MARCIA A. TESTA, ALEXANDER TURCHIN, OLUKAYODE A. SOSINA, DONALD C. SIMONSON, Boston, MA

Health Care Delivery/Economics Presented on Sunday, June 7, 2015 12:00 PM

Author(s):  MARCIAA. TESTA, ALEXANDER TURCHIN, OLUKAYODE A. SOSINA, DONALD C. SIMONSON, Boston, MA

We previously developed a clinical treatment algorithm to personalize HbA1c goals and establish high benchmark probabilities (HBPs) for achieving glycemic control using clinical, demographic and social markers from a pooled clinical trials database. To test the model in clinical practice, we used a 10-year electronic health record (EHR) database of 2,477 patients with T2DM (58% female, 57.3% White, 16.8% Black, 16.6% Hispanic) who were recently diagnosed, medication naïve, with mean ± SD age=59±13 yrs, HbA1c=8.7±1.8%, BMI=33±7 kg/m², and median income =$51,296. Initial monotherapy was metformin (MET; 77.8%) or sulfonylurea (SU; 22.2%). The prototype logistic regression model applied to the EHR yielded clinical performance probabilities (CPPs) for achieving HbA1c < 7.0% and < 8.0% using inputs of initial HbA1c, sex, age, BMI, DM duration, race/ethnicity and income. Both HBP and CPP models revealed covariate-dependent treatment effectiveness heterogeneity, p < 0.001. The predictive validity of the HBP model (ROC area=0.91) was higher than the CPP model (ROC area=0.73). HBPs for a reference White male, age=50 yrs, BMI=30, FPG=150 mg/dl, HbA1c=9.0%, and DM duration=1 yr were 0.06 and 0.48 for achieving HbA1c < 7% and < 8% without treatment, compared to 0.51 and 0.94 on SU. The corresponding CPPs were 0.46 and 0.62 for HbA1c < 7% and < 8% on SU, but were higher on MET {CPPs =0.66 and 0.79 [OR (95% CLs): 1.8 (1.5, 2.3) and 2.2 (1.7, 2.8) for HbA1c < 7% and < 8%; p < 0.001]}. HBPs for a reference Black female, age=50 yrs, BMI=36, FPG=150 mg/dl, HbA1c=9.5%, DM duration=1 yr were 0.04 and 0.18 without treatment, and 0.41 and 0.78 with SU. Corresponding CPPs were 0.42 and 0.55 with SU, and 0.57 and 0.73 with MET. Standard glycemic control targets are typically static, population-based, and seldom benchmarked against a high quality reference. In contrast, a personalized benchmark using clinical and sociodemographic data reflecting treatment heterogeneity might yield more realistic patient-centered glycemic goals for T2DM.

Disclosure: M.A. Testa: None. A. Turchin: None. O.A. Sosina: None. D.C. Simonson: None.

To view this Abstract and e-Poster, click on the hyperlink and then the View e-Poster Icon    View e-Poster -Click Here 

To open View e-Poster it is best to use Chrome or Mozilla. If your browser opens automatically in Windows Explorer copy and paste direct link below in Chrome or Mozilla browser: https://ada.scientificposters.com/index.cfm?k=3vssynswvf

Note:  You may search and find all e-Posters by clicking on the Main Page General View E-Poster – Click Here Then click on the “Search” tab menu. Set “Poster Topic” = ALL (default), Year = “2015” or any year,  and set “Enter under any search terms, (e.g., 1135-P, health, smith)”  = e.g., 1330-P or any search term (author, topic, etc.).  Click on the “Find” icon to your right.   The abstract title will be shown.  Click on the “e-Poster” icon (or anywhere on the title)  and the abstract will be displayed.  On the abstract page, click on the “View ePoster” icon (found under the title) and the Poster will be displayed.  You may use the zoom in, zoom out and reset frame icons in the lower right corner of the poster to view any part of the ePoster with greater resolution.  All ADA Meeting Content including abstracts, webcast oral presentations and ePosters can be found by going to

http://professional.diabetes.org/CONTENT/PREVIOUS-SCIENTIFIC-SESSIONS-ABSTRACTS-POSTERS-AND-WEBCASTS