[HSCI 51] Module 3

Hill’s Criteria for Causation

1. strength

2. consistency

3. specificity

4. temporality

5. biological gradient/dose response

6. plausibility

7. coherence

8. experimental evidence

9. analogy

Strength

• Relative risk greater than 1 

• Takes account relative prevalence of other exposures 

Consistency

Same exposure, similar results from several studies

Specificity

Exposure must be specific to disease

Temporality

Effect has to occur AFTER the cause

Biological Gradient or Dose Response

Increase in exposure = increase in contracting disease

Plausibility

• Consistent possibility of exposure causing the outcome 

• Physical evidence

Coherence 

Exposure is consistent with natural history of disease

Experimental Evidence

Effect of exposure proven through RCT

Analogy

Exposure and disease relationship drawn from other similar relationship

Study design

 systematic way of establishing association between exposure and outcome

1. Exposure

2. Outcome

3. Statistical comparison

3 features of study design

Experimental & Observational

Types of study design

Experimental

has assigned exposures 

RCT, Quasi-experimental

Experimental study design types

RCT

random allocation of groups

Quasi-experimental

no random allocation

Observational

no assigned exposure

Descriptive & analytical

Observational study design types

Descriptive

no comparison group

Analytical

has comparison group

Cohort, case-control, cross-sectional

Analytical study design types

Cohort study design

exposure → outcome

Case-control

outcome → exposure

Cross-sectional

• Exposure and outcome at same time 

• Can be both analytical or descriptive

Systematic review, meta-analysis of RCT

Evidence level 1

RCT

Evidence level 2

Quasi-experimental, systematic review of BOE, mixed methods

Evidence level 3

Case-control or cohort

Evidence level 4

Systematic reviews of descriptive and qualitative

Evidence level 5

Descriptive or qualitative studies

Evidence level 6

Opinion, reports, and lit reviews

Evidence level 7

Descriptive

• Describes outcome 

• Useful for further investigation by generating hypothesis

1. Individual-level

2. Group-level

3. Cross-sectional

Descriptive types

Individual-level

record certain characteristics (previously undescribed syndrome or disease, unexpected association of exposure and outcome)

Case report, case series

Individual-level types

Case Report

one case

Case series

many similar cases

Group-level

• Ecologic studies 

• Compares population

Ecological fallacy

Popu-level analysis applied at indiv-level

Cross-sectional

• Snapshot of outcome

• Can measure prevalence 

Cross-sectional

• Health planning 

• Less resources

Cross-sectional advantages

• No temporality

• Recall bias

Cross-sectional disadvantages

Case-control

• Recruit popu based on outcome (retrospective)

• identify which exposure increased the odds of developing outcome

Cases

outcome

Controls

without outcomes

Case-control

Ideal rare outcome, multiple exposure

Case-control advantages

• Not for rare exposures

• Not for incidence rate

• Selection and recall bias

Case-control disadvantages

Cohort

Based on exposure and follow them to assess whether they develop outcome

• For rare exposure, long outcomes

• Established temporality

• Measures incidence

Cohort advantages

• Not for rare outcomes

• Large resources 

• Loss to follow-up

Cohort disadvantages

Prospective Cohort

Measures exposure at beginning. Outcome has NOT occurred 

Fixed recruitment

same time

Dynamic Recruitment

• Different times

• Measures incidence density

Prospective Cohort

Retrospective Cohort

• Measures exposure at beginning. Outcomes already occurred 

• Measured from historical data 

Retrospective Cohort

true, quasi-experimental

Interventional studies

True Experimental

• RCTs 

• Gold standard

Experimental

receives intervention

Control

given a placebo, std treatment, or none at all

Placebo

no clinical effect but gives perception of getting treated

True Experimental

• Dose response

• Specificity & plausibility 

• Meets stat assumptions

• Blinding

True Experimental advantages

• Large amt of resources

• Unethical to withdraw

• Affect validity

True Experimental disadvantages

Single blinding

subject, does not know exposure 

Double blinding

subject and observer

Triple blinding

subject, observer, data analyst/statistican

Internal Validity

Robustness of study methods

Differential attrition

higher dropout rates of participants with specific characteristics 

External Validity

• Generalizability of results to popu 

• Limited exclusion and inclusion criteria

Clinical, community trials

True Experimental types

Clinical Trials

• Individual analysis 

• Example: Test efficacy of drugs

Community Trials 

• Community analysis 

• Example: Testing effectiveness of prevention program in reducing prevalence of tobacco use

Quasi-Experimental

NOT randomly selected

1. Ethical considerations

2. Logical difficulty

3. Small population

Reason for nonrandomization of quasi-experimental

Quasi-Experimental

• Less strength causation

• Confounders

Quasi-Experimental disadvantages

• Rule out alternative

• Conduct stat analysis

Quasi-Experimental solutions

nonequivalent groups, prettest-posttest, interrupted time-series

Quasi-Experimental types

nonequivalent group

 not randomly assigned design

prettest-posttest

dependent variable measured before and after intervention 

interrupted time-series

measurements taken multiple times 

prct, natural experiments

other interventional variants

PrCTs

• Not includes blinding 

• Randomly assigned

Natural Experiments

• No manipulation of intervention because it is naturally occurring 

• Blinding is NOT possible 

systematic reviews, meta-analysis, evidence-based guidelines

Critical appraisal

Systematic Reviews

• Synthesis of literature 

• Uses standardized method and specific inclusion criteria 

• Find answers to specific research question

• One of the strongest study designs in proving evidence

Meta-analysis

• Systematic reviews w/ stat analysis 

• Weighted average of intervention effect 

• Typically conducted on RCTs

Evidence-based guidelines

• “Clinical practice guidelines” 

• Synthesize available evidence on management and care of health conditions 

• Creates standards in clinical practice,

Incidence

Rapidity of disease occurrence and risk of contracting disease

Baseline/Background Risk

• Incidence of unexposed population 

• Risk of developing disease in absence of exposure 

Unexposed individuals

NOT 0 background risk due to multifactorial nature of disease 

Cumulative Incidence

New Cases at specific time period / Population-at-risk at the start of period

Incidence density

New Cases at specific time period / Total person-in-time

Prevalence

• Proportion of population with disease at a point in time 

• Presents “snapshot” of all cases within a given point (new + old) 

Point Prevalence

• Disease at certain point in time 

• One popu

• People died PRIOR do NOT count

Period prevalence

• Disease at any point during a specified time period

• Average of two or more popu

• Existing case if developed/had outcome and died DURING the time period