lecture 4, research design (slides)

Upload: justden09

Post on 07-Apr-2018

224 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 lecture 4, Research Design (slides)

    1/25

    Designing a study I

    Research Methods

    Dent 313

  • 8/3/2019 lecture 4, Research Design (slides)

    2/25

    Choosing a research question First step of research design

    Identifying a research question

    Determine the unknowns in your field

    What do you wish you knew in your field? What does the available literature lack? Your capacity and experience

    Research question Descriptive Analytic

  • 8/3/2019 lecture 4, Research Design (slides)

    3/25

    Descriptive research question Explaining clinical phenomena

    Prevalence of disease E.g. (Prevalence of caries among school children)

    Survival trends E.g. (% of men w prostate cancer alive at 5 years)

    Health service utilization E.g. (% seniors receiving H1N1 vaccination)

    Clinical test characteristics E.g. (mean value of LDL among patients w IHD)

  • 8/3/2019 lecture 4, Research Design (slides)

    4/25

    Analytic research question

    Comparative

    E.g. (Is prevalence of caries higher among private orgovernmental school children?)

    More significant than descriptive questions Answering enable to

    develop intervention to prevent disease Target intervention to particular population

    Descriptive questions must be answered first

  • 8/3/2019 lecture 4, Research Design (slides)

    5/25

    Choosing a research questions

    Specify the population

    Determine the length of the study and your

    willingness towards completion Median time from enrolling subjects to

    publication was found to be 5.5 (J.P. Ioanndis)

  • 8/3/2019 lecture 4, Research Design (slides)

    6/25

    Choosing a research questions

    1st criterion

    Choose a question that keeps youexcited all the way through

    Identify obstacles to performing andpublishing research Subjects, ethical approval, collaborators, lab

    capacity, review committees, editors, missing data

  • 8/3/2019 lecture 4, Research Design (slides)

    7/25

    Choosing a research question

    2nd criterion

    Choose a question that will have animpact on the health and well being ofthe population

    Difficult to fully appreciate the impact a studywill have before doing it Modifications, groundbreaking results,

    discoveriesetc

    Purpose of clinical research is to improvehealth not for grantsmanship, publication andpromotion

    Enrollment of sufficient number of subjects

  • 8/3/2019 lecture 4, Research Design (slides)

    8/25

    Choosing a research questions

    3rd criterion

    Consider what questions you are readyto answer based on

    The prevalence of the disease in your area Your prior experience Your colleagues Your community contacts

  • 8/3/2019 lecture 4, Research Design (slides)

    9/25

    Choosing a research questions

    4th criterion

    Be sure it has not already beenanswered unless if you can do better

    Computerized literature searches Consulting the others in the field Attending conferences

    Scientific / academic relations

    Not all conference abstracts are electronicallyaccessible

  • 8/3/2019 lecture 4, Research Design (slides)

    10/25

    Choosing a study design

    There is no best study design

    Determine the best study design to

    answer your question Feasibility, cost, length of time, risks and

    benefits on participants

    Broad categories of study design Randomized studies Observational studies

  • 8/3/2019 lecture 4, Research Design (slides)

    11/25

    Differences between randomized and

    observational studies

    Randomized Investigator manipulates the condition or

    group assignment

    One group receives a treatment Another receives a different treatment or placebo

    Observational Investigator assesses a population without

    altering the condition or group assignment ofthe population

  • 8/3/2019 lecture 4, Research Design (slides)

    12/25

    Advantages & disadvantages of

    randomized studies

    Better at dealing with confounding andbias

    Less generalizability Slower to conduct

    More expensive

    Cannot answer as broad a range ofquestions as observational studies

  • 8/3/2019 lecture 4, Research Design (slides)

    13/25

    Confounding

    Association between a risk factor and anoutcome is affected by the relationship of

    a third variable (confounder) to the riskfactor or the outcome

    Confounder

    Associated with the risk factor

    Causally related to the outcome

  • 8/3/2019 lecture 4, Research Design (slides)

    14/25

    Confounding

    Risk factor

    Confounder

    Outcome

    Randomized groupassignment

    Potentialconfounder

    Outcome

    With randomization there should be no relationship confounder and group assignment

  • 8/3/2019 lecture 4, Research Design (slides)

    15/25

    Eliminating confounding

    Randomizing subjects

    Randomization should be unbiased

    Randomized groups will be equal with respectto confounders

    Randomization eliminates known and unknownconfounders

    Other techniques for minimizing knownconfounders Matching, stratification, multivariable adjustment

  • 8/3/2019 lecture 4, Research Design (slides)

    16/25

    Minimizing bias

    Randomization eliminates confoundingonly when unbiased

    Group assignment should be done at the time of enrollment not before or after

    No group assignment change by personnel by someone with no contact with the subject using a random number table or generator

  • 8/3/2019 lecture 4, Research Design (slides)

    17/25

    Minimizing bias

    In observational studies, investigatorsand subjects usually know which group

    the subject / examiner were assigned to Investigators may have some expectation onwhat the outcome would be (possibility ofbias)

    Subject may leave the study if they know theybelong to control (not treatment) group

  • 8/3/2019 lecture 4, Research Design (slides)

    18/25

    Minimizing bias

    Blinding in randomized trial

    Preventing both the investigator and thesubject from knowing group assignment

    Double blinded vs. single blinded Double blinding is impossible with

    observational studies

  • 8/3/2019 lecture 4, Research Design (slides)

    19/25

    Generalizability

    Generalizability refers to the ability toapply the outcomes of a study to

    populations other than the study sample The results of a trial apply only to

    populations that resemble the studysample

  • 8/3/2019 lecture 4, Research Design (slides)

    20/25

    Generalizability and randomized

    studies

    Generalizability is more a problem with randomizedstudies

    Randomized subjects are different from the generalpopulation because they carry more burdens Selection, blinding, previous exams and blood testsetc.

    Trial conditions are different from those in clinical practice Trial subjects receive more attention

    Treatment works under tight research protocol Treatment efficacy vs. treatment effectiveness Efficacy: how well an intervention works in a research setting Effectiveness: how well an intervention works in a clinical setting

  • 8/3/2019 lecture 4, Research Design (slides)

    21/25

    Observational studies and

    treatment effectiveness

    Closer together than in randomizedstudies

    Still some differences Observational study patients receive

    additional educational or testing than normalclinical patients

    Observational study patients may changetheir behavior (Hawthrone effect)

  • 8/3/2019 lecture 4, Research Design (slides)

    22/25

    Length of time to conduct

    Observational studies are faster toconduct than randomized studies

    Especially with an existing database and theuse of case-control design

  • 8/3/2019 lecture 4, Research Design (slides)

    23/25

    Minimizing expenses

    Observational studies are less expensive Especially with an existing database and the use of

    case-control design

    Less markedly with prospectivecohort design Observers are just observing the outcomes

    Randomized control trials are expensive

    Paying for all of the interventions e.g. medicines, tests, appliances..etc.

  • 8/3/2019 lecture 4, Research Design (slides)

    24/25

    Addressing a broader range of

    questions

    Observational studies are able to answera broader range of questions

    Many situations where it is unethical orimpractical to randomize subjects (you cannotrandomize persons to smoke or not to smoke)

    Randomized control studies are rarely helpful

    in identifying causes of disease outbreaks

  • 8/3/2019 lecture 4, Research Design (slides)

    25/25

    Indications of observational

    studies

    Instances where it is unethical orimpractical to perform a randomized

    study Time is of the essence in obtaining the

    results