Overview of research methodology Flashcards
What is research?
Old French term “Re-cerchier” meaning to “search intensively”: a process by which we aim to discover new knowledge. Uses a systematic process Scientific Method
Why is critical appraisal required
- Published research is not always reliable or relevant.
- We need a systematic framework to conduct research, interpret research and assess quality.
Why is study design important?
- Specific questions need specific design (feasibility, ethics)
- Different designs have different causal strengths
- Determines nature of bias and confounding (validity)
- Affects resources needed: sample size, staff, expertise, access to records, time! efficiency
- Ethical issues vary
- May influence measurement accuracy (e.g. prospective vs retrospective data collection)
Descriptive study
No comparison group. Answers who, what, when, where.
E.g. case study, case series, non-analytic cohort
advantages of descriptive study
Hypothesis generation
Health service planning & management
disadvantages of descriptive study
Cannot establish causality
Cross sectional study
Exposure and outcome measured at one point in time
= snap-shot
= prevalence study
Advantages of cross sectional study
-Relatively cheap & easy
-Measure multiple variables at the same time
Disadvantages of cross sectional study
- Cannot measure temporality
-Cannot measure incidence
-Information bias (recall bias); Selection bias
Cohort study
“A group of people who share a common characteristic or experience within a defined period”.
Advantages of cohort study
–Temporality
-Can calculate incidence
-Useful for rare exposures
-Can study multiple outcomes (of single exposure
Disadvantages of cohort study
-Loss-to-follow-up
-Detection bias
-Time consuming, expensive
retrospective
Study begins when outcome has already occurred
Participants grouped on exposure (NOT outcome status)
prospective
Outcome has not yet occurred
Participants grouped on basis of exposure
Follow up to see if outcome occurs
case control study
Identify cases (those who have the outcome/disease)
Choose controls (those without outcome/disease)
Measure exposure in both cases and controls
Common sources hospital, neighbourhood, family members. “Matched controls”
Advantages of case control study
Efficient
Relatively quick
Good for rare outcomes
Can study multiple exposures
Disadvantages of case control study
Selection bias (selection of controls)
Information bias: Recall
Can only calculate odds ratio
No temporality
Randomised Control study
Similar to cohort, but exposure
assigned randomly & always prospective
Randomisation is a way to deal with confounders (known & unknown)
Participants have equal chance of receiving exposure/intervention as the control.
Advantages of RCT
Minimal bias/confounding
Strong evidence
Disadvantages of RCT
Ethical issues
Time-consuming, expensive
Loss-to-follow-up
Not suitable for all research questions
What is blinding or masking in research studies?
Blinding or masking involves concealing information about the intervention received (e.g., treatment or placebo) from participants, clinicians, field workers, lab personnel, or statisticians involved in the study. This is done to prevent bias in the assessment of outcomes.
What are the types of blinding?
Blinding can be single-blinded, where either participants or researchers are unaware of the intervention received, or double-blinded, where both participants and researchers are unaware. Triple-blinding involves concealing information from participants, researchers, and outcome assessors.
Why is blinding used in research studies?
Blinding is used to minimize bias that could result from knowledge of the intervention. It helps ensure that participants behave consistently and that researchers assess outcomes objectively, reducing the risk of bias in study results.
Ecological studies
Work with population or group-level variables. Previous examples were all looking at individual level. Study design itself can vary: but mostly cross-sectional ecological study
Advantages ecological studies
Hypothesis generation
Relatively quick and cheap
Disadvantages of ecological studies
Ecological fallacy
Confounding, bias
Cannot determine causality
What is occurrence?
Describes the burden of disease (how much there is).
When do we use measures of occurrence or frequency
Quantify the extent of an outcome(e.g. occurrence of disease) populations in order to motivate for response/ resources.
E.g How many cases of diabetes were diagnosed in South Africa in 2022?
Measures of occurrence/ frequency
Prevalence
Incidence (Incidence Proportion & Incidence Rate)
Odds
What is association?
Compares the occurrence of disease in different exposure groups. Explores the relationship between exposure(s) and an outcome(s) of interest and whether it could be causal.
When do we use measures of association
We want to understand how exposures and outcomes are related, if one causes the other, if we want to modify exposure to increase or decrease the outcome.
E.g. How strongly is smoking related to lung cancer?/
How many cases of lung cancer could we prevent by implementing tobacco control measures?
measures of disease association
-Relative measures
Prevalence ratio
Incidence: Risk ratio or rate ratio (RR)
Odds ratio (OR)
-Absolute measures
Proportion / rate and risk differences
Number needed to treat etc.
cases/ counts
Individuals in a population who have the disease, health disorder, or suffers the event of interest. Usually forms the numerator for measures of disease frequency (A)
population at risk (N)
The population from which cases come from. Usually forms the denominator in calculations of disease frequency
ratio
Comparison of two groups (numerator and denominator can be related)
rate
A rate is a ratio with a time component / change over time
proportion
A ratio in which the numerator is included in the denominator (A/N)
odds
A ratio in which the numerator is not included in the denominator (A/N-A)
prevalence
The number of cases of disease in a population at one point in time, as a proportion of the total number of persons in that population i.e. proportion of a population that has the disease/event at a specific point in time
no of disease/ total population at point in time
incidence proportion (risk)
no of new cases in a given time period/ no at risk at beginning of time period
incidence rate
no of new cases in a given time period/ total person- time of observation or risk during that period
Used if incomplete follow-up – i.e. loss to follow-up/open population (participants entering and leaving)
IR takes into account the:
Size of population (number of individuals in a population that become ill)
length of time contributed by all persons during the period they were in the population.
aims of measures of association
Measures of association allow us to compare measures of disease frequency
Quantify the association
relative measures of association
Measures of disease frequency are divided by one another
Ratios of two measures of disease frequency
Absolute measures of association
Measures of disease frequency are subtracted from one another
Tells us about the Public Health Impact
relative measures of association: ratio
-prevalence ration
- cumulative incidence ratio
- incidence rate ration
- odds ratio
Prevalence ratio
Prevalence of disease in exposed / prevalence of disease in unexposed
Cumulative incidence ratio
Risk (incidence proportion) of disease in exposed / risk (incidence proportion) of disease in unexposed
Also called risk ratio or relative risk
incidence rate ratio
Rate (incidence rate) of disease in exposed / rate (incidence rate) of disease in unexposed
Also called rate ratio or relative rate
odds ratio
Odds of exposure in disease / odds of exposure in non-diseased*
Interpretation of Relative Measures of Association: <1 (less than 1):
The numerator is smaller than the denominator
The exposure is “protective”/disease is less likely in the exposed
Interpretation of Relative Measures of Association =1
The numerator and denominator are equal
There is no difference in disease occurrence between the exposed and unexposed
Interpretation of Relative Measures of Association >1 (greater than 1):
The numerator is larger than the denominator
The exposure is “harmful”/disease is more likely in the exposed
Interpretation of Relative Measures of Association = Significant?
The relative measure of association is an estimate, and should be reported and interpreted with a 95% Confidence Interval (generated by statistical software)
If the 95% confidence interval includes the value of 1, then not significant
risk ratio
ratio of two probabilities
probability of an event occurring in treatment group / probability of the event occurring in control group
risk in exposed/ risk in unexposed
based on incidence- cohort studies, RCT
odds ratio
ratio of two odds
odds of exposure occurring in cases group/ odds of exposure occurring in control group
odds in cases/ odds in controls
case control
prevalence ratio
prevalence exposed/ prevalence unexposed
for cross sectional studies
Absolute measures of association
- prevalence difference
- incidence proportion difference
- incidence rate difference
prevalence difference
Prevalence of disease in exposed - prevalence of disease in unexposed
=The number of cases that could be eliminated if the exposure was eliminated
incidence proportion difference
Risk (incidence proportion) of disease in exposed - risk (incidence proportion) of disease in unexposed
incidence rate difference
Rate (incidence rate) of disease in exposed - rate (incidence rate) of disease in unexposed
Population Risk or Rate difference: PRD = Rt – Ru
Population risk difference gives you the excess number of cases in the whole population that is associated with the exposure
Can use incidence rate / incidence proportion or prevalence
-Rt = Incidence rate OR incidence proportion OR prevalence in the whole population
-Ru = Incidence rate OR incidence proportion OR prevalence in the unexposed
Attributable Proportion:
This calculates the proportion of disease that is attributable to the exposure, also known as attributable risk (or fraction)
Can calculate attributable proportion in the exposed or in the population
Attributable Prop in Exposed
APe=[(Re – Ru) / Re]
APe is the attributable proportion in the exposed
Re is the incidence rate, cumulative incidence or prevalence in the exposed
Ru is the incidence rate, cumulative incidence or prevalence in the unexposed
APt is the attributable proportion in the total population
Rt is the incidence rate, cumulative incidence or prevalence in the total population
Attributable Prop in Population
APt=(Rt – Ru) / Rt
APe is the attributable proportion in the exposed
Re is the incidence rate, cumulative incidence or prevalence in the exposed
Ru is the incidence rate, cumulative incidence or prevalence in the unexposed
APt is the attributable proportion in the total population
Rt is the incidence rate, cumulative incidence or prevalence in the total population
Number needed to treat
1 / [risk in exposed) – (risk in unexposed)
The number needed to be treated or the exposure to be prevented in order to prevent one case of disease
Interpretation of Absolute Measures of Association
If difference=0 there is no association between exposure and disease
Difference will lie between -100% and 100%, or -1 and 1 when expressed as proportions.
E.g. Prevalence difference = 0.08
“ In this study, smokers had 8 additional cases of TB per 100 people compared to non-smokers”