SOURCES OF ERRORS IN EPIDEMIOLOGICAL STUDIES

SUPERVISOR: DRMALIMU

SOURCES OF ERRORS IN EPIDEMIOLOGICAL STUDIES

Objectives

At the end of this session each student should be able to;

Define errorMention sources of errorDescribe types of errorState the ways of minimizing error

DefinitionError Defined as a mistake which can occur during the study.

Sources of error• Selective survival• Selective recall• Incorrect classification of subjects with regard to their

disease and/or exposure status.

Types of error

Random error Systematic error

A. random error Define as that part of our experience that we cannot

predict. From a statistical perspective, random error can also

be conceptualized as sampling variability .

The major strategies for reducing random error are:

Increase sample size A larger sample, other things being equal, will yield

more precise estimates of population parameters .

Improve sampling procedures A more refined sampling strategy .

Reduce measurement variability by using strict measurement protocols, or averages of multiple measurements.

Use more statistically efficient analytic methods Statistical procedures vary in their efficiency, i.e., in the

degree of precision obtainable from a given sample size.

B. Systematic error (bias) Is a difference between an observed value and the true

value due to all causes other than sampling variability. It can arise from different sources, including factors

involved in the choice or recruitment of a study population and factors involved in the definition and measurement of study variables

Types of bias

I. Selection bias Occurs as a result of errors in identifying the study

population. Sources of selection bias Sampling bias Systematically excluding or over-representing certain

groups.

Allocation bias Systematic differences in the way which subjects are

recruited into different groups for a study. For example, a study may fail to use random

sampling, the first 20 patients who arrive at a clinic are allocated to a new treatment, and the next 20 patients are allocated to an existing treatment.

However, the patients who arrive early may be better-

off, on the other hand the doctor may have asked to see the most seriously ill patients first.

Minimizing selection bias

• Clear definition of study population• Choose cases and controls from same population• Selection of exposed and non-exposed without

knowing disease status.

II. Information bias (also called misclassification bias) Is caused by systematic differences in data collection,

measurement or classification. Sources of Information bias Recall bias People suffering from a disease may have spent more

time thinking of possible links between their past behavior and their disease than no sufferers.

Interviewer bias. Interviewers may phrase questions differently for

different subjects, or write down their own interpretations of what subjects have said.

Follow-up bias In studies that follow up subjects at intervals, people

from certain groups may tend to be lost to follow-up, or a disproportionate number of exposed subjects may be lost to follow-up compared with non-exposed subjects.

Validity Is the extent to which a measurement measures what it

is supposed to measure. • Internal validity refers to absence of systematic error

that causes the study findings (parameter estimates) to differ from the true values as defined in the study objectives.

• External validity refers to the extent to which a study's findings apply to populations other than the one that was being investigate.

Relationship of reliability and validity

Validity" is used as a general term for accuracy or correctness

For example, a study designed to estimate the

prevalence of smoking in a population may select

subjects for interview in a number of locations.

If the interviews are only conducted on weekdays, the

study is likely to under-represent people who are in

full-time employment, and include a higher

proportion of those who are unemployed, off work or

mothers with children.

Recording bias. Medical records may contain more information on

patients who are 'cases‘.

Minimizing information bias• Standardise measurement instruments • Administer instruments equally to cases and controls

(exposed/unexposed) • Use multiple sources of information– Questionnaires– Case records

Confounding Confounding occurs when a separate factor (or

factors) influences the risk of developing a disease, other than the risk factor being studied.

To be a confounder, the factor has to be related to the

exposure, and it also has to be an independent risk factor for the disease being studied. (Third variable problem).

To be a confounding factor, 2 conditions must be met

For example, if a study assesses whether high alcohol consumption is a risk factor for coronary heart disease, smoking is a confounding factor (also called a confounder) .

This is because smoking is known to be related to alcohol consumption, and it is also a risk factor for coronary heart disease.

High alcohol consumption Coronary heart disease

(Possible risk factor) (Disease)

Smoking (confounder)

Minimizing confounding Randomization Adjustment (Age-adjustment)

Conclusion Identification of possible bias is a difficult exercise

but is crucial to improve validity. Bias can’t usually be totally eliminated.

It must be to keep it to a minimum, to identify those

biases that cannot be avoided, to assess their potential impact and to take this into account when interpreting the results.

References

Schoenbach, J. V. (2001). Sources of error.

Stewart, A. (2002). Basic statistics and epidemiology. A practical guide.

Validity Is the extent to which a measurement measures what it

is supposed to measure. • Internal validity refers to absence of systematic

error that causes the study findings (parameter estimates) to differ from the true values as defined in the study objectives.

• External validity refers to the extent to which a study's findings apply to populations other than the one that was being investigate.