automated use of clinical laboratory results
DESCRIPTION
These slides are from the Dartmouth Jones Lecture of May 2008 by Benjamin Littenberg. They describe the development and evaluation of the Vermedx Diabetes Information SystemTRANSCRIPT
Automated Use of Clinical Laboratory Results in Adults
Benjamin Littenberg, MDUniversity of Vermont
andVermont Clinical Decision Support, LLC
Disclosure
I have an equity position in Vermont Clinical Decision Support, LLC, (Vermedx) along with The University of Vermont and other
faculty inventors.
www.Vermedx.com
Agenda
• Two problems in diabetes– Quality of individual care– Public health impact
• A novel strategy for using information
• Clinical and public health impacts
• Economic impact
Problem Characteristics
• Lab results are central to diabetes management– Good consensus on at least some aspects of care
• Keeping track of results is difficult– Doctors love flow sheets – but hate to keep them up!– Patients use many different laboratories
• Interpreting results is difficult– Diabetes is among the most complex problems a primary care
provider faces• Patients get “lost to follow-up”
– No reminder systems• Doctors and nurses care for individual patients
– Nobody has a “population view”
Guidelines• Hemoglobin A1C measures average blood sugar over last 6-8 weeks
– Low: <7.0%; Medium 7-9%; High: >9%– Frequency: 3 months if high or medium; 6 months if low
• LDL-Cholesterol measures “bad cholesterol” and vascular disease risk– Low: <100 mg/dl; Medium: 100-130; High >130– Frequency: 3 months if high; 6 months if medium; 12 months if low
• Creatinine measures current kidney function– High depends on age and sex (and race and size and....)– Frequency: annual
• Urine protein ratio measures risk for future kidney failure– Low: <30; Medium: 30-300; High: >300– Frequency: 12 months if low; never again if high or 2 mediums or if on
certain medications
And that’s just (some) of the lab tests!
Registry Approach
• Integrate data from multiple sources– Multiple laboratories
• Feedback to providers about individual patients with “value-added reports”
• Feedback to patients• Aggregate to population level
It works for infections. Why not diabetes?
Design Criteria
• Low cost per case• Low technology investment at the practice
– With or without electronic medical record
• Little change in practice flow• Little disruption of patient-provider
relationship• Accommodate multiple data sources• High face validity (accuracy)
Vermedx Diabetes Information System (VDIS)
• Automatic reports to providers and patients
• Based on clinical laboratory results
• Daily feeds from clinical labs
• Electronic messages or faxes to providers (doctors, nurses, physician assistants)
• Letters to patients
VDIS Architecture
VDIS Computer
Primary Care
Provider
Patient
ClinicalLab
ClinicalLab
Electronic medical record
or fax
MailSecure Network
Public HealthSurveillance
VDIS Clinical Outreach
• Flow sheet updates after every lab result
• Reminders to providers
• Reminder letters to patients
• Clinical alert letters to patients
• Quarterly population reports to providers
All products based on national guidelines
Flow sheet
• A flow sheet of pertinent labs, with decision support recommendations
– sent to provider whenever a pertinent test is done
– use in the visit to decide what to do
– follow trends– decide when to re-test– remember odd testing
intervals– possible handout for the
patient
Patient Reminder
• Letter from provider with practice address and telephone
– Explains which tests are due and when
– Asks patient to call office to set up testing
– Motivates patients to stay involved
– Reminds them that the practice cares about their long-term health
– Sent 30 days after the test is overdue (grace period)
Provider Reminder
• A reminder to the practice when patient is overdue
– Use this to keep patients from getting lost to follow up
– Sent 30 days after the test is overdue (grace period)
Patient Alert
• Letter from provider alerting patient to abnormal results
– Sent only for high results (A1C>8 or LDL>130)
– Asks patient to call provider’s office to set up further care (if they haven’t already)
Population Report
– Lists provider’s roster of diabetes patients, their most recent result and overdue status
• Makes it easy to find lost or out-of-control patients for quality improvement
– Provides “report card” for the practice, compared to all other providers and the top 10% performance
– Delivered every 3 months by mail
– Confidential – not shared with anyone else!
• VDIS is not recommended as a “Pay for Performance” system
Clinical Impact: The VDIS Trial
• Randomized by practice• Active practices get all 5 products• Control practices get none• 32 months• Outcomes:
– On-time testing– Blood sugar (A1C), Cholesterol– Costs
• Supported by NIH (R01 DK61167)
Participants
• 64 General Internal Medicine or Family Practices
• 128 Primary Care Providers– MD, DO, NP, PA
• 7,412 adults with diabetes confirmed by provider
• 1,006 randomly selected for home survey
1.17
1.39 1.40
1.74
11.
52
2.5
3O
dds
Ra
tio a
nd
95%
CI
A1C Cholesterol Creatinine Urine Protein
Adjusted for baseline testing and clustering n = 7,412On time testing
Other patient outcomes
• No change in– A1C levels– Cholesterol levels– Renal function– Blood pressure– Functional status– Body mass index
• Improvement in self-care (exercise)
Health Care UtilizationOutcome Control Intervention
AdjustedEffect* P
Hospital days/y 1.89 1.18 -1.01 0.047
Emergency room visits/y
0.72 0.55 -0.23 0.020
Primary care visits/y 2.86 2.04 -0.81 0.010
Specialty visits/y 0.23 0.15 -0.08 0.044
Costs $/y $4937 $3202 -$2426 0.033
*Linear regression adjusted for age, sex, marital status, education, health literacy, race, insulin use, comorbidity, hospital, and clustering within practices.
Health Care UtilizationOutcome Control Intervention
AdjustedEffect* P
Hospital days/y 1.89 1.18 -1.01 0.047
Emergency room visits/y
0.72 0.55 -0.23 0.020
Primary care visits/y 2.86 2.04 -0.81 0.010
Specialty visits/y 0.23 0.15 -0.08 0.044
Costs $/y $4937 $3202 -$2426 0.033
*Linear regression adjusted for age, sex, marital status, education, health literacy, race, insulin use, comorbidity, hospital, and clustering within practices.
The VDIS registry with patient outreach saves over $2,400 per patient per year.
Study Conclusions
• A registry-based clinical outreach program in primary care:– is feasible– improves diabetes care– saves money
What about using it for Public Health?
Registries in Public Health
• Inexpensive surveillance data
• Population view
• Analytic uses
• Clinical outreach (sometimes)
Laboratory results fit the bill.
What can a registry do for public health surveillance?
• Current overall status • Outcomes in subgroups
– Age, sex, geography, provider
• Trends over time• Maps• Combination with other data sources
– Census, surveys, hospital discharges, claims
Data for policy, persuasion and programs
80
10
01
20
14
0M
ea
n L
DL
-Ch
ole
ste
rol (
mg
/dl)
0 20 40 60 80 100Percent of Practices
71 practices 7,512 patients 105.8 patients/practice
Cholesterol by Practice
46
81
0A
1C (
%)
A B C D E F G H I J K LHospital Service Area
Blood Sugar Control by Hospital Service Area
6.50
6.75
7.00
7.25
7.50
7.75
Mea
n A
1C
2001 2002 2003 2004 2005 2006 2007
Monthly Averages
Linear Fit
A1C is falling by 0.036% per year (P<0.001)A1C over Time
6.50
6.75
7.00
7.25
7.50
7.75
Mea
n A
1C (
%)
2001 2002 2003 2004 2005 2006 2007
Vermont
New York
VT: -0.068 per year NY: +0.003 per year (P<0.05)Population Trend: A1C over Time
Clinton, NY
Franklin, NYSt. Lawrence, NY
Grafton, NH
CaledoniaChittenden
EssexFranklin
Gra
nd I
sle
Lamoille
Orange
Orleans
Rutland
Washington
Windham
Addison
Ben
ning
ton
Windsor
Red = High (P<0.05)
Blue = Low (P<0.05)
Grey = Non-outlier
White = No data
Glycemic Control by County
Clinton, NY
Franklin, NYSt. Lawrence, NY
Grafton, NH
CaledoniaChittenden
EssexFranklin
Gra
nd I
sle
Lamoille
Orange
Orleans
Rutland
Washington
Windham
Addison
Ben
ning
ton
Windsor
Red = High (P<0.05)
Blue = Low (P<0.05)
Grey = Non-outlier
White = No data
Glycemic Control by County
Current Vermedx RegistriesVermont New York City San Antonio
Started 2002 2006 May 2008
Sponsor Integrated provider group, insurer
Health Department
Health Department
Patients ~3,000 ~600,000 ~210,000?
Scope A1C, lipids, renal A1C only A1C only
Outreach Yes Pilot phase No
Surveillance No Yes Yes
Patient consent Opt-out None None
Benefits of Population-Based Decision Support
• Improved clinical outcomes for patients
• Help with management of chronic illness for physicians
• Cost savings for the health care system
• Improvement in Public Health