harmful medical care.1,2

More insight on the amount of inappropriate variation may be gained by exploring spending variation levels in the commercial sector. While private payers must also deal with high healthcare spending growth, the tools for cost containment are different in the public and private sectors. In general, public payers are more likely to leverage their ability to control prices, whereas private payers are more likely to use utilization controls, such as disease management programs and prior authorization.

The research in this area is limited, but 2 of the more robust studies found greater spending variation in the commercial sector.3,4 While these findings support the notion of Medicare pricing policies playing a role in cost control, the research is still far from being able to provide definitive explanations for the differences in public and private efficiencies. Studies have also found positive correlations between hospital utilization in Medicare and privately insured patients,3,5 but with levels of hospital utilization variation somewhat less in the private sector,4 thereby supporting the notion of how the private sector manages care.

In an earlier study, stimulated by the significant difference in per capita Medicare spending between the Texas counties Hidalgo (which includes the city of McAllen) and El Paso,6 we found that overall per capita spending for the populations insured by Blue Cross Blue Shield of Texas (BCBSTX) in these 2 demographically similar communities was quite similar.7 The significant and somewhat surprising findings in the Hidalgo/El Paso paper prompted an expanded study across the state of Texas to examine the extent to which the earlier findings could be generalized. We obtained data similar to the earlier study from BCBSTX on 32 regions in Texas and compared the Texas-wide spending and inpatient utilization variation of BCBSTX with that of Medicare.



Blue Cross Blue Shield provided aggregated spending and utilization data for 16 geographic areas and 16 counties in Texas. Each county and geographic area had at least 2000 BCBSTX members in each of 4 age categories and 2 gender categories we requested. The 16 geographic areas were broadly defined as contiguous counties that represent healthcare markets for BCBSTX and were obtained by building around Metropolitan Statistical Areas to approximate pricing markets and satisfy general population requirements. The 16 counties were large counties that satisfy general population requirements. Data for the counties were also included in the data for the geographic units in which each county is located (see Figure 1 and Table). In order to maximize the number of regions used in the analyses, we separately identified spending and utilization data for the counties for which data were provided and the surrounding regions. For example, we had data on Harris County, Galveston County, Ford Bend County, and Montgomery County in the greater Houston area. In the analyses, we used 5 regions for the greater Houston area: the 4 counties (Harris, Galveston, Fort Bend, and Montgomery) and the aggregate of the remaining counties (listed as Houston Surrounding Region in the Table). We had a total of 32 regions for the analyses. A list of counties included in each region and regions’ characteristics are presented in the Table. We used the same regions for the Medicare population.


The private sector data were for 2,440,806 BCBSTX subscribers, which includes all their Texas subscribers in 2008 with the exceptions described below. This included fully insured lives as well as lives for which BCBSTX does not bear financial risk, but provides administrative services. Included were members of fee-for-service (FFS) plans, preferred provider organizations, and point-of-service plans, but subscribers in health maintenance organizations and indemnity plans were excluded because data for those plans were not easily comparable. Similarly, we excluded subscribers to the Federal Employee Program, which has non-standard financial agreements, and the Texas Health Insurance Pool, which represents a different risk pool. These 2 programs represent 9% of subscribers. Members who are primarily insured through Medicare were also excluded.

Expenditures were reported on a per-member-per-month basis and were based on allowable charges, which include any deductible, co-pay, and coinsurance paid out of pocket by the patient, as well as the payments made by BCBSTX to the providers. BCBSTX provided estimates for total spending amounts, excluding prescription drug spending, and for inpatient hospital admissions. Data were provided in aggregate by gender for 4 age cohorts (0-20, 21-44, 45-54, and 55-64 years).

Medicare Data

Total Medicare spending amounts for 2008 were taken from county-level estimates derived from the Office of the Actuary at the Centers for Medicare & Medicaid Services (CMS).8 Each BCBSTX region is based on county boundaries, so we were able to aggregate the Medicare county-level data accordingly to match the BCBSTX units of analysis. Medicare inpatient utilization data were derived from the Medicare Provider and Analysis Review file. Only discharges for short-term acutecare hospitals and critical access hospitals were included in the analysis to better match the data on the under—65-years-old population, who are less likely to be admitted to post–acutecare facilities such as long-term–care hospitals.

As opposed to the BCBSTX data, the Medicare spending figures do not include beneficiary cost-sharing amounts. This is not expected to bias the results as we are mainly interested in relative spending comparisons (as opposed to differences in spending levels) and Medicare cost-sharing is fairly standard throughout the program. Only data from enrollees of the traditional Medicare FFS program are included, so Medicare managed care (ie, Part C) and outpatient pharmacy, and mail order drugs (ie, Part D) are excluded. Also, enrollees who became eligible for Medicare because of a disability or endstage renal disease are excluded from the Medicare data. Data from 1,921,377 Medicare FFS enrollees were included in the analysis.

Price Adjustment

In order to capture differences across the regions for the costs of providing care, we used the Medicare hospital wage index for inpatient and outpatient hospital expenditures and the Medicare geographic adjustment factor, based on geographic practice cost indices, for professional expenditures.9 We similarly price adjusted expenditures for both BCBSTX and Medicare. The price adjustment is only adjusting for input prices. Variation in market prices above differences in input prices remain.

Adjustments for Age and Gender

Ideally, we would like to control for the underlying risk severity for the Medicare and BCBSTX populations across each region. However, because of data limitations, we could not apply a robust risk adjustment methodology for both payers. Nevertheless, because we previously found age and gender differences in BCBSTX health spending,7 we adjusted the BCBSTX data for gender and age group by computing spending and utilization in each region standardized to the Texas average distribution of gender and age groups for BCBSTX subscribers. We did not have age and gender aggregations to make similar adjustments to the Medicare data. However, we performed an analysis to see how sensitive our results were to underlying health status differences in the Medicare populations. We used a weighted average of the CMS hierarchical condition category (HCC) scores for the Medicare FFS population by county to adjust for health status at the regional level. The HCC scores adjust for prior disease histories as well as for age and gender impacts on spending. We found that the correlations between Medicare and BCBSTX spending were similar when using figures adjusted or unadjusted for health status. We report results using the unadjusted Medicare values in the paper, as we believe they are more comparable to the BCBSTX figures which do not account for prior disease histories.

Statistical Analysis

We conducted statistical analyses on spending and utilization variation within BCBSTX and Medicare and between the 2 payers. Statewide payer-specific measures of variation were weighted by the regional enrollment figures for the respective payer. For measures that include both BCBSTX and Medicare data, for example in computing correlations, we weighted by each region’s total population. We used several measures of variation for average total spending per enrollee/ member and average number of admissions per 1000 enrollees/ members. The simplest measure is the variance; however, the variance is dependent on the average levels, which are considerably higher for the over-65 population than for younger age groups. Therefore we computed the coefficients of variation (standard deviation [SD] divided by the mean) for BCBSTX and Medicare to adjust for levels. To indicate how each region differed from the overall Texas mean, we used the SD from the mean (representing the number of the SDs away from the overall Texas mean) and the index of variation (obtained by dividing each observation by the overall Texas mean). Finally, we computed the kernel density estimates for SDs from the mean and indices of variation for expenditures and utilization by BCBSTX and by Medicare. The kernel density estimates provide a graphical depiction of the expenditure and utilization distributions across the regions. Because the findings were similar for analyses using SDs from the mean and indices of variation, we report only results for SDs from the mean.


Average total spending was $849.07 (SD $92.20) per Medicare enrollee per month and $258.30 (SD $22.17) per BCBSTX member per month. The deviations from the mean for BCBSTX and Medicare total spending in each region are presented in the Table and Figure 1. The deviations from the mean presented a pattern of regions with high Medicare expenditures tending to also have high BCBSTX expenditures. In 22 of 32 regions, total spending deviated from the mean in the same direction for Medicare and BCBSTX. There were some exceptions—in particular, in Hidalgo County and the surrounding counties (described as the Valley Region in the Table), where Medicare spending was the highest in Texas, while BCBSTX spending was among the lowest. Coefficients of variation for total spending are similar within payers, though slightly lower for BCBSTX (0.09) than for Medicare (0.11). However, when we exclude Hidalgo and the Valley Region, the spending coefficient of variation drops from 0.11 to 0.07 for Medicare and from 0.09 to 0.08 for BCBSTX.

Because we had found differences in BCBSTX spending variations by age group in the Hidalgo (McAllen)/El Paso comparison,7 we investigated BCBSTX spending variations across the 32 regions by age group and found very similar coefficients of variation in each age group. In particular, in the 55 to 64 age group the coefficient of variation was 0.10, which is close to 0.09 for the whole sample. (Results by age group are available from the authors upon request.)

Figure 2

In , a scatterplot of the SD from the mean for BCBSTX and Medicare spending for each region visually presents the correlation between the 2 payers. Hidalgo and the Valley Region were clear outliers. The correlation between BCBSTX and Medicare spending is low but negative (-0.09). However, the negative correlation can be explained by the inverse relationship in spending in Hidalgo and the Valley Region. If we exclude those 2 regions, the correlation becomes positive at 0.54.

Inpatient utilization is also similar across Texas for BCBSTX and Medicare. Coefficients of variation for inpatient utilization are similar within payers, though slightly lower for BCBSTX (0.09) than for Medicare (0.11). The correlation between BCBSTX and Medicare inpatient utilization is positive (0.32 and 0.31, excluding Hidalgo and the Valley Region). In 20 of 32 regions, admission rates deviated from the mean in the same direction.

Figure 3

Inwe present the kernel density estimates of deviations from the mean for total spending and inpatient utilization in BCBSTX and Medicare. Each data point underlining the kernel estimate is the difference between the regional level mean and the overall Texas mean. While the kernel density estimates are based on only 32 regions and the visually observable differences are hard to interpret, the figures indicate that the variations in total spending and inpatient utilization are similarly distributed in BCBSTX and Medicare.


Variations in total per capita spending for BCBSTX and Medicare across the state are quite similar. Thus, at the state level, the findings from the Hidalgo (McAllen)/El Paso comparison, which showed significantly higher variation in Medicare spending but similar BCBSTX spending, were not replicated. The statewide results indicated that Hidalgo was an outlier in terms of Medicare spending, but not for BCBSTX spending. In fact, Hidalgo had the highest Medicare spending in the state and was somewhat below average in BCBSTX spending, while El Paso had below average spending for both BCBSTX and Medicare. Our results indicated that, when we excluded Hidalgo County and the Valley Region, there was a positive correlation between BCBSTX and Medicare overall spending, indicating that patterns of geographic variations in the privately insured population and Medicare population were similar.

These results somewhat contrast with previous research that found greater levels of spending variation in the commercial sector.3,4 Partly explaining this may be the fact that the previous studies used data encompassing multiple private payers, which could have added additional variation. Also, in this case, the public sector variation figures are slightly inflated by Hidalgo and the Valley Region, as the Medicare coefficient of variation drops below the BCBSTX coefficient when excluding those counties. Previous research found that Medicare and private insurance plans exhibited a slight negative correlation in spending across Hospital Referral Regions,3 while we found a positive correlation when excluding the 2 outlier regions.

We also did not see significant differences in the variation of inpatient utilization for BCBSTX and Medicare across the state. In fact, inpatient utilization was positively correlated across the 2 payers. These findings are consistent with Baker et al,5 who reported that hospital discharges in the last 2 years of life for chronically ill patients were positively correlated (correlation 0.66) for Medicare FFS and privately insured patients in California, and with Chernew et al,3 who reported a 0.59 correlation between Medicare and privately insured inpatient days.

There are many limitations with these data. Given the aggregated data, the small number of regions, and the lack of outcomes or clinical data, we could not adequately adjust for socioeconomic status and for health status across the state, which can significantly affect spending and utilization levels. Also, it is not clear how much we can generalize these results to the rest of the country and to the entire under-65 population, given that our sample is limited to the BCBSTX enrollees in Texas. In addition, the aggregated spending data include post—acute-care spending, such as for home-health services, which are much more prevalent for the Medicare population. Ideally, we would have excluded these services from the analysis or at least investigated them separately. Home-health spending was found to be a significant contributor to the Hidalgo (McAllen)/El Paso comparison in our previous analysis. In fact, when looking at Medicare homehealth spending across all Texas Hospital Referral Regions,10 McAllen and El Paso are again shown to be high and low outliers, respectively.

We should note that in explaining Medicare variation in per capita expenditures, much of the variation can be attributed to variation in utilization since Medicare prices are relatively fixed. In contrast, variation in BCBSTX per capita expenditures are more difficult to explain since the variation is attributable to both price and utilization variation. Given that BCBSTX cannot set prices in the way that Medicare does, it would not be surprising if prices resulting from negotiations between providers and BCBSTX were to vary significantly depending on the relative bargaining power of the respective parties. There is substantial anecdotal evidence indicating such price variation,11 and the notion is consistent with the results found in previous studies.3,4 Moreover, for private payers, it is possible that the variation in prices and the variation in utilization can offset each other, dampening the overall variation in expenditures. Without detailed pricing information on the private sector, which our data do not provide, it is not possible to explain the underlying causes of private variation.

However, the findings from this study are important in comparing overall spending variations in the Medicare andprivately insured populations since, from a societal perspective, it is the total per capita spending that represents the economic burden and the potential need for interventional policies. While the underlying drivers of variation are important for developing specific policies to mitigate unwarranted variation and potentially lower healthcare expenditures, with our limited data, it is not possible to make specific policy recommendations. Consequently, at this point, it is only possible to note that unlike the McAllen/El Paso comparison, overall spending patterns appear to be similar across Texas for Medicare and BCBSTX. As private and public payers make available more data on costs, prices, and utilization (ideally at the claims level), future research should determine the relative contributions to total cost variation. More research is also needed to identify which factors explain price and utilization variation, such as competitive market forces. This research may be able to provide guidance for specific policies.


The authors thank J. Darren Rogers, Gary S. Brantz, Donald W. Hatfield, and Prashant Nayak for providing Blue Cross Blue Shield of Texas data.

Author Affiliations: From University of Texas School of Public Health (LF, OIM), Houston, TX; The Commonwealth Fund (MZ), New York, NY; Engelberg Center for Health Care Reform (IC, SS), The Brookings Institution, Washington, DC; The Centers for Medicare & Medicaid Services (JDS), Baltimore, MD.

Funding Source: No external funding.

Author Disclosures: Mr Smith reports employment with The Centers for Medicare & Medicaid Services. The other authors (LF, OIM, MZ, IC, SS) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (LF, OIM, MZ, SS); acquisition of data (LF, OIM, MZ, JDS); analysis and interpretation of data (LF, OIM, MZ, IC, SS, JDS); drafting of the manuscript (LF, OIM, MZ, IC, SS, JDS); critical revision of the manuscript for important intellectual content (LF, OIM); statistical analysis (LF, MZ, IC); and administrative, technical, or logistic support (LF).

Address correspondence to: Luisa Franzini, PhD, Director, Division of Management, Policy, and Community Health, University of Texas School of Public Health, 1200 Pressler Dr, Houston, TX 77030. E-mail: Luisa.Franzini@uth.tmc.edu.

1. Fisher E. More care is not better care. National Institute for Health Care Management. Expert Voices. 2005;7. http://www.nihcm.org/pdf/ ExpertV7.pdf. Accessed June 1, 2011.

2. Fisher E, Bynum J, Skinner J. The Policy Implications of Variations in Medicare Spending Growth. A Dartmouth Atlas Project Brief Report. http://www.dartmouthatlas.org/downloads/reports/Policy_Implications_Brief_022709.pdf. Published February 27, 2009. Accessed June 1, 2011.

3. Chernew ME, Sabik LM, Chandra A, Gibson TB, Newhouse JP. Geographic correlation between large-firm commercial spending and Medicare spending. Amer J Manag Care. 2010;16(2):131-138.

4. Philipson T, Seabury S, Lockwood L, Goldman D, Lakdawalla D. Geographic Variation in Health Care: The Role of Private Markets. Paper for Brookings Papers on Economic Activity (preliminary version). Published March 2010.

5. Baker LC, Fisher ES, Wennberg JE. Variations in hospital resource use for Medicare and privately insured populations in California. Health Aff (Millwood). 2008;27(2):w123-w134.

6. Gawande A. The cost conundrum. The New Yorker. June 1, 2009.

7. Franzini L, Mikhail OI, Skinner JS. McAllen and El Paso revisited: Medicare variations not always reflected in the under-sixty-five population. Health Aff (Millwood). 2010;29(12):2302-2309.

8. Centers for Medicare & Medicaid Services. Medicare Advantage —rates & statistics. http://www.cms.gov/MedicareAdvtgSpecRateStats/05_FFS_Data.asp#TopOfP. Accessed June 1, 2011.

9. Medicare Payment Advisory Commission. Measuring regional variation in service use. http://medpac.gov/documents/Dec09_Regional Variation_report.pdf. Published December 2009. Accessed June 1, 2011.

10. The Dartmouth Atlas of Health Care. Selected Medicare reimbursement measures. http://www.dartmouthatlas.org/tools/downloads. aspx#reimbursements. Published 2011. Accessed June 1, 2011.

11. Personal communication from Darren Rodgers. President BCBSTX. May 2011.

Related Videos
Carrie Kozlowski
Carrie Kozlowski, OT, MBA
Carrie Kozlowski, OT, MBA
kimberly westrich
Bruce Sherman, MD
Dr Jeffrey Sippel
Lalan Wilfongd, MD, US Oncology Network
Related Content
© 2024 MJH Life Sciences
All rights reserved.