Epidemiology, ‘Medical

Statistics’ and Public Health

Medunsa, Pretoria
14-18 February 2011

Prof. Dr. Joost Weyler

Epidemiology and Medical Statistics
Epidemiology en Social Medicine, University of Antwerp
 Scheme

• Introduction to Epidemiology
- Basic concepts and uses in Public Health
• Inferential Statistics:
- Theoretical Population and Sampling
- Estimation or Hypothesis testing
• Basic epidemiological study designs
- Diagnostic, Prognostic, Interventional, Etiognostic
- Error: precision and validity
• Summarizing and presenting data in tables
and graphs

1
 Scheme, continued
• Inferential statistics for one group, two groups
and more than two groups
- Parametric and non-parametric
• Regression, multiple regression basic
concepts:
- Linear
- Logistic (Binary-)
- Cox
• Advanced Statistics as input for research
design
• Intervention Research and RCT

2
Introduction to Epidemiology:
Basic concepts and uses of
Epidemiology in public Health

Medunsa, Pretoria
14-18 February 2011

Prof. Dr. Joost Weyler

Epidemiology and Medical Statistics
Epidemiology en Social Medicine, University of AntwerpA
 Scheme
• History of Epidemiology:
- Public Health (?)
- Causality and Prevention
• What is ‘Epidemiology’?
- Occurrence research
- Examples
• Study questions in Medical Science
- Diagnostic, Prognostic, Interventional, Etiognostic
- In Public Health

4
 History of Epidemiology
Where do we come from?

John Snow?

What about William Farr, Ignaz Semmelweis,

James Lind and so many others?

What about Ancient History?

Public Health orientation

5
 History of Epidemiology
William Farr (1807-1883)

Hygology (Hygiene)
Vital Statistics

Cholera epidemic

6
 History of Epidemiology
William Farr (1807-1883)

‘Miasma Theory’

Association of elevation of housing

above the river Thames and cholera
mortality
so:…

7
 History of Epidemiology
John Snow (1813-1858)

‘Germ Theory’

8
 History of Epidemiology
John Snow (1813-1858)

Mapping

9
 History of Epidemiology
John Snow (1813-1858)

Pump in the Broad

Street

‘Southwark and
Vauxhall company’

10
 History of Epidemiology
John Snow & William Farr

Analysis by Farr
district water supply Population Deaths due to Cholera mortality
cholera pro mille
Southwark & Vauxhall Cy 167.654 844 5,0
Lambeth Cy 19.133 18 0,9
Both companies 300.148 652 2,2

Problem: Ecological fallacy

11
 History of Epidemiology

district water supply Population Deaths due to Cholera mortality

cholera pro mille
Southwark & Vauxhall Cy 167.654 844 5,0
Lambeth Cy 19.133 18 0,9
Both companies 300.148 652 2,2

Water supply individual Population Deaths due to Cholera mortality pro

households cholera mille
Southwark & Vauxhall Cy 98.862 419 4,2
Lambeth Cy 154.615 80 0,5

What is striking?

12
History of Epidemiology

13
 History of Epidemiology
John Snow (1813-1858)

Explanation:

14
 History of Epidemiology

A drop of London Water, 1850 in Punch

15
 History of Epidemiology
Janet Lane-Claypon (1877-1967):

‘case-control study’ (1926) causes (in pregnancy and

breast feeding) of breast cancer

500 cases 3.4 children

500 controles 5.3 children

16
 Epidemiology, nowadays…
• Epidemiology is no longer the exclusive
activity for the Public Health Professional

• Clinical Epidemiology
• Genetic Epidemiology
• …

• Increasing role of statistics (medical statistics,

biostatistics,…)

17
 Epidemiology, Quo vadis?
• Epidemiology: where are we going?
• EBM

18
 Evidence Based Medicine (EBM)
BMJ 1996;312:71-72 (13 January)
David L Sackett, William M C Rosenberg, J A Muir Gray, R Brian Haynes, W
Scott Richardson
Editorials
Evidence based medicine: what it is and what it isn't
Evidence based medicine is not restricted to randomised trials and meta-
analyses. It involves tracking down the best external evidence with which to
answer our clinical questions. To find out about the accuracy of a diagnostic test,
we need to find proper cross sectional studies of patients clinically suspected of
harbouring the relevant disorder, not a randomised trial. For a question about
prognosis, we need proper follow up studies of patients assembled at a uniform,
early point in the clinical course of their disease. And sometimes the evidence
we need will come from the basic sciences such as genetics or immunology.

19
 EBM

It is when asking questions about therapy that we should try to avoid the non-
experimental approaches, since these routinely lead to false positive conclusions
about efficacy. Because the randomised trial, and especially the systematic
review of several randomised trials, is so much more likely to inform us and so
much less likely to mislead us, it has become the "gold standard" for judging
whether a treatment does more good than harm. However, some questions about
therapy do not require randomised trials (successful interventions for otherwise
fatal conditions) or cannot wait for the trials to be conducted. And if no
randomised trial has been carried out for our patient's predicament, we must
follow the trail to the next best external evidence and work from there.

20
 EBM
Levels of evidence
Level Therapy (prevention) Prognosis Diagnosis
Aetiology (harm)
1 RCT (SR/ind.) Inception cohort study Level1 diagnostic
(SR/ind.) study/ cohort study

2 Cohort study (SR/ind.) Retrospective cohort Level2 diagnostic

study (SR/ind.) study/ exploratory
cohort study
3 Case control study (SR/Ind.) Non consecutive study

4 Case series (poor cohort or case Case series (poor cohort Case control study
control study) study)

5 Expert opinion Expert opinion Expert opinion

Oxford Centre for Evidence-based Medicine Levels of Evidence (May 2001)

21
 List of study design features
Q-
RC RC NR CB PC RC HC NC
T T CT A S S T C CC XS BA CR/CS
Was there a comparison:
between two or more groups receiving
different interventions? Y Y Y Y Y Y Y Y Y Y N N
with the same group over time? P P N Y N N N N N N Y N
Were participants allocated to groups by:
Concealed randomization? Y N N N N N N N N N na na
Quasi-randomization? N Y N N N N N N N N na na
By other action of researchers? N N Y P N N N N N N na na
Time differences? N N N N N N Y N N N na na
Location differences? N N P P P P P na na na na na
Treatment decisions? N N N P P P N N N P na na
Patient preferences? N N N P P P N N N P na na
On the basis of outcome? N N N N N N N Y Y P na na
Which parts of the study were prospective:
Identification of participants? Y Y Y P Y N P* Y N N P P
Assessment of baseline and treatment
allocation? Y Y Y P Y N P* Y N N na na
Assessment of outcomes? Y Y Y P Y P P Y N N P P
Generation of hypotheses? Y Y Y Y Y Y Y Y P P P na
On what variables was comparability
between groups assessed:
Potential confounders? P P P P P P P P P P N na
Baseline assessment of outcome
variables? P P P Y P P P N N N N na

Y=Yes; P=Possibly; P*=Possible for one group only; N=No; na=not applicable. NB. Note that ‘possibly’ is used in the table to indicate cells
where either ‘Y’ or ‘N’ may be the case. It should not be used as a response option when applying the checklist; if uncertain, the response
should be ‘can’t tell’.
RCT=Randomized controlled trial; Q-RCT=Quasi-randomized controlled trial; NRCT=Non-randomized controlled trial; CBA=Controlled
before-and-after study; PCS=Prospective cohort study; RCS=Retrospective cohort study; HCT=Historically controlled trial; NCC=Nested
case-control study; CC=Case-control study; XS=Cross-sectional study; BA=Before-and-after comparison; CR/CS=Case report/Case series

22
 The Principles of The Cochrane
Collaboration

1.Collaboration, by internally and externally fostering good communications, open decision-making

and teamwork;
2.Building on the enthusiasm of individuals, by involving and supporting people of different skills
and backgrounds;
3.Avoiding duplication by good management and co-ordination to maximize economy of effort;
4.Minimizing bias, through a variety of approaches such as scientific rigour, ensuring broad
participation, and avoiding conflicts of interest;
5.Keeping up to date, by a commitment to ensure that Cochrane reviews are maintained through
identification and incorporation of new evidence;
6.Striving for relevance, by promoting the assessment of healthcare interventions using outcomes
that matter to people making choices in health care;
7.Promoting access, by wide dissemination of the outputs of the Collaboration, taking advantage
of strategic alliances, and by promoting appropriate prices, content and media to meet the needs
of users worldwide;
8.Ensuring quality, by being open and responsive to criticism, applying advances in methodology,
and developing systems for quality improvement;
9.Continuity, by ensuring that responsibility for reviews, editorial processes and key functions is
maintained and renewed;
10.Enabling wide participation in the work of the Collaboration by reducing barriers to contributing
and by encouraging diversity.

23
 Epidemiology, Quo vadis?
• Epidemiology: where are we going?
• EBM
• Poor concepts
• Illness, disease, sickness
• Frequencies
• Study object
• Study design
• Model analysis versus data analysis

24
 Epidemiology, Quo vadis?
• Epidemiology: where are we going?
• EBM
• Poor concepts
• Urge for publications and PhD-theses
• Quantity vs Quality
• …

25
 Epidemiology, Quo vadis?
• Epidemiology: where are we going?
• EBM
• Poor concepts
• Urge for publications and PhD-theses

• Epidemiology: where should we go?

26
 ‘Epidemiology, Quo vadis?’
Eur J Epidemiol. 2004;19(8):713-8.
Miettinen OS.

Department of Epidemiology and Biostatistics, Faculty of Medicine, McGill University, Montreal,

Canada.

In our etiologic research, we epidemiologists need to leave behind the concepts of 'cohort' study
and 'case-control' study and adopt that of the etiologic study as the singular substitute for these.
We then need to realize that the etiologic study is well suited to be viewed as paradigmal for
intervention studies. We finally need to become serious about object design before methods
design in both etiologic and intervention research. Once these developments have occurred,
we'll be ready for truly meaningful research to advance the knowledge base of both types of
causality-oriented 'gnosis' in the practice of clinical medicine, etiognosis and intervention-
prognosis; and descriptive-prognostic study we'll see as inherent in any intervention-prognostic
study. As for diagnostic research, then, we need to come to see it as nothing but a special case
of our familiar descriptive prevalence research. Because of this readily attainable theoretical
readiness peculiar to us research epidemiologists, and for other reasons besides, only we can
assume the central role in the production of the knowledge base for scientific medicine. We
consequently have the obligation to assume this larger and higher, meta-epidemiologic mission--
and some even higher ones besides.

27
 Object design before study design

Meta-epidemiology:

Epidemiology is nowadays so ill defined that a logic taxonomy

of its research topics becomes impossible

Miettinen: It is now the time for a new approach (meta-

epidemiology), starting from applied medical research

Proposition for meta-epidemiological types of research

questions:

DIA GNOSTIC
ETIO GNOSTIC
PRO GNOSTIC

28
 Object design before study design

• Diagnosis
• Etiognosis
• Prognosis

Questions:
• Symptomatic (sick) person: is illness
present ?
• Person with illness: why ?
• Person with illness: will he be cured ?

29
 Object design before study design
Examples:

Suppose you want to compare blood pressure among males

and females.

Question: I wonder whether the blood pressure in males

differs from the blood pressure in females...

How would you study this question?

Measuring
Database
Statistical test (t-test)

What about the study object?

What is the mathematical equation (model)?

30
 Object design before study design
Examples:

Study of the occurrence of various values of blood pressure

among people
descriptive parameter: median of the pressure (again, not
a constant of nature)
blood pressure depends on age (and other characteristics)
of individuals

A rule of thumb used to be:

blood pressure , in mm Hg, is 100 plus age in years

This is a regression model (a regression function) of the

form:
P = B0 + B1 x Age

31
 Object design before study design

P is the median of systolic pressure

B0 = 100 mm Hg
B1 = 1 mm Hg/yr

The correct phrasing of the rule of thumb is therefore

blood pressure , in mm Hg, is 100 mm Hg plus 1 mm

Hg/yr times age in years

Age is a determinant of the parameter, but cave...

How would you conduct a study to learn about this?

32
 Object design before study design

How would you conduct a study to learn about this?

Sampling
What is sampling?
How would you sample?

Measuring

Analysis

Presentation of results

33
Object design before study design

Mean of
systolic BP
(mm Hg), P
150
P = 100mmHg + (1 mm Hg/yr).A

100

50

0
0 20 40 60
Age (yr), A

34
 Object design before study design
LEEFTIJD vs. SBD
SBD = 97,775 + ,55625 * LEEFTIJD
Correlation: r = ,47883
180

160

140
SBD

120

100

80 Regression
15 25 35 45 55 65 75 95% confid.

LEEFTIJD

35
 Object design before study design

Presentation of results

Discussion: What have we learned?

36
 Object design before study design

Presentation of results

Discussion: What have we learned?

Be careful: over-interpretation of results

When object design is on blood pressure and aging, study

design different

37
 Object design before study design
‘Study objects’ (questions):

What are the questions (clinical, public health) that need for scientific
answers (abstract, objective, rational):

Is the patient suffering from a specific illness?

Is the asymptomatic person suffering from the illness for which a screening
test has been performed?

Will the patient still be alive in five years from now, when choosing for this
intervention?

Should we prefer this diagnostic/screening test above the other

What are the causes of relapse/death in this patient and what was the
contribution of the intervention?

38
 Directionality
‘Study Objects’ (Questions):

PROGNOSIS
Future, taking into account the present, the past en when interventional
conditional on part of the future

DIAGNOSIS
Present, taking into account the present and the past

ETIOGNOSIS
Present, taking into account the past

39
 Epidemiological outcomes
‘Study methods’ (type):

PROGNOSIS
Cumulative incidence
Surviving probability
Incidence density

DIAGNOSIS
Prevalence
Comparison: ROC, area under the curve

ETIOGNOSIS
Incidence density, (Cumulative incidence, Prevalence?)

Do not over-simplify...

40
 Directionality
‘Study methods’ (type):

PROGNOSIS
Cumulative incidence
Paradigm: RCT: Experimental cohort study

DIAGNOSIS
Prevalence study
Cross-sectional study

ETIOGNOSIS
Incidence density study, Prevalence study
Case-control study

Do not over-simplify...

41
 Object design before study design
Meta-epidemiological study objects:
DIAGNOSTIC
Current prevalence (T0) as a descriptive function of diagnostic
indicators (current, past)
ETIOGNOSTIC
Current incidence (T0) as a causal function of exposure in the past
(risk-indicators)
PROGNOSTIC
Interventional: Future (post-T0) incidence/prevalence as a causal
function of future intervention
Otherwise: Future (post-T0) incidence/prevalence as a descriptive
function of prognostic indicators (current)

Type study object:

What is the functional relation between the objects under study ?
Y = f ( X1, X2, X3, …, Xn)

42