The HeartDecision Computer Decision Support Pilot Study

The HeartDecision Computer Decision Support

Pilot Study

Matthew C. Tattersall D.O.Adjhaporn Khunlertkit Ph.D.

Peter Hoonakker Ph.D.Jon G. Keevil M.D.

Disclosures

Tattersall: No Disclosures Khunlertkit: No Disclosures Hoonakker: No Disclosures Keevil: Founder/Owner HealthDecision, LLC – a zero

revenue company building decision support tools.

Background

Current 2010 AHA/ACC guidelines recommend calculation of absolute cardiovascular risk. (Class I LOE: B) • “All adults ≥ 40 y/o should know their absolute

risk of developing coronary heart disease” The level of cardiovascular risk

determines corresponding lipid goals. The level of current lipid goals determines

the need for pharmacotherapy.

AHA Guidelines for Primary Prevention of Cardiovascular Disease and Stroke: Circulation 2002;106;388-391

ATP III JAMA. May 16 2001;285(19):2486-2497

Background

Importance of Cardiovascular Risk Assessment:• Clinicians over and under-estimate risk (as

high as 76% of patients)

• Initial errors in risk assessment lead to inappropriate use of pharmacotherapy

• A recent meta-analysis displayed CHD risk assessment improves patient outcomes with no harm.

Friedmann PD, et al. Differences in generalists’and cardiologists’ perceptions of cardiovascular risk and the outcomes of preventive therapy in cardiovascular disease. Ann Intern Med. 1996;124:414 –21.

Grover SA, et al. Do doctors accurately assess coronary risk in their patients? Preliminary results of the coronary health assessment study. BMJ. 1995;310:975– 8.

Sheridan SL et al., Does the routine use of global coronary heart disease risk scores translates into clinical benefits or harms? A systemic review of the literature. BMC Health Serv Res. 2008;8:60.

Background

Methods used to calculate risk:• Pad and Paper• Hand-Held Calculators• Online Calculators

Overall risk calculation is not being performed.• McBride et. al.: Only 17% of primary care

physicians routinely calculate cardiovascular risk.

Clinician Barriers

Time consuming Where to find a calculator to calculate

risk? Which risk model to use? Multi-staged, dynamic guidelines with

changing lipid goals Which evidence-based pharmacotherapy

should be used?

Computer Decision Support Tools (CDST)

CDST Barriers

While CDST’s improve:• Diagnosis• Prevention • Management of chronic diseases

Many CDST’s Fail:• Poor integration into clinician workflow:

– AHRQ GLIDES study: Clinician workflow integration significant barrier.

– Very little field testing of CDST’s. – Previous studies focus solely on performance.

HeartDecision CDST Pilot Study

Multi-disciplinary collaborative pilot study with two aims:• To address usability, integration into work flow

and field testing.• To assess impact of the CDST since launch

date. (2-1-2010)

HeartDecision Pilot Systems Engineering Initiative for Patient

Safety (SEIPS) part of the UW College of Engineering.• Previously developed a work system design

model integrating– Human factors engineering– Healthcare quality models

HeartDecision Pilot

Hypothesis #1: Application of the SEIPS model will help identify and characterize the enablers and barriers to the integration of the HeartDecision CDST into primary care clinician workflow.

HeartDecision Pilot: Methods

Human Factors Engineering Field Testing:• 8 Physicians from 5 WREN DFM clinics.• Clinic encounter with standardized patient

from UW School of Medicine with mock EMR.• Data collected/analyzed via SEIPS qualitative

methods using time study, observation and post-encounter interviews.

HeartDecision Pilot: Results

Time (in minutes) spent in HD

0.08

4.03

5.45

7.068.34

9.56 9.55

12.52

0

5

10

15

20

25

30

Start HD running Risk Page Goals page Ideal page Handoutspage

Summarypage

End

Observation 1

Observation 2

Observation 3

Observation 4

Observation 5

Observation 6

Observation 7

Observation 8

Average

On an average, the physicians spent 13 minutes using the HD tool

Time Study of the HeartDecision CDST

Facilitators

“The tool is intuitive” “The tool presents patient assessment in logical

sequence” “Data is automatically populated” “The risk level (low, moderate, and high) is clear to

the patient” “The graphical display helps with communication

with patient” “Hand out provides good information for patient”

Barriers

Clinician Work Flow• Time pressure: “Patients with multiple

conditions” Work Environment

• “Cannot print educational PDF files and graphs”

• “Cannot open PDF files on Winterms” Program Interface

• “20 second delay upon opening the program” Program

• “No pharmacotherapy recommendations”

• “Nice to have patient peer comparisons”

Web-Based Survey

To further delineate barriers and enablers a web-based survey was sent to clinicians within the Department of Family Medicine and the Department of Medicine

73 respondents (50%) from Department of Family Medicine, 71 respondents (49%) from Department of Medicine.

Web-Based Survey

Barrier Survey Results Means: (N=66)(1=strongly disagree, 5=strongly agree)

Time: “I have too little time to use the tool.”

2.3

Work Environment:“I cannot use some functions of the HeartDecision tool because of lack of support from the computer-, Winterms-, or Health Link-

system.”

2.9

This tool fits well in my workflow. 3.8

Program Interface:“When I open up the tool, the tool is occasionally delayed.”

3.1

Program:“It would be helpful to add medication recommendations to treat

cholesterol in patients with certain risk in the tool.”

4.0

“It would be helpful if the (Ideal) Graph could be printed.” 4.1

Field Testing Conclusions

Work flow barriers exist with the HD CDST.• Time• Work Environment improvements: (Printing,

speed).• Post encounter patient handouts/chart

documentation.• Need for specific treatment recommendations.

Assessing Early Impact

Hypothesis #2: Since implementation of the HeartDecision CDST into the UW electronic medical record the frequency of cardiovascular risk documentation has increased.

Measuring Impact Retrospective Pre-Post Chart Review 6 WREN Physicians at 5 different clinics Patients identified by CDST use Compared two time periods:

• 1-1-2009-1-31-2010 versus 2-1-2010 -3-11-2011 Assessed rate of cardiovascular risk documentation

pre-HD and post-HD• Compared rates using an exact McNemar’s Chi Squared• Compared Physician rate changes using Fisher’s Exact

test.

Inclusions/Exclusions

Inclusions: • No high risk conditions (CVD, PVD, DM2)• Must have at least one visit in each time

period with provider 62 patients met inclusion criteria

• 27 male (44%)• 35 Female (56%)

Descriptive Statistics

Characteristic Mean (SD)

Age (years) 54.2 (10.7)

Total Visits (N) 4.1 (2.2)

LDL-C (mg/dL) 139.0 (40.5)

Systolic Blood Pressure (mmHg) 124.3 (15.8)

Total Cholesterol (mg/dL) 220.2 (40.9)

Framingham Risk Score (%) 5.3 (5.8)

HDL (mg/dL) 47 (13.3)

CV Risk Documentation

3.2%(95% CI 0.4-11.2%)

50% (95% CI 37-63%)

Post HD rate not dependent on Physician p=0.42 (Fishers Exact)

P<0.0001

Impact Assessment Conclusions

The rates of CV risk documentation improved in this small selective physician cohort.

Hypothesis generating

Conclusions

Workflow barriers exist with use of HD CDST• Time constraints• Need for more treatment recommendations• Printer friendly graphs• More patient education tools

Since incorporation of HD into Epic• In a small, selective group of physicians CV risk

documentation rates have improved since HD CDST incoporation.

Overall, hypothesis generating • Will the use of a CDST that is well integrated into clinician

workflow improve CV measures of performance in a large cohort of physicians.

Thank You