A Measurable Difference: Clinical Decision Support Improves Care and Safety

1
A Measurable Difference Clinical Decision
Support Improves Care and Safety

• Scott Weingarten, MD
• President and CEO, Zynx Health
• Clinical Professor of Medicine
• UCLA School of Medicine

2
Convergence of Ideas

• Executives, physicians and Board members will
  require an ROI for Clinical Information Systems.
• Patient-specific evidence-based information,
  provided to clinicians at the time of care (CDS),
  can improve care.
• CDS maintenance, updates require clinical
  knowledge management.
• Implementation is key to success.

3
CIS - ROI

• HIMSS Analytics
• 46 (1,904 / 4,115) hospitals have bought CPOE
• Expensive
• Requires CFO, CEO, Board approval
• Someone promised improvements in quality, safety,
  and efficiency
• Someone will remember

4
CIS Potential Side Effects

• Studies
• Introduction of errors

Pediatrics 20051161506-1512
5
CIS Potential Side Effects

• Interviews from 176 hospitals
• CPOE
• Median 5 years (6 months to 25 years)
• J Am Med Inform Assoc 200714415-23

Concern
More work 95
System demands 92
New errors 91
Users with emotional reactions 90
6
CIS Potential Placebo Effect

• Ambulatory EHRs
• 2003, 2004
• 17 ambulatory quality indicators
• 1.8 billion EHR ambulatory visits
• As implemented, EHRs were not associated with
  better quality ambulatory care.
• In selecting an EHR, physician practices should
  carefully consider the inclusion of clinical
  decision support to facilitate quality care
• Arch Intern Med 20071671400-5

7
Physician Decisions 17 Year Lag
Practice guidelines
Print Sources
Clinical pathways
Physician order
Patient
Algorithms / maps
Textbooks
Other referentialcontent sources
8
CIS PotentialReferential CDS
Time to Conduct Search Time to Conduct Search Time to Conduct Search
Source Median time (minutes) Information very or somewhat helpful
Up-to-Date 15 91
Medline 20 59
Clinical Evidence 3 23
Cancer Net 7 4
Total 42 minutes per patient
JGIM 200419402-9
9
CIS Potential Referential CDS

• 23 physicians, 46 questions
• 6.5 minutes searching per question
• 1.8 resources per question
• 39.1 answers correct before searching
• 42.1 answers correct after searching

McKibbon KA, et al. J Am Med Inform 200613653-9
10
Six Sigma
Sigma Level Error rate per million opportunities Health Care Outside of Health Care
6 3.4 Publishing misspelled words
5.4 Deaths from anesthesia
5 230 Airline fatalities
4 6,210 Airline baggage handling Restaurant billing
3 210,000 Antibiotics for colds
2 580,000 Depression adequately treated
1 790,000 Beta blocker treatment after acute myocardial infarction
Mark Chassin Milbank Quarterly
11
CIS Potential

• Improves legibility
• Microsoft Word also improves legibility

12
CIS Potential
Errors of Commission Errors of Commission Errors of Omission Errors of Omission Errors of Omission
Eliminate Hospital ADEs Eliminate Ambulatory ADEs 5 Preventive Care Services DM for Diseases Prevention and Management of Chronic Conditions
Cost Reduction (per year) 1 billion 3.5 billion 1.74 billion -3.65 billion 28 billion 139 billion
Deaths Avoided (per year) 39,900 to 83,800 400,000
Hillestad R, et al. Health Affairs
2005241103-17
13
CIS Potential Personal Order Sets

• 1,100 physicians
• 560 physicians with POS
• 2,247 POS
• 10,123 unique orders
• Each physician had an average of 4 POS
• Include SL nifedipine, droperidol

AMIA 2003 Symposium Proceedings page 1031
14
CIS Potential -Personal Order Sets

• At QMC, there are more than 10,000 unique orders
  in POS. This is an unwieldy number to administer
  centrally. Additionally, these POS contain
  orders that are no longer considered best medical
  practice and in some cases may be dangerous.
  Such disadvantages need to be considered when
  implementing POS.
• If orders are reviewed once a year
• 40 unique orders need to be reviewed daily

AMIA 2003 Symposium Proceedings - page 1031
15
CIS Potential Personal Order Sets

• When order sets areinadequately maintained,
  they become templates for efficiently practicing
  outdated medicine on a widespread basis
• A. Bobb et,al. JAMIA v14 Jan/Feb07 pp 41-47

16
CIS - ROI

• Advisory Board Observations
• Project goals abandoned in light of overwhelming
  workload sole focus becomes completion.
• Far fewer CIOs focusing on meeting intended
  business value of the technology.
• IT Pursuit of Clinical Transformation. Blueprints
  for Foundational IT. 2006
• The potential for significant expense without a
  measurable return on investment could result in
  costly mistakes, ultimately hindering
  profitability and balance sheet growth.
• Fitch Ratings

17
HCIT Consultants

• Selection of CPOE system

• Implementation of CPOE

• Try to find benefits

18
The Worlds Literature
Meta-analysis 68 effective, (n 6,000
clinicians), (n130,000 patients)
Predictors of Success Adjusted odds ratio
Automatic provision of decision support as part of workflow 112
Provision of decision support at the time and location of decision-making 15
Provisions of recommendations rather than assessments 7
Computer-based decision support 6
N70 studies

• 75 of interventions succeeded when the decision
  support was provided to clinicians automatically,
  whereas none succeeded when clinicians were
  required to seek out the advice

Kawamoto K, et al. BMJ 20053301065
19
CIS ROI

• Meanwhile in the trenches, we are struggling to
  extract the anticipated value from the hefty bets
  we are making on clinical systems. Taken alone,
  clinical information technologies like
  bar-coding, smart pumps, nursing documentation,
  pharmacy and physician order entry systems dont
  sufficiently improve clinical practice to justify
  these investments. The systems must be
  supplemented by embedding intelligence into the
  clinical workflow. Decision support is the key
  to driving high quality and fail safe care.

True North Hardwiring The Evidence The Advisory
Board Company, Washington, DC
20
Strategies for Deploying Patient-Specific
Clinical Information

• Electronic
• Order sets
• Plans of care
• Alerts
• Default settings
• Structured documentation
• Default settings

21
Future Physician Decisions
Practice guidelines
Clinical pathways
Physician order
Patient
Algorithm / map
Referential content sources
22
Evidence-based Order sets

• Pediatric Inpatient Asthma
• 790 patients
• Pediatr Allergy
  Immunol 200617199-206

Measure Order Set Before p-Value
Pulse oximetry 94 75 lt0.001
Systemic corticosteroid use 56 48 lt0.02
Use of MDIs 91 79 lt0.001
23
Evidence-based Order Sets

• The integration of evidence-based treatment
  recommendations as computerized order sets within
  an inpatient CPOE system can improve compliance
  with evidence-based treatment recommendations.
  Our study provides further evidence to support
  the argument that expansion of the use of CPOE
  with integrated order sets can lead to general
  improvements in quality of care, consistency of
  care, and hopefully, outcomes.

Pediatr Allergy Immunol 200617199-206
24
Evidence-based Order Sets

• 259 hospitals
• OPTIMIZE HF
• 48,612 patients
• Tools
• Practice algorithms
• Customizable admission order sets
• Customizable discharge order sets

Arch Intern Med 20071671493-1502
25
Evidence-based Order Sets

• Pre-printed Order Sets and Discharge Checklists
• 259 hospitals
• 48,612 patients
• largest national hospital-based program
  dedicated to quality-of-care improvement for
  patients hospitalized with HF in the United
  States
• Process of care improvement tools included
  customizable admission and discharge order sets,
  evidence-based algorithms
• Arch Intern Med 20071671493-1502

26
Evidence-based Order Sets
plt0.001
plt0.002
plt0.001
plt0.001
Eligible Patients Treated,
Arch Intern Med 20071671493-1502
27
Evidence-based Order Sets

• Acute MI care (n2,857 patients)Standardized
  orders, pocket guideline, discharge toolIncrease
  in standard orders from 19.8 to 45.5 (plt0.001)
  

Evidence-based order sets Usual order sets
Hospital mortality 10.4 13.6
30-day 16.7 21.6
1-year 33.2 38.3
21 to 26 reduction in mortalityJ Am Coll
Cardiol 2005461242-8
28
Evidence-based Order Sets

• 52 of denials for order tests for conditions
  that are not allowable
• Insurance denial reduction
• MRI/MRA 37
• Colonoscopy 29 AMIA Symposium Proceedings
  2005, page 1151

29
CDS Benefits

• C-section, vaginal deliveries
• 290 malpractice claims
• Compliance with evidence-based clinical pathways
  associated with an almost 6 fold decrease in
  malpractice claims
• Risk of not following evidence-based clinical
  pathway (6.1 claims for 1000 non-compliant
  deliveries)
• Ransom S, et al. Ob Gyn 2003101751-5

30
Patient-specific evidence-based information
31
Professionalism With Decision Support

• Pilots require extensive knowledge, training and
  experience
• Utilize decision support

32
Clinical Knowledge Management

• 20,000 biomedical journals
• 500,000 indexed in PubMed annually
• gt150,000 articles per month
• 6,000 articles a day
• gt3,000 molecular diagnostic testsMedical
  References Services Quarterly 2007261-19

33
Clinical Knowledge Management
GAP

• 6,000 articles/day
• 150,000 articles/month
• 300,000 RCTs
• 20,000 biomedical journals

200 MB capacity
Dr. Frank Davis
Ann Intern Med 2001135309-12
34
Clinical Knowledge Management

• Finish medical school and residency knowing
  everything
• Read and retain 2 articles every single night
• At the end of 1 year
• 1,225 years behind

W Stead. JAMIA 200512113-20 Alper BS, Hand JA,
Elliott SG, et al. J Med Libr Assoc
200492429-37.
35
Clinical Knowledge Management
Total Unique
Active Performance Measures 2268 1302
Active Regulatory Measures 887 312
Active P4P Measures 154 98
36
Clinical Knowledge Management
Before Clinical Decision Support Before New Evidence New Evidence Results Clinical Decision Support After
For every 20 patients treated with albumin, 1 excess death2 Do not prescribe albumin alert Study with 5-times sample size3 Albumin does not increase mortality Remove alert
No significant benefit of adding dipyridamole to aspirin4 Dipyridamole not on order set Larger study published5 Composite reduction in cardiovascular outcomes with ASA and dipyridamole 1 per year Add dipyridamole to order set

1. Ann Intern Med 2007147224-233
2. Cochrane Database 2002CD0011208
3. N Engl J Med 20043502247-56
4. BMJ 200232471-86
5. Lancet 20063671665-73

37
Clinical Knowledge Management
CPOE readiness components Readiness score (mean range)
Organizational Structure Function 70 (35-94)
Information Technology Composition 64 (24-94)
Information Technology Infrastructure 63 (47-73)
Organizational Leadership 59 (21-95)
External Environment 57 (45-82)
Access to Information 56 (30-90)
Order Management Process 53 (25-78)
Organizational Culture 51 (27-93)
Care Standardization 44 (25-100)
JCJQS 200329336-44
38
Clinical Knowledge Management

• Updates, maintenance
• Most hospitals will want
• 250 to 1000 order sets
• Update at least once per year
• 1 to 4 order sets updated per business day

39
Clinical Knowledge Management

• Internally developed
• Notify physician when blood sugar is gt40
• Administer O2 at ___ liters per min via ___ (if
  PAO2gt60mm Hg or spot check pulse-ox sat.gt90)
• Raise HOBgt30

40
Clinical Knowledge Management

• Iterative process
• of the 48 order sets that are actively used in
  POE, nearly all have been changed multiple times
  since initial release for clinical use. One
  particular order set has been revised 44 times
• J Biomed Inform 2007

41
Disease Specific - CAP
42
What Works?

• Clinician
• Awareness
• Familiarity
• Agreement
• Time to perform
• Computer
• Couple reminder with order on same screen
• Distinctive color screen to highlight
• Making override more difficult
• Default set to order
• Repetition
• BMC Med Inform Decis Mak 2006Feb 1, 6

43
What Works?Computer Default Settings
44
What Works?Default Values

• Hospital
• Pneumococcal immunization
• Influenza immunization
• Removal of all urinary catheters within 72 hours
  unless otherwise indicated
• Intensive care units
• Elevated head of bed
• Physician office
• HbA1C for diabetic patients
• Retinal examinations for diabetic patients
• N Engl J Med 20073571340-4

45
8020 Rule

• Today
• 80 development
• 0 maintenance
• 20 marketing, selling order sets to physicians
• Result --- under-utilization
• Tomorrow
• lt5 development
• 50 maintenance, updates
• 50 marketing, selling to physicians
• Result --- better utilization

46
CDS Framework
Discipline
Genomics
Medical
Nursing
Pharmacy
RT
PT
Other
Critical Care
Hospital Care
Venue
Ambulatory Care
Emergency Care
Home care
47
Future Hypothetical Clinical Case

• Mr. Jones is a 69-year-old white man with a
  history of diabetes who was hospitalized with a
  temperature 101.20F, non-productive cough, and a
  left-lower-lobe infiltrate. His room air O2 sat
  was 85. I was planning on performing blood
  cultures, administering oxygen, and prescribing
  doxycycline

48
Future Hypothetical Clinical Case

• Clinical decision support has been run against
  the patients demographic, genomic, hemodynamic,
  physiological, history, physical, radiographic,
  and laboratory findings. The patients
  probability of mortality can be reduced by 12
  and length of stay reduced by 8 by prescribing
  an antipneumococcal fluoroquinolone. The
  patients mortality may be reduced by an
  additional 3 by using the hyperglycemia order
  set. Also, a fall reduction program will reduce
  the patients risk of a fall leading to prolonged
  stay by 11.

49
Conclusions

• CIS have great potential for delivering clinical
  and ROI benefits
• Executives, physicians and Board members will ask
  for the ROI.
• Patient-specific evidence-based information,
  provided to clinicians at the time of care, can
  provide the benefits.
• Clinical knowledge management required for CDS.
• Implementation is key to success
• Marketing and selling CDS to clinicians

50
Questions???