BIAS ERROR

BIAS ERROR

1. Selection

   • Referral (Volunteer)
     • difference between those that volunteered vs. the general population
       • Healthier
       • Higher Risk
       • More likely to comply
   • Length-Biased (Prognosis Selection)
     • Does the screening selectively identify cases with better prognosis? (Longer they are in pre-clinical, the more likely we can detect in preclinical)
       • Fix with INTERVAL cases (compare screened in pre-clinical with unscreened in pre-clinical)
   • Overdiagnosis
     • Detect preclinical diseases that do not progress or regress (maybe too slow to produce disease before death)
       • Similar to Length Bias
       • Standardize screened and unscreened
       • Creates false impression of increased rates of detection and diagnosis
   • respondents and non-respondents are systematically different
   • Exclusion
     • exclusion criteria for controls may differ from cases (need to have consistency with criteria amongst the groups)

2. Information

   • Lead Time
     • When comparing survival, screening may have just identified early but not decrease risk of dying (increase interval from death to diagnosis)
       • Use measures of morbidity, quality of life measures etc.
       • Maybe less prone if we look at a mortality rates rather than survival times
       • estimate lead time to make allowance
   • wording, who asked, mode, time of interview
   • Misclassification
     • Nondifferential
     • Differential
     • Cases may be classified as controls and vice versa
   • Recall
     • Cases recall better
   • Surveillance
   • Reporting

3. ERROR

   • Sampling (Random) ERROR
     • Choice of Study Subjects
     • Direction that it changes parameters is unknown
     • Small sample size

4. Confounding (Analytic Studies Only)

   • Associated with the outcome, cause, and is NOT a step on the causal pathway

5. Any systematic difference between the control and the cases group