S - Lecture 2 - Basics of Research Design.pdf

§ Identify the basic terminologies and language used in research. § Describe the general principles of research designs used to gather and provide drug information. § Differentiate the kinds of research designs used in healthcare research, including their advantages and disadvantages. § Identif...

§ Identify the basic terminologies and language used in research. § Describe the general principles of research designs used to gather and provide drug information. § Differentiate the kinds of research designs used in healthcare research, including their advantages and disadvantages. § Identify and describe the role of bias in research design. § VARIABLES § Independent variable § Predictors / antecedent events / treatments / interventions – selected by the experimenter § Dependent variable § Outcome variable /Results § Objective of most research is to know whether the independent variable is causing the outcome § Extraneous variables § Controlled to minimize their influence on the association under study § Research hypothesis § Guides the study § A proposed relationship between the independent and the dependent variables and whose veracity is tested during the experiment. § INDEPENDENT VARIABLE (IV) § Manipulated § Explicitly and precisely defined or controlled § Often examine > 1 level § E.g., the dose of a drug § Effects of several IVs may be examined in a single research study. § Multivariate regression (recall Biostatistics) § Examples § ? § DEPENDENT VARIABLE (DV) § Reflects objective and measurable effects of the independent variable § Primary outcomes § Findings the investigators are most interested in examining because they provide the most pertinent evidence § Secondary outcomes § Research results of less interest that might provide supporting evidence. § Examples § EXTRANEOUS VARIABLES § Extraneous variables § Also known as confounding variables § Can affect the relationship between the independent variable and the dependent variable. § Extraneous variables are controlled so that their effects are minimized. § For example, differences in study subjects may be reduced by matching individuals on known factors such as age and gender or by assigning subjects to experimental conditions randomly. § Other examples - ? § CONTROL GROUP § To control the effect of extraneous variables § Members of a control group = experimental group in terms of extraneous variables. § For example, both groups have same age (if age is an extraneous variable in the study) § Difference between Experimental group and Control group: § Same attention / procedure / etc. as the treatment/experimental group, EXCEPT for the Treatment. § Descriptive § Focus on novel or unusual signs, symptoms, or events § Observational § Involve examining participants expressly because they exhibit either a specific characteristic or a specific outcome § Experimental § Involve the random assignment of participants to experimental conditions § DESCRIPTIVE § Case Study § A description of an individual with a novel or unusual condition § Case Series § A summary of the health status of several individuals who all display a similar novel clinical finding or condition § The main purpose of these communications is to notify the healthcare community of an unusual clinical event, e.g., adverse events § DESCRIPTIVE § Do not provide any inferential (i.e., cause and effect relationship) evidence § Do not have any control group § Do not account for any confounding or extraneous variables. § Descriptive statistics are only used for descriptive research § Inferential statistics are not used in descriptive research  TABLE : Strengths and Weaknesses of Case Reports and Case Series Design Strengths Weaknesses Case Reports Identifies rare occurrences No causal inference can be made Identifies delayed ADRs Potential reporting bias Hypothesis generation No statistical analysis Requires minimal resources Potential for reporting false results Case Series Study results are closer to those of No causal inference can be made routine clinical practice May be useful when a randomized Susceptible to selection and controlled trial is challenging to measurement bias conduct High external validity An absolute risk cannot be calculated Cost-effective and time saving Data collection may be design incomplete § OBSERVATIONAL § 2 most important types: - Cohort Study - Case control study § Both of these studies can evaluate exposure – outcome causal relationship if properly done. § You might find some books write that case control studies can not evaluate causal relationship, but cohort study can do it. – This is not a fair statement § There can be many different types of biases present in these two study designs. § These study designs do not use randomization § CASE-CONTROL STUDY § Cases are matched to controls § Controls should be selected from same population that gives rise to cases § Example of Sources of controls : Neighborhood controls, Hospital or clinic based controls, Friend controls, Family controls (e.g., twins) § Matching cases and controls is done to make sure that the two groups differ only with respect to the suspected exposure variable, thus excluding the roles of confounding and nuisance variables in the outcome. § Ideally cases and control should be identical in all aspects (including risk of exposure) except that cases have the outcome and controls do not have the outcome § In a case-control study, you should report the effect of exposure on outcome in terms of Odds Ratio Strengths Efficient design for rare outcomes, long exposure outcome intervals Comparatively inexpensive and easy to conduct Can be completed quickly Weaknesses Susceptible to many different forms of bias It is difficult to match cases and controls on all confounding variables except exposure May be difficult to determine time relationship between exposure and outcome At best, evidence indicates an association between exposure and outcome § COHORT STUDY § Long period of observation § Participants do not have the outcome of interest to begin with § Selected based on exposure status of the individual § Prospective or retrospective § Examples = Framingham study & Swiss HIV study § In case of a cohort study, you should report the effect of exposure on outcome in terms of Relative risk Strengths Can be used to study harmful exposures Can establish incidence (i.e., frequency of cases) of outcome Exposure effects can be examined retrospectively or prospectively Best with rare exposures Weaknesses Susceptible to confounding Groups cannot be matched on all variables except exposure Inefficient for long exposure–outcome intervals May be expensive Requires more patients than a case-control study § SELECTION BIAS § This kind of bias arises from procedures used to select subjects and from factors that influence study participation. § In presence of selection bias, the relation between exposure and outcome is different for those who participate in a study and for all of those who should have been eligible to participate in the study, including those who actually did not participate § True estimate of exposure –outcome effect is dependent on who is participating in the study and who is not participating. § Self-selection: Sometimes subjects refer themselves to the study. This might create a problem if the reasons for self-referrals are associated with outcomes of the study. § INFORMATION BIAS § Once subjects enter into study, information is collected from them § Bias in estimating an effect can be caused by measurement errors in collecting information, hence it is called Information Bias. § For discrete (categorical) variables, this type of measurement error can lead to Classification error or Misclassification § E.g., Male is misclassified as Female § Misclassification happens a lot in Smoking related studies, where smokers are classified as heavy smoker, moderate smoker, non-smoker § RECALL BIAS § This refers to the finding that, because of different life experiences, cases and controls may have different recollections of events. § Example: case-control studies of birth defects. It often happens that a couple with unfortunate incident of having a baby with birth defect are motivated (by guilt or concern) to report more of fewer adverse exposures than a couple with normal infant. § Recall bias is a possibility in any case-control study that relies on subjects’ memory for getting history of exposures. § How Biases can impact study result(s)? § By increasing or decreasing the true measure of the effect of exposure on outcome. § It is difficult to predict the direction of the effect in presence of bias, hence studies should be designed very carefully to take care of those bias. § Prospective § A study which looks forward, looks to the future, follows a condition or disease into the future § Retrospective § Study design that looks back in time to study events that already occurred § Case series, Case study, Case-control study all are retrospective in nature § Cohort study can be retrospective or prospective § EXPERIMENTAL DESIGN § Researcher/ investigator controls the application of the exposure (e.g., drugs) among two or more groups of individuals to assess its effect on outcome variables § These are always prospective in design § Randomized controlled trial is the classic example  Data Methods Classification and Sources for Clinical Research Citation: Chapter 3 Research Design and Methods, Aparasu RR, Bentley JP. Principles of Research Design and Drug Literature Evaluation, 2e; 2020. Available at: https://accesspharmacy.mhmedical.com/content.aspx?bookid=2733&sectionid=226710702 Accessed: August 13, 2021 Copyright © 2021 McGraw-Hill Education. All rights reserved § Reliability § How consistently a method measures something § Example - ? § Validity § Internal validity § Can be threatened by bias & confounders § External validity / generalizability VALIDITY TYPES § Face validity § Whether a test appears to measure what it’s supposed to measure (is it relevant and appropriate for what it’s assessing?) § Content validity § Extent to which a measurement contains the required domains or areas to accurately measure a concept § Construct validity § Extent to which an instrument measures the underlying construct that it purports to measure. § Convergent validity § Extent to which similar constructs are correlated with one another. § Criterion validity § Ability of an instrument to correlate well with a particular criterion or standard. § Discriminant validity § Extent to which an instrument purporting to measure a construct that is different from theoretically unrelated constructs.