Would supply-side policies reduce regional variation in drug spending?

Based on one study from Sweden, the answer is ‘no’.

A paper by Johansson and Svensson (2022) use data from Sweden on regional variation in drug spending. First, some background on the Swedish approach for prescription drugs spending.

The Swedish national health service offers universal coverage for all residents. The system is decentralized in 21 regions with responsibility to finance and provide health care. The regions subdivide into 290 municipalities with responsibility (among other things) for long-term care. The provision of health care is carried out by a mix of public and private providers and all providers are reimbursed at the same rate through public funds (regional and municipal income taxation).
Health care is subsidized at point of service with relatively small out-of-pocket prices for health services, identical across providers within the same region (private and public). The cost-sharing scheme for prescription drugs is identical for all regions and takes the form of a deductible with multiple thresholds, where the patient annually pays a maximum of €224 (1€ = 10.5 SEK, year 2019). The patient out-of-pocket price for prescription drugs is the same irrespective of whether the physician is employed in public or private. The national Dental and Pharmaceutical Benefits Agency (TLV), regulates which drugs are included in the national pharmaceutical benefits scheme based on health need, disease severity, and cost-effectiveness…If a drug is approved, it is sold at private pharmacies throughout the country at the fixed price agreed by the producer and TLV. In 2009, the pharmacy market was deregulated from a single state-owned pharmacy to allow for multiple private owners, which lead to an increase by 22% in the number of pharmacies…

Despite this relatively centralized system, there is regional variation in drug spending. The authors used data from the National Board of Health and Welfare’s register of prescribed drugs which is merged to Statistics Sweden’s population registers based on individual identification numbers. For the 21 Swedish regions, the authors find that average regional drug spending per capita per year varies from −7% to +28% around the national mean; across 290 municipalities, variation in drug spending ranges from
−28% and +103% around the national mean.

To determine whether this variation in cost is due to demand or supply-side factors, the authors use a two-way fixed effect model with individual and year fixed effects, and an event study specification, similar to the approach from Finkelstein et al. (2016). The approach identifies this decomposition by looking at changes in drug spending when individuals to migrant to different regions within Sweden.

Based on this approach, the authors find that:

…the place effect [i.e., supply-side] accounts for about 5%–10% of the variation in drug spending. The results indicate that most of the variation in drug spending is caused by individual-level demand factors

In short, patient-specific demand factors (e.g., how sick they are) explain the vast majority of geographic variation in drug spending, and policies targeted to providers or supply side aspects are likely to be ineffective (in Sweden at least).

Looking at the broader published literature, the authors note that Moura et al. 2019 find that 28% of regional variation in drug spending is due to supply side factors; this is a higher percentage than Johansson and Svensson (2022) , but still much lower than 50%. Salm and Wübker (2020) estimated a place effect of about 10–20% when examining geographic variation in outpatient spending in Germany. On the other hand, Finkelstein et al. (2016) and Godøy & Huitfeldt (2020) use data from the US (Medicare) and Norway respectively and find supply-side effects explain close to 50% of variation in drug spending.