Clinical Research Oriented Workshop (CROW) Meeting: January 21, 2016

Present:  Kairn Kelley, Ben Littenberg, Charlie MacLean, Gail Rose, Connie van Eeghen

Start Up:  PowerPoint trick in presentation mode: “B” will black out the screen.

1.                  Discussion: CDC with Graham Center, who contracted with DARTNet: what data elements can predict DM; 8 sites providing data for analysis; UVM-MC is one site.  Rodger made a test request for data (see first slide/page of handout) and needs to improve the request.

a.       Goal: can we identify patients at risk for further developing pre-DM or DM that are not known to have DM (no dx or lab values that indicate DM)

                                                  i.      Who are documented as DM with no follow up

                                                ii.      Who are not documented as DM but has an indicative test result

                                              iii.      Who have no test results, but have risk factors that indicate likelihood of later DM à this is the CDC’s focus: currently unidentified pre-DM and DM

1.      How to predict the lab value of a patient today

                                              iv.      Who have a low test result, but have risk factors that indicate likelihood of later DM: predict the lab value of a patient in the future (interesting and different, but not the CDC’s focus)

b.      Risk factors of interest are known (see second slide in handout)

                                                  i.      Missing: Waist size

                                                ii.      Study: machine learning, using one of a variety of methods

                                              iii.      Question as to whether it is likely to produce a useful predictor, especially since the screen is simple, fast, and cheap (A1c)

c.       UVM-MC has clarified its process for data requests

                                                  i.      Preferred: patient record data

                                                ii.      Alternative: summary records – but these are vulnerable to misinterpretation of the analytic request

                                              iii.      Many requests are developed with the help of Allison Holmes

                                              iv.      Philosophically, better to get all patient records (DM or not), rather than looking for specific risk factors up front (e.g. don’t just look for A1c results, or other clinical outcomes that depend on an assumption about whom to collect those data from)

d.      Next steps:

                                                  i.      Start practicing yoga

                                                ii.      Tell Wilson that these data are accessible from UVM-MC

                                              iii.      Think about who in DARTNet can communicate with the data extractor at each site

2.                  Next Workshop Meeting(s): NOTE NEW TIME: Thursdays, 1:30 p.m. – 2:30 p.m., at Given Courtyard South Level 4.   

a.       January 28, 2016: Kairn’s paper

b.      February 4, 2016: Ben’s paper

c.       February 11, 2016

Recorder: Connie van Eeghen