Using computerized decision support in hospitals

1
Using computerized decision support in
hospitals

• Kirsty Buising FRACP
• Infectious Diseases Physician and Clinical
  Research Fellow
• Victorian Infectious Diseases Service and NHMRC
  Centre for Clinical Research Excellence in
  Infectious Diseases

2
Computerised decision support what and why

• Definitions
• Using technology to aid decision making
• Drivers for decision support
• Knowledge performance gap
• Growth of medical knowledge
• Pressures to use knowledge
• Evidenced based medicine, clinical governance,
  cost

3
Implementation is the key

• Guidelines are useless if no-one uses them
• Implementation
• We have well established criteria to create a
  guideline, but dont know how to encourage
  guideline uptake
• Implementation is the Achilles heel for
  guidelines
• We need to
• Make the information usable
• Present it in a form that fits workflow and suits
  the context

4
Types of Computerised Decision Support
ALERTS TRIGGERS e.g Allergy alerts Out of range
test flagging
ACTIVE Patient specific Interactive Dynamic Can
be very advanced eg neural networks
PASSIVE Simply presenting guidelines e.g paper
based docs stored on intranet
5
Does CDS work in healthcare?

• Yes,
• Reduced medication errors improved drug dosing,
  fewer allergy mismatches (Level I)
• Increased concordance with guidelines (Level I)
• Reduced drug costs, test ordering, length of
  stay
• (Systematic reviews Walton BMJ 1999 Hunt
  JAMA 1998 Johnston 1994)
• Examples
• Warfarin, aminoglycoside, heparin dosing
• Preventive vaccination prompts
• Screening
• Vancomycin use

6

• Limitations
• Publication bias - a few major centres heavily
  computerised, well integrated
• US paradigm not applicable to Australia
• There are reports of failed systems
• must understand context, fit workflow, suit
  users
• Rousseau BMJ 1998, Thursky 2006

7
The 10 commandments of effective decision support

• Primary determinant of user satisfaction is speed
  
• Should anticipate needs and deliver in real-time
• Integrated with clinical practice and user
  workflow
• Usability is very important
• Physicians will often override reminders/suggestio
  ns if they have strong beliefs about the
  medication or clinical situation.
• Simple interventions work the best
• Additional information should not be requested
  from the user unless necessary.
• The impact should be monitored
• The systems should provide incentives for use
• Maintain the knowledge-based systems
• (Bates et al, JAMIA 2003 Shiffman, JAMIA
  1999)

8
Antibiotic stewardship

• Large volumes of antibiotics are prescribed, up
  to 50 of usage may be inappropriate
• wrong dose, wrong drug, wrong diagnosis, therapy
  too prolonged etc. House Lords review 1995, EU
  Copenhagen recommendations 2002
• Impact upon rates of multiresistant pathogens
• affect patient outcomes, more costly drugs,
  longer lengths stay etc
• Need for strategies to improve prescribing
  practices

9
Local CDS projects

• Focus on antibiotic stewardship
• Driven by VIDS and pharmacy, 2000-present
• Pilot - cAAS
• Pilot - ADVISE
• GUIDANCE DS
• Evaluations of iApprove and CAP guideline

10
ADVISEAntibiotic Decision Support for
Victorian Infectious Diseases Service

• Project 2000-2002 funded by DHS
• ICU chosen as pilot site
• Make information available at point of care
• Microbiology browser
• Access to core knowledge
• Known and predicted sensitivities
• Adjusted recommendations based on site, allergies

11
ADVISE Pre and post implementation studyThursky
et al IQHC 2006

• 986 ICU patients, 6 months pre and post ADVISE
• Change in pattern of antibiotic use
• - 3rd gen cephalosporins OR 0.58 0.52 0.79, p
  0.001
• - carbapenems OR 0.62 0.39 0.97, p 0.037
• - vancomycin OR 0.69 0.45 0.99, p 0.047
• Increased de-escalation (4.6 to 10.1, p0.013)
• Cost savings 20,000/year
• Popular- accessed 6000 times in 6 months
• 3 times per patient per day

12
GUIDANCE DS

• Commonwealth Biotechnology Innovation fund grant
• Internet based, .NET framework, eDSML
• Integration with existing hospital databases -
  extracts and presents relevant information
• Launched Jan 2005 at RMH

13
3 modules of GUIDANCE

• iApprove - a Restricted Drug Approval system
• iGuide - diagnostic/ management guidelines
• iMicro - recommendations in response to
  microbiologic isolates (like ADVISE)

14
Evaluation of iApprove

• Uptake number of uses and coverage
• Drug usage patterns of consumption of broad
  spectrum agents
• Resistance patterns of local bacteria
• Patient outcomes
• eg gram negative bacteraemia, mortality and
  length of stay
• Usability- users opinions

15
iApprove

• Uptake
• Currently 250-300 approvals/ month
• 70-100 coverage in gen med/ surg wards
• Usability
• Independent Evaluation Monash Uni
• 115 participants
• 80 believe iApprove easy to use
• Zaidi 2007

16
Impact of iApprovereduced use of broad spectrum
antibiotics
17
iGuide Why would CDS improve guideline uptake?

• The user gets something back
• Fast
• Simple/ makes sense of complex documents
• Educational
• Accessible 24 hours
• Transparent - layers of background info
• Locally endorsed
• Always up to date
• Make it easy to do the right thing

18
Presenting guidelines as computerized decision
support

• Need to modify the guideline to exploit the
  benefits of CDS
• Make it patient specific - algorithmic design
• Extract and present data from other sites
• Layered presentation - allow user to explore
• Local information
• Web enabled - external guidelines
• References provided

19
Evaluation of iGuide

• Impact of different implementation strategies for
  community acquired pneumonia (CAP) guideline
• 740 patients with over 4 years
• Passive dissemination 62 concordance (at 1
  year)
• Academic detailing 68 concordance (at 2-3
  years)
• CDS 89 concordance

20
Concordance with guideline
21
Capacity building

• Guidance DS already at
• RMH
• PMCI
• Barwon Health
• and soon to be at
• Eastern Health
• Alfred Hospital
• Tasmania (statewide)
• Ability to share guidelines between hospitals and
  modify content (taking ownership)

22
Summary

• CDS offers a new implementation strategy for
  evidence based medicine
• Staff are adopting it
• Must fit workflow and understand the context
• MH are leaders in this field

23
Acknowledgments

• Medseed Computing
• Ryan Warrener
• Russell Beattie
• Bron Gondwana
• Hugo Stephenson
• Michael Mahemoff
• All Clinicians and Pharmacists of RMH
• Declaration VIDS staff are employees of
  Melbourne Health and have no financial ties to
  Medseed

• VIDS/ CCREID
• Karin Thursky
• Jim Black
• Alan Street
• Michael Richards
• Graham Brown
• Lachlan Macgregor
• Renu Shansugamundaram
• Department of Pharmacy
• Marion Robertson
• Intensive Care Unit
• Jeffrey Presneill
• Jack Cade
• Emergency department
• - Marcus Kennedy