A Disease-Based Price Index (DBPI) measures changes in the average price level to treat an episode of specific disease. A DBPI is calculated by estimating the average expenditure for all medical services used to treat a specific medical condition. It is a good indicator of medical prices in health care because an average customer is more interested in the total costs of treating a disease than the price for a single medical service such as one visit to doctor's office. See the complete description of our methods to construct disease based price indexes.
Currently published official medical price indexes track the price changes of a fixed market basket of medical goods and services while DBPIs encompass all services related to a specific medical condition over a period of time. DBPIs reflect not only changes in the price of the medical services, but also changes in the average utilization of the services for treating a disease. Traditional service based indexes have difficulty capturing technological improvements that allow diseases to be treated using fewer medical services or treatments moving to less intensive settings, e.g. inpatient to outpatient.
Taking a recommendation from National Research Council (2010), the U.S. Bureau of Economic Analysis (BEA) introduced a Health Care Satellite Account(HCSA), in which spending is reported by disease rather than by medical goods and services. BEA has also generated the Medical Care Expenditure (MCE) indexes for broad 18 categories of diseases in the HCSA. The HCSA comes with two versions: MEPS account and Blended account.
Since both BEA and BLS (with a 3-year lag) obtain the service quantity information from the Medical Expenditure Panel Survey (MEPS) and use the same index calculation formulas, differences in the indexes should be attributable to the price component. While BEA directly takes into the calculation the average expenditure to treat a disease with all medical services from MEPS, BLS instead uses current monthly CPI/PPI indexes for each medical service to publish timely indexes. In general, the average service prices per encounter grow faster than the service prices per procedure, resulting in a faster growth in BEA’s MEPS aggregate MCE index than BLS’s aggregate DBPI.
To address volatility issues due to relatively small sample sizes in MEPS data, BEA incorporates large claims data such as MarketScan Data, Medicare claims, and The Medicare Current Beneficiary Survey as supplements to create the so-called blended account. The MCE indexes for each disease category in the blended account are the same as those in the MEPS account; however, aggregate indexes across all diseases are different because of re-calculated expenditure weights with a variety of databases. Thanks to the increased sample sizes in the blended account, BEA recently released MCE indexes at a finer level - from 18 diseases categories to more than 260 specific medical conditions.
The file below contains charts and the monthly and annual history of the various disease based price indexes from January 1999 to the latest month in 2020. It contains not only the utilization adjusted disease based price indexes but also the indexes that are computed under traditional methods (the Lowe Indexes).
All diagnoses are classified into 19 disease categories by the ICD-9 manual (please see the instructions document above for detail). We produce and update our experimental DBPIs for each category on a monthly and/or an annual basis. The monthly indexes are available as both one month relative price change and cumulative versions from the base period- January 1999. The all-disease indexes are calculated by aggregating the individual disease series.
There are 4 options available for each disease index: fixed quantities (Lowe) or adjusted quantities (disease based), dental combined or dental separated, with or without comorbidity adjustment, and smoothed or unsmoothed. In the fixed quantities indexes, quantities are fixed at the base year level throughout the series. In contrast, quantities are updated annually in the adjusted quantities series. The “dental separated” and “dental combined” option refers to whether the dental diseases are separated from diseases of the digestive system (Category 9). In the comorbidities unadjusted index, if a physician treats two diseases in the same visit, one visit will be allocated to the treatment bundle for each disease. In the comorbidities adjusted index, a fraction of the physician visit is assigned to each disease. The treatment bundle for each disease is updated annually. There are the unsmoothed indexes where all the yearly quantity updates are done in January of each year, which causes a “jump” in the January index, and the smoothed indexes where the annual quantity change is spread over the entire year (1/12 of the yearly quantity adjustment is applied to each month).
The indexes are calculated at the disease level and are then aggregated to form an all-disease index. Figure 1 below compares three different all disease price indexes. The first is the all disease index computed under the traditional fixed-basket method where the medical utilization for each disease is fixed at base period (1999) levels. We call this the Lowe Index and it only captures the changes in the prices of medical goods and services. The next two are disease-based price indexes where one makes an adjustment for comorbidities and the other does not. The comorbidity adjustment has a minimal impact on the Lowe index, so only the unadjusted version is presented. Similarly, separating the dental diseases only has a small effect on the aggregate index, so only the dental separated version is presented. From 1999 to 2018, the disease-based price indexes on average grew less rapidly than the traditional Lowe index. However, there were periods when the reverse was true, particularly from 1999 to 2007. This was a period when health insurance coverage shifted from health maintenance organizations to more generous preferred provider policies. In recent years, the disease-based price indexes have grown more slowly than the Lowe index as average utilization for many diseases has decreased.
Figure 2 shows the effect of using different price indexes to calculate real medical expenditures in 2015. Real expenditure in 2015 is calculated by deflating nominal expenditures by the traditional Lowe Index without comorbidities, DBPI without comorbidities, and DBPI with comorbidities, respectively. Using a disease-based price index results in higher real medical care expenditures for 2015 than using the traditional Lowe price indexes. This also increases real GDP.
The file below contains the results from a decomposition of the growth in nominal expenditures by disease into the parts that come from inflation growth, population growth, and prevalence growth. The DBPIs and Lowe indexes used to deflate nominal expenditures here are dental separated, with and without comorbidities adjustments, and with smoothed quantities indexes.
In 2017, healthcare accounted for 17.9 percent of U.S. Gross Domestic Product (GDP). Because healthcare is such a large sector, it is important that we measure its output and prices correctly. If published healthcare inflation rates are too high, then measured real output growth is too low and consumers are getting more for their healthcare dollar than the published estimates suggest. Similarly, if published healthcare inflation rates are too low, measured real output growth would be too high.
The Bureau of Labor Statistics (BLS) is committed to producing and publishing the most accurate medical price indexes possible. BLS has constructed experimental disease-based price indexes to find a better way to estimate inflation, real medical output, and real consumption.
Federal statistical agencies currently report medical data for goods and services. The National Health Expenditure Accounts (NHEA), the National Income and Product Accounts (NIPA), the Producer Price Index (PPI), and the Consumer Price Index (CPI) all report their medical statistics for physician services, hospital services, pharmaceuticals and other types of medical goods and services. However, many economists and others who analyze healthcare data believe this is not the best way to report medical statistics. In 1967, the U.S. Department of Health, Education, and Welfare noted:
"...the average consumer of medical care is not as interested in the price of a visit or hospital day as he is in the total cost of an episode of illness."
Starting with the pioneering work of Anne Scitovsky (1967), many analysts found that reporting medical statistics on a disease basis rather than a goods and services basis could provide better information on well-being. There can be large differences between the two methods because reporting on a disease basis can account for new technology that changes the use of medical resources. For example, in the 1990s a new generation of antidepressants could treat depression with fewer therapy visits. A disease-based price index for depression could account for this change in treatment, but indexes produced under the traditional approach of using medical goods and services could not.
Studies completed in the 1990's and early 2000's compute price indexes for cataracts, heart disease and depression. These studies find that their disease-based price indexes grow less rapidly than indexes based on goods and services. The reason is that innovations changed how medical goods and services are used to treat these diseases. As a result, the Committee on National Statistics (CNSTAT) in 2002 published a recommendation that BLS create experimental disease-based price indexes. This recommendation calls on BLS to use medical claims data to determine the quantity of physician visits, hospital visits and other inputs and use these quantities as weights in the construction of disease-based price indexes. The prices for these indexes would continue to come from the current price-collection system. While BLS would continue to generate monthly experimental disease-based price indexes from its monthly price collection system, the quantities would only be updated every year or two. The information on this page results from the CNSTAT recommendation.
When BLS set out to implement the CNSTAT recommendation, we established several criteria. First, the indexes had to be timely. Second, they needed to have a cost-of-living basis. Third, they could be used as an input for the All-Items Consumer Price Index. Fourth, there could be no additional costs or any disruption to existing statistical programs when constructing these indexes. Finally, the methods must be transparent.
Because of the criterion for no additional costs, BLS could not use medical claims for inputs because medical claims data are expensive. Instead, we use the publicly available Medical Expenditure Panel Survey (MEPS). We then get a blended data result, with prices from the BLS price index programs and quantities from MEPS.
One challenge in constructing disease-based price indexes is the choice of a method that accounts for comorbidities. Comorbidities occur when a physician office visit or a hospital visit treats a patient for more than one disease. We construct two types of disease-based price indexes that account for comorbidities differently.
Similar to BLS's currently published Lowe medical indexes, the experimental disease-based price indexes need a representative sample of medical transaction prices. The sampling of medical prices is a challenging task. Respondent participation in our price-collection programs is voluntary, and the reimbursement rates negotiated between insurers and medical providers often are proprietary. These rates are not posted for all customers to observe in the same way as, say, coffee prices in a grocery store. This puts more burden on respondents for the medical providers and on the BLS field economists who collect these prices. BLS has reduced respondent burden, and we are trying to reduce it even more. We appreciate the cooperation of the medical providers who participate in our price-collection program.
It is a great accomplishment to release these indexes in timely manner without increasing costs or disrupting our current statistical programs. BLS has found a way to use our existing products better.
Yet, there is still much to do. Patients consume medical goods and services to heal or be protected from disease. However, there currently is no reliable data source on the healing and prevention outcomes from medical spending. Many data users have suggested that BLS adjust our healthcare price indexes to reflect changes in the quality of the treatment outcomes that result from new technology. There are many challenges to quality adjustment, and we outline them in our methods.
Disease-based price indexes are in their infancy. We regard them as experimental because we still need to learn more from the research that we and others will conduct. As we learn and improve these indexes, BLS hopes that they will greatly enhance our understanding of the healthcare sector.
We list below additional research about healthcare price indexes. Not all the authors of the research papers and conference presentations are affiliated with BLS. We provide this information for your convenience, and this research does not necessarily reflect the views or policies of BLS.
 We use the MEPS to get the medical spending totals and the most current year is 2015.
 This is an estimate from the National Health Expenditure Accounts (NHEA) by the Centers for Medicare & Medicaid Services, CMS.
 US Department of Health, Education and Welfare (1967), A Report to the President on Medical Care Prices, U.S. Government Printing Office, page 13.
 For heart disease, see Cutler et. al. (1998). For depression, see Berndt et. al. 2002. For cataracts, see Shapiro and Wilcox (1996).
 This is recommendation 6.1 in Mackie and Schultze (2002).
Last Modified Date: March 5, 2020