Clinical Decision Support to Improve Diabetes Care: A Search for Unintended Consequences

1
Clinical Decision Support to Improve Diabetes
Care A Search for Unintended Consequences

• Randall D. Cebul, M.D.
• Center for Health Care Research Policy
• Case Western Reserve University
• MetroHealth Medical Center
• Cleveland

Supported by grant R01-HS015123, AHRQ
2
Purpose/Overview

• Describe aspects of a trial of real-time clinical
  decision support (CDS) to improve care and
  outcomes of patients with diabetes.
• Describe analyses/preliminary results of
  Secondary Aim 4 to examine general and
  intervention-specific unintended consequences of
  our intervention.

3
The Search for Unintended Consequences

• Has our focused attention on diabetes and
  glycemic control been associated with
• Higher rates of clinically important
  hypoglycemia?
• Inattention to/reduced quality for other
  conditions?
• Has the CDS been associated with Alert
  Fatigue? - do clinicians want to keep or
  eliminate the CDS at the end of the trial?
• Chronic Alert Fatigue Syndrome (CAFS)
  is an insidious disorder, characterized by
  mental exhaustion and exasperation secondary to
  frequent, involuntary exposure to displays of
  alerts and reminders in computerized clinical
  information systems. Greengold NL. P T. 2005
  30 506-511.

4
Diabetes Improvement Group Intervention Trial
(DIG-IT) Overview

• Commercial EMR-facilitated
• study design (cluster randomized trial, or CRT)
• real-time clinical decision support
• performance measurement
• Nurse Case Management
• Two health care systems 24 sites, 200 PCPs
• Cleveland Clinic, MetroHealth

Funded, in part, by grant R01-HS015123 Agency
for Healthcare Research and Quality
5
8,000 Diabetics, by Race 100 PCPs 10 Group
Practices MetroHealth System (MHS)
6
Design Random Assignment of Practices to CDS -
Disease Management (DM2) - MHS
7
Table 1. Pre-trial Practice Characteristics
8
Clinical Decision Support DM2

• Alerts and Linked Order Sets
• Patient Lists/Registry
• Practice panel performance feedback
• Nurse Case Management

Practice Panel Tools
9
Illustrative encounter-based Alerts
Links to Automated Order Sets

• What do we know about this patient?
• She has diabetes and is visiting her PCP
• Her kidneys are leaking protein, she has an
  elevated LDL, she is not on an ACE inhibitor or
  ARB, nor a statin, and has no documented
  allergies to them, and she has no other
  contraindications.
• There are several alternative drugs/doses

10
SmartSet Linked to ACE/ARB Alert
Patient name
Patient name
11
Practice Performance Feedback
12
Individual Practice Registry
Physician name
Patient name, hosp number and phone
13
Real-time Access to Nurse Case Management
Encounter-based Alerts
Staff Msg Solicitations

• Alert HbA1c is over 9 and increasing

Priority Patient Heres how I can help
SmartSet Refer
PCP Refers
Nurse Case Management
14
Unintended consequences

• Has our focused attention on diabetes and
  glycemic control been associated with
• Higher rates of clinically important
  hypoglycemia?
• Outpatient visits with documented hypoglycemia?
• ED Visits or hospitalizations for hypoglycemia?
• Inattention to other conditions?
• Preventive services mammograms and pap smears?
• Diagnosis or management asthma and CHF?

15
Were there more clinically important episodes of
hypoglycemia in the CDS group?
(P0.01)
(P0.96)
(Documented glucoselt60mg/dl)
(DM-related hypoglycemia)
16
Were there fewer indicated mammograms, Pap
smears, or evaluations for CHF in the CDS group?
(All pts with dx CHF)
(Women 18-75)
(Women 50-75)
(P0.21)
(P0.55)
(P0.78)
17
Were there more ED visits or hospitalizations for
asthma (unrelated) or CHF?
(P0.70)
(P0.03)
18
Planned Safety Comparisons Related to Alerts and
Order sets

• Differences in proportions of patients
• Prescribed ACE inhibitors or ARBs in face of
  relative contraindications (K, renal function)
  or documented allergy
• Prescribed statins with relative C/I or allergy
• Obtaining appropriate follow-up testing for Rx
  side effects (kidney, liver function)

19
Alert Fatigue?PCP Satisfaction with CDS (n51)

• Re 1) Alerts 2) Practice Panel tools and 3)
  Nurse Case Management
• How often do you use?
• How useful do you find?
• Should X remain in Epic after the trial?

20
Early Returns (71 response)Should X remain
after the trial?
Yes
21
Summary Work-in-Progress

• CDS intervention appears to be associated with
• Significant although small increases
• ED visits for hypoglycemia - unintended but not
  unexpected.
• ED visits for CHF - ?related to use of TZDs
• No significant differences
• Preventive services (mammograms, pap smears)
• Appropriate testing among patients with CHF
• ED visits for unrelated condition (asthma)
• CDS-related alert fatigue did not occur, at least
  as measured by PCPs desires to retain the Alerts
  as the trial period ends.

22
Unintended Consequences Closing Thoughts

• Dont assume that well-intentioned and
  well-designed CDS is without unintended adverse
  consequences.
• Unintended consequences may relate
• directly from the activity that we are trying to
  improve (e.g., hypoglycemia ? CHF from TZDs?)
• to seemingly unrelated conditions (e.g.,
  screening)
• Controlled trials are the best way to test for
  these side effects, and can be facilitated by
  using EMR-facilitated balancing of sites before
  assignment.

23
(No Transcript)
24
Illustrative Performance MeasuresADA Scores
Measured Every Week
Clinical Outcome Change in Scores Ace/ARB 1 Pnv
x 1 Eye Exam 1 LDLlt100 1 A1Clt7 1 BMIlt30
1 Non-Smker 1 SBPlt130 1 0-8 points
MD-centric Measures
25
Changes in ADA Scores for MHS Experimental Group
Patients (n5288)
of 4 MD-Centric Measures Met
of All ADA Measures Met