Data Brief #2 September 2015 Taking the pulse of health care KEY FINDINGS Health care markets are more complex markets than some may have expected The price indices suggest inpatient and A discussion of the Healthy Marketplace Index (HMI) price and utilization outpatient price levels vary across and metrics within markets. Introduction amounts paid by insurers including High price areas the patient cost share component) High price areas—Boulder ,CO, El Paso, The Health Care Cost Institute (HCCI), compared across geographies. Even TX, Bridgeport, CT, Dallas, TX, Milwau- with grant funding from the Robert less has been reported on compari- kee, WI, Philadelphia, PA, Denver, CO, Wood Johnson Foundation (RWJF), sons of prices within geographies. and Fort Collins, CO. has developed a set of economic met- Examination of the HMI metrics reveal Low price areas rics, collectively referred to as the that health care markets are more Low price areas—Tucson, AZ, St. Louis, Healthy Marketplace Index (HMI). complex than some may have ex- MO, New Orleans, LA, Peoria, IL, and Lou- These metrics are intended to provide pected. For example, to date, much of isville, KY. baseline measurements of health care the research on health care services market performance related price, markets has been focused on inpatient Differing inpatient and outpatient productivity, and competition. This price areas hospital markets.2 The price and utili- HCCI Data Brief discusses patterns in zation indices demonstrate that the Low inpatient—high outpatient price the price and utilization indices re- features of inpatient hospital markets levels include: Corpus Christi, TX, Green ported in the 2015 Healthy may not appropriately characterize Bay, WI, Miami, FL, Lakeland, FL, and Marketplace Index Report. The full re- Trenton, NJ. markets for other types of services. In port also includes numerous other particular, the price indices suggest High inpatient—low outpatient price measures regarding resource use, that prices for inpatient and outpa- levels include: Orlando, FL, and Jackson- health, and provider concentration as tient services vary within markets; ville, FL. well as a detailed description of the there are often different price levels methodology.1 Prices do not predict utilization for health care services even within a Only recently has administrative single geography. This implies there No statistical relationship was found be- claims data for the privately insured may actually be different markets for tween the price and utilization indices. population become more accessible different types of services and/or that for research and reporting purposes. the geographic boundaries of the mar- not always use more of the higher priced Historically, many studies have used kets differ.3 It is also likely that prices services. data from Medicare, which has admin- differ between primary care and spe- istered prices and generally covers an cialists or insurance product type (e.g., The HMI, one of the first of its kind, is a set of older and less healthy population. HMO, PPO, etc.). Within the broad cat- comprehensive economic performance met- Much of what is known about the mar- egories of inpatient and outpatient rics on the privately insured population. The kets for health care services for the services there may also be differences HMI metrics were developed using actual privately insured population has been depending on the types of services prices from the administrative claims data of based on case studies of individual (e.g., orthopedics, cardiology, oncolo- a commercially insured population. The markets or subsets of the privately gy, etc.). Furthermore, the amount of measures are intended to provide economic insured population, which are not nec- services used within a geography is performance reference points (e.g., prices, essarily representative of the overall not necessarily indicative of the price resource use, hospital concentration, etc.) population. For example, little was level. Areas with higher prices may for researchers, policy makers, employers, known about how prices (i.e., actual and health care industry leaders. As demon- www.healthcostinstitute.org 1 strated in this brief, the metrics offer across markets. Instead, the utilization ence between the maximum and mini- unique insights at a national and local index allows the mix of services to vary mum index values within the states was level. and measures how the amount of ser- 38% for inpatient prices and 25% for vices used, given a fixed price, influence outpatient prices. For example, the Fort The HMI allows for comparisons of expenditures. A utilization index greater Collins, CO inpatient price index (1.47) price and utilization across markets than 1.00 suggests more high-priced was 43% larger than the Colorado and over time services were used in a given CBSA rela- Springs, CO inpatient price index (1.03). All HMI metrics were calculated using tive to the mix of services used in the The variation in utilization indices be- HCCI’s administrative claims database, total population. In CBSAs where more tween CBSAs in these states was slightly which includes membership and claims lower cost services were used, relative less. The average difference in inpatient data for over one quarter of the national to the total population, the utilization utilization between the minimum and under 65 ESI population. HMI metrics index values was less than 1.00. If the maximum index value was 12%; the were calculated for the population of costs of the mix of services in a CBSA average outpatient utilization difference individuals with employer sponsored was equal to the total population the was 19%. insurance (ESI) for the years 2011 utilization index value was 1.00. through 2013 in 41 Core Based Statisti- There are, however, “neighboring” cal Areas (CBSAs), which is referred to The HMI includes separate indices of CBSAs with similar index values. For as the “total population.”4 This brief re- inpatient and outpatient price and utili- example Atlanta, GA and Augusta, GA ports on the three year averages of met- zation. The CBSA-level three year aver- have similar inpatient price indices ric values by CBSA. Although, the met- age index values are reported in Table (0.92 and 0.94, respectively) and outpa- rics discussed in this brief are price and 1.5 Patterns in the CBSA-level price and tient price indices (0.99 and 0.95, re- utilization indices, any or all of the met- utilization index values are discussed in spectively).7 Even within FL and TX rics can be used to make comparisons this brief, including: similarities as well there are CBSAs with similar indices. across markets or over time. as differences across geographies in For example, the outpatient price index The HMI price and utilization indices price and utilization and relative rank- value in North Port, FL was 0.98 and it were constructed from a fixed set or ings of CBSAs differed by inpatient ver- was 0.97 in Tampa, FL. El Paso, TX and “basket” of health care services allowing sus outpatient price indices; moreover, Austin, TX had inpatient utilization indi- for consistent comparisons. The price no statistical relationship between price ces of 0.97 and 0.98, respectively. While index holds the set of services fixed and and utilization was found. there is a vast and growing literature allows prices for those service to vary examining geographic variation, addi- Although many factors such as practice tional understanding of health care mar- between regions. Therefore, differences patterns, population health status, kets may be achievable through investi- between markets can be attributed to technology, and competition, differ gations of geographic similarities. prices rather than the types or amounts across markets, consistent price and of services used. A CBSA index value of The HMI price indices suggest price utilization levels were observed for 1.00 indicates that, on average for a bas- levels within markets are not consist- some markets within states ket of services, the prices in the CBSA ently higher or lower for all services were equal to those of the total popula- Consistent with a large body of litera- A positive, statistically significant corre- tion. CBSAs with higher than average ture regarding geographic variation in lation was found between inpatient and prices have index values larger than health care prices and utilization, varia- outpatient price indices in each analysis 1.00, and CBSAs with lower than aver- tions were observed across CBSAs with- year, ranging from 0.532 to 0.597.8 This age prices have index values less than in the same state.6 The variation is ex- results suggest inpatient and outpatient 1.00. emplified by the differences between prices are related. However, further ex- CBSAs in Florida, Texas, and Colorado. amination of the CBSA-level index val- The utilization index uses the same bas- Each of these states had HMI metrics ues demonstrated that making generali- ket of services as the price index but reported for multiple CBSAs (8 in FL, 6 zations about overall price levels from a holds the price of each service constant in TX, and 4 in CO). The average differ- www.healthcostinstitute.org 2 single index may be inappropriate. For Orlando, FL and Jacksonville, FL have with the comparison of price indices, example, Miami, FL had relatively low lower outpatient and higher inpatient particular CBSAs of interest can be iden- inpatient price levels and high outpa- prices. tified on the basis of price and utiliza- tient price levels, while the opposite tion index values. More detailed analyses of prices in mar- pattern of price levels was observed in kets where prices diverge by service Markets of particular interest for inter- Orlando, FL. location may also be useful for obtaining ventions are likely those where both As seen in Figure 1, a scatter plot of the insights with broader implications to utilization and price levels are higher average HMI inpatient and outpatient health care markets. There are most than average. These are also the mar- price indices, many of the CBSAs have likely differences in market structures kets where health care expenditures are either higher inpatient or patient prices between those markets with differing likely to be higher than average given levels. The figure is divided into four prices levels and those with similar that expenditures are the products of quadrants, numbered clockwise starting price levels. Efforts to characterize both prices and quantities. The CBSAs with from the upper right. Quadrant I are types of markets may lead to a better three year average price and utilization CBSAs with higher than average inpa- understanding of the demand and sup- index values of 1.05 or greater are la- tient and outpatient prices. Quadrant II ply of health care services, as well as beled in Figure 2 and Figure 3. Denver, contains CBSAs with inpatient price in- identify additional areas for more de- CO has price and utilization index values dex values over 1.00 but outpatient tailed investigations. 5% higher than the total population for price index values less than 1.00. CBSAs both inpatient and outpatient services. Within markets, some health care ser- with inpatient and outpatient price indi- Greensboro, NC, and Palm Bay, FL also vices may be available in both settings ces less than 1.00 appear in Quadrant had price and utilization index values (inpatient and outpatient). There are III. In Quadrant IV, CBSAs have lower greater than 1.05 for inpatient services. many services that are available in only than average inpatient prices and high- Corpus Christi, TX, Fort Collins, CO, one or the other. This is similarly true er than average outpatient prices. Boulder, CO, Philadelphia, PA, and Tren- for the labor and capital needed to pro- ton, NJ had outpatient services price and There are noticeable outliers with high vide the services. Where individuals utilization indices over 1.05. inpatient and outpatient prices receive services, however, is dependent (Boulder, CO, El Paso, TX, Bridgeport, on their health care needs as well as Other markets of interests to research- CT, Dallas, TX, Milwaukee, WI, Philadel- market factors. A better understanding ers as well as policy makers, employers, phia, PA, Denver, CO, and Fort Collins, of the complexities of markets will also, and health care leaders may be those CO), as well as outliers with low inpa- hopefully, result in appreciation of how markets where both utilization and tient and outpatient prices (Tucson, AZ, the health of the population influences price indices are below average. For St. Louis, MO, New Orleans, LA, Peoria, the market as well as how the market example, Miami, FL, and Trenton, NJ had IL, and Louisville, KY). High price and performance affects health. low index values for inpatient services low price areas are often compared and for both price and utilization and Au- No consistent relationship was found both are of interest to researchers and gusta, GA and Peoria, IL had low price between prices and utilization of policy makers trying to understand the and utilization index values for outpa- health care services across markets factors influencing prices. tient services. The HMI results found no statistically The findings, that receive seemingly less significant correlation between price Interestingly, of the CBSAs identified in attention in research and policy, were and utilization indices. Figure 2 and Fig- Figures 2 and 3, Boulder, CO, Philadel- differences between inpatient and out- ure 3 provide visual representation of phia, PA, and Denver, CO were identified patient price levels within the same the price index relative to the utilization as CBSAs with higher than average pric- market. For example, Corpus Christi, TX, index, respectively, for inpatient and es in both inpatient and outpatient pric- Green Bay, WI, Miami, FL, Lakeland, FL, outpatient services.9 The results suggest es. However, only Denver, CO had simi- and Trenton, NJ have lower than aver- at an aggregate level, prices cannot be larly higher utilization index values for age inpatient prices but higher than av- used to make generalizations about uti- both service types. Additionally, Tren- erage outpatient prices. Conversely, lization or vice versa. However, just as ton, NJ, which had above average outpa- www.healthcostinstitute.org 3 tient price and utilization indices, had a lation and includes claims from all 50 produce a weighted mean market bas- below average inpatient price and utili- states and the District of Columbia for ket price for each CBSA and the total zation for inpatient services. over 40 million individuals per year.10 population in each year. The CBSA The HMI analysis cohort includes all weighted mean market basket prices Conclusion adult ESI members ages 18 – 64 for the were divided by the total population There are two main conclusions from years 2011, 2012, and 2013. Core Based weighted mean market basket price this descriptive analysis of the HMI Statistical Areas (CBSAs) were used to from the same year to create CBSA-level price and utilization indices. First, better define market areas for the HMI because index values. performing markets can be identified they contain sufficiently large popula- Endnotes based on the HMI metrics. For example, tions for analyses and are clearly de- some CBSAs have lower prices or lower fined by collections of counties. The 1. HCCI (Health Care Cost Institute). prices and a less costly mix of services. metrics were developed using data from September 2015. 2015 Healthy Second, although CBSAs with better or Marketplace Index Report., DC: HCCI. CBSAs where the total 18 – 64 ESI popu- Web. worse HMI metrics can be identified, a lation was greater than 100,000 and 2. For many reasons, geo-political ranking of market “health” is nearly im- HCCI data included at least 25% of the boundares such as state, county, or possible. Given that price levels varied membership.11 ZIP-code are often used as the basis by type of service within a CBSA and for discussions of health care The price and utilization indices were service areas and/or markets. On- there was no statistical relationship be- built using a set of DRG codes common going HCCI research is also tween prices and utilization it is difficult to the inpatient setting and Current Pro- examining characteristics of to determine how all of those should be markets with more intricate cedural Terminology or Health Care incorporated into a single ranking. definitons based on patient flows. Common Procedure Coding System 3. Gaynor, M. and Town R., While the HMI is a source of perfor- (CPT/HCPCS) codes in frequently ob- "Competition in Health Care mance measurement data, analysis of served in the outpatient setting. The Markets," Handbook of Health HMI metrics also reveals the need for inpatient market basket of services in- Economics, vol. 2 Ed. Pauly, MV, additional research to better under- Mcguire, TG, and Barros, PP. cluded the 100 most commonly occur- stand the dynamics underlying econom- Waltham, MA North Holland 2012 ring DRGs on the basis of the number of ic behavior in health care markets. pages: 499–637. admissions for each DRG in the year 4. United States Census Bureau. The indices discussed in this brief and 2012. Counts of codes were limited to "Metropolitan and Micropolitan," the other HMI metrics provide a com- the 41 CBSAs included in the analyses. Available at: http:// mon set of measures to employers, www.census.gov/population/ For the price index, code level weights metro/. health plans, and providers within their were calculated by dividing the number 5. Index values are reported by year communities. Offering this type of infor- of observations for a given code by the in the HMI report for the years mation to individuals and organizations 2011—2013. total number of codes observed in the who are active participants in health 6. For a comprehensive discussion claims from the total population. A mean and analysis of geographic care financing and delivery will hopeful- price for each code was calculated from variation, see IOM (Institute of ly lead to an improved understanding of the 2012 claims within each CBSA and Medicine). 2013. Variation in Health the markets they operate in and provide for the full population. For the utiliza- Care Spending: Target Decision a basis for sensible and meaningful poli- Making, Not tion index, an average price for each cies that improve the value of health Geography.Washington, DC: The code was calculated using the total pop- care services. National Academies Press. ulation. Code weights were calculated 7. For convenience, CBSAs are Data and methods within each CBSA and for the total popu- referred to throughout the Brief by HCCI’s data is comprised of HIPAA com- lation. For both sets of indices, for a giv- the first city and state in the CBSA name. Complete CBSA names are pliant, statistically de-identified admin- en year, the mean price of each code was listed in Table 1. istrative claims data for approximately multiplied by its respective weight. The 8. HCCI (Health Care Cost Institute). 27% of the national under 65 ESI popu- weighted mean prices were summed to September 2015. 2015 Healthy www.healthcostinstitute.org 4 Marketplace Index Report., DC: HCCI. Web. 9. There may be statistical relationships between price and utilization for particular services. The correlations were performed using the HMI index values. 10. The percentage estimate is based on HCCI's 2012 membership data compared to The U.S. Census Bureau's American Community Survey 3-yr (2010—2012) estimate of the ESI population. 11. The total ESI population was as- sumed to be the ESI population esti- mated by The U.S. Census Bureau's American Community Survey. Author Eric Barrette ebarrette@healthcostinstitute.org 571-257-1584 Copyright 2015 Health Care Cost Institute, Inc. Unless explicitly noted, the content of this report is licensed under a Creative Commons Attribution Non-Commercial No Derivatives 4.0 License Health Care Cost Institute, Inc. 1100 G Street NW, Suite 600 Washington, DC 20005 202-803-5200 www.healthcostinstitute.org 5 Table 1. Three year averages of HMI price and utilization indices Price Utilization CBSA name Inpatient Outpatient Inpatient Outpatient Atlanta-Sandy Springs-Roswell, GA 0.92 0.99 1.01 1.15 Augusta-Richmond County, GA-SC 0.94 0.95 1.06 0.84 Austin-Round Rock, TX 1.01 1.04 0.97 1.11 Beaumont-Port Arthur, TX 0.98 1.08 1.14 1.01 Boulder, CO 1.19 1.13 1.01 1.11 Bridgeport-Stamford-Norwalk, CT 1.23 1.16 0.95 0.70 Cape Coral-Fort Myers, FL 1.03 1.04 1.05 0.92 Cincinnati, OH-KY-IN 0.99 0.89 1.03 1.01 Colorado Springs, CO 1.03 1.04 1.03 0.99 Columbus, OH 1.03 1.05 1.03 0.88 Corpus Christi, TX 0.83 1.08 1.06 1.10 Dallas-Fort Worth-Arlington, TX 1.13 1.11 1.01 0.97 Dayton, OH 1.18 1.00 1.04 0.93 Denver-Aurora-Lakewood, CO 1.05 1.23 1.05 1.12 El Paso, TX 1.16 1.21 0.98 1.00 Fort Collins, CO 1.47 1.32 0.97 1.06 Green Bay, WI 0.95 1.09 1.07 0.90 Greensboro-High Point, NC 1.07 0.98 1.08 0.98 Houston-The Woodlands-Sugar Land, TX 0.99 1.22 1.00 1.03 Jacksonville, FL 1.08 0.88 1.01 1.03 Kansas City, MO-KS 0.89 0.96 1.00 0.99 Lakeland-Winter Haven, FL 0.94 1.11 1.01 0.94 Lexington-Fayette, KY 0.83 0.96 1.04 0.95 Louisville/Jefferson County, KY-IN 0.76 0.85 1.04 1.16 Miami-Fort Lauderdale-West Palm Beach, FL 0.93 1.06 0.95 1.02 Milwaukee-Waukesha-West Allis, WI 1.09 1.07 1.04 0.83 New Orleans-Metairie, LA 0.83 0.80 1.02 0.80 North Port-Sarasota-Bradenton, FL 0.93 0.97 1.06 0.97 Omaha-Council Bluffs, NE-IA 1.00 1.02 1.00 0.81 Orlando-Kissimmee-Sanford, FL 1.22 0.93 0.96 1.06 Palm Bay-Melbourne-Titusville, FL 1.14 1.01 1.06 0.89 Peoria, IL 0.80 0.91 1.04 0.96 Philadelphia-Camden-Wilmington, PA-NJ-DE-MD 1.17 1.09 0.95 1.15 Phoenix-Mesa-Scottsdale, AZ 1.00 0.94 0.98 1.09 St. Louis, MO-IL 0.78 0.76 1.02 0.80 San Antonio-New Braunfels, TX 0.86 0.99 1.01 1.21 Tampa-St. Petersburg-Clearwater, FL 1.03 0.98 1.00 0.91 Trenton, NJ 0.92 1.30 0.94 1.10 Tucson, AZ 0.72 0.69 1.03 0.88 Tulsa, OK 0.82 1.00 1.09 1.06 Washington-Arlington-Alexandria, DC-VA-MD-WV 0.99 0.92 0.95 1.15 Source: HCCI, 2015. Note: Values reported are the averages of 2011 through 2013 CBSA-level indices from the 2015 Healthy Marketplace Index Report. www.healthcostinstitute.org 6 Figure 1. Scatter plot of CBSA-level inpatient and outpatient price indices www.healthcostinstitute.org Source: HCCI, 2015. Note: Values reported are the averages of 2011 through 2013 CBSA-level indices from the 2015 Healthy Marketplace Index Report. 7 Figure 2. Scatter plot of CBSA-level inpatient price and utilization indices www.healthcostinstitute.org Source: HCCI, 2015. Note: Values reported are the averages of 2011 through 2013 CBSA-level indices from the 2015 Healthy Marketplace Index Report. 8 Figure 3. Scatter plot of CBSA-level outpatient price and utilization indices www.healthcostinstitute.org Source: HCCI, 2015. Note: Values reported are the averages of 2011 through 2013 CBSA-level indices from the 2015 Healthy Marketplace Index Report. 9