A Simulated Diabetes Learning Intervention Improves Provider Knowledge and Co...
A Simulated Diabetes Learning Intervention Improves Provider Knowledge and Confidence in Managing Diabetes HILLEN
1. A Simulated Diabetes Learning
Intervention Improves Provider
Knowledge and Confidence in
Managing Diabetes
JoAnn Sperl-Hillen, MD
Co-director of Center for Chronic Care Innovation
HealthPartners Research Foundation, Minneapolis, MN
Wednesday May 2, 2012 8-9:30am
18th Annual HMO Research Network Conference
Seattle, WA
Accelerating excellence in health performance
through education, advocacy, and collaboration
2. Team Members
JoAnn Sperl-Hillen Steve Asche
Patrick O’Connor George Biltz*
Heidi Ekstrom Deb Curran
William Rush Paul Johnson*
Omar Fernandes Andrew Rudge
Jerry Amundson Todd Gilmer**
Deepika Appana
HealthPartners Research Foundation and HealthPartners
Institute for Medical Education, Minneapolis, MN;
* Carlson School of Management, University of
Minnesota, Minneapolis MN;
** Department of Family and Preventive Medicine, University
of California, San Diego, La Jolla, CA
3. Presenter Disclosures
NIH research support
Listed inventor on a U.S. patent application filed
related to simulation technology
HPRF has recently entered into a royalty-bearing
license agreement with a third party to commercialize
the simulated learning technology for the purpose of
broader dissemination.
Non-paid director on the board of directors for that
licensee (SimCare Health)
4. Why is provider training
needed?
Provider performance varies, even
within the same clinic populations
Clinical inertia is common, particularly
for insulin treatment
Provider knowledge varies
The cognitive processes and tasks
related to diabetes are complex
5. Barriers to Provider Training
Time constraints
Lack of continuity experiences
Relatively limited ambulatory
experience in residency training
Complicated diseases with need for
personalization of care
Experts & opinion leaders are often not
available or affordable, and teaching is
difficult to standardize
6. What is simulation?
―Simulation is a technique—not a technology—to replace
or amplify real experiences with guided experiences that
evoke or replicate substantial aspects of the real world in
a fully interactive manner.‖ Gaba (2004)
History of Simulation
Aviation
NASA
Military
Medical
1960s First Mannequin: Resusci-Annie
1960s-70s Computer-assisted learning program in
medicine
1990 High fidelity mannequins become available
7. What are the advantages to
simulation?
Efficient & cost effective
Sustainable & standardized
In line with adult learning principles
Personalized (case-based). Case-based
simulations provide a context for learning
People are more likely to remember learning
and replicate in real-world situations.
Capture the importance of continuity of care
Proven satisfaction & effectiveness
8. Elements needed to create a
simulated learning program
1. Identify the learning needs and
create a library of case scenarios
2. Create an interactive web-interface
3. Model and program the physiology
4. Program the feedback – to critique
action the provider takes between
encounters
9. Demo of SimCare available at www.simcarediabetes.org
Patient “snapshot” screen shot
12. Early SimCare Study
Funding through R01HS10639, Physician Intervention to Improve Diabetes Care
57 consented PCP’s and their 2,020 patients.
Randomized to one of 3 groups:
(A) no intervention
(B) learning intervention (SimCare) consisting
of 3 simulated learning cases (1 hr)
(C) SimCare + physician opinion leader
Results:
SimCare reduced risky prescribing of
metformin in patients with renal impairment
(p=0.03).
Group B (SimCare alone) achieved slightly
better glycemic control than A or C (p=.04)
13. SimCare Version 2
Funding through R01DK068314, Reducing Clinical Inertia in Diabetes
Eleven clinics with 41 consenting PCP’s
Randomized to receive or not receive an
improved version of SimCare (12 cases
assigned based on profiled ―needs‖, 3 hrs)
Results: Patients of intervention providers
with baseline A1c > 7% had significantly
greater A1c reduction (-.19%) relative to
patients of non-intervention providers.
14. SimCare Version 3
Funding through R18DK079861, Simulated Diabetes Training for Resident Physicians
19 eligible residency programs
linked to 723 residents
382 residents did
not consent
341 residents consented
Intervention – Early learning 10) Control – Late learning (9)
Residents (177) Residents (164)
Completion rates
Completion rates
Learning cases (142)
Assessment cases (135)
Assessment cases (97)
Knowledge survey (128)
Knowledge survey (92)
Evaluation (94)
15. Implementing the learning program
Residents at 19 programs were given a brochure that
we provided and asked to sign up online
Resident participation was voluntary.
Time commitment – 18 cases, 1 hour/month for 8
months if randomized to the early intervention group
Incentives - $50 Target gift card on completion of the
assigned tasks
Promotions – 4 iPad raffle promotions and 1 Target
gift card promotion to achieve acceptable learning
and assessment case completion rates
16. Baseline characteristics of residents
Intervention Control
P-value
(n=92) (n=128)
% female 48% 57% 0.31
% white 48% 58% 0.41
Age (median) 29 29 0.69
Specialty
Family Medicine 34% 49%
Internal Medicine 54% 42% 0.15
Med-Peds 8% 7%
Other 4% 2%
Post graduate year
35% 34%
1
36% 34%
2 0.70
28% 28%
3
1% 4%
4
17. Example Knowledge Question
2. A 77 year old black man is seeing you for follow up. He has a 13 year history of type 2
diabetes, coronary heart disease (CABG at age 58), chronic stable angina, and
dyslipidemia. He has been eating out a lot and gaining weight. His current medications are
metformin 1000 mg bid, atenolol 50 mg qd, and simvastatin 40 mg qd. His BMI is 37, BP is
165/86, A1c 9.3%, Cr 2. 2 mg/dl, eGFR 28, LDL 94 mg/dl, HDL 36 mg/dl, and TG 278
mg/dl.
Which of the following would be your MOST likely recommended action?
A. Start basal insulin and treat to an A1c goal of < 7%. No change in other glycemia
medications.
B. Discontinue metformin and start basal insulin. Follow up with patient for insulin
adjustments with an A1c goal of < 7%.
C. Start basal insulin and follow up with the patient for insulin adjustments with an A1c goal
of < 8%. No change in other glycemia medications.
D. Discontinue metformin and start basal insulin. Follow up with patient for insulin
adjustments with an A1c goal of < 8%.
E. No change now because I would address other patient problems
Correct answer D (59% intervention, 26% control)
18. P-
Q# Knowledge topics covered Early Late
value
1 Screen for DM (using an A1c) 75.0 75.8 .894
2 Basal insulin start, individualized A1c goal < 8% 58.7 25.8 <.0001
Check ketones in newly diagnosed symptomatic patients &
3 31.5 28.1 .586
start insulin
Reduce basal insulin due to nocturnal hypoglycemia
4 64.1 70.3 .333
(Somogyii)
5 Relax A1c target due to hypoglycemia unawareness 57.6 32.8 .0002
6 Start insulin in a newly diagnosed symptomatic patient 33.7 11.7 <.0001
Use of a loop diuretic rather than thiazide in patient with renal
7 insufficiency. Fenofibrate not beneficial in addition to statin. 44.6 19.5 <.0001
DC metformin due to renal contraindication.
Initiate BP tx (without confirmatory testing) if BP > 180/100.
8 59.8 44.5 0.026
Statins may be helpful for most patients with DM.
9 Start a statin, screen for depression, basal insulin start 66.3 57.0 .164
Geriatric polypharmacy concerns, depression screening,
10 46.7 41.4 .431
hypoglycemia management, statin use in the elderly
19. Results of Knowledge Testing
Number of items
Intervention Control
correct out of 10
0-4 29% 66%
5-7 60% 32%
8-10 11% 2%
Mean score 5.31 4.10
p < .001
(95% CI) (4.87-5.75) (3.69-4.50)
N=220 completers of knowledge survey
20. Results of self-rated confidence and
knowledge about diabetes management
Topic Intervention Control P-value
How knowledgeable are you about how to
use all available drug classes to manage 61 25 <.001
patients with diabetes?
How knowledgeable are you about how to
83 45 <.001
start and adjust insulin?
How knowledgeable are you about
interpreting patient self-monitored glucose 85 59 .009
values (SMBGs)?
How knowledgeable are you about setting
individualized treatment goals for people 83 44 <.001
with diabetes?
How confident are you in managing patients
79 44 <.001
with diabetes?
21. Evidence for learning transfer
to actual patient care
77% applied learning to actual patients
63% shortened visit intervals
78% more likely to add or increase drugs if patient is
not at goal
92% more confident about insulin use in actual
patients
….and results of two trials had demonstrated improved outcomes
of actual patients of practicing providers who used earlier
versions of SimCare
22. Study limitations
Voluntary participation, not all completed the learning
program and completed evaluations
No outcome data on non-completers
Survey completion rates were lower in the intervention
(52%) than the control groups (78%)
No actual patient data to evaluate
Assessment case outcomes not yet available
23. Thank you! For additional questions, please
contact…
JoAnn Sperl-Hillen:
joann.m.sperlhillen@healthpartners.com
Patrick O’Connor:
patrick.j.oconnor@healthpartners.com
Heidi Ekstrom:
heidi.l.ekstrom@healthpartners.com
24. Talk References
Simulated Physician Learning Intervention to
Improve Safety and Quality of Diabetes Care: A
Randomized Trial
O’Connor PJ, Sperl-Hillen JM, et al. Simulated physician
learning intervention to improve safety and quality of diabetes
care: A Randomized Trial. Diabetes Care. 2009;32(4): 585-590.
Simulated Physician Learning Program Improves
Glucose Control in Adults with Diabetes
Sperl-Hillen JM, O’Connor PJ, Rush WA, Johnson PE, Gilmer
TP, Biltz G, Asche SE, Ekstrom HL. Simulated Physician
Learning Program Improves Glucose Control in Adults with
Diabetes. Diabetes Care. 2010;33(8): 1727-1733.
Editor's Notes
Knowledge - Knowing the right thing to doComplexity of cognitive processes and tasks related to diabetesPrioritization - Addressing competing clinical prioritiesAbility to anticipate treatment effects & determine the best follow-up intervalSafety and monitoring routines - Avoiding potential safety hazardsLack of guideline consensus on goals makes is also a reason
High clinical inertia ratesProvider performance variability
In general terms, simulation is a technique or device that attempts to create characteristics of the real world. In health care, simulation may refer to a device representing a simulated patient or part of a patient. Such a device can respond to and interact with the actions of the learner. Simulation also refers to activities that mimic the reality of a clinical environment and that are designed for use in demonstrating procedures and promoting decisionmaking and critical thinking. In health care education, simulation can take many forms, from relatively simple to highly complex Medical simulation has its foundations in aviation, the military and industry.A flight simulator was developed in the 1920s to create an easier, safer and less expensive way to learn how to fly, but it wasn’t until after several catastrophic events in the ‘30s that the military purchased several of these “Link trainers”. Military needs during WWII increased usage & initiated the development of additional simulators. In the 1950s the FAA required simulation recertification to maintain commercial pilots’ licenses. In the 1970s, NASA used simulations to bring the Apollo 13 crew home. Later in the 1970s, human factors were recognized as the source of many events and the concepts of cockpit or crew resource management were formed. Use of simulation in these industries have continued to grow. Modern medical simulation started much later, in the 60’s with the first resuscitation mannequin. By the end of the decade, Harvey, a cardiology mannequin was launched. A rudimentary computer program was also developed for medical education. Since the 1990s simulation in healthcare has exploded and new simulators are in development.
Simulation allows health care providers to develop their skills without endangering the patient or affecting their own self confidence. Because a situation or encounter is simulated, it can be re-run, stopped or altered for improved educational value. This creates a learning environment where multiple options can be practiced and cognitive or psychomotor skills mastered.Simulation-based education has been proven to be effective in industry, aviation, the military and healthcare. There are a plethora of publications describing its application in healthcare in the areas of emergency management, procedural skills, obstetrics, surgery, teamwork & communication. Little has been published on its use in chronic disease management.So why do it? In some applications, like this interactive computer-based simulation program, it is efficient & cost effective…Case-based simulations provide a context for learningPeople are more likely to remember learning and replicate in real-world situations.Provides chart encounters of simulated patients in a primary care environment.Simulation depicts some aspect of reality where the learner can identify and study the effects of change.
The design of the intervention concentrated on four elements described in more detail below: 1. A comprehensive library of case scenarios of patients with type 1 and type 2 diabetes.A team of experts that included physicians proficient in current diabetes guidelines and treatment, a professor of medical physiology, and a professor of decision-making science collaborated regularly to design a clinical content map of diabetes management that would comprehensively address care skills deemed important for resident providers to learn. Table 1 is a list of the learning principles that emerged on the map within the domains of glycemia, lipids, blood pressure, and safety and monitoring. Next, 18 cases were developed with initialized scenarios and patient states that would facilitate provider learning pertaining to all learning topics on the clinical content map.2. An interactive web-based interface display for providers to manage the simulated cases. A web-based interface was designed to mimic an electronic health record. It allowed for complex medical scenarios to come alive through narrative and information displayed in the patient’s “chart,” and for users to perform multiple clinical functions such as review the chart history, prescribe medications, start and adjust insulin with each meal and at bedtime, order labs and diagnostic tests such as electrocardiograms, chest x-rays and sleep studies, make referrals, give patient advice, view self-monitored blood sugar (SMBG) results and change SMBG frequency, and schedule phone or visit follow-up at any desired frequency. See figure 1 for a screen shot of the interface.3. A physiologic model (or engine) to calculate effects of provider actions.Formulas derived from published literature and clinical experts were modeled to compute physiologic responses to provider actions. For example, the simulated patient’s blood sugar state was represented by 8 self-monitored blood sugar (SMBG) values or “patient states” throughout the day. Short and long-acting Insulin and oral drug effects were based on pharmacokinetic curves that distribute the SMBG effects of the drugs over time. Other clinical states that were modeled included blood pressure, pulse, weight, lipids, renal function, potassium, microalbuminuria, depression, drug and behavioral adherence. The collection of patient clinical state predictors were stacked in a predefined way and could be added to in a plug and play fashion to extend to new domains and patient states. The simulation engine was developed using a Java application running on Linux servers, and the data was stored in an Oracle Database. 3. A physiologic model (or engine) to calculate effects of provider actions.Formulas derived from published literature and clinical experts were modeled to compute physiologic responses to provider actions. For example, the simulated patient’s blood sugar state was represented by 8 self-monitored blood sugar (SMBG) values or “patient states” throughout the day. Short and long-acting Insulin and oral drug effects were based on pharmacokinetic curves that distribute the SMBG effects of the drugs over time. Other clinical states that were modeled included blood pressure, pulse, weight, lipids, renal function, potassium, microalbuminuria, depression, drug and behavioral adherence. The collection of patient clinical state predictors were stacked in a predefined way and could be added to in a plug and play fashion to extend to new domains and patient states. The simulation engine was developed using a Java application running on Linux servers, and the data was stored in an Oracle Database.
Example of a patient with type 2 diabetes of 5 years duration
Knowledge test was completed by 52% of early and 78% late intervention participants
We ran an analysis looking at change in these questions from baseline to follow up (to account for the higher ratings on self-efficacy at baseline in the intervention group compared to the control group) and had similar findings. Everything remained significantly improved in the intervention group.
simCare: 57 consented docs and 2000 of their patients. 2001. Better glycemic control and reductions in risky prescribing of metformin in patients with renal impairment-DM inertia: 41 consenting docsand 3000 patients, Conducted 2006-07, better mean A1c and higher % of patients at goal. Costs trended lower -$71 per patient relative to non-intervention patients.