Effect of Computerized Decision Support on Percent of Asthma Patients with Asthma Action Plans
1. EFFECT OF COMPUTERIZED DECISION SUPPORT
ON PERCENT OF ASTHMA PATIENTS
WITH ASTHMA ACTION PLANS
Yiscah Bracha, MS, PhD
Assistant Vice President Data/Analytics
James M. Anderson Center for Health Systems Excellence
Cincinnati Childrens Hospital Medical Center
yiscah.bracha@cchmc.org
Research performed at Hennepin County Medical Center, Minneapolis MN, in
fulfillment of PhD requirements at the University of Minnesota, Division of
Health Services Research and Policy
3. • Generate Qs
• Perform studies
Researchers &
Manufacturers • Generate evidence
• Review evidence
• Summarize results
Expert Panels • Issue guidelines
Known gap:
Evidence-based
recommendations & Informticsts • GAP
physician behavior
• Hear about guidelines (maybe)
Practicing • Practice medicine
Physicians
4. • Generate Qs
• Perform studies
Researchers &
Manufacturers • Generate evidence
• Review evidence
• Summarize results
Expert Panels • Issue guidelines
• Represent guidelines as
Hypothesis: computerized decision support (CDS)
CDS can “close the Informaticists • Place CDS at point of care
gap”
• Invoke CDS tool while delivering care
• See real-time guideline recs
Practicing
Physicians • Practice medicine
5. Types of CDS
• Critiquing mode • Guided mode
• Alerts, order sets • Decision trees at multiple
• Provides: branching points
• Reminders • Provides support for:
• Warnings • Complex, cognitive tasks
• Constrained choices (reduce • Managing chronic disease
variability)
• Delivery: Typical EHR
• Delivery: Thru EHR
system cannot deliver
• Common
• Uncommon
6. Approaches to CDS for chronic disease
• Document-centric • User-centric
• Goal: Convert guidelines • Goal: Help docs perform
into code work more effectively
• “Customers”: Those who • “Customers”: Physician-
want docs to behave as users
guidelines recommend
7. HIT Asthma Project
A user-centric approach to asthma CDS
• Initial development supported by AHRQ
• Clinician requirements.
Produce a patient-specific Asthma Action Plan (AAP)
Support clinical decisions necessary to populate the AAP
Be reachable from the local EHR
Make it easy to retrieve previously created AAPs
Auto-place med order in the EHR
• Implementation sites as of December 2011
• Hennepin County Medical Center. July 2009
• Fairview Health Systems. March 2010
• Altru Health System. May 2011
• User stats: January – November 2011
• 6000+ AAPs; ~ 700 AAPs/month
• ~ 500 active clinical users
8. HIT Asthma Project
A user-centric approach to asthma CDS
• Initial development supported by AHRQ
• Clinician requirements.
Produce a patient-specific Asthma Action Plan (AAP)
Support clinical decisions necessary to populate the AAP
Be reachable from the local EHR
Make it easy to retrieve previously created AAPs
Auto-place med order in the EHR
• Implementation sites as of December 2011
• Hennepin County Medical Center. July 2009
• Fairview Health Systems. March 2010
• Altru Health System. May 2011
• User stats: January – November 2011
• 6000+ AAPs; ~ 700 AAPs/month
• ~ 500 active clinical users
9. EFFECT ON
PROPORTION OF PATIENTS WITH
ASTHMA ACTION PLANS
Results from implementation at HCMC.
Weekly rates for current AAPs from March 2008 – February 2010
10. Questions and analytic strategy
• Questions
• What effect did CDS tool have on % of patients
w/current AAPs?
• If there was no effect, was this because:
• Docs resisted the technology OR
• Docs were not creating AAPs?
• Strategy
• Targeted sample for chart review (to find paper AAPs)
• Calculate weekly rates for current AAPs using:
• Paper template (manual)
• CDS tool (automatic once decision support complete)
• Either paper or CDS
11. Patients Contributing Data:
Four Clinics in Two Age Groups
Age on 01MAR10
Clinic School age Adult
Where Patients Received CAre (5-14 years) (ge 21 years)
1st Family Medicine Clinic (FM1) 93 185
2nd Family Medicine Clinic (FM2) 77 119
1st Pediatrics Clinic (PED1) 160 .
2nd Pediatrics Clinic (PED2) 434 .
Total sample is 899 patients.
Sample selected for chart review (to detect presence of paper AAP)
Sums across clinics exceed 899; some patients received care at multiple clinics.
Responsibility for current asthma action plan is to all clinics where patients had at least
one visit (any reason) from March 1, 2007 – February 9, 2010.
12. Weekly Prevalence Rates
for Current Asthma Action Plans
Pediatric and Adult Asthma Patients at Four HCMC Primary Care Clinics
13.
14. OBVIOUS EFFECTS
Adults at FM1
Paper AAP forms available
Clinic culture emphasized need for kids
Several docs start using tool to create AAPs for adults
Kids & Adults at FM2
No paper AAP forms available
Docs began using CDS tool as newly available support
15. NON-OBVIOUS EFFECTS
All pediatric patients.
Plans generated by paper decline as plans generated electronically increase
Continued generation using CDS tool
Overall effect (black line) AAPs unclear … “special cause”?
19. “Special Cause” detection challenges
• Data are serially correlated
• Special Cause trend rules
• Assume independence among successive observations
• Apparent trends could be “random walks”
20. Box-Jenkins Interrupted Time Series
• Auto-Regressive Integrated Moving Average (ARIMA)
• Explicit model for serially correlated observations
• Inclusion of “interruption” (e.g. intervention) term
• Permits statistical analysis of effect of intervention compared to
auto-correlated background noise
Results for Children (age 5-11) With Visits at:
Effect of the FM1 Peds1 Peds2
intervention Coeff P-val Coeff P-val Coeff P-val
Initial effect -0.601 0.49 -2.25 0.12 1.89 0.01
Attenuation -0.91 <0.0001 -0.74 <0.0001
0.934 <0.0001
(mult lags) 0.67 0.08 -0.64 <0.001
Initially positive,
Interpretation No effect No effect
oscillating thereafter
21. NON-OBVIOUS EFFECTS
All pediatric patients.
AllPlans generated by paper decline as plans generated electronically increase
pediatric patients.
Plans generated by paper declinetoolplans generated electronically increase
Continued generation using CDS as
Continued generation using CDS tool
Overall effect (black line) AAPs. Analysis using Interrupted TS shows:
Overall effect (black line)PEDs1 unclear … “special cause”?
No effect at FM1 or AAPs
Effect in PEDs2, oscillating over time
22. Interpretation: Effect of tool on AAPs
• Parsimonious theory:
• Susceptible patient will receive an AAP if
• Physician is “inclined” to create one AND
• Support exists. Computerized tool preferred to paper template
• Quality improvement implications
• To increase rates of tool-generated plans, target docs who don’t
create plans at all.
• Provide user-centric computerized decision support
23. CO-INVESTIGATORS
Gail M. Brottman, MD (Hennepin County Medical Center)
Angeline Carlson, PhD (Data Intelligence)
Kevin Larsen, MD (Hennepin County Medical Center)
24. ACKNOWLEDGEMENTS
Brendon Cullinan, MD (HealthEast)
Jennifer Rodlund, RN (Hennepin Faculty Associates)
Cherylee Sherry and Thouk Touch (Minneapolis Medical Research Foundation)
Susan Ross, RN (Minnesota Department of Health)
The IT and EHR team at Hennepin County Medical Center
Donald Uden, PharmD (University of Minnesota)
Robert Grundmeier, MD (The Children’s Hospital of Philadelphia)
Tim Michalski (Point of Care Decision Support)
Robert Mayes, RN (AHRQ and American Medical Informatics Association)