The Stew BLOG

Rethinking the Bradley Ratio

Jack Homer, Senior Modeling Consultant | 07/06/2016

A 50-state study on the relationship between social service expenditures, health care expenditures, and health outcomes was published recently in the journal Health Affairs. Conducted by Elizabeth Bradley and her Yale School of Public Health colleagues, the study found that “states with higher ratios of social to health spending had better health outcomes one and two years later, compared to states with lower ratios.” But the authors caution: “Although it may be tempting to interpret this finding as a demand for the reallocation of state monies from health care to social services and public health, it is important to reiterate that we have reported statistical associations and could not infer causality.”

As a fellow public health researcher, I applaud the authors for their work on this important topic, and I certainly agree with the main conclusion about the great potential of social spending for improving population health. I’m also glad that the authors have stated that their statistical finding does not necessarily imply causation.

A possibly unclear message

But I am concerned that they have not given enough emphasis to this caveat, and that they do in fact end up appearing to advocate for a policy of state budget reallocation. I offer the following pieces of evidence for this concern:

  1. A feature article in USA Today, drawing heavily on the Health Affairs paper, states: “Yale University public health professor Elizabeth Bradley, lead author of the study, is urging more efficient—not more—government spending.”
  2. In a current report from Rogan and Bradley, for the Milbank Memorial Fund, the authors state: “Evidence suggests that the paradox of high spending and poor health outcomes may be related to an overemphasis on health care…rather than on health…”
  3. In a current report from Canavan, Rogan, Bradley, and others, for the BlueCross BlueShield of Massachusetts Foundation, the authors find that Massachusetts performed rather well on measures of health despite its having a lower ratio of social to health spending than most other states. Their comment: “Massachusetts appears to be somewhat anomalous.”

Preventive and chronic care improve health outcomes, too

In fact, rather than being anomalous, Massachusetts may be telling us something important about the potential weakness of the Bradley ratio as a guide to policy. The denominator of the ratio, total health spending, includes not only spending for acute and post-acute conditions (typically in hospitals and nursing homes), but also spending on preventive and chronic care (typically in physician offices plus purchases of pharmaceuticals). In the National Health Expenditure Accounts (NHEA), a little less than 30% of all spending on personal health care goes for physician office visits and pharmaceuticals.

What if it were possible to improve population health outcomes not only by increasing social spending, but also by improving preventive and chronic care? In this case, one might achieve good health despite having a lower Bradley ratio. Perhaps this is the case in Massachusetts, a state known for delivering excellent office care.

A simple spreadsheet model illustrates the need for caution

This idea is illustrated with a simple yet plausible algebraic spreadsheet model I have built based on data for the US overall, including NHEA spending data. In this model, health (an index from 0 to 1) is positively impacted by social spending and by preventive and chronic care. Better health means less need for acute and post-acute health spending. The model produces a health index nicely in line with national measures of health-related quality of life, as well as a Bradley ratio close to the 1.06 that Bradley found as the mean across the 50 states for 2000-2009.

This spreadsheet model replicates the basic Bradley result, that by increasing social spending, one will improve health and increase the Bradley ratio. But it is not necessarily the case that a shift in spending from preventive and chronic care to social spending will yield a net improvement in health. In particular, it is possible to adjust the model’s coefficients to produce the Bradley result but with preventive and chronic care being more cost-effective in improving health than social spending is. Under these conditions, the Massachusetts result no longer appears anomalous, and it would actually be a mistake to shift preventive and chronic care spending to social spending!

A more nuanced view

Note that this simple model is static and contains no time lags, accumulations, nonlinearities, or feedback loops of the sort one would see in a more nuanced dynamic model of the situation. Examples of such dynamic models of the health system include the HealthBound model developed for the CDC, and the more recent ReThink Health Dynamics model developed for the Rippel Foundation.

Our modeling at ReThink Health reinforces the main message of the Bradley work, which is the need for smarter investments. We agree that social spending (in the form of programs to promote healthier behaviors and programs to fight poverty) can help to significantly improve a regional health system. But our analysis also shows the importance of several initiatives on the healthcare side of the ledger, especially those that would improve primary care and reduce the waste and excess spending on specialty and hospital care.

Let’s avoid zero-sum thinking

Moreover, the ReThink model shows that there is only a limited amount that can be accomplished by trying to shift among fixed state funds for maximal effect. Boosting those funds—for example, with sin taxes (on tobacco, alcohol, sugary beverages, gambling, fossil fuels, etc.)—can be useful but has its limits. The model suggests there is greater potential in an Accountable Communities for Health (ACH) scheme that would (a) negotiate with insurers to set aside a portion of the savings from the waste-reducing initiatives, and (b) funnel a portion of those shared savings back into a variety of health-improving initiatives. Such a scheme could, within several years, generate the funding necessary to simultaneously pursue both social and healthcare programs, without having to choose between them.

I would hate to see policymakers falling into the zero-sum trap, claiming, for example, that it is necessary to cut state Medicaid eligibility in order to fund social spending. It would be a shame if they cited the work of Bradley and her colleagues to justify such a decision. The Bradley analysis helps us to understand the importance of social spending for health, but it does not allow for a direct comparison with preventive and chronic care. It is likely we would benefit from smart expansions of both.