Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
InfectionControl.tipsJoin. Contribute. Make a Difference
Systematic Reviews Andrew Duong, M.Sc
InfectionControl.tipsJoin. Contribute. Make a Difference
The Systematic Review
Process
Identify Research Question/Objective
Development of Research
Question/PICO question
Development and Execution of Search
Strategy
Data AbstractionPresentation and
Synthesis of Results
Assessment of Study Quality and Publication Biases
InfectionControl.tipsJoin. Contribute. Make a Difference
1. Developing the Research
Strategy
Start with your research question
Population
Intervention(s)
Types of study design
Each database/systematic review may require a
different search strategy
Medline, EMBASE, PubMed, CENTRAL
(Cochrane), CINAHL
InfectionControl.tipsJoin. Contribute. Make a Difference
1. Developing the Research
Strategy
When doing background research, identify
key words and terms that are important
Exclusionary terms are just as important
InfectionControl.tipsJoin. Contribute. Make a Difference
2a. Electronic Database
Searching
Search optimization
◦ Truncation ($, *)
◦ Operators (AND, OR, ADJn)
◦ Syntax (title, abstract, subject heading, publication type)
◦ Subject headings
Automatic explosion (.exp) expands on topic and includes all predesignated subheadings
Searches should be kept broad
InfectionControl.tipsJoin. Contribute. Make a Difference
2a. Electronic Database
SearchingMEDLINE EMBASE Pubmed
1. exp Femoracetabular
Impingement/
2. femoral adj acetabular adj
impingement.mp.
3. femoroacetabular
impingement.mp.
4. femoracetabular
impingement.mp
5. Hip impingement.mp
6. Pincer impingement.mp
7. Cam Impingement.mp.
8. 1 or 2 or 3 or 4 or 5 or 6 or 7
9. Limit 8 to English language
July 31, 2015: 1070 results
1. exp Femoracetabular
Impingement/
2. femoral adj acetabular adj
impingement.mp
3. femoroacetabular
impingement.mp
4. femoracetabular
impingement.mp
5. Hip impingement.mp
6. Pincer impingement.mp
7. Cam Impingement.mp.
8. 1 or 2 or 3 or 4 or 5 or 6 or 7
9. Limit 8 to English language
July 31, 2015: 1800 results
1. Femoracetabul*
Impingement/
2. Femoro*acetabul*
impingement/
3. Femoral acetabul*
impingement/
4. Hip Impingement/
5. Pincer Impingement/
6. Cam Impingement/
7. 1 or 2 or 3 or 4 or 5
or 6
8. Limit 7 to 2015 [year]
October 21, 2015: 207
results
InfectionControl.tipsJoin. Contribute. Make a Difference
2b. Manual Searching
Manual search of journals
Consulting experts
Conference proceedings
New, unpublished data
Data not necessarily peer-reviewed
Dissertations and theses
InfectionControl.tipsJoin. Contribute. Make a Difference
2b. Manual Searching
Ongoing clinical trials
https://clinicaltrials.gov/
Websites
Useful for treatment guidelines American Medical Association, Center for
Disease Control
Guidelines may be anecdotal
Always review the primary literature
InfectionControl.tipsJoin. Contribute. Make a Difference
Biases
Systematic errors that consistently result in misattribution of findings
Avoid by establishing clear criteria a priori that do not discriminate based on the evidence presented Publication Bias
Misrepresentation of the population of studies
Reporting Bias Attrition Bias Detection Bias
InfectionControl.tipsJoin. Contribute. Make a Difference
3. Screening
The development of inclusion/exclusion
criteria is important for this process
Always done in duplicate
Title Screening Abstract Screening
Full Text Screening
Screening is typically done in
spreadsheets (Microsoft Excel)
InfectionControl.tipsJoin. Contribute. Make a Difference
3. Screening
Code your screening 1 = Include
2 = Maybe include
3 = Exclude (non-randomized control trial)
4 = Exclude (duplicate article)
5 = Exclude (cadaveric study)
Etc.
InfectionControl.tipsJoin. Contribute. Make a Difference
3. Screening
Before screening, test a limited number of
articles (~30) to ensure that both
screeners are clear about what is to be
included, and what is not.
If there is a lot of disagreement, it is owrth
going back to the original
inclusion/exclusion criteria and revising
this.
InfectionControl.tipsJoin. Contribute. Make a Difference
Statistical Analysis
Agreement: Calculation of variability
during screening
Kappa (Cohen’s Kappa, κ)
Intraclass Correlation Coefficient (ICC)
InfectionControl.tipsJoin. Contribute. Make a Difference
Statistical Analysis
Improve agreement by
Developing a clear research question a priori
Developing inclusion/exclusion criteria a priori
Testing screening agreement with a few articles to determine if both reviewers are interpreting criteria in the same way
Disagreements should be resolved at this point, and criteria should be changed to be more clear
InfectionControl.tipsJoin. Contribute. Make a Difference
Agreement Calculation of variability during screening
Kappa (Cohen’s Kappa, κ)
Statistic for the measure of inter-rater
agreement
Systematic review: Level of agreement
between screeners (dichotomous only)
Indirect measure of quality of
inclusion/exclusion criteria
Κ = Total agreements – expected agreements
Overall total – expected agreements
InfectionControl.tipsJoin. Contribute. Make a Difference
Agreement
Intraclass Correlation Coefficient (ICC)
Inter-rater reliability based on quantitative values (continuous variables)
Useful for evaluations where scores are given
E.g. radiographic analysis can deem a patient to have a score of 1-10, which can vary depending on radiologist
Detsky quality assessment score based on multiple variables evaluating quality
InfectionControl.tipsJoin. Contribute. Make a Difference
4. Data Abstraction and
Analysis
Following screening, data should be
abstracted from full text articles in
duplicate
Similar to screening, test a limited
number of articles at first to ensure that
the data abstraction form is clear
Analysis of information can then be
performed from this subset
InfectionControl.tipsJoin. Contribute. Make a Difference
4. Data Analysis
Meta-Analysis
Integration of a number of studies to
create a more comprehensive
conclusion
Reduce heterogeneity among studies
Pooled studies must be sufficiently similar to
allow for integration of results
InfectionControl.tipsJoin. Contribute. Make a Difference
4. Data Analysis
Meta-analysis
Forest Plot
Heterogeneity: Variation in outcomes between studies Cochran’s Q: Weighted
difference between individual studies and the pooled effect of all studies
I2 statistic: Percentage of variation across all studies due to heterogeneity (not chance)
InfectionControl.tipsJoin. Contribute. Make a Difference
5. Presenting the Data
Magic formula for systematic reviews
Figure 1: Screening results (PRISMA
Guidelines)
Table 1: Study characteristics
Appendix 1: Search strategy
InfectionControl.tipsJoin. Contribute. Make a Difference
Records identified through database searching
(n = )
Screen
ing
Included
Eligibility
Iden
tification
Additional records identified through other sources
(n = )
Records after duplicates removed(n = )
Records screened(n = )
Records excluded(n = )
Full-text articles assessed for eligibility
(n = )
Full-text articles excluded, with reasons
(n = )
Studies included in qualitative synthesis
(n = )
Studies included in quantitative synthesis
(meta-analysis)(n = )
InfectionControl.tipsJoin. Contribute. Make a Difference
5. Presenting the Data Table 1: Summary of the studies,
highlighting general study information
Author, year, number of patients, mean age, gender proportion, general study guidelines, overall results
InfectionControl.tipsJoin. Contribute. Make a Difference
5. Preparing the Data
Appendix 1: Search Strategy
InfectionControl.tipsJoin. Contribute. Make a Difference
6. Writing the Manuscript
Outline
This is the most important part of the paper
Plan out the road map of the paper,
including details of what needs to be
introduced when, what results will be
discussed, and how figures are presented.
Edits at this stage are extremely easy
The more detail that is put in here, the easier the
rest of the writing process will be
InfectionControl.tipsJoin. Contribute. Make a Difference
6. Writing the Manuscript
Introduction
Background information, current gap in
knowledge, and purpose of study
Methods
Be as specific as possible, especially
search strategy development, screening
process, descriptive statistics,
InfectionControl.tipsJoin. Contribute. Make a Difference
6. Writing the Manuscript Results
Describes all the meaningful results
from the study
Figures and tables are complementary
to the results. Not a reiteration
InfectionControl.tipsJoin. Contribute. Make a Difference
6. Writing the Manuscript
Discussion Synthesis of results, which aims to answer
the research question First paragraph Statement of all major findings
Include all important findings from analyzing the results Do not restate the results
Strengths and limitations Include aspects that affect the quality(both
positively and negatively)
InfectionControl.tipsJoin. Contribute. Make a Difference
6. Writing the Manuscript
References
Use a reference manager
Mendeley reference manager
Be sure to cite using the journal’s stated
format
http://scholar.google.com/
InfectionControl.tipsJoin. Contribute. Make a Difference
6. Writing the Manuscript
Figures and Tables
Ensure that all figures and tables are
original
Typically, figures should be 300 d.p.i. or
higher
Abstract
Presented first, but written last
InfectionControl.tipsJoin. Contribute. Make a Difference
Tips for Performing
Systematic Reviews
Purpose of a Systematic Review is to
objectively summarize the data
Do not present every detail of what is
happening in each study
Raw data is to be avoided, especially
when descriptive statistics can be used
in its place
InfectionControl.tipsJoin. Contribute. Make a Difference
Tips for Performing
Systematic Reviews
Create a well defined research question, and refer to literature to help you refine it
Search strategy should be reviewed by all authors, and results should be screened to ensure appropriate articles are included
Searches should be as broad as possible
A priori inclusion criteria can be modified as you go. Test it against multiple articles before screening on a full scale
InfectionControl.tipsJoin. Contribute. Make a Difference
Tips for Performing
Systematic Reviews
All screening and data abstraction steps should be done in duplicate.
Disagreement can be resolved by a third party
Keep records of every step, including number of agreements, why articles were excluded, type of information abstracted and analyzed, and what methods/software were used for data and statistical analysis
InfectionControl.tipsJoin. Contribute. Make a Difference
Tips for Performing
Systematic Reviews
Prior to writing/creating figures, create a
paper outline
What will be included in each section,
and what order will things be presented?
InfectionControl.tipsJoin. Contribute. Make a Difference
Tips for Performing
Systematic Reviews
Try to put everything into a figure (when it
makes sense)
Figures/Tables and text should try to be
mutually exclusive
Figures/tables are supplementary to
understanding the information and the
focus should be to SUMMARIZE the
data
InfectionControl.tipsJoin. Contribute. Make a Difference
Tips for Performing
Systematic Reviews Discussion
First paragraph should summarize the major findings
Be explicit with limitations of the study
Improve quality
Methods to reduce bias, increase agreement, broaden search, perform steps in duplicate, robustness of article databases
Decrease quality
Limited number of studies, lack of meta-analysis, deviance from protocol, quality of evidence abstracted, limited scope, other limitations
InfectionControl.tipsJoin. Contribute. Make a Difference
Thank you