1.3 IL Health problem and research
question
The funnel approach
- Situation: context, magnitude of disease, current
knowledge, current care
- Complication: problem, gap in knowledge, suboptimal care
- Question: What do we need to address the complication
- Answer: Hypothesis, solution of the complication, possible
answer, gain
FINE criteria
Used to evaluate a research question.
F(easible) Adequate sample size? affordable in time and money?
I(mportant) Does it improve clinical care or public health?
N(ovel) New findings? Or confirms, refutes or extends previous findings?
E(Ethical) Meets already set criteria?
1.6 LE Study population and
data collection
Random error: occurs due to chance.
Systematic error: occurs due to measurement
system (or person measuring).
Information bias (= measurement bias):
(Systematic) measurement error in exposure and/or
outcome à misclassification of study objects.
In which phase does it occur?
- design - improper case or exposure definitions
- data collection – imprecise or invalid measurement instrument(s)
What is the impact on the study results?
- non-differential misclassification – bias toward unity (equal misclassification per group)
- differential misclassification – over- or underestimation (different misclassification per group)
How is it minimized?
- Blinding
Target population
domain (research question)
Accessible population
source population
Intended study sample
Sample population (study plan)
, Actual study sample
Study subjects (actual study)
Selection bias:
Comparability of study populations (case-control) that should come from the same population. So
the same association should be in the intended sample as in the true study sample.
In which phase does it occur?
- design – improper choice of unexposed/control group
- recruitment / data collection – non-participation / non-response
- follow-up – attrition bias (selective/systemic drop-out)
- analysis – complete case analyses (excluding missing values)
What is the impact on the study results?
- internal validity / external validity (generalizability)
How is it minimized?
- Randomization
Type I error = false-positive (alpha)
Type II error = false-negative (beta)
Power = 1 – Beta
Effect size: The minimum size of the difference in the two groups being compared that the
investigator wishes to have a reasonable chance of detecting.
Variability: The amount of spread in a measurement, usually expressed as a standard deviation
How to prevent confounding?
Randomisation : Through design or recruitment (only in intervention study)
Matching: Link persons in groups, so that distribution of important confounders are the same
between groups
Restriction: Limit study to specific groups or exclude subjects with (strong) confounders
How to remove confounding?
Correction or adjustment for confounding during data analysis
- stratification (1/2 confounders): creating two or more categories/subgroups in which the
confounding variable either does not vary or does not vary very much.
- multivariable regression techniques (2< confounders): e.g. linear regression, logistic regression
IL 2.1 Clinical Trials
question
The funnel approach
- Situation: context, magnitude of disease, current
knowledge, current care
- Complication: problem, gap in knowledge, suboptimal care
- Question: What do we need to address the complication
- Answer: Hypothesis, solution of the complication, possible
answer, gain
FINE criteria
Used to evaluate a research question.
F(easible) Adequate sample size? affordable in time and money?
I(mportant) Does it improve clinical care or public health?
N(ovel) New findings? Or confirms, refutes or extends previous findings?
E(Ethical) Meets already set criteria?
1.6 LE Study population and
data collection
Random error: occurs due to chance.
Systematic error: occurs due to measurement
system (or person measuring).
Information bias (= measurement bias):
(Systematic) measurement error in exposure and/or
outcome à misclassification of study objects.
In which phase does it occur?
- design - improper case or exposure definitions
- data collection – imprecise or invalid measurement instrument(s)
What is the impact on the study results?
- non-differential misclassification – bias toward unity (equal misclassification per group)
- differential misclassification – over- or underestimation (different misclassification per group)
How is it minimized?
- Blinding
Target population
domain (research question)
Accessible population
source population
Intended study sample
Sample population (study plan)
, Actual study sample
Study subjects (actual study)
Selection bias:
Comparability of study populations (case-control) that should come from the same population. So
the same association should be in the intended sample as in the true study sample.
In which phase does it occur?
- design – improper choice of unexposed/control group
- recruitment / data collection – non-participation / non-response
- follow-up – attrition bias (selective/systemic drop-out)
- analysis – complete case analyses (excluding missing values)
What is the impact on the study results?
- internal validity / external validity (generalizability)
How is it minimized?
- Randomization
Type I error = false-positive (alpha)
Type II error = false-negative (beta)
Power = 1 – Beta
Effect size: The minimum size of the difference in the two groups being compared that the
investigator wishes to have a reasonable chance of detecting.
Variability: The amount of spread in a measurement, usually expressed as a standard deviation
How to prevent confounding?
Randomisation : Through design or recruitment (only in intervention study)
Matching: Link persons in groups, so that distribution of important confounders are the same
between groups
Restriction: Limit study to specific groups or exclude subjects with (strong) confounders
How to remove confounding?
Correction or adjustment for confounding during data analysis
- stratification (1/2 confounders): creating two or more categories/subgroups in which the
confounding variable either does not vary or does not vary very much.
- multivariable regression techniques (2< confounders): e.g. linear regression, logistic regression
IL 2.1 Clinical Trials
