Quasiexperimental study designs in evaluating medicines use interventions

1
(No Transcript)
2
Quasi-experimental study designs in evaluating
medicines use interventions
3

• Lloyd Matowe 1 Craig Ramsay 2
• 1 Faculty of Pharmacy, Kuwait University
• 2 HSRU, University of Aberdeen

4
BACKGROUND

• The RCT is seen as the gold standard methodology
  for evaluating medicine use interventions
• There are situations where it is impractical to
  carry out RCTs
• For example in situations where
• It is impractical to randomize subjects
• Difficult to find an appropriate control
• Intervention has been retrospectively implemented
• Quasi-experiments are often used as an
  alternative

5
Simple before and after studies
Time
6
Interrupted time series analysis
Time
7
Controlled before and after design
Before
After
8

• There is suspicion that quasi-experiments are
  often conducted and analysed inappropriately

9
OBJECTIVE

• The objective of this study is to evaluate the
  use of quasi-experimental designs in medicines
  use interventions

10
METHODS

• A rigorous systematic review
• Subset of a large study on the effectiveness of
  various policy implementation strategies
• Search strategies
• EPOC, MEDLINE, EMBASE, HEALTHSTAR, SIGLE
• Bibliographies of previously published reviews
• Studies included CBAs ITS
• Interventions policies/ guidelines to improve
  medication use
• Outcomes- Objective measures of provider
  behaviour and/ or patient outcome
• Time series regression used to reanalyse serial
  data where reanalysis was necessary

11
RESULTS - 1

• Search strategy identified 150 000 hits
• 5 000 hits identified as potentially relevant
• 285 studies included
• 44 studies were on medicines use
• 17 were CCTs or RCTs
• 27 studies were quasi-experiments
• 15 ITS
• 12 CBAs
• 24 studies were on multiple interventions
• Study areas included Antibiotics,
  antipsychotics, analgesics, GIT drugs and
  anti-hypertensives

12
RESULTS - 2

• Time series
• Overall methodological quality poor
• 8 of 15 studies analysed inappropriately
• Effect size different in all 8 studies
• Common errors included
• Failure to test for autocorrelation and trend
• Analysis using non time series methods
• Failure to ascertain intervention as independent
  from other interventions

13
Trends
Time
14
RESULTS-3

• Controlled before and after studies
• Overall methodological quality poor
• 7 of 12 had unit of analysis errors
• Other common errors included
• Failure to protect against contamination
• Failure to take baseline measurements
• Failure to perform power calculation

15
Controlled before and after design
Before
After
16
SUMMARY

• Quasi-experiments are often designed analysed
  inappropriately
• Methods and design are often inadequately
  reported
• Study methodological weaknesses are often
  inadequately reported

17
RECOMMENDATIONS

• For ITS studies we recommend
• Use of time series regression
• Pre-testing for autocorrelation and trend
• Ascertain that the intended intervention was
  independent of other interventions
• For CBA we recommend
• Protection against contamination
• Measurement of baseline characteristics
• Power calculation
• For both ITS and CBAs methodological weaknesses
  should be reported

18
ACKNOWLEDGEMENTS

• I would like to thank Professor Jeremy Grimshaw,
  from Ottawa Health Research Institute and Center
  for Best Practices, University of Ottawa, Canada
  for giving me access to the studies used for this
  evaluation