The Gartner Group recently hosted its 2018 BI bake-off event, aimed at getting Business Intelligence vendors to showcase their products capabilities analyzing data as part of the Data for Good movement, using public data sets that support a social cause. After analyzing traffic fatalities, college costs and homelessness in previous years, the 2018 bake-off was on data about the U.S. opioid epidemic. (See also our previous post on the Drug Overdose Epidemic)
From the bake-off blog post:
There were some consistent findings in all the demos on the descriptive side:
- The opioid crisis is getting worse, not better.
- New York had the most over dose deaths associated with opioids specifically.
- When taking into account population density, West Virginia had the most over dose deaths due to overdoses. It’s hard to say for sure if Opioids are the highest cause as the particular drug is not always captured.
Four panelists provided their respective analysis and dashboards: Microsoft, MicroStrategy, Qlik and Tableau.
Each vendor’s demonstrations were both illustrative of their specific features / capabilities as well as interesting findings on the trends of this epidemic. For example, scatter-plots make it easy to pinpoint outliers, in some cases leading directly to doctors who are now under investigations for prescribing exorbitant amounts of drugs. In many cases, there are both short (5-min) summary videos as well as links to publicly available dashboards to discover underlying data patterns interactively.
Findings from the MicroStrategy article:
(1) Pharmaceutical payments to doctors do not affect prescription rates. We were curious whether the data would show a correlation between pharmaceutical company payments to doctors and their rates of prescription. In fact, we found no such correlation. In other words, there is no evidence that direct payments to doctors result in higher prescription rates.
(2) State GDP and opioid usage are inversely correlated. There seems to be a correlation between state GDP and opioid usage as measured by Medicaid claims. States with lower GDPs tend to have higher opioid usage. This confirms that the opioid epidemic has economic factors.
(3) State college education rates and opioid usage are inversely correlated. There also seems to be a correlation between college education and opioid use as measured by Medicaid claims, but not a similar correlation for high school graduation rates. In other words, states with higher rates of upper education (college or beyond) tend to have lower rates of opioid use, but this doesn’t apply to high school education.
For anyone interested in understanding the opioid epidemic in more detail, having access to its public data as well as various interactive visualizations is valuable. Data for Good, indeed.
End Note: The data used in this bake-off does not align exactly with that published by various sources (CNN, NYT, WaPo) linked to in our previous post. Instead, it appears to be offset by 1 year, that is to say that the data 2015 – 2017 in the bake-off aligns fairly closely with that reported previously as for the years 2014 – 2016.