Public healthcare databases have always, at a crude level, been good for two things: they show us what we are doing right and where we are making progress on big healthcare issues. And they show us what we continue to miss and where we need to continue to focus our attention.

One of my favourite public databases is the Behavioral Risk Factor Surveillance Survey (BRFSS), which is the largest continuously conducted survey in the world. The objective of this annual survey conducted in all 50
states is to collect uniform, state-specific data on US adults’ health-related risk behaviours, chronic health conditions and use of preventive services. But a word of caution: the BRFSS database, and others like it, are cross-sectional designs which, by their very nature preclude our ability to make causal inference between variables. A cross-sectional designed survey or study simply looks at a single point in time and cannot tell us whether the exposure came first, or the outcome came first (unless the exposure is fixed like gender or eye colour).

So, for example, if we were interested in the association between income and overall health, we could not draw a conclusion between the exposure (income) and the outcome (overall health), because we don’t know whether it is the income that contributes to individuals’ overall health or whether it’s individuals’ overall health that drives their income. And, perhaps equally important, many of these databases rely on self-reported data which, of course, is subject to many potential biases.

With this said, let’s take a look at some of the data from the BRFSS. An easy place to start is with the question: how is your general health? And, if you’re like me, you are interested in those whose response to this question is ‘poor’. The states with the highest prevalence of respondents who report themselves being in poor health are all conspicuously concentrated in one geographic band. The beauty of the interactive BRFSS website is that it allows us to change our filters and we can actually look all the ‘poor’ responses by gender or age or educational attainment. So, if you’re interested in baseline health status, this is an interesting place to start.

But what if you’re interested in access to healthcare services for example. Well, take a look at Figure 1 which also comes from the BRFSS interactive website. This is a vitally important question for policymakers and for society at large: was there a time you needed to see a doctor but couldn’t because of cost? Again, the results are interesting but perhaps more so for the prevalence numbers than the geography. In other words, one would not imagine that one-fifth of people in certain states could not see a doctor due to cost barriers. And the other beauty about the BRFSS database and others like it is that you can go back and compare the current year being reported (2016) to a previous year to check any changes. And in this case, there haven’t been any changes going back to 2013. Cost is still a barrier and it hasn’t budged. And what if you’re simply interested in demographics like annual household income or internet usage in the past 30 days. The results are staggering. In some parts of the country close to 50% of survey respondents live on less than $15,000 per year. And similarly, in some parts of the US, between 20% and 40% of survey respondents have not used the internet for 30 days. That is not a misprint. 30 days.

Public databases are largely for researchers and epidemiologists to use. They run complex databases and produce complex results. They are used as the basis for publishing fascinating trends in healthcare among defined populations and for looking at social determinants of health such as income and education. But they also provide a snapshot of health-related issues that force us to open our eyes. Perhaps the greatest value of these public databases is that they draw our attention to disparities in health.

You don’t need a background in health policy or epidemiology to be stunned by some of the data I’ve shared above. And you can surely argue that there are confounders and biases that mitigate some of the raw results we observe. But that would be missing the point.