Like many others, we struggle with the idea of defining “recidivism”, or the preferred term, “recurrence”. HUD won’t do it, and no one that I’ve heard has come out and said, “I GOT IT- this is the definition.” It is impossible to code a great report on this without the answers to some of the questions that come out of these “definition” discussions. These discussions often get mired in questions like “what is an episode?” How long does a person need to be out of the system before we consider their return a recurrence? One day? One week? Further, what is the goal of looking at recurrence? Do we count entering into a permanent housing program after a shelter stay as recurrence or is it strictly only returning to shelters and transitional housing? How do we account for shelters with restrictive return policies? If clients are homeless, but aren’t showing in recurrence data because they exceeded some locally-imposed quota and so are not allowed to re-enter, is it fair to compare recurrence data for that kind of shelter to recurrence data of shelters that are actively placing people into permanent housing? And on and on!!!
So, when, in our Advanced ART training last month, they covered a very easy way of showing recurrence, it sort of enticed me to play around with that and see where it could get us. The basic design of the report is this:
Built on the Entry Exit universe, there are two queries. Both pull in basic information like Provider, Client ID, Entry Exit ID, Entry and Exit dates. The first tab is called Period 1 (the date range for this can be any set of dates). It uses Provider, Period 1 Start, and Period 1 End Plus One Day prompts, pulling in Stayers and Leavers. Nothing mysterious about this query.
The second query is very similar, except for Client ID, you pull in Client ID from the Year 1 tab by using “results from another query”. The prompts are the same as the first query, but you make it read Period 2 Start and Period 2 End Plus One Day. You can either include the Provider prompt in the second query, as was done in the training class, or you can take it out to see all the Entry Exits a client had during the second period, regardless of which provider it was.
When you run it, you make Period 1, say 1/1/2012 to 1/1/2013, and Period 2, 1/1/2013 to now. Choose whichever provider you’d like to know their recurrence rate, and that should be it for the prompts.
When it loads, you should get two tabs, one that shows all the clients from 2012 for the chosen provider, and the second one that shows any clients that were served by that provider in 2012 who also had another Entry Exit in 2013. If you took the provider prompt out of the second query, then you will see all stays any of those clients had, even if they were in permanent housing, rapid rehousing, or whatever.
I haven’t really played with this report further, but when I do, I will be pulling in the Provider Type Code so that I can show how many of the clients that recurred were actually returning to homelessness, since that is typically the purpose for pulling recurrence data.
We ran this report on multiple shelters in our CoC and the percentages of returns to homelessness varied widely, but in sort of expected ways, based on what we know of the shelters we picked.
SOME PROBLEMS with this report:
1. It doesn’t let you account for any length of time NOT homeless in order to call it a true return to homelessness, instead of really more of the same episode. So, a client could have exited in late December and re-entered somewhere on January 1st, and that client would show as having recurred, even though *some* might call that just more of the same episode of homelessness, depending on how recurrence was defined.
2. It misses instances of recurrence where a client may have been homeless in February 2012, and then again in November 2012, since both of those stays would have been accounted for in the first tab. Granted, you can adjust your dates to say, 6 months at a time, or a quarter at time, or you could even make the first time period a month, and then look at the next two years. Really it all depends on how you run the report.
When you are wanting to study recurrence for the first time in your CoC, these problems can be acknowledged and adjusted for in the expectations set by the CoC. We are not fully decided on how to move forward on recurrence, but this very simple report could make its way into our performance measurements as a very rudimentary look at recurrence.
How are you measuring recurrence?