At Last: The Data To Routinely Discuss Health Spending By Medical Condition

In a paper from the January 2015 issue of Health Affairs, Abe Dunn and colleagues from the U.S. Bureau of Economic Analysis (BEA) introduce readers to their innovative new Health Care Satellite Account (HCSA) and use the data to provide insights about the recent slowdown in health spending growth rates. Rather than focusing on health spending by type of service or product (hospital care, physician services, prescription drugs, etc.), the HCSA presents spending by medical condition. It combines this with information on numbers of persons being treated so that spending growth rates can be decomposed into a volume effect (persons treated) and a price effect (spending per person treated) for each of an all-inclusive set of conditions.

This should be very exciting to those involved with population health, public health, and clinical care as they can now regularly track health spending in terms that relate so much more directly to their areas of concern — the prevalence of disease and the patterns and costs of treatment.

The Health Spending Dialogue Is Shaped By Available Data

Discussions of health spending trends are constrained by available data. The National Health Expenditure Accounts (NHEA), maintained by the Centers for Medicare and Medicaid Services (CMS), presents spending by type of service or product and source of funds. As a result, their annual release analyzes changes primarily in these terms. The BEA National Income and Product Accounts (NIPA) include health sector spending broken out by service/product categories that are similar to those in the NHEA.

Each month, our Center releases a series of health spending reports in which we combine these, and other data sources, to report on health spending, prices, and employment by these same product/service categories. Naturally, we focus on these categories in our discussion of the data each month. For example, in our recent trend report we note that the jump in the health spending growth rate in 2014 was largely due to prescription drugs but the additional increase in 2015 was due to health care services such as spending on hospitals and physicians.

Thanks to the HCSA, we are now in a position to analyze health care spending in terms of the medical conditions being treated, and to determine how much of the growth in spending on a particular condition is due to changes in the proportion of the population being treated for the condition (treated prevalence) versus the average cost per person being treated (cost per case or price). The data reveal which broad categories of medical conditions contributed most to spending growth rates and, for each, the relative importance of changes in treated prevalence and cost per case.

The HCSA Opens The Health Spending Dialogue To Those With Training In Population Health, Public Health, And Clinical Care

This decomposition is especially profound, I believe, because it links health spending and topics such as prevention, population health (and its many non-clinical determinants), and the cost and effectiveness of technological advancements in the treatment of specific medical conditions. In effect, the health spending dialogue that has previously been limited mainly to health economists is now open to perspectives from the clinical, public health, and population health communities.

For example, the HCSA data reveal that spending on circulatory conditions (such as heart disease and stroke) and respiratory conditions (such as emphysema and pneumonia) grew much more slowly than on other broad categories over the 10 year period; slow (and even negative) growth in treated prevalence played a significant role in both cases. A large share of the credit for these slow growth rates is likely shared between past public health successes in reducing cigarette smoking and improved treatments for hypertension and hyperlipidemia, both important risk factors for heart disease and stroke. This type of discussion does not naturally arise when health spending is presented in the usual categories of hospitals, physicians, and prescription drugs.

As another example, the chart below from the Dunn paper shows a slowdown in per capita spending on hypertension, broken into its treated prevalence and cost per case components. Dunn et al relate the slowdown in cost per case to the emergence of less expensive generic substitutes for prescription drugs used to treat hypertension. They also mention that the slowdown in treated prevalence may not reflect trends in the actual prevalence of hypertension, but rather a slowdown in the proportion being treated. Again, the discussion naturally shifts to terms that are familiar and meaningful to those with backgrounds in population health, public health, and clinical care.

Regular Updates And Enhancements Are Key To The Usefulness Of The HCSA

The HCSA is not the first time health spending has been presented by medical condition and broken out by treated prevalence and cost per case. Ken Thorpe and colleagues pioneered work in this area with a series of papers of papers starting in 2005. Subsequent studies include a 2011 paper I wrote with David Rousseau, and a paper by Martha Starr and colleagues from 2014. Data and methods varied considerably across these studies and the final products were limited to the published manuscripts and supplementary materials on methods. The importance of the HCSA is that it promises to be an ongoing, publicly available, government data set with regularly scheduled updates and enhancements. This is essential for a broad ongoing national dialogue about health spending by medical condition and the relative importance of prevalence and cost per case.

Recommendations For Improvement

I have two recommendations for future HCSA enhancements. The first is to expand the data to include spending by institutionalized populations such as nursing home residents and long-term psychiatric patients. Exclusion of this population leads to a significant underestimate of the share of health spending on mental disorders. BEA has been doing research in this area and I look forward to it being incorporated.

The second is to expand the level of detail in the medical conditions reported. The publicly available data are limited to major diagnostic categories such as circulatory conditions and musculoskeletal conditions, while most consumers of the data would be interested in the subsets of these broad categories such as heart disease, stroke, osteoarthritis, and back problems. The HSCA’s blended account, which combines survey data with large claims data sets, is an important advancement over the survey-based estimates from earlier studies and increases the reliability of annual estimates at a greater level of medical condition detail.

Consumers Of HCSA May Misinterpret Treated Prevalence

Treated prevalence is an important measure but must be interpreted carefully. Casual consumers of HCSA data could easily assume it measures the extent of disease in the population (“true” prevalence) when it actually measures only those who are being treated, leaving out the untreated group. As Matt Daly and I have shown in a recent paper, this can lead to the conclusion that certain diseases are spreading rapidly through the population when, in fact, it is mainly the proportion being treated that is rising.

The chart below, taken from that paper, plots treated prevalence and true prevalence for three medical conditions: diabetes, hypertension, and hyperlipidemia. For each condition, treated prevalence is significantly less than true prevalence but the gap is closing over time. The average annual growth rate in treated prevalence far exceeds the growth in true prevalence in all three cases. For hypertension, the average annual growth in treated prevalence is four times as large as that for true prevalence (4 percent versus 1 percent).

Mistaking the growth in treated prevalence for that of true prevalence would result in a highly misleading picture of the spread of hypertension in the population. For hyperlipidemia, treated prevalence is an even more misleading indicator of true prevalence, as it increased by more than 7 percent annually while true prevalence was essentially flat.

Wanted: Time Series Estimates Of True Prevalence By Medical Condition

As someone who has worked in this area for some time, I know how much time and effort has gone into the development of the HCSA, and I hope BEA will continue to invest in future updates and enhancements. It will be important for BEA to continually emphasize that treated prevalence is not true prevalence and that their rates of growth can be dramatically different.

As the HCSA matures, I believe it will well-used and appreciated. It will also create an increasing appetite for data on true prevalence across the full set of reported medical conditions so that population health can be monitored more accurately and untreated portions of the population can be measured and assessed. Creating such data will be no small task, but neither was the HCSA.

Author’s Note

The author receives funding from the Department of Commerce’s Bureau of Economic Analysis (BEA) to update estimates of spending by medical condition for comparison with their Health Care Satellite Account.

from Health Affairs Blog http://ift.tt/1T2NYmF