Impact of an EHR-based Diabetes Management Form on Quality and Outcomes of Diabetes Care in Primary Care Practices

Select to access the PowerPoint® presentation (1.5 MB).

Slide 1

Impact of an EHR-based Diabetes Management Form on Quality and Outcomes of Diabetes Care in Primary Care Practices

Slide 2

Investigators

Jeph Herrin, PhD^{1, 2}
Phil Aponte, MD³
Briget da Graca, JD, MS³
Greg Stanek, MS³
Terianne Cowling, BA³
Cliff Fullerton, MD, MSc⁴
Priscilla Hollander, MD, PhD³
David J Ballard, MD, MSPH, PhD³
 

¹ Department of Medicine, Yale University, New Haven CT
² Health Research and Educational Trust, Chicago IL
³ Institute for Health Care Research and Improvement, Baylor Health Care System, Dallas, TX
⁴ HealthTexas Provider Network, Baylor Health Care System, Dallas, TX

AHRQ grant: R21 HS20696-02

Slide 3

Background

• Electronic Health Records (EHRs) may:
  • Improve communication between patient and physician.
  • Provide clinical decision support.
  • Provide registry-type functionality for tracking care.
  • Facilitate physician performance measurement.
• Some or all of these may lead to improved care of patients with chronic conditions.

Bodenheimer, T. 2003. "Interventions to Improve Chronic Illness Care: Evaluating Their Effectiveness." Disease Management 6 (2): 63–71.

Slide 4

Background

• Evidence is limited:
  • Evaluations of tailored EHRs.
  • Evaluations of commercial EHRs on a small scale.
• And conflicting:
  • No impact on chronic care.
  • Some impact on chronic care.
• No large studies of commercially available EHRs...

Slide 5

Background

...until recently^†

.

• We looked at 14,501 diabetes patients at 34 practices.
• Our outcome was "Optimal Care" (HbA1c≤8 percent; LDL cholesterol <100 mg/dl; blood pressure <130/80 mmHg; not smoking; and documented aspirin use in patients 40 years of age).
• We found a difference of 9.2% (95% CI: 6.1, 12.3) in the final year between patients exposed to the HER (higher rate of optimal care) and those not exposed to it.
• Also improved processes of care (eye exams, foot exams, labs).

^†Herrin, J., Nicewander D, Fullerton C, Aponte P, Stanek G, Cowling T, Collinsworth A, Fleming NS, Ballard DJ. "The effectiveness of implementing an electronic health record on diabetes care and outcomes." 2012. Health Serv Res 47(4): 1522-1540.

Slide 6

Objective

• Hypothesis:
  • The effect of the EHR on the care and outcomes of diabetes patients was due in part or in entirety to the incorporation of a "Diabetes Management Form" (DMF), a component of the EHR designed to manage the care of diabetes patients.

Slide 7

Setting

• HealthTexas Provider Network (HTPN):
  • Is the ambulatory care network affiliated with the Baylor Health Care System, a not-for-profit integrated healthcare delivery system serving patients throughout North Texas.
  • Comprises >100 practices, with 450 physicians, and has >1 million patient encounters annually.
• The current study incorporates all practices which include physicians specializing in Internal Medicine (IM) or Family Medicine (FM), with EHR implemented prior to Jan 1 2006.

Slide 8

Setting

Image: A map of Texas divided into counties shows the locations of HTPN Service Areas.

Slide 9

Data Collection

• What made this study possible is the contemporaneous collection of data on diabetes patients.
  • In 2007 HTPN established and began populating a retrospective diabetes prevalence cohort database using the AMA Physician Consortium Adult Diabetes Performance Measure set.
  • Each cohort was defined by the claims-based algorithm used by the Centers for Medicare & Medicaid Service (CMS).
  • All patients with ≥2 ambulatory care visits ≥7 days apart with a diabetes-related billing code (CMS National Measurement Specifications Diabetes Quality of Care Measures [2002]: ICD-9-CM Diagnosis Codes 250.xx) during the preceding 12 months were identified from administrative data.

Slide 10

Study Population

• All patients who:
  • Were 40 years or older.
  • Had at least 2 diabetes related visits in 2007.
  • Had no DMF "exposure" in 2007 or prior.
  • Had at least 2 diabetes related visits in 2009.

Know: age, sex, insulin usage, number of visits.

Slide 11

Intervention

Images: Overlapping screenshots show forms for data entry.

Slide 12

Intervention

Key element—last dialog box

Image: A screenshot shows the last dialog box. Titled "Centricity," this box offers therapeutic recommendations based on the data entered on the previous form pages.

Slide 13

Outcomes

• Primary Outcome: Optimal Care Bundle:
  • HbA1c≤8 percent.
  • LDL cholesterol <100 mg/dl.
  • Blood pressure <130/80 mmHg.
  • Not smoking.
  • Documented aspirin use.
• All criteria met = optimal care (yes/no).

Slide 14

Outcomes

• Secondary:
  • Clinical:
    • HbA1c=8 percent.
    • LDL <100 mg/dl.
    • BP <130/80 mmHg.
    • Not smoking.
    • Documented aspirin use.
    • Triglycerides <150.
    • Total cholesterol <100.
  • Process:
    • HbA1c checked.
    • Lipids checked.
    • Microalbumin checked.
    • Eye exam done.
    • Foot exam done.
    • Flu vaccine.
    • Smoking status assessed.
    • Smoking cessation.

Slide 15

Design

• Design Considerations:
  • Not all patients have measurements in both 2007 and 2009.
  • DMF exposure in 2009 might effect outcomes in 2009.

Slide 16

Design

Naïve Design: Use all available data

• 2007: Baseline →
• 2008: No DMF / DMF →
• 2009: Followup.

Slide 17

Design

Naïve Analysis:

logit(Pr[Y_ij]) = β₀ + BX + β_TT + β_DMFDMF + β_XDMF * T + ζ_ij + ν _j

T is time (baseline vs followup)
β_X is the interaction effect
ζ_ij + ν _j are random effects at patient, practice level to account for repeated measures on patients, within practices

Slide 18

Patients

Slide 19

Results

Naïve Results: Unadjusted

Slide 20

Results

Naïve Results: Adjusted

Slide 21

Design

Improved Design: Only Patients with both 2007 & 2009 measurements!

• 2007: Baseline →
• 2008: No DMF / DMF →
• 2009: Followup.

Slide 22

Design

Main Model:

logit(Pr[Y_ij]) = β₀ + BX + β_TT + β_DMFDMF + β_XDMF * T + ζ_ij + ν _j

T is time (baseline vs followup)
β_X is the interaction effect
ζ_ij + ν _j are random effects at patient, practice level to account for repeated measures on patients, within practices

Slide 23

Patients

Slide 24

Results

Slide 25

Results

Slide 26

Limitations

• Observational trial.
• Difficult to disentangle exposure and measurement:
  • Sicker patients may be more likely to be measured.
  • Sicker patients may be more likely to be "exposed."
• DMF "exposure" includes no measure of fidelity:
  • DMF may merely be opened and closed.
  • DMF may be used incorrectly.
• Incremental effect on top of EHR effect may be difficult to detect.

Slide 27

Conclusion

While EHR improved care and outcomes of diabetes patients (prior study), evidence here is that the incremental effect of a Diabetes Management Form is negative or mixed.

Definitive inferences may require randomized trial