Transcript The Contribution of Clinical Epidemiology to Modern Health

Bringing Evidence to the Bedside
in Critical Care
Allan S. Detsky
Physician-in-Chief
Mount Sinai Hospital
We need information
If asked:
• We need it twice a week
• We get it from textbooks, journals
and our colleagues
We really need information
If shadowed:
• We need it twice for every 3 outpatients
and 5 times for every patient
• But we rarely get what we need from the
resources that we use
– Colleagues
– Textbooks
– Journals
Ann Intern Med 1991;114:576-81
Evidence of Care Gaps
•
•
•
•
Beta blockers in post MI patients
Statins in post MI patients
ACE-I in patients with CHF
Antithrombotic therapy in patients with
nonvalvular atrial fibrillation
HOW CAN WE BRIDGE THIS GAP?
Bringing evidence to the point of care
• Need it within seconds if it is to be incorporated
into busy clinical rounds
• Focus on users’ needs and important clinical
outcomes
• Our initial attempts to bring the best evidence to
a busy clinical team caring for over 200 patients
per month
JAMA 1998;280:1336-8.
Kinds of Questions
1
.
Therapeutic Effectiveness
2.
Diagnostic Accuracy
3.
Prognostic Information
Therapeutic Effectiveness
For patients with health state X,
does Rx A do more good than
harm (compared to Rx B)?
Diagnostic Accuracy
Does test A help us separate those
with disease from those without
for patients that look like X?
Prognostic Information
For patients with health state X,
was the risk of an adverse
outcome over a specified period
of time?
What do clinicians want on
PDAs?
• Clinical bottom line from preappraised
resources
• Management algorithms
• Drug dosages and interactions
• Numerical summaries of risks and benefits
• They don’t want traditional clinical practice
guidelines
Examples
Systems
Computerized decision
support systems (CDSS)
Synopsis
Clinical Evidence
Syntheses
Studies
Cochrane Reviews
Original published
articles in journals
Internet Available Resources
www.cebm.utoronto.ca
www.eboncall.co.uk
How To design
1. Learn how to recognize questions.
2. Learn how to quantify answers.
Therapeutic Effectiveness
Pc = event rate in control group
Pt = event rate in experimental group
Pc - Pt = ARD
(Pc - Pt)/Pc = PRD
Pt/Pc = RR
1
Pc - Pt
= NNT
10% - 5% = 5% = ARD
(10% - 5%)/10% = 50% = RRD
5%/10% = .5 = RR
1
10% - 5%
= 20 = NNT
DIAGNOSTIC TESTS
• Sensitivity
• Specificity
• Likelihood ratios
LR > 1
LR = 1
LR < 1
Read and appraise individual
studies, systematic reviews
Then go to websites to see how
results are presented.
Next - start a simple research
project.
Medical Consults
What is the validity of a
cardiac risk index in
our patients?
• Start slowly
• Progress will take time
• It’s harder than you think