Professional Documents
Culture Documents
Teerapon Dhippayom
PharmD, MClinPharm, PhD
Scope
Application of pharmacoepidemiology Validity of research finding Bias and confounding Issues to be considered in pharmacoepidemiological study
Application of Pharmacoepidemiology
To promote the rational use of medicine To investigate the safety of medicine To investigate the effectiveness of medicine
Bias
Bias in research process Literature review and variable selection Sample selection Intervention or treatment Exposure and outcome measurement Data analysis Results interpretation Paper publication
Bias
Selection bias The relation between exposure and outcome is different for those who participate and those who do not participate Information bias A distortion in measuring exposure or outcome data that results in different quality or frequency of information between comparison groups
Bias
Bias
Selection bias
Sampling bias
Information bias
Measurement bias Recall bias
Confounding
Confounding A situation in which a measure of the effect of an exposure on risk is distorted because of the exposure with other factor(s) that influence the outcome under study
Confounding
Aspirin
GI bleeding
NSAIDs
Confounding control
Confounding control
Study design
Restriction Matching
Data analysis
Stratification Multivariate analysis
Issues to be considered
What to look for in observational studies Is selection bias present? Is information bias present? Is confounding present? If none of the above is presented, could they be the results of chance If all errors are excluded, the results might be real and worthy of note
Issues to be considered
Cohort study Bias and confounding Measurement and outcome Follow-up time period
Issues to be considered
Case-control study Selection bias Case selection Control selection Sources of control group Measurement and outcome
Issues to be considered
Case selection Well defined (inclusion criteria) Represent population at risk Incident case Control selection Risk of developing outcome case Chance of exposing to risk factor case
Issues to be considered
Sources of control group Community based Reduce referral bias High cost and low compliance Ideal, generalization Hospital based Convenience and low cost May not represent the population at risk
Issues to be considered
Measurement and outcome Recall bias Interviewer bias Case and control group must be assessed for exposure in the same way
Application: an example
Abstract
Background
Several studies suggested the beneficial effects of selected CHF drugs Older patients were generally not included in HF trials Older patients may not received appropriate HF drugs like younger patients
Objectives
Compare outpatient CHF drug utilisation and hospital re-admission patterns in patients 75+yrs with those age <75 yrs Evaluate the potential benefits of HF medications
Methods
Design Nested case-control Setting A French teaching hospital Study period 12-month period (in 2000)
Methods
Study population All adults admitted to eight departments Diagnosed as HF at discharge Data sources Standardised form? Claims for reimbursement under French health insurance system
Methods
Outcome measured CHF drug utilisation patterns
Medication prescribed at discharged Medication prescribed during follow-up
Results
Patient characteristics
Results
Exercise
Odds ratio: calculation
Cases Noncases
Exposed
Not exposed
OR =
AD BC
Exercise
Exposure to risk factor
Yes No Time
Disease conditions
Sample with disease (cases)
Yes No
Exercise
What is case HF patients age 75+ years What is control HF patients age 75 years or younger What is exposing factors HF medications
Exercise
Calculate OR for prescribing of ACE inhibitors
> 75 years < 75 years
ACE inhibitors
48
63
No ACE inhibitors
102
68
OR =
48 x 68 63 x 102
= 0.51
Results
Results
Issues considered
Study design: Nested case control Begin with a record of factors interested Follow-up a group of people (cohort) Identify case and control from a cohort Look back for the expose of risk factors in each group
Issues considered
Selection bias A French teaching hospital Loss of data during follow-up Only 50% have their EF measured Any different among group?
Number and type of concomitant medication Appropriate dose among group were not measured
Issues considered
Information bias Appropriateness of diagnosis
Data from 8 different departments Criteria for discharge diagnosis was unknown
The completeness of the information recorded It is not cleared if the re-hospitalisation was for HF related
Issues considered
Confounding bias Confounding variables were adjusted Other potential confounder were not addressed
Post-MI Stroke Cardiac surgery etc
Issues considered
Other issues Hospitalisation among younger patients appeared not different The difference between two age group not prescribed HF drugs were obvious, but yet not statistically sig. Other factors may contribute to the longer duration in older patient not prescribed the drug
Conclusion
The descriptive results on utilisation pattern in this study maybe useful The implication on analytical part is questionable It is crucial to evaluate the method used before utilising the outcomes of a study
Questions