1 / 26

Choosing study subjects

Choosing study subjects. Prof. Rodney Ehrlich School of Public Health and Family Medicine. Learning Objectives. 1. To be able to define the population to which you seek to generalise your results 2. To be able to identify your sampling frame (or recruitment pool)

keldridge
Download Presentation

Choosing study subjects

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Choosing study subjects Prof. Rodney Ehrlich School of Public Health and Family Medicine TRRM - 2005/6

  2. Learning Objectives 1. To be able to define the population to which you seek to generalise your results 2. To be able to identify your sampling frame (or recruitment pool) 3. To be able to choose a sampling strategy (or randomisation strategy) 4. To understand the distinction between random sampling error and systematic error (bias) and how to minimise these TRRM - 2005/6

  3. Population • Statistical term – any coherent group to which you wish to generalise your results • Typically share some characteristic(s) of common interest: • Geography (e.g. Cape Town) • Health relevant characteristic (“Tik” users) • May be defined by health services use fundamental distinction between “patient” and “population” based studies. TRRM - 2005/6

  4. What are your patients representative of? • Compared to population at large, i.e. with respect to population level characteristics? • Compared to patients with same condition or diagnosis who are not hospitalised (or who do not attend that health service).

  5. Selecting the population you want to study Inclusion and exclusion criteria: Demographic: age, gender, race, socioeconomic status Clinical: risk factor or outcome of interest, stable, no aggravating factors, contraindications, etc. Administrative: availability vs difficulty; Ethical: ability to consent, record access, etc. Always trade off between efficiency and generalisability TRRM - 2005/6

  6. Population measures, not health service measures. “20% of patients attending Dermatology OPD have eczema” = proportion (not prevalence) “10% of admissions to ICU were for acute MI” = proportion (not incidence). Digression on terminology: prevalence and incidence in hospital patients TRRM - 2005/6

  7. Why sample? • Can’t afford to study whole population • No need to study whole population, as sampling is an efficient (i.e. lower cost) way to get the same information TRRM - 2005/6

  8. Which sampling strategy? • Depends on: • study design (in turn dependent on the question) • availability of subjects/information on subjects • resources TRRM - 2005/6

  9. Different types of sampling • Convenience (or availability) sampling • Probability sampling TRRM - 2005/6

  10. Convenience sampling • “Haphazard” – no logic other than availability of subjects • Volunteer • Consecutive TRRM - 2005/6

  11. Probability sampling • Simple random • Stratified random • Systematic random • Cluster random • Important when want truly representative estimates of population  mainly in descriptive studies or surveys • Basis of statistical procedures TRRM - 2005/6

  12. Simple random sampling • Starts with sampling frame (e.g. list, archive) representing population • “Random” – precise statistical term: each subject has known chance of being selected • Use table or programme or other procedure to generate random numbers TRRM - 2005/6

  13. Stratified random sampling • Divide sampling frame into strata (i.e. subjects in strata share common characteristic) • Random sampling within each stratum • Can oversample some strata (remember to reweight if want an estimate for the whole) TRRM - 2005/6

  14. Systematic random sampling Sampling frame is “consecutive” (e.g. records in a filing system, or patients arriving) Use random number start Then choose every nth subject, where n= sampling interval (e.g. every 10th subject), until sample size is reached TRRM - 2005/6

  15. Cluster random sampling • Mainly for population/community surveys • Take advantage of naturally occurring administrative units (e.g. community health centres, schools, blocks of houses). Each unit is a cluster • Can also create clusters, as in Extended Programme on Immunisation (EPI) • Have to take “cluster effect” into account in the sample size calculation and the analysis Cluster sampling: A sampling method in which each unit selected is a group of persons rather than an individual - Last TRRM - 2005/6

  16. Sampling error • If random with respect to outcome of interest = random error • If systematic or biased with respect to measurement interest = systematic error or bias TRRM - 2005/6

  17. Random sampling error • As likely to be in one direction as another • Expected value (e.g. mean) is unbiased •  wider confidence interval or larger p-value (i.e. greater uncertainty) • Control of random error: increase sample size TRRM - 2005/6

  18. Systematic sampling error / bias • In one direction more than another • Expected value (e.g. mean or %) is biased • Minimise systematic error by improving selection process and measurement instruments TRRM - 2005/6

  19. Types of systematic error / bias • Selection bias - arises before data collection • Measurement bias – arises during data collection TRRM - 2005/6

  20. Randomisation • Form of assignment, not sampling • NOT the same as random sampling • Purpose is to distribute potential confounders equally between groups TRRM - 2005/6

  21. Exercise 1 • Discuss how patients at the UCT linked hospitals/facilities may differ from: a. the general population of the Western Cape; b. patients from Cape Town with the same diagnosis and condition not attending these facilities. TRRM - 2005/6

  22. Exercise 2 • You want to do a case-control study of the possible causes of childhood leukaemia . Discuss: • a. how you would select your cases; • b. how you would select controls (without leukaemia). TRRM - 2005/6

  23. Exercise 3 You do a telephone survey of patients waiting for elective orthopaedic surgery. You are aware of selection bias arising from the fact that some patients do not have telephone access. Also, the response rate (those available and willing to be interviewed) among those that do have telephone numbers is 65%. a. Would increasing the sample size help to minimise the potential selection biases? b. What other ways could help to reduce these biases? TRRM - 2005/6

  24. Exercise 4 You are planning a study of health professionals’ attitudes to performing elective terminations of pregnancy within the provincial administration health services. Devise a sampling strategy. TRRM - 2005/6

  25. Exercise 5 You are planning to conduct a randomised controlled trial of a new oral anticoagulant for chronic atrial fibrillation. a. How would you recruit patients? Is sampling a consideration here? b. To whom would you like to generalise your findings? c. What would your inclusion and exclusion criteria be? TRRM - 2005/6

  26. Exercise 6 You are planning to test the performance of a new rapid screening test for prostatic specific antigen against the standard laboratory test. a. How would you recruit patients? Is sampling a consideration here? b. To whom would you like to generalise your findings? c. What would your inclusion and exclusion criteria be? TRRM - 2005/6

More Related