Based on analysis from National Income and Product Accounts (NIPAs), multifactor productivity (MFP) in the health care and social assistance sector declined by 0.4% per between 1987 and 2018. On the other hand, the economy as a whole experience a 0.9% increase in MFP. With new technology, telehealth, gene therapies, treatments going generic, is it really the case that productivity in health care declined?
There are a number of reasons why actual productivity in healthcare may differ from how productivity has been traditionally measured. First, the key output of healthcare is not a good to be purchased, but rather improvements in patient health. Measuring changes in health, however, is often very difficult to measure. Standard goods have a similar issue that quality of a product may be difficult to measure. Another issue is that in national accounts, each provider type (e.g., hospitals, physicians, pharmaceuticals) is treated as a separate industry. This creates problems in cases where new innovation-for instance in pharmaceuticals–may lead to decreased hospitalizations; however the pharmaceutical firms in the national accounts does not get credit for productivity improvement related to hospitalization cost savings.
To solve this problem, a paper by Cutler et al. (2022) creates a satellite account for health using data from the 1999-2012 Medicare Current Beneficiary Survey (MCBS) on health outcomes and cost and data from National Health and Nutrition Examination Survey (NHANES) to measure disease prevalence. How do satellite accounts work? The authors explain:
The key design innovation in our account is to use medical conditions as the industries, not providers of care. For example, there is an industry for heart disease and a second one for lung cancer. Our account makes no distinction based on the type of care provided; all that matters to people is how much they spend on care and their resulting health
One challenge is identifying a limited but relatively comprehensive set of diseases. The authors identified highly prevalent conditions within each ICD chapter as their own conditions and then create a residual “other” category for the remaining diseases identified within the ICD chapter.
The key challenge with this approach is competing risks. Many individuals–especially in the Medicare population–have multiple comorbidities. How does one disentangle the impact of treatment on the diseases separately? The authors use propensity score models to compare people with each condition to a similar group without the condition. The results are then rescaled so that the totals match national data on mortality and spending. While the data does cover the most common 80 diseases, the rescaling is needed as many conditions are omitted from the analysis simply due to the need for parsimony (In fact, there are over 60,000 medical conditions in the ICD-10 coding).
Based on this approach, the authors find:
Over this time period we examine, productivity growth in medical care as a whole is relatively rapid: 1.5 percent annually. This is in contrast to official data, which show virtually no change in productivity over this interval. The high rate of productivity growth is a reflection of the large increase in QALE [quality-adjusted life expectancy] for the elderly population…QALE at age 65 rose by one year. However, medical treatment changes accounted for an even greater increase, roughly 1.7 years, as overall QALE was held back by obesity and other nonmedical factors.