Follow
Publications: 22 | Followers: 0

An Intervention to Improve HIV Medication Adherence among ...

Publish on Category: Birds 0

H676 Week6-MA in Context- PlanforToday
NancyBorensteinet al., 2009 Chapter 30 - Publication biasWilliamson et al., 2005 - Outcome selectionbiasCriteria for judging/rating biasesBrianPageet al., 2015 -Handling of multiplicity of resultsBorensteinet al., 2009 Chapters 41-43Reporting the results of a MACumulative MACriticisms of MA
1
Nancy
PublicationbiasBorensteinet al., 2009 Chapter30Outcome selection biasWilliamson et al.,2005Criteria for judging/rating biases
2
Brian
How to handle multiplicityofresults to avoid biasPageet al.,2015Reportingthe results of a MABorensteinet al., 2009Chapter 41Cumulative MABorensteinet al., 2009Chapter 42CriticismsofMABorensteinet al., 2009Chapter 43
3
How can multiplicity of results lead to bias?
Discussion/comparison of the studies in your own meta-analysesVolunteer note-taker on thewhiteboardWhat kinds ofmultiplicity are there?What are implications of each kind for bias?How common is multiplicity?
4
How can multiplicity of results lead to bias?
Multiplicity of results is common49% in medical research reviewedProbably even more in some kinds of studiesHow about yours?Multiple kinds of multiplicityMultiple outcomesMultiple ways of measuring individual outcomesDifferent measures have different quality (reliability & validity)Multiple times or analytical samplesAdjusted for pretest values on not?Intention to Treat vsPer ProtocolMultiple arms in trialsWhich arm provides best estimate of effects?
5
What are the implications for your MA?
6
Implicationsforyour MA
Make decisions up frontDocument in your protocolDocument any changes in your protocolChanges can lead to bias due to selective inclusionBe consistent across studiesGo back and recode earlier studies if changes occur part-way through a reviewReport types and levels of multiplicity
7
Chapter 41 – Reporting results of a MA
What is yourfocus?Main effects or dispersion and/or moderator resultsI2= proportion of observed dispersion that reflects differences in the true effectT2= amount of heterogeneity on an absolute scaleWhichcomputational model(s) did you use?Fixed-effect or random-effectsmodel (and reasons for your choice)Include a forestplotEspecially with boxes proportional to study weightWhatsensitivity analysesdid you conduct?E.g., RCTs vsQETs, different ES measures, missing data, publication biasSometimes moderator analyses cover these
8
Chapter 42 – Cumulative MA
A mechanism for displaying a series of different analyses representing the accumulation of studiesOften over timeBut also across other properties, e.g., study qualityLook at extent to which effect size stabilizesFor display or educational purposes onlyAppropriate analysis is meta-regression
9
Chapter 43 – Criticisms of MA
Close your books and devicesWhat are some of the criticisms of MA?Small-group discussionsReport out
10
Chapter 43 – Criticisms of MA
One number cannot summarize a research fieldThe file-draw problem invalidates MAIt can be mixing apples and orangesGarbage in, garbage outImportant studies can be ignoredMA results can disagree with RCT resultsMA are often poorly performed/reportedCannot easily include qualitative dataMaybe a narrative review would be betterAn exercise in mega-silliness (Eysenck, 1978)?
11
Criticisms
Responses
12
Criticisms of MA andsome responses
One number cannot summarize a research fieldThe file-draw problem invalidates MAIt can be mixing apples and orangesGarbage in, garbage outImportant studies can be ignoredMA results can disagree withnew RCTresultsMA are often poorly performed/reportedCannot easily include qualitative dataMaybe a narrative review would be betterAn exercise in mega-silliness (Eysenck, 1978)?
Synthesis – average ES or dispersionAnalyze publication biasMA asks questions about fruit as a groupInclusion criteria as process of waste managementDevelop protocol and make criteria transparentIt’s the accumulation of effects that mattersSimilar to primary studies, problems need to be correctedCan use qualitative data to supplementThey also suffer from all of the above issuesBoth narrative reviews and MAs should combine practical expertise and quantitative results in a way that enhances both.

0

Embed

Share

Upload

Make amazing presentation for free
An Intervention to Improve HIV Medication Adherence among ...