Improving Office Care for Chest Pain
Thomas D. Sequist, MD MPH
Associate Professor of Medicine and Health Care PolicyBrigham and Women’s Hospital, Division of General Medicine
Harvard Medical School, Department of Health Care PolicyHarvard Vanguard Medical Associates
Why Chest Pain?
• Chest pain is a common symptom– Increasing burden in primary care
• Frequent missed diagnosis of acute MI
• Excess utilization of resources
Patient Care Model
Primary care visit
Home without further testing
Home with further testing
Emergency Department
Discharged
Chest Pain Unit
Inpatient
ICU
Patient Care Model
Primary care visit
Home without further testing
Home with further testing
Emergency Department
Discharged
Chest Pain Unit
Inpatient
ICU
Primary Care Challenges
• Low risk population– Limit excess resource utilization– Avoid missed diagnoses
• Time-limited care– Cannot usually observe over several hours
• No immediate cardiac stress testing
• No immediate cardiac enzymes
Can the Framingham Score Help?
• Main utility is to raise awareness
• FRS variables are generally available
• FRS compares favorably with exercise stress testing
Defining High Risk Patients
FRS Cutoff Sensitivity Specificity
≥ 5% ≥ 10% ≥ 20%
96
85
54
61
75
86
Sequist et al. Arch Intern Med 2006.
Study Questions
1.Can risk score alerts within an EHR improve risk-appropriate care for patients with chest pain?
2.What are the additional opportunities to improve the efficiency of chest pain care?
Harvard Vanguard Medical Associates
• Multi-specialty group practice
• Integrated electronic health record
• 15 ambulatory health centers
• 175 primary care physicians
• 300,000 adult patients
Randomization Scheme292 Primary Care Clinicians7,083 patients (≥ 30 years old)
Intervention Group149 clinicians3,634 patients
Control Group143 clinicians3,449 patients
High Risk717 patients
Low Risk2917 patients
High Risk610 patients
Low Risk2839 patients
Intervention Design
• Identification of patients with chest pain– Medical assistant training
• Automated calculation of Framingham Risk Score
• Delivery of risk-appropriate recommendations via electronic alerts
Risk Appropriate Recommendations
• High risk patients (FRS ≥ 10%)– Electrocardiogram performance– Aspirin therapy
• Low risk patients (FRS < 10%)– Avoidance of cardiac stress testing
Entry of Chest Pain Complaint
High Risk Patient Alert
Low Risk Patient Alert
SmartLink (.frsdetail)
Baseline Patient Characteristics
Intervention(n = 3,634)
Control(n = 3,449)
p value
Mean age, years
Female, %
Insurance Commercial Medicare Medicaid Uninsured
Framingham Risk Score < 10% ≥ 10%
49.7
63
761483
8020
48.6
65
771193
8218
0.001
0.03
0.01
0.03
Clinical Care and Outcomes
High Risk(n=1327)
Low Risk(n=5756)
p value
Evaluation and treatment Electrocardiogram Aspirin therapy Cardiac stress test
Follow up care Home Hospitalized
Diagnoses Acute myocardial infarction*
501917
917
1.1
437
10
963
0.2
<0.001<0.001<0.001
<0.001<0.001
0.01
* Among 26 cases of AMI, 10 (36%) represented missed diagnoses
Impact of Electronic Alerts
51
20
10
48
18
9
0
10
20
30
40
50
60
70
80
EKGPerformance
AspirinTherapy
Cardiac StressTesting
% R
ecei
ving
Intervention Control
High Risk Patients Low Risk Patients
Clinician Views on Intervention
5
4047
8
0
20
40
60
80
100
Always Often Sometimes Rarely orNever
% R
epor
tin
g
Is the Framingham Risk Score a valid tool for evaluating chest pain?
Clinician Views on Intervention
12
81
7
0
20
40
60
80
100
Too high About right Too low
% R
epor
tin
g
Is a Risk Score Cutoff of 10% to identify high risk patients….
Conclusions
• Acute MI is uncommon among primary care patients with chest pain
• Missed diagnosis of acute MI is common, while many low risk patients undergo cardiac stress testing
• Electronic risk alerts do not change care patterns
Implications
• Failure to change care patterns– Is it lack of belief in the risk assessment tool?– Is it failure to deliver information effectively?– Do we need more comprehensive efforts?
• Electronic health records represent one piece of a multi-component program
Improving Efficiency of Chest Pain Care
• Map flow of patients from primary care
• Evaluate cost implications for varied evaluation and management strategies
• Analyze variation in care patterns
Patient Care Model
Primary care visit
Home without further testing
Home with further testing
Emergency Department
Discharged
Chest Pain Unit
Inpatient
ICU
Estimated Average Costs Per Patient
Primary care visit
Home without further testing
$293
Home with further testing
$442
Emergency Department
Discharged
$1,087
Chest Pain Unit
$3,192
Inpatient
$17,562
ICU
$47,575
55%
40%
5%
37%
47%
13%
3%
Estimated Average Costs Per Patient
Primary care visit
Home without further testing
$293
Home with further testing
$442
Emergency Department
Discharged
$1,087
Chest Pain Unit
$3,192
Inpatient
$17,562
ICU
$47,575
55%
40%
5%
37%
47%
13%
3%
Physician Level Clinical Variation
Cardiac Stress Testing*
Emergency Department Triage*
% of patients referred for care within physician practices
3.8% 26.7%
1.3% 14.9%
* p<0.01 for random effects of physician level variation.
0% 50%
10.8%
4.7%
Legend95%
Lower CI95%
Upper CI
Average
How Can the EHR Improve Efficiency?
• Increasing awareness of pre-test probability– All variation is within low risk patients
• Focus on low value emergency department referrals
• Peer to peer education
Clinical Process Flow
Primary care visit
EKG
Stress ECHO
Stress Nuclear
ETT
Cardiology
Home
Triage Emergency Dept
ICU
Inpatient
Chest Pain Unit
Triage
Triage
Home