Becoming ADEPT

1
Becoming ADEPT

• Introduction and Description of the Course
• Andrew Booth, Suzy Paisley

2
Why ADEPT?

• librarians need to be able to support Evidence
  Based Healthcare
• users need to be able to do their own focused
  searching
• only a small proportion (about 1) of the medical
  literature is clinically relevant

3
ADEPT - methods

• SEE-ONE, DO-ONE, TEACH-ONE
• Scenarios
• Focused questions
• Small group teaching
• The Group as a resource
• Problem based learning

4
Topics

• Types of question - primary studies
• Diagnosis
• Etiology
• Prognosis
• Therapy

• Types of article - secondary studies
• Reviews
• Economic Evaluations
• Guidelines

5
The Process

• Introduction to study design
• Clinical Scenario
• Focusing the question
• Translating into MeSH terms
• Conducting the literature search
• Applying the methodological filters
• Specialist resources

6
Resources

• About ADEPT
• Becoming ADEPT Project Overview - HLR Dec 1997.
• Becoming ADEPT delivering distance learning on
  EBM to Librarians - HLR Sept 1998.

• For each study
• User Guides from JAMA
• ACP Journal Club or EBM MEDLINE Guides
• Institute of Health Sciences Filter page

7
Introduction to Study Design

• Andrew Booth

8
Scenario

• You are a postgraduate medical librarian in
  Kleptovale NHS Trust. You have bid for a Nicksnix
  Library security system. The postgraduate Dean
  wants some evidence that it is effective. There
  is no published evidence so you will need to
  conduct an original study. What are you going to
  do?

9
Answer. A Case-control study

• You call up the Regional Library Unit. Details of
  10 libraries with book losses of gt5 (cases) and
  of 10 libraries with losses of lt1 (controls).
• Phone all 20 libraries and ask Do you use a
  library security system? Yes/No

10
You compile a 2 by 2 table.
11
Oh no! Statistics

• Only 40 of libraries with heavy losses had a
  security system.
• 60 of the controls had a security system

• The Odds Ratio
• Odds of Library with heavy losses A/C 4/6
  0.66
• Odds of Library w/o heavy losses B/D 6/4 1.5
• A/C 0.66 0.44
• B/D 1.5

12
But........

• Controls might be underreporting their losses to
  make system look good.
• People with security systems may be more security
  conscious (using other methods)
• Bias (flaw in research method) or confounding
  (some other factor not taken into account).

13
Case-control studies Retrospective

• Advantages
• Quick and cheap
• Its the only method for rare events
• Requires fewer subjects than other designs

• Disadvantages
• Relies on recall/records
• Open to confounding
• Selection of controls is difficult
• Potential bias in recall or selection.

14
Answer A Cohort Study

• Convince 110 members of HLG to take part in a
  prospective cohort study. Over a two year period
  you record who uses the system Exposed and who
  doesnt Non-Exposed. At end of two years you
  calculate rate of heavy losses within both groups.

15
Another 2 by 2 table
16
But.....

• OK so you may have eliminated recall and
  reporting bias through baseline data collection,
  but
• Libraries with security systems may be in richer
  areas and less prone to theft anyway.

17
Cohort studies

• Advantages
• Subjects can be matched
• Eligibility criteria can be standardised
• Easier to run than an RCT.

• Disadvantages
• Blinding is difficult
• Randomisation is not present
• Requires a large sample and longer follow up.

18
Answer An RCT

• You go for a sponsored Randomised Controlled
  Trial. You recruit another group of HLG members.
  You randomly assign half with a free NicksNix
  system and the others must go without. After a
  year you have the following results.

19
Yet another 2 by 2 table
20
Finally....

• After three years and thirty minutes and three
  separate studies the Postgraduate Dean believes
  he has the right answer-
• No difference, No Nicksnix,
• No comment!!!

21
Randomised Controlled Trials

• Advantages
• Unbiased distribution of confounders
• Blinding more possible
• Randomisation facilitates statistical analysis

• Disadvantages
• Expensive in time and money
• Volunteer bias
• Ethical problems, sometimes
• May create artificial behaviour

22
Postscript

• Practically, you wouldnt go to all that effort.
  Maybe use a cross-over design where one Library
  (n-of-1 study) or more went through periods
  without and then with a system. You would need to
  make sure that there was a washout period. Also
  you would have to guard against a placebo effect.

23
Hierarchy of research evidence

• Systematic Review
• Randomised Controlled Trial
• Cohort study Prospective
• Case-control Retrospective (May be contemporary
  or historical)
• Case series
• Individual case study