Docstoc

Sampling

Document Sample
Sampling Powered By Docstoc
					           Bias
EPIET Introductory Course, 2011
   Lazareto, Menorca, Spain



             Update: S. Bracebridge
             Sources: T. Grein, M. Valenciano, A. Bosman
        Objective of this session


• Define bias

• Present types of bias

• How bias influences estimates

• Identify methods to prevent bias
         Epidemiologic Study


An attempt to obtain an epidemiologic
 measure
• An estimate of the truth
         Definition of bias

  Any systematic error in the design or
   conduct of an epidemiological study
resulting in a conclusion which is different
                from the truth


  an incorrect estimate of association
  between exposure and risk of disease
Main sources of bias


1. Selection bias

2. Information bias

3. [Confounding]
Should I believe the estimated effect?


Mayonnaise                   Salmonella
              RR = 4.3




 True association?       Bias?
                         Chance?
                         Confounding?
                   Warning!

• Chance and confounding can be evaluated
  quantitatively

• Bias is much more difficult to evaluate
  - Minimise by design and conduct of
    study
  - Increased sample size will not eliminate
    bias
              1. Selection bias


• Due to errors in study population selection

• Two main reasons:

  - Selection of study subjects

  - Factors affecting study participation
             Selection bias

• At inclusion in the study

• Preferential selection of subjects
 related to their

  - Exposure status (case control)

  - Disease status (cohort)
     Types of selection bias

• Sampling bias

• Ascertainment bias
  - surveillance
  - referral, admission
  - diagnostic

• Participation bias
  - self-selection (volunteerism)
  - non-response, refusal
  - survival
  Design Issues

Case-control studies
                    Selection of controls
 Estimate association of alcohol intake and cirrhosis
                            Cases          Controls A
                        liver cirrhosis   trauma ward


Heavy alcohol use            80               40
                                                        OR = 6


Light/no alcohol use         20               60


   How representative are hospitalised trauma patients
     of the population which gave rise to the cases?
               Selection of controls

                            Cases         Controls A   Controls B
                       liver cirrhosis   trauma ward   non-trauma


Heavy alcohol use            80              40            10


Light/no alcohol use         20              60            90


                                          OR = 6          OR = 36

Higher proportion of controls drinking alcohol in
      trauma ward than non-trauma ward
                                                                a b
                                                                    c d
        Some worked examples

• Work in pairs

• In 2 minutes:
  - Identify the reason for bias
  - How will it effect your study estimate?
  - Discuss strategies to minimise the bias
  Oral contraceptive and uterine cancer
You are aware OC use can cause breakthrough bleeding

                               Cases
                                            Controls
                           uterine cancer

      Takes oral
                                 a             b
      contraceptives

      Does not take oral
                                 c             d
      contraceptives


        • OC use  breakthrough bleeding  increased
          chance of testing & detecting uterine cancer

a b     • Overestimation of “a”  overestimation of OR
        • Diagnostic bias
c d
        Asbestos and lung cancer
  Prof. “Pulmo”, head specialist respiratory referral
 unit, has 145 publications on asbestos/lung cancer

                       Cases admitted
                                          Controls from
                        and diagnosed
                                          surgical wards
                       with lung cancer
    Contact with
                              a                 b
    asbestos

    No contact with
                              c                 d
    asbestos


      • Lung cancer cases exposed to asbestos not
        representative of lung cancer cases
a b
c d   • Overestimation of “a”  overestimation of OR
      • Admission bias
Selection Bias in
 Cohort Studies
            Healthy worker effect
Association between occupational exposure X and
                   disease Y
                  Exposed    General
                  workers   population

   Deaths             50       7,000
   Person-time
                    1,000     100,000
   in years

   Mortality
                     0.05       0.07     RR=0.7
   (cases/year)
             Healthy worker effect

                                 General population

              Exposed                  Non-
                        Workers                       Total
              workers                 workers

Deaths           50      4,500         2,500          7,000
Person-
               1,000    90,000        10,000     100,000
time

Mortality
                0.05      0.05          0.25           0.07
(cases/yr)
  Prospective cohort study- Year 1

                           lung cancer
                          yes       no



             Smoker       90      910    1000



             Non-smoker   10      990    1000



      90     10
RR              9
     1000   1000
       Loss to follow up – Year 2

                               lung cancer
                              yes       no



              Smoker           45       910   955



              Non-smoker       10       990   1000



      45    10
RR             4.7
     955   1000            50% of cases that smoked
                               lost to follow up
       Minimising selection bias

• Clear definition of study population

• Explicit case, control and exposure
  definitions

• Cases and controls from same population
  - Selection independent of exposure

• Selection of exposed and non-exposed
  without knowing disease status
       Sources of bias



1. Selection bias

2. Information bias
             Information bias

• During data collection

• Differences in measurement

  - of exposure data between cases and controls

  - of outcome data between exposed and unexposed
            Information bias

• 3 main types:
  - Reporting bias
    • Recall bias
    • Prevarication

  - Observer bias
    • Interviewer bias

  - Misclassification
                 Recall bias
Cases remember exposure differently than controls
           e.g. risk of malformation
                             Mothers of

                   Children with
                                      Controls
                   malformation

Took tobacco,
                        a                 b
alcohol, drugs


Did not take            c                 d


• Mothers of children with malformations
  remember past exposures better than mothers
  with healthy children
  • Overestimation of “a”  overestimation of OR
                   Prevarication bias
Exposure reported differently in cases than controls
               e.g. isolation and heat related death
                                  Relatives

                        Elderly dead          Controls


    Isolated                 a                   b


    Not isolated             c                   d




        • Relatives of dead elderly may deny isolation

         • Underestimation “a”  underestimation of OR
              Interviewer bias
Investigator asks cases and controls differently
                about exposure
           e.g: soft cheese and listeriosis
                     Cases of
                                  Controls
                    listeriosis

 Eats soft cheese       a            b

  Does not eat
                        c            d
  soft cheese



      • Investigator may probe listeriosis cases about
        consumption of soft cheese (knows hypothesis)
       • Overestimation of “a”  overestimation of OR
      Misclassification
        Measurement error leads to assigning
         wrong exposure or outcome category
     Non-differential                   Differential
• Random error                 • Systematic error
• Missclassifcation exposure   • Missclassification exposure
       EQUAL                          DIFFERS
between cases and controls     between cases and controls
• Missclassification outcome   • Missclassification outcome
       EQUAL                          DIFFERS
between exposed & nonexp.      between exposed & nonexposed
   => Weakens measure            => Measure association
      of association                distorted in any direction
   Nondifferential misclassification
True Classification
                      Cases           Controls          Total
Exposed                100               50              150
Nonexposed             50                50              100
                       150               100             250

     OR = ad/bc = 2.0; RR = a/(a+b)/c/(c+d) = 1.3

Nondifferential misclassification - Overestimate
exposure in 10 cases, 10 controls – bias towards null
                      Cases           Controls          Total
Exposed                110               60             170
Nonexposed              40               40              80
                       150              100             250


      OR = ad/bc = 1.8; RR = a/(a+b)/c/(c+d) = 1.3
        Differential misclassification
True Classification
                      Cases          Controls             Total
Exposed                100             50                 150
Nonexposed             50              50                 100
                       150             100                250

    OR = ad/bc = 2.0; RR = a/(a+b)/c/(c+d) = 1.3


Differential misclassification - Underestimate exposure
for 10 controls

                      Cases         Controls          Total
Exposed                100             40                 140
Nonexposed              50             60                 110
                       150            100                 250

      OR = ad/bc = 3.0; RR = a/(a+b)/c/(c+d) = 1.6
       Differential misclassification
True Classification
                      Cases         Controls          Total
Exposed                100             50                 150
Nonexposed             50              50                 100
                       150             100                250

    OR = ad/bc = 2.0; RR = a/(a+b)/c/(c+d) = 1.3


Differential misclassification - Underestimate exposure
for 10 cases

                      Cases         Controls          Total
Exposed                 90             50                 140
Nonexposed              60             50                 110
                       150            100                 250

     OR = ad/bc = 1.5; RR = a/(a+b)/c/(c+d) = 1.2
   Minimising information bias

• Standardise measurement instruments
  - questionnaires + train staff

• Administer instruments equally to

  - cases and controls
  - exposed / unexposed

• Use multiple sources of information
      Summary: Controls for Bias

• Choose study design to minimize the
  chance for bias
• Clear case and exposure definitions
  - Define clear categories within groups (eg age
    groups)

• Set up strict guidelines for data collection
  - Train interviewers
     Summary: Controls for Bias

• Direct measurement
  - registries

  - case records

• Optimise questionnaire

• Minimize loss to follow-up
             Questionnaire

• Favour closed, precise questions
• Seek information on hypothesis
  through different questions
• Field test and refine
• Standardise interviewers’ technique
  through training with questionnaire
      The epidemiologist’s role

1. Reduce error in your study design

2. Interpret studies with open eyes:

  • Be aware of sources of study error

  • Question whether they have been

    addressed
       Bias: the take home message

 • Should be prevented !!!!
    - At PROTOCOL stage
    - Difficult to correct for bias at analysis stage

• If bias is present:
   Incorrect measure of true association
   Should be taken into account in interpretation of
  results
  •Magnitude = overestimation? underestimation?
        Objective of this session


• Define bias

• Present types of bias

• How bias influences estimates

• Identify methods to prevent bias
                  References

Rothman KJ; Epidemiology: an introduction.
Oxford University Press 2002, 94-101


Hennekens CH, Buring JE; Epidemiology in
Medicine. Lippincott-Raven Publishers 1987, 272-
285

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:0
posted:3/20/2012
language:
pages:40
yaohongm yaohongm http://
About