Partners IS and the Benefits of Clinical Decision Support

1
Partners IS and the Benefits of Clinical Decision
Support

• David W. Bates, MD, MSc
• Medical Director of Clinical and Quality
  Analysis, Partners Healthcare
• Chief, Division of Gen Medicine, Brigham and
  Womens Hospital
• Associate Professor of Medicine,
• Harvard Medical School

2
Goals

• Describe Partners and its information systems
• High performance medicine initiatives
• Discuss how clinical decision support can be used
  to improve
• Quality, safety, efficiency
• Inpatient
• Outpatient
• Some new directions
• Future of clinical systems at Partners IS

3
To Err is Human

• Errors are common
• Errors are costly
• Systems cause errors
• Errors can be prevented and safety can be improved

4
The Chasm Report

• Indeed, between the health care that we now
  have, and the health care we could have, lies not
  just a gap, but a chasm. Crossing the Quality
  Chasm A New Health System for the 21st Century

5
Crossing the Quality Chasm

• Care should be safe
• Care should be effective
• Based on sound knowledge
• Care should be patient-centered
• Respectful, responsive to individual preferences,
  needs and values
• Care should be timely
• Unnecessary waits should be reduced

6
Crossing the Quality Chasm

• Care should be efficient
• Care should be equitable
• Should not vary in quality because of patient
  characteristics, such as ethnicity, or geographic
  location

7
Safety and Systems
Safety is a systems property.
Chasm Report

• Every system is perfectly designed to achieve
  exactly the results it gets.

Donald Berwick
8
(No Transcript)
9
Handwriting example
10
Rates of Adverse Events Around the World

• 3.7 of hospitalizations in New York
• 58 preventable
• 2.8 Colorado-Utah
• 16.6 in Quality in Australian Health Care study
• 10.8 in UK
• Similar results now from Canada, New Zealand
• Rate in most countries appears to be about 10

11
Partners Healthcare

• Large integrated delivery system in eastern
  Massachusetts
• Includes Brigham and Womens Hospital,
  Massachusetts General, 4 smaller hospitals
• Many community-based providers
• Total about 7000 physicians
• 3,000,000 patients in enterprise master patient
  index (EMPI)

12
Scale of the Partners Clinical Information Systems

• 56,000 user accounts
• 3,000,000 patients in the Partners MPI
• 350,000,000 results in the Clinical Data
  Repository and growing at a rate of 100,000
  transactions per day
• 80,000,000 images archived
• 26,000 inpatient orders are written on an average
  day, across Partners, using CPOE

13
Partners Why Are the High Performance
Initiatives So Important?
These initiatives position PHS providers to
succeed under pay-for-performance contracts.
EMR dissemination (TAHP)
Information Infrastructure
Progress on Leapfrog criteria (HPHC, TAHP)
Safe Med Administration (BCBS)
Patient safety
Diabetes improvement (HPHC, TAHP, BCBS)
Quality metrics (Diabetes IP Cardiology)
Signature Initiative Teams
Contract Target Areas
Improvement of hospital cardiology care (TAHP)
Disease Mgmt (CHF and High Risk)
Hospital use (HPHC, TAHP, BCBS)
Rate of rise in Rx trend (HPHC, TAHP, BCBS)
Trend Mgmt. (Rx Radiology)
Rate of rise in radiology trend (TAHP, BCBS)
14
High-Performance Medicine Initiative 1
Investing in Quality and Utilization
Infrastructure

• Why Information systems have been inadequately
  applied to health care
• What Build on our leadership position by
  accelerating deployment of
• The electronic medical record (EMR)
• Computerized provider order entry systems (CPOE)
• Clinical decision aids
• Goal Deploy a more fully developed EMR and CPOE
  across the system reaching a physician
  penetration of 55 for EMR and 75 for CPOE after
  one year - and 80 and 100 in 3 years

15
Partners EHR Effort

• Now rolling out ambulatory EHRs to all outpatient
  physicians
• Ambitious goals for PCPs
• Specialists also using but accept lower
  levelse.g. uploading notes, e-prescribing

16
EHR Use At Partners 2006 Statistics

• PCP EHR use rate 75
• 96 in AMCs
• 65 in community (was 9 in 2003!)
• Specialist EHR use rate 59
• 86 at AMCs
• 17 in community
• Vendors
• 2350 LMR
• 210 GE
• When at scale, likely 75 LMR, 15 GE, 10 other
  (On-Call, Epic at CHA)

17
Development/Implementation of CPOE at BWH

• Home-grown application
• Developed beginning in 1993
• Highly multidisciplinary approach
• Went up in staged fashion
• Have gradually layered on decision support
• Sold commercial rights to a large vendor

18
Event Monitor Architecture
Applications (new data)
Patient database
Applications (new data)
Applications (new data)
page, email, write to file, real time message
Annun-ciators
Inference engine (decisions)
Rule editor
Knowledge base
Coverage List
19
Functions of CDSSs

• Alerting--high lab value
• Reminding--mammogram
• Critiquing--rejecting an order
• Interpreting--interpreting an ECG
• Predicting--risk of mortality using severity
  score
• Assisting--tailoring antibiotic choices
• Diagnosing--ddx in CP
• Suggesting--for adjusting mechanical ventilator

Randolph et al, JAMA 1999, from Pryor, 1990
20
Underpinnings

• Alerts, reminders, critiques often simple if-then
  rules
• (sometimes other Boolean operators)
• Alerts use event monitors, evaluate streams of
  data
• Finding right person hard
• Reminders notify patients of tasks to be done
  before event occurs
• Critiques--alternative suggestions, evaluate plan
  after started
• Interpreting/predicting/diagnosing/assisting and
  suggesting are higher order
• Harder to program, require more data

21
Types of Decision Support

• Structuring
• Information display
• Generic across patients
• Patient-specific
• Calculations
• Reminders
• Alerts
• Guidelines/Algorithms

22
Key Domains of Benefit

• Availability
• Communication
• Operational savings
• Decision support
• Reducing errors
• Improving guideline compliance
• Reducing costs
• Quality measurement
• Satisfaction
• Efficiency

23
Ways IT Can Improve Safety

• Prevent errors and adverse events
• Facilitating a more rapid response after an
  adverse event has occurred
• Tracking and providing feedback about adverse
  events

Bates and Gawande, NEJM 2003
24
Main Strategies for Preventing Errors and AEs
Using IT

• Tools to improve communication
• Making knowledge more readily accessible
• Requiring key pieces of information
• Assisting with calculations
• Performing checks in real time
• Assisting with monitoring
• Providing decision support

Bates and Gawande, NEJM 2003
25
Specific IT Applications

• Computerized physician order entry
• Smart pumps
• Smart monitoring
• Computerized notification about critical test
  results
• Computerized ADE monitoring
• Tracking abnormal test results

26
How to Prioritize with Decision Support in CPOE?

• ROI of inpatient CPOE evaluated
• Cumulative net savings were 16.7 million over 10
  years, and net operating 9.5 million
• Leading contributors
• Renal dosing guidance
• Tools to help nurses
• Specific drug guidance
• Adverse drug event prevention

Kaushal, JAMIA 2006
27
Improving Drug Ordering With Order Entry

• Serious medication error rate fell 55
• Streamline, structure process
• Doses from menus
• Decreased transcription
• Complete orders required
• Give information at the time needed
• Show relevant laboratories
• Guidelines
• Guided dose algorithms
• Perform checks in background
• Drug-allergy Dose ceiling
• Drug-drug Drug-patient characteristic
• Drug-laboratory

28
Non-Missing Dose Medication Error Rate
29
Effect of Changing Default Dosing Frequency for
Ceftriaxone
30
NEPHROS study

• Effect of real-time decision support for patients
  with renal insufficiency
• Of 17,828 patients, 42 had some degree of renal
  insufficiency
• Interv Control
• Dose 67 54
• Frequency 59 35
• LOS 0.5 days shorter

Chertow et al, JAMA 2001
31
Guided Prescription for Geriatric Inpatients

• Elderly patients frequently get dosages that are
  too high, especially initial dosages
• Performed RCT of decision support around dosing
  for psychoactive drugs
• Was associated with
• More frequent recommended dose (29 vs. 19)
• Lower fall rate (2.8 vs. 6.4 falls/1000 pt days)
• Lower frequency of 10-fold overdose (2.8 vs. 5)
• No difference in mental status change, LOS
• Clearly beneficial to suggest starting with lower
  dosage but more room for improvement

Peterson, Arch Intern Med, 2005
32
Clinical Consequences Of a Cart Fill Error
33
Background Epidemiology of Dispensing Errors

• Dispensing errors are relatively common in
  hospital pharmacies because of the high volume of
  medications dispensed.
• More than 44,000 errors occur per year in a
  735-bed hospital (6 million doses/yr).
• Many dispensing errors have the potential to harm
  patients.
• More than 9500 errors with potential to harm
  patients occur per year in a 735-bed hospital.
• Only 1/3 of these serious errors intercepted
  prior to administration.

Cina, Gandhi, Churchill, Fanikos, McCrea, Mitton,
Rothschild, Featherstone, Keohane, Bates, Poon.
Jt Comm J of Qual Safety. Jt Comm J of Quality
and Safety, Feb 2006
34
Dispensing Errors and Potential ADEs Before and
After Barcode Technology Implementation

• Projections for errors prevented per year at
  study hospital
• gt13,500 medication dispensing errors
• gt6,000 potential ADEs

plt0.0001 (Chi-squared test)
35
Vancomycin RCT

• Initiation, renewals both targeted
• Vancomycin use was reduced by intervention
• Bigger effect on renewals than on initiation
• Magnitude of overall decreases
• Vancomycin-days/prescriber 37 lower
• Duration of therapy 17 lower

Shojania et al, JAMIA, 1999
36
Low Yield Critique
37
Alternate Exam
38
Partners LMR Cost-Benefit Model

• Created a typical patient panel for a PCP
• Created a model using base case assumptions
• Analyze costs savings
• All cost and benefit figures are per PCP per year
• Created variants of model for other feature set
  configurations
• Performed a 1-way sensitivity analysis

Wang S, Middleton B, Prosser L, et al. A
Cost-Benefit Analysis for Electronic Medical
Record Systems in Primary Care. AJM 2003
39
Costs of EHR vs. Benefits
Wang et al, Am J Med 2003
40
(No Transcript)
41
Center for IT Leadership Mission

• Produce timely, rigorous market-driven technology
  assessments which
• Help providers invest wisely
• Help IT firms understand value proposition
• Help shape public policy
• Established at Partners HealthCare in partnership
  with HIMSS

C!TL Improving Healthcare Value
42
ACPOE Costs The Upshot

• Cost rapidly increases with system functionality
• Advanced ACPOE costs essentially include license
  for ambulatory EMR
• Adopting practice size has large implications for
  ACPOE costs
• Huge economies of scale for sophisticated ACPOE
  by spreading costs over more providers

43
Clinical Impact of ACPOE

• Per average provider, Advanced ACPOE systems
  would prevent
• 9 ADE/yr
• 6 ADE visit/yr
• 4 ADE admission/5yr
• 3 life-threatening ADE/5yr

44
ACPOE Financial Benefits

• Cost Savings
• Using national average capitation rate of 11.6
• Save 28,000 per average provider per year
• Revenue Enhancements
• Eliminate more than 10 in rejected claims per
  outpatient visit
• Address drug, procedure and coding issues through
  advanced clinical decision support
• Productivity Gains
• Neutral effect on provider time with improved
  staff productivity

45
Per Average Provider Annual Cost Saving
Projections
46
Advanced Systems Produce Superior Returns

• For example, Advanced ACPOE costs nearly
• 4x as much as Basic, but
• Generates over 12x more financial returns
• Produces nearly ten-fold greater reduction in
  number of ADEs
• Provides IT infrastructure for core clinical
  computing the outpatient EMR which produces
  additional benefits
• Pays for itself within first two years

47
US Would Benefit from Healthcare Information
Exchange

• Nationwide implementation of standardized
  healthcare information exchange would
• Save 337B over 10 years
• Save the US 78B annually at steady state
• Cumulative breakeven during year five of
  implementation
• There is a business case for standardized
  healthcare information exchange and
  interoperability

48
What Does It Take to Achieve High Levels of
Quality in Outpatients? A Lesson from Harvard
Vanguard

• Good electronic health record with solid clinical
  decision support
• Registry function
• Should be easy for providers
• Case management including HIT tools for case
  managers

49
Diabetes Report
50
Quality Dashboard
51
Ten Commandments for Effective Clinical Decision
Support

• Speed is everything
• Anticipate needs and deliver in real time
• Fit into the users workflow
• Little things can make a big difference
• Physicians resist stopping

• Changing direction is fine
• Simple interventions work best
• Asking for information is OKbut be sure you
  really need it
• Monitor impact, get feedback, and respond.
• Knowledge-based systems must be managed and
  maintained.

Bates DW et al, JAMIA 2003
52
Dilbert
53
Results Manager Home Page
54
(No Transcript)
55
Research Patient Data Repository

• Includes rolled-up data from every encounter
• Can get deidentified data readily to get counts
  of patients will diseases
• Need IRB approval to get identified cohorts
• Includes wide array of clinical data, more being
  added all the time
• Tremendous resource

56
(No Transcript)
57
Trajectories that Will Shape the Next Five Years

• Healthcare context
• Movement of care to outpatient/non-acute settings
• Managing inpatient capacity
• Growing dominance of the treatment of the
  chronically ill in the healthcare cost discussion
• Gradual movement to payment based on quality
• Increased patient service and participation
  expectations
• Technology context
• Growing presence of mobile technologies
• Ubiquitous wireless
• Improved interoperability between systems

58
Key Clinical IS Over the Next Five Years

• Computerized provider order entry
• Electronic medical record
• Knowledge repositories and management
• Physician-to-physician consultation
• Patient-provider communication/monitoring
• Care analysis
• Integration of clinical systems

59
Key Lessons

• Physicians are happy to change direction
• Much less willing to stop after action started
• Respond well to quality-related suggestions
• But even if low-yield utilization identified may
  proceed
• Satisfaction with these efforts good
• Success depends on integration with practice flow
• Developers must think speed, speed, speed...

60
Conclusions

• Providers want immediate individual feedback
• Computerized decision support can decrease costs,
  improve quality, by
• Pointing out redundancies
• Suggesting alternatives
• Identifying errors of omission
• Emphasizing important abnormalities
• Making guidelines accessible
• Have to manage knowledge
• Quality measurement vital as we strive to provide
  the same or better care at lower cost
• Most future QM will be done by computer as a part
  of routine care

61
(No Transcript)