1. Interpretation of laboratory and epidemiological evidence
Lecture notes
Yvan Hutin and Aftab Jasir

2. Conducting a collaborative epidemiology- laboratory investigation
Formulating the objectives
Drawing conclusions
Planning
Analysing
Data analysis
Lab analysis
Data
Specimens
Preparing
Collecting
Instruments
Sampling strategy
Data
Specimens
When faced with the need to interpret evidence, bear in mind why the investigation was conduced

3. Dictionary definition of the verb: Interpret
“Explain the meaning of…”

4. Description and interpretation in neurophysiology
Occipital cortex, visual zone:
I see concentric circles
Pre-visual zone:
This is a TARGET!

5. Description and interpretation in a relationship
S/he did not call back 
S/he is angry
S/he is shy
S/he is too in love
There is someone else
I said something wrong
S/he lost my number
I was too insistent
The parents disagree
Horoscopes do not match
Probably not enough data to conclude….

6. Description and interpretation in clinical practice
Situation
Description
Interpretation
Chest X-ray
Alveolar opacity
Systematized, lobar opacity
Consolidation (Pneumonia?)‏
Dermatology
Copper papules
Soles and palms
Desquamation
Secondary syphilis 6

7. Language and interpretation
(1) Describe and (2) interpret: What is this?
Pay attention to the language you used
If you used the word “ear” you make it a rabbit
If you used the word “beak”, you make it a duck
Jastrow

8. Interpretation of descriptive epidemiological data to generate hypotheses
Narrow epidemic curve
Time
Hypothesis: The public tap was contaminated for a brief duration and caused the outbreak
Cluster of public tap
Place
Case patients used the tap
Person

9. Interpretation of analytical epidemiological data and additional investigations to test an hypothesis
Strong association water drinking / illness OR
Epidemiological evidence supported the hypothesis of the tap as the source of the outbreak
High attributable fraction
AFP
Water analysis
Water positive for S. Typhi
Sewage contaminated the tap with S. Typhi and caused the outbreak
X-contamination sewage / water supply
Sanitary assessment

10. Elements to consider before interpreting association as causation
Chance
Bias
Confounding factor
Causation
Strength of the association
Dose response
Consistency
Biological plausibility
Exposure/ outcome sequence 10

11. Casting the net and pulling it up
Descriptive epidemiological data generates hypotheses
Analytical epidemiological data tests hypotheses
Can you guess why two different fishermen? 11

12. Language used for data description and data interpretation
Description: Results
Interpretation: Discussion
Cases started to occur at 5AM, peaked at 7AM and decreased with a last case at 10AM
Cases clustered around cooling tower A
Malaria rates were high in all age groups
Cases were more likely than controls to lack health insurance
The shape of the epidemic curve suggested a point source outbreak
We suspected that tower A was the source of outbreak
Unstable transmission does not lead to population immunity
Access to health care may increase the risk of illness

13. Interpretation of data in a discussion section of a paper+ = 13

14. Example of integration of various pieces of evidence into an interpretation
Outbreak of cutaneous anthrax
Beef slaughter in West Bengal, India
Cohort study
Contact with meat is a risk factor
Null hypothesis:
Eating meat does not cause cutaneous anthrax

15. Attack rate of anthrax by exposures, Sarkarpara, Murshidabad, WB, India, 2007
AR in exposed (%)‏
AR in unexposed (%)‏
Association RR
95% CI
Age > 20 18 11
1.6
0.9-3
Female sex 17 14
1.2
0.7-2
Slaughter 83 9
8.7
6-13
Handling 26 10
2.6
2-4
Carry skins
100 15
6.7
5- 9
Eating
Undefined
This slide summarizes the results of the cohort study in the first village.
Risk factors significantly associated with illness included slaughtering, handling skins and handling meat.
17% of persons who ate beef developed anthrax, compared with 0% of those who did not eat.
All case-patients who had eaten meat had also other exposures 15

16. Other elements of evidence beyond epidemiological data
Elements available before the outbreak
Heat inactivate spores
Infected meat causes intestinal disease
Meat involved in intestinal outbreaks was poorly cooked (e.g., Kebabs)‏
Elements from the outbreak investigation
The beef meat was boiled 16

17. Testing the hypothesis that eating meat could cause cutaneous anthrax+ = 17

18. Always consider other hypotheses
Avoid:
We found that…
This could be due to... [this real phenomenon]
Prefer:
The results are …
Two possibilities
This could be due to this real phenomenon
This could be an artifact of the study
Examine both options
See what the data support and conclude

19. Dealing with an un-expected finding
One unexpected exposure is associated with outcome
Absence of context
No other studies
No biological rationale
Treat as a hypothesis generation:
This association should be examined in other studies
Do not force an explanation/ rationalization
“This may be due to…”

20. Risk factors for post-traumatic stress disorder (PTSD), Indian Tsunami, 2005
Unexpected effect modification:
Single woman more PTSD than married
This may be due to the fact that they are alone…
Coding error: It’s the converse that is true!
Married women have more PTSD
This may be due to the fact that they have to deal with their whole family…
Do not force interpretations
Propose further studies to look into it

21. Take home message Interpretation has a subjective component
Requires a careful, documented approach
We raise hypotheses with descriptive epidemiology and test them with analytical epidemiology
Findings acquire a meaning in the context of what was known before

22. Conducting a collaborative epidemiology- laboratory investigation
Formulating the objectives
Drawing conclusions
Planning
Analysing
Data analysis
Lab analysis
Data
Specimens
Preparing
Collecting
Instruments
Sampling strategy
Data
Specimens
When faced with the need to interpret evidence, bear in mind why the investigation was conduced

23. Possible objectives of joint laboratory epidemiology investigations
Test a hypothesis (Qualitative outcome)
Test a hypothesis
About an etiologic agent (e.g., Is West Nile virus the cause of the outbreak?)
About the relatedness of isolates (e.g., Are the cases caused by an identical pathogen?)
Measure a quantity (Quantitative outcome)
Estimate a quantity
Prevalence
Incidence

24. Using laboratory evidence to confirm a diagnosis during an outbreak
Short list potential etiologic agents (Hypothesis generating) according to:
Epidemiological characteristics
Clinical characteristics
Setting
Test for agents short listed (Hypothesis testing)
Positive test
Negative test
Use predictive values positive and negatives
Q fever,

25. Case scenario 1 Viral Hemorrhagic Fever (VHF)
Bleeding disorders
Progress to high fever
Shock
High case fatality

26. Virus families causing VHF sensitive or specific? Short listing?
Diverse group of animal and human illnesses that may be caused by five distinct families of RNA viruses
Arenaviridae, (Lymphocytic choriomeningitis virus, Lassa virus, Argentine, Bolivian, Brazilian and Venezuelan hemorrhagic fevers viruses)
Filoviridae (Ebola virus and Marburg virus)
Bunyaviridae (Hantaviruses, Crimean-Congo hemorrhagic fever)
Flaviviridae (dengue, yellow fever, WNV)
Rhabdoviridae (Lyssavirus)

27. Interpreting positive tests results during an outbreak
Use the predictive value positive that depends upon:
The frequency of the disease
The specificity of the test +++
Elements that support the hypothesis of a true positive
The disease is frequent (GAS)
The test is specific ( emm typing)
Elements that support the hypothesis of a false positive
The disease is rare (WNV)
The test is not sufficiently specific (CFT)
Q fever, GAS
The viral hemorrhagic (or haemorrhagic) fevers VHFs) are a diverse group of animal and human illnesses that may be caused by five distinct families of RNA viruses: the families Arenaviridae, Filoviridae, Bunyaviridae, Flaviviridae, and Rhabdoviridae. All types of VHF are characterized by fever and bleeding disorders and all can progress to high fever, shock and death in many cases. Some of the VHF agents cause relatively mild illnesses, such as the Scandinavian nephropathia epidemica, while others, such as the African Ebola virus, can cause severe, life-threatening disease

28. Interpreting negative tests results during an outbreak
Use the predictive value negative that depends upon:
The frequency of the disease
The sensitivity of the test +++
Elements that support the hypothesis of a true negative
The disease is rare (WNV)
The test is sensitive (IgM ELISA)
Elements that support the hypothesis of a false negative
The disease (condition) is common (GAS, Q fever)
The test is not sufficiently sensitive(Gram staining/PCR)

29. A test was negative only for the pathogens that were looked for
If the culture on a specific medium was not done, the test cannot be interpreted as negative for the specific pathogen
If you did not ask for Campylobacter culture, the “negative” stool culture is not really “negative” for Campylobacter

30. Host-pathogen relationship
Presence of an organism may have different interpretation according to the context
Immune system
Immunocompetent patient
Opportunistic pathogens may be innocent
Immunocompromised patient
Opportunistic pathogens may be the cause of the infection
Age (pertussis)
Physiological status (e.g., Urinary infection in pregnancy)

31. Possible objectives of joint laboratory epidemiology investigations
Test a hypothesis (Qualitative outcome)
Test a hypothesis
About an etiologic agent (e.g., Is West Nile virus the cause of the outbreak?)
About the relatedness of isolates (e.g., Are the cases caused by an identical pathogen?)
Measure a quantity (Quantitative outcome)
Estimate a quantity
Prevalence
Incidence

32. Using laboratory evidence to confirm the relatedness of isolates
Generate hypotheses using epidemiological evidence
Studies allowing the use of statistical tests (Large sample size)
Studies not allowing the use of statistical tests (Small sample size)
Test hypotheses using laboratory evidence
Use typing technique adapted to:
Hypothesis
Pathogen

33. Nosocomial IGAS infections, Skåne, Southern Sweden, 2012
43 cases of invasive Group A Streptococcus (iGAS) between 9 January and 29 April 2012
27 cases between 3 January and 24 April 2011
Thirteen of the 43 cases in 2012 were treated in an Intensive Care Unit. One case, a 84 year old already hospitalised prior to iGAS diagnosis, died
Hospital rejects the hypothesis of nosocomial infections

34. IGAS cases typing results, Skåne, Sweden, 2012
Typing (EMM/PFGE) N %
emmst1 / P7 20
47%
emmst3 / P1 15
35%
emmst28 / P6 2 5%
emmst81 / P2
emmst89 / P5
emmst4 / P3 1 2%
emmst1 / P4

35. Cases of IGAS by date of onset, Skåne, Sweden, 2012

36. Toxin (superantigen) pattern, IGAS, Skåne, Sweden, 2012
PHASE 1 emmst3 / P1
PHASE 2 emmst1 / P7
other subtypes
Total
# (n=15) %
#(n=20)
# (n=8)
# (n=43)
SPE_A 15
100% 20 1
13% 36
84%
SPE_B 8 43
SPE_C 3
38% 38
88%
SPE_F
SPE_G 14
93% 19
95% 7 40
SPE_H 5
33%
63% 29
67%
SPE_I 0%
SPE_J
47%
15% 13
30%
SSA

37. Possible objectives of joint laboratory epidemiology investigations
Test a hypothesis (Qualitative outcome)
Test a hypothesis
About an etiologic agent (e.g., Is West Nile virus the cause of the outbreak?)
About the relatedness of isolates (e.g., Are the cases caused by an identical pathogen?)
Measure a quantity (Quantitative outcome)
Estimate a quantity
Prevalence
Incidence

38. Interpreting prevalence and incidence
A study estimating the frequency of a disease on the basis of a laboratory test (e.g., serological survey) must be interpreted according to:
Predictive value positive
Predictive value negative
These will depend upon:
The test used (sensitivity and specificity)
The frequency of the disease

39. Be careful about what the manufacturer may say about the predictive values
The manufacturer may report values of
Sensitivity
Specificity
These probably come from panel testing
Be careful with values of predictive values positive and negative reported by manufacturers
These values depends upon specific prevalence settings
They may come from a combination of a positive and negative panels that generate an artificial prevalence of 50%

40. Take home message: Interpret epidemiological and laboratory evidence as a team
Positive tests are likely to rule in the diagnosis if the test is specific and the disease is common
Negative tests are likely to rule out the diagnosis if the test is sensitive and the disease is uncommon
Emergent pathogens are discovered in the laboratory and assessed according to additional studies
Laboratory investigations of relatedness must be based on hypotheses developed on the basis of the epidemiology
Interpret incidence and prevalence indicators according to predictive values positive and negative