1. 4th Annual Healthcare Informatics Symposium
Design and Implementation of a Diabetes Medication Computer Assisted Decision Support (CADS) System
4th Annual Healthcare Informatics Symposium
April 29th, 2011
Richard ‘RJ’ Kedziora
In partnership with several leading researchers and endocrinologists (Rodbard &Vigersky) at Walter Reed Army Medical Center with support and funding from the Army Telemedicine and Advanced Technology Research Center (TATRC), we have developed and published a CADS application for the management of patients with Type 2 Diabetes who are on oral agents. The CADS program simplifies the work of the primary care provider by automatically making medication recommendations based on an established consensus algorithm integrating the essential information necessary including the patient’s blood glucose data upload from home, current and previous medications, diagnoses, and current laboratory data. Medication regimen recommendations include modifying the existing regimen, increasing or decreasing the current medication(s) used and/or adding additional oral agents to better control the patient’s diabetes. A one-year multi-site IRB-approved, randomized controlled trial is scheduled to begin the summer of 2011 to test the clinical efficacy of the system. Additional work is scheduled to expand the rules to include Type 1 Diabetes and multiple daily insulin doses (MDI). This presentation will provide an overview of the system and discuss the process of designing and implementing the clinical algorithms within the Comprehensive Diabetes Management Program (CDMP), an open source disease management application. Particular attention will be focused on the decisions and challenges in implementing the complex clinical rules and the technology chosen to implement the solution.

2. Funding / Disclosures Founding Partner/Owner - Estenda Solutions
Funding from
U.S. Army Medical Research and Materiel Command (USARMC) AMEDD Advanced Medical Technology Initiative (AAMTI) program.
Congressionally Directed Medical Research Programs administered by Air Force in partnership with University of Pittsburgh Medical Center - sponsored by the honorable U.S. Representative John P. Murtha
PI on grant COL. Robert Vigersky M.D. at Walter Reed Army Medical Center

3. The Problem
Not enough endocrinologist to treat patients with diabetes – most care is managed by primary care doctors
Most patient’s not goal (A1C 6.5 – 7%)
SMBG Issues
SMBG not used effectively by patients / providers
SMBG perception is that it is not used to adjust medications
Large number of drug and combinations
Therapy is not adjusted frequently enough

4. Potential Medication Combinations
Drug classes include: biguanide, DPP-4 inhibitor, GLP-1 agonist, secretagogue, TZD, AGI, and basal insulin
Excluded: Colesevelam and Bromocriptine
68 potential treatment combinations
8 mono
26 dual
31 triple
3 quadruple
Dual : (M + DPP-4, M + TZD, TZD + GLP-1)
Triple : (M + DPP-4 +TZD, M + GLP-1 + AGI)

5. The Solution - CADS
Designed for primary care doctors to assist in better decision-making in modifying patient’s drug regimen to bring their blood glucose into better control.
Currently Type 2 (Type 1 planning)
Idea, concept and rules developed by
COL. Robert Vigersky, M.D. - Director, Diabetes Institute, Endocrinology Service, Department of Medicine, Walter Reed Army Health Care System, Washington DC
David Rodbard, M.D. – Biomedical Informatics Consultants, LLC, Potomac, Maryland

6. Journey
Multiple facilitated group clinical chart reviews to reach consensus
Initial standalone prototype development using CLIPS and Microsoft ASP – early 2000s
Experimented with DROOLS moved to table-driven algorithm coded in Java
Production system coding and integration with CDMP
Clinical Trial 2011 – 2012 and beyond
FDA Validation

7. Input Age, Gender, Type of Diabetes
Self-managed blood glucose data (SMBG)
Current and past medications
Adverse Reactions
Labs (A1C, ALT, Creatinine)
Significant Diagnoses
Renal, Hepatic, Gastrointestinal, Cardiac
Target A1C

11. SMBG Testing Protocol For 3 months
Twice daily (or more depending on DR. discretion)
Once a week
Before meals (x3) and bedtime = 4 tests
Once a month
Before and 2 hours after meals (x3), bedtime and night at approximately 3AM = 8 tests

12. Pre-Analysis Availability of SMBG
SMBG correlation with most recent A1c
Identification of problem time-frames based on SMBG data
Hypoglycemia
Hyperglycemia
Variability
Time frames – before breakfast, after lunch

13. Analysis Overall quality of glycemic control
Effectiveness of SMBG testing
Inappropriate medication combinations
Existing Medication Contraindications
Age, Gender, Labs, Diagnoses
Based on SMBG profile analysis and medication effectiveness
First, address Hypoglycemia
Then address Hyperglycemia
Still make recommendations if not enough data, but mark as suspect

14. Algorithm for Treatment of Type 2 Diabetes
Diet and Exercise
If A1C > 6.5%
Monotherapy or Combination Therapy
Adequate
Not adequate
Follow-up q 3 mo
Other Oral Combinations
Adequate
Not adequate
Follow-up q 3 mo
Oral Agent Plus Insulin at Bedtime
(Glargine or NPH)
Adequate
Not adequate
Follow-up q 3 mo
Split-Mixed Insulin
or Lispro or Aspart qac
+ Glargine or NPH qhs

15. Recommendations
Modify the existing regimen because of contraindications
Increase or decrease the dosage of current medication(s)
Add additional oral agents/basal insulin
5+ medications or 4 with hyperglycemia - recommendation to consult endocrinologist
Contraindications related to age, labs and diagnoses
Increase only if patient is not already on ½ max of FDA recommended dosage.

16. Additional Output Where testing can be improved FDA Warnings
Rosiglitazone use has been severely restricted by the FDA because of concerns that it causes an increased number of cardiovascular events. Continued use requires your patient be enrolled in a risk evaluation and mitigation strategy program established by GlaxoSmithKline. You should consider switching this patient to pioglitazone at an equivalent dose…
SMBG Profile by Time Period
Min, Max, Average, Standard Deviation
% high, % low based on thresholds

17. Sensitivity to being authoritarian
Suggestion vs. Command
Multiple suggestions
Education of clinicians

23. Algorithm Development
Started with Expert Rules System
Initially used CLIPS
Migrated to DROOLS
Final solution - Table-driven logic with algorithm coded in Java
Number of combinations
Ability for versioning and customization by individual non-rule experts

25. Next Steps
A one-year multi-site IRB-approved, cluster-randomized controlled trial
Expand rule base to include
Insulin dependant Type 2
Type 1
Expanded pattern recognition and treatment plans
Post-prandial fluctuations
Trends during day or night
Hypoglycemia followed by rebound "Somogyi reaction“
Dawn Phenomenon
cluster-randomized, controlled, clinical trial involving 30 PCPs who will each recruit approximately 19 patients from their respective geographic site.
There will be up to 570 patients and 30 primary care providers in 3 geographic regions
Trial – recently extended to a second year – controls become test subjects

26. Publication
Rodbard and Vigersky, Design of a Decision Support System to Help Clinicians Manage Glycemia in Patients with Type 2 Diabetes
Journal of Diabetes Science and Technology,
Volume 5, Issue 2, March 2011

27. Thank you!
Richard ‘RJ’ Kedziora Office: (610) Cell: (610)

28. CDMP Background
Complete customizable web-based clinical application for management of patients with chronic disease.
Based on the Chronic Care Model, it was originally designed for military healthcare to better manage patients with diabetes.
Evolved into a generalized chronic disease and population health management system supporting the Patient Centered Medical Home model. 
For details visit: