1 of 1 people found this helpful
I work in the quality management department of a hospital/physician practice system and regularly deal with these issues. Unfortunately, I've seen very little that's available on dealing with readmissions other than a screenshot here or there, from my own perspective the work to generate a useful data set has been too much of a mountain to tackle.
One of my hopes for the Tableau Customer Conference is that we can use the healthcare meetup session to make some connections and share some best practices.
Not knowing your data, I can't say exactly how to approach this, but let's just assume that each row in your raw data is a patient visit with a visit ID, medical record number (that crosses visits), an admit date, and a discharge date. The first thing I'd do is develop some Custom SQL to identify the readmissions and associate them back to the index admits, so each row that Tableau gets has an "index" admission and potentially data about the readmission. While you can do this in Tableau with table calculations, further aggregations get more complex, so it's better to flag the readmissions in SQL. Different organizations have different rules about what constitutes an index admission and a readmission (Medicare uses at least 7 different readmission measures that I know of).
Once you have the Admit Date and Discharge Date, you can make a row level calculated field like [Discharge Date]-[Admit Date]. However, you also need to think about when and where you will round. By default, that number will be a decimal number of days that Tableau can format, but your organization might have rules about when the Length of Stay goes from 0 to 1 days, etc. Once you have that number as a continuous Measure, then you can drag it into views and do sum, avg, etc. aggregations.
Besides Length of Stay, I typically have two other measures that I use. I use SUM([Number of Records]) to be the # of Discharges, and then something like SUM(IF NOT ISNULL([Readmit Admit Date]) THEN 1 ELSE 0 END) for the # of Readmits. With those two measures in place, then your Readmit Rate is [# of Readmits]/[# of Discharges]. And when you have all that, then you can start using other dimensions in the data (like the floor/unit/office, diagnosis, discharge disposition, etc. and create pictures like this, which is looking at the readmission rate per hour of discharge date for a set of patients, with a discrete dimension on the Row Shelf to create the stacked small multiples, another discrete Dimenson on the Color Shelf to create the three colors, the number of discharges is on the Size Shelf for the circles, and a polynomial trend line is turned on. The trend lines are actually unecessary to get a sense of what's going on in this view, which is that discharges later in the day for this cohort have an increase in readmission rates, particularly for the brown circles.
Hopefully this will help you get started. If you'd like to set up a webex at some point, contact me at jonathan (dot) drummey (at) gmail.com and I can try to help you through any sticky spots.
Most went over my head. My hospital data can be quite complex when dealing with dates, patients, re-admissions, DRGs and so on. Do you recommend for me to get some Tableau training? Which courses? I don't know SQL so I am at a disadvantage.
I know what you mean about complexity, I deal with the same kind of data - literally hundreds of dimensions (patient class, financial class, admit source, discharge disposition, DRG, primary and secondary diagnoses, etc.) with very few measures to start with.
In your case, I categorize the "what you need to know to get the job done" into three areas:
- How to use Tableau to build a view in general - this is knowing about blue pills and green pills, aggregated and non-aggregated fields, the basics of calculated fields, etc., also known as the "learn how to walk before you run." You can get what need in this area from the online training, the knowledge base, the Tableau documentation, outside books like the manuals written by Freakalytics or Larry Keller, or in-person training, from Tableau, Interworks, VIA Insight, or another consultancy (these are all US based ones I know of).
- How to get your data properly shaped for Tableau - I don't know what format your data is in, it might be really ready to go or might not be. Here, having someone with experience with hospital data is helpful, and certainly knowing SQL can be helpful. When the data is really ready for Tableau, building views is very often as easy as the marketing videos show, whereas when it's not there can be a lot of head scratching and frustration, so (for me, at least) it's worth taking the time to get the data ready. For resources, there's the Tableau forums, Tableau tech support, folks like myself who do some independent consulting, Tableau's professional consulting, and consultancies that focus on healthcare, including VIA Insight, Vizual Outcomes, and USEReady. Also, since we work with patient data, there's an extra burden in terms of HIPAA compliance on top of the usual confidentiality requirements whenever we are sharing data. Entering into a contract with a consultant can help you around that.
- How to build a set of readmissions views. This is an area where, as I'd noted in my first post, there is little to no public information at this time. However, once you get the first two items taken care of, the views will come together very quickly. I'd be happy to privately share some of what I've done, I'm not able to post it publicly.
Not knowing anything about your data, timeframes, or resources, it's hard for me to give more suggestions, I hope this is helpful to you!