LO: Describe and compare the design, strengths, limitations and appropriate use of common research methodologies (RCTs, quasi-experimental studies, cohort studies, case-control studies, systematic reviews and meta-analyses) and explain how these relate to the hierarchy of evidence.

Interventional studies

• Participants allocated to receive one or more interventions (new drugs, device or behaviour change) to compare outcomes
• Randomised controlled trials (RCTs) - main interventional study design

Randomised controlled studies (RCTs)

• Randomly allocate to groups - control and comparator group
• Interested in a particular group
• CONSENT!!!!!

PICO

• Useful for summarising studies

Experiments

• Evidence of causation
• Confounder - variable that influences both dependent (exposure) and independent variable (outcome), causes a spurious relationship

flowchart TD
	Weather --> A[Ice cream sales]
	Weather --> B[Homicide rates]
	A <-.-> |Does not cause|B

• In this example, weather is the confounder.

Randomisation

• Distribute confounders equally so balanced across arms
• Large sample needed
• Difference found is more likely to relate only to the intervention
• May not be ethical - withholding treatment
• Logistically infeasible
• Resource intensive

RCT limitations

• Bias - systematic deviation from truth
• RCTs subject to bias

Volunteer bias

• Difference between those who choose and choose not to participate
• Reduces generalisability - application to larger population from subset of population
• Solution: reduce barriers to participation - incentivisation

Selection bias

• Groups differ in characteristics due to the way of selection
• Benefits of randomisation lost or reduced - increased confounding
• Solution: allocation concealment - protocols

Performance bias

• Differences in care provided/participant behaviour separate from intervention
• Obscures impact of intervention
• Solution: blinding (masking)

Lost to follow-up bias

• Difference between drop outs within groups
• Jeopardises comparability of groups
• Solution: minimise dropout - intention to treat

Detection bias

• Difference in outcome assessment
• Affects measuring true effect
• Solution: blinding of outcomes assessor

Reporting bias

• Selective outcome reporting
• Focus on positives - biased literature
• Solution: publish protocols - statistical analysis

Where bias can occur

flowchart TD
	A[Study population] --> B{Selection criteria applied}
	B --> C[Potential participants]
	B --> D[Non-participants]
	C --> E[Invitation to participate]
	E --Volunteer bias--> Participants
	E --> D
	Participants --Selection bias--> Randomisation
	Randomisation --> Treatment
	Randomisation --> Control
	Treatment --Performance bias--> Outcome/s
	Control --Lost to follow up bias--> Outcome/s
	Outcome/s --Detection bias--> F[Published data]

	F --Reporting bias--> Public

Internal validity

• Strength of evidence for causality - degree to which results are attributable to the intervention than other factors
• Strong in RCTs with robust randomisation and blinding
• RCTs good for measuring efficacy
  

RCT weaknesses

• Expensive (t and £)
• Short follow-ups
• Not always feasible
• Not always ethical
• Generalisability/external validity

External validity

• Degree to which study results can be applied to wider populations and settings
• RCTs often have lower external validity
• Selection criteria lead to particular participant types recruited
• Focus on efficacy rather than effectiveness

Quasi-interventional studies

• Interventional study without random assignment
• 'Quasi-' means lacks
• Allocation - characteristic, natural experiments, pre-post comparison
• Pros - potentially large samples, feasible where RCTs aren't
• Cons - no randomisation leading to confounding, other naturally occurring events can impact measuring effect