Causation and Validity in Research Design

• Research design is the critical backbone for a sound scientific investigation
• Good design increases the validity of research findings
• Flawed design could raise questions on the credibility of those findings
• Poorly designed studies have limited scientific value and can be harmful to patients
• Understanding research designs is essential to evaluate and apply research evidence
• Research design refers to the overall plan that allows researchers to gather answers to study questions and test study hypotheses
• The decision to use a particular study design hinges on the ability of that design to provide valid results
• Validity reflects the accuracy of study results

Research Design Validity

• Validity can be distinguished into internal validity and external validity
• Internal validity reflects the extent to which clinical outcome of interest (dependent variable) in a study is caused by the treatment (independent variable)
• External validity refers to the extent to which the results of a study can be generalized to other settings

Research Design Classification

• Research designs can be classified using different sets of criteria, which includes study purpose, time orientation, experimental setting etc
• These classifications capture different dimensions of research approaches but are interlinked and sometimes can overlap

Research Design Classification - Purpose

• The purpose of the study could be either descriptive or analytical
• A researcher may decide on a particular design to just describe a phenomenon (descriptive) or provide interpretation of relationship and/or causal mechanism that may exist between two or more variables an existing phenomenon (analytical)

Research Design Classification - Purpose - Descriptive studies

• Descriptive studies describe or summarize information about the disease or events without making any causal inferences
• Descriptive studies incorporate five important elements pertaining to a new disease or event: who, what, why (proposed possibility, not proven), when and where

Research Design Classification - Purpose - Analytical studies

• Analytical studies are aimed at understanding the relationship and/or causal mechanism that may exist between two or more variables
• Correlation does not imply causation

Causation

• Causation is one of the most commonly used terms in scientific literature
• Cause-and-effect relationship or causality exists if there is a causal relationship between the treatment (cause) and clinical outcome (effect)
• Establishing cause-and-effect relationships can help develop rational solutions to clinical problems
• In 1965, Sir Austin Bradford Hill listed a set of nine criteria that should be fulfilled for relationship between two variables to be considered as potentially causal which are called "Bradford Hill" criteria
• As acknowledged by Hill, the fulfillment of nine criteria does not automatically imply causation but rather assist researchers in making decisions regarding the presence or absence of causal mechanisms

Research Design Classification – Time Orientation

• Research designs may be classified as prospective or retrospective

Research Design Classification – Time Orientation - Retrospective studies

• Retrospective studies involve evaluation of data of past events or existing data such as medical records to achieve the research objective
• In retrospective research, the event of interest has already occurred, and researchers go backwards in time to determine the relationship between cause and event
• The main advantage of retrospective design is that they are minimally resource intensive as it only involves analysis of existing data or past events
• Limitations: the researcher cannot control past events or data collection methods and they have to rely on existing data or previous events without any say on what variables are needed and how they are defined and collected

Research Design Classification – Time Orientation - Prospective studies

• Prospective studies are where the researcher collects the data after the study onset by following individuals over a period of time
• The main strength of this design is to determine and define the research variables and prospectively collect relevant data to achieve the objectives
• The main limitation of prospective design is that they are resource (time and cost) intensive

Types of Research Designs - Experimental Setting

• Research designs can be broadly categorized into two types: observational and experimental designs
• In observational designs, the researcher merely observes the interplay of independent variable (exposure or treatment) with the dependent variable (outcome)
• In experimental designs, the researcher controls the independent variable (exposure or treatment) to see if it will have an impact on the dependent variable (outcome) and are also referred to as “interventional studies"

Experimental Designs

• In experimental designs, the researcher controls the treatment (independent variable) that is likely to have an impact on clinical outcome (dependent variable)
• There are three essential elements of a true experiment:
  • The independent variable (treatment) must be manipulated by the researcher
  • The study subjects are randomized into groups
  • A control or comparison group must be incorporated within the design
• Other criteria such as temporality, strength, biological gradient, etc. can easily be achieved in experimental studies to draw causal conclusion Basic researchers often rely on experimental designs to test their research hypothesis
• The most commonly used experimental design in clinical research is the Randomized Controlled Trial (RCT)
• RCT's are also referred to as Randomized Clinical Trials
• The RCT is considered to be the gold standard in evaluating the safety and efficacy of an intervention
• RCTs in general have high internal validity
• Since RCTs are conducted in tightly controlled clinical settings, RCTs tend to have low external validity (generalizability)

Observational Designs

• A researcher observes the relationship between the study variables, mainly independent (exposure or treatment) and dependent (outcome or disease) in a natural setting
• Unlike experimental studies, there is no manipulations of the independent variable or randomization of participants into experimental and control group
• The variations in exposure and outcomes are observed to evaluate their relationship or association
• Investigators collects data regarding exposure and outcomes using primary data techniques like interviews, surveys, etc., or use data collected previously for other purposes (secondary data) like medical charts
• Observational studies include the following: case reports, case series, cross-sectional studies, case-control studies, and cohort studies
• A case report involves a study of a single case of a new disease or manifestation, while a case series involves a study of multiple similar cases
• Cross-sectional are those wherein the exposure and/or outcome of interest are measured at the same point in time and can be purely descriptive such as prevalence studies or are used to ascertain the relationship between variables of interest, such as through analytical cross-sectional surveys
• Case-control studies involve comparison of exposure status among individuals with the disease or outcome of interest (cases) and those without (controls)
• Cohort studies are observational studies wherein two groups, i.e., exposed and unexposed, are followed over a period of time until the development of outcome of interest and can be divided into two types: prospective and retrospective In prospective cohort study, the exposed and unexposed groups are classified at baseline and then followed in future to determine the occurrence of outcome of interest in the two groups
• In retrospective cohort study, a researcher uses previously collected (historical) data to identify exposure status and occurrence of outcome in the study group