There are a wide variety of open data sources for examining post-market surveillance of drugs. Although one data source is only one dimension of analysis, becoming familiar with how to access data for your own analytics is not as dismal as you may think. I use a combination of MedDRA and FDA Adverse Event Reporting System (FAERS) data.
You have options of submitting a Freedom of Information Act (FOIA) request or accessing the public dashboard although you have limited granularity. There are a few easily navigated options for viewing reports and the scale of outcomes--serious reports (excluding death) or death reports. A little background on relational databases allows a little more digging.
Excel isn't the sexiest BI tool on the market but when you need to manage large files it is quite utilitarian. Especially if the friendliest format available for queries is XML or ASCII. I have no idea what happened to the option of readily importing XML so I opted for ASCII. No big deal, you just need to tell Excel how the columns are separated. You may know that CSV is comma delimited--ASCII is $ so I entered that in the screenshot below.
 |  |
I prefer to clean up a spreadsheet in Excel before uploading to Tableau but my current beta version of Tableau 10.5 needed a little exploring. You can also open directly from Dropbox avoiding the need to download
You will need to read the Data Element descriptions found in the ASC_NTS.pdf files before cleaning your data and customizing for better analyses. If you are familiar with Tableau you can also use the native file data reshaper. Also always a good idea to familiarize yourself where the types of data are located.
Not a trivial task but you will need to clean up date formats, create a data model for dealing with missing data, and learn how to clean up the column headings so they are labeled with meaningful labels.
After joining all of the individual tables from the FAERS spreadsheet in Excel it is possible to begin your hypothesis testing. This is a general and perhaps oversimplified visualization but at a glance you can view the drugs with the highest frequency of outcomes reported to FAERs, a snapshot of overall outcomes reveal the incidence of hospitalization and other serious outcomes requiring intervention. The lower right graphic is filtered to a single outcome--death.
Once you begin curating your data it is possible to measure many other variables for example, individual drug reported role in reported events (primary suspect, secondary suspect, concomitant, and interacting for example), route of administration, cumulative dose to first reaction, and who originated the report, consumer or health professional.
Learning to explore data is sort of like a recipe. There are certain skills you need but like anything else--the outcome is better with practice--or consultation.
Learning to explore data is sort of like a recipe. There are certain skills you need but like anything else--the outcome is better with practice--or consultation.
