Statistical and machine-learning tools leveraged from the data science community are critical to an organization’s ability to accurately report all populations served. It allows the question, “how many clients were served last year?” to be answered, a basic but often impossible task using HMIS alone. Rhode Island’s Continuum of Care will serve as an example in describing the problem associated with reliance on HMIS for data reporting.
From afar, one might imagine a coordinated network of service providers who seamlessly manage a single database. From this imagined database, reports could be pulled that provide information about all clients served, across and within programs. Dashboards could be created that show the number of people in shelter, or the number of chronically homeless in a CoC. Right now, HMIS is this imaginary database.
HMIS has a high barrier to entry, and as a result, its data is often not comprehensive of the homeless population within a CoC or sometimes even within a single provider. Programs who are not required to use HMIS have little incentive to do so, resulting in their clients missing from HMIS-based analyses. Other programs simply cannot exclusively use HMIS, such as those serving protected populations (i.e. domestic violence programs) or those who also serve the non-homeless (i.e. managed properties).
The near-ubiquity of HMIS facilitates within-CoC standardization and collaborative longitudinal tracking, but its omnipresence may lead to systematic bias resulting in under-reporting of marginalized and at-risk homeless populations. This beginner-level presentation will describe the field of service providers in Rhode Island’s CoC – all of the places that interface with people broadly experiencing homelessness – and the range of databases and reporting methods each uses. The goal is to identify the places where data is being lost and to recognize the impact of this on individual populations.
Presenter(s): Elizabeth McDonnell, Data Scientist, Crossroads