Introduction to Epidemiology: a tour

Epidemiology is the study of distribution and determinants of diseases in populations and use of this knowledge for the improvement in public health

Epidemiology in triads

Epidemiology triad

• Distribution of diseases
• Determinants of diseases
• Use of the information

Distribution of diseases

• Person
• Place
• Time

Distribution of diseases person

Distribution of diseases with place

Distribution of diseases over time

Measures of disease distribution

• Prevalence (proportion)
• Incidence (rate)
• Ratio (standardised mortality ratio)

Concept of prevalence

• total cases of a disease in the population OVER
• total number of people there
• MULTIPLIED by factor of 100 (percent), 1000 or 10, 000
• At a FIXED PERIOD of TIME

Example of prevalence

Advantage of prevalence

• Simple snapshot of a health condition in the population
• For a single population, at one point in time
• You can use it to compare two or more populations

Limitation of prevalence

• Does not provide any information about how the disease spreads in the community
• Is the disease increasing?
• Is it getting worse?
• Is it getting better?

Question for the class - what data do we need?

• What additional data do we need for that?
• (Question for the class, what do you think?)

Concept of Incidence

• How many NEW cases of the disease in the community
• How many people were AT RISK?
• Over WHAt period of time?

Concept of Person-time

• If you follow 1 person for 1 year,
• You get 1 person-year
• If you follow 100 people for 1 year,
• You get 100 person-year
• If you follow 100 people for 10 years,
• You get 1000 person-years

Question for the class

Let's say we follow 1000 people for 5 years, which of the following is correct?

• (X) We have 5000 person-years
• (Y) We have 1000 person-years
• (Z) We have 5 person-years

Concept of incidence rate

• Incidence is a rate because it has TIME as DENOMINATOR
• Number of NEW CASES OVER Person-years
• REQUIRES FOLLOW UP of people

Example of incidence rate

Question for the class 2

Based on the data presented below, is the incidence

• (A) increasing or
• (B) Falling off

Incidence of the disease in the data

Advantage of Incidence rates

• Used for charting epidemics
• Helps you to understand whether an epidemic is getting better or worse
• Helps to chart data in real time

Example of incidence rate in real life

Question for the class - 3

How can we use the information on the COVID Epidemic Curve?

• (1) Test whether the infection is rising or falling
• (2) Get an idea when to introduce lockdown or other containment measures
• (3) Find out whether lockdown or containment measures are working or not
• (4) All of the above

Limitation of Incidence

• We need longitudinal data
• Without follow-up data we cannot estimate incidence
• Incidence is too simple if we want to compare different populations that differ in age groups

Question for class - which population has higher incidence?

Concept of age-standardised rates

Question for class - which population has NOW higher incidence?

Summary for distribution of diseases

• Three measures: prevalence, incidence, and standardised ratios
• Prevalence is used for static time
• Incidence is used in the context of person-year as denominator
• Incidnece is used for measuring how disease increases or decreases over time
• Standardised ratios used to compare two different populations

Break for 10 minutes

Determinants of diseases

What does determinants of diseases in populations mean?

• This means what cause diseases in populations?
• Examples of some questions:
• Does cigarette smoking cause lung cancer? How do we know?
• Does long term sitting and sedentary lifestyle lead to heart diseases?
• Do particular food items such as chicken salad lead to gastroenteritis?
• How do we know?

How do we know that X causes Y?

Observe facts –> Frame theories –> Test with new facts

What is meant by exposure and outcome

• Consider a disease, any disease, call it O
• We all Y an OUTCOME (health outcome)
• Examples: Diabetes, High blood pressure, lung cancer, so on
• Consider something to which people are exposed
• Example: air pollution, smoking, drinking, gambling, reckless driving
• All of these are called EXPOSURE, give it a name E

What is association?

• If high levels of E leads to high levels of O, then
• We call that a positive association, or RISK
• That is, those with high levels of E will have high incidence of O
• Example: people who smoke lots of cigarettes end up with lung cancer
• We say cigarette smoking is associated with lung cancer
• Or, we say Smoking is a RISK for Lung Cancer

How do we measure associations?

• Risk Ratio (RR) Or
• Odds Ratio (OR)
• Risk Ratio = Risk of Disease among Exposed OVER Risk of Disease among non-Exposed
• Odds Ratio = Odds of Exposure among Diseased OVER Odds of Exposure among those without the disease

Decisions about RRs and ORs

• If RR > 1, the risk is high, OR, association is positive
• Otherwise, if RR = 1, then we cannot say anything
• Else, if RR < 1, there is BENEFICIAL effect!

Question for the Class - Is this high risk?

You conducted a study on cigarette smoking and risk of lung cancer, and found RR = 2.50; what would you say?

• (1) Cigarette smoking increases the risk of lung cancer
• (2) Cigarette smoking has no effect on lung cancer

Establishment of cause and effect

• If we have to show exposure E is a cause of disease O, then
• Show that E has TRUE or REAL association with O
• And show that,
• that Association is one of CAUSE and EFFECT

Four things to consider

• Valid Association: did not occur due to chance (rule out chance)
• Observed association could not be due to biases (eliminate biases)
• Observed association cannot be due to a third factor (confounding)
• Examine causal factors using Hill's criteria
• Examine counterfactual theories of causation

Rule out the play of chance for exposure disease relationship

• Test with hypothesis testing
• Null Hypothesis: that there is equal chances of disease with and without exposure
• Alternative hypothesis: risk of disease is higher with exposure
• Always test with null hypothesis
• Test p-values and 95% confidence interval

What is meant by null hypothesis? (Example: smoking and lung cancer)

• Exposure and disease outcomes are unrelated
• People who smoke and who are non-smokers get lung cancer at the same rate
• Rate of lung cancer among smokers = Rate of smoking among non-smokers
• All epidemiological research is about disproving the null hypothesis

Question for the class - what is the correct null hypothesis

Imagine you are investigating whether smoking cause lung cancer. What is the correct null hypothesis?

• (X) Cigarette smoking DOES NOT cause Lung Cancer
• (Y) There is NO ASSOCIATION between cigarette smoking and lung cancer
• (Z) The RISK of Lung Cancer is same for Smokers and Non-smokers

p-value and 95%Confidence Interval

• You completed a study of an Exposure E and outcome O
• You were to repeat the study a 100 times!
• You found a measurement of RISK (RR) of 2.5
• You found a p-value of 0.02
• You found 95% Confidence Interval of 2.1 - 4.2
• What do these things mean?

Question for the class - what does RR of 2.5 mean?

• (A) E is high risk for O
• (B) E has no association with O

Concept of p-value

• Assuming that the null hypothesis is TRUE,
• and suppose you ran this study 100 times,
• Then only in 2 out of those 100 studies,
• you might get the kind of high risk RR you got
• You can reject the null hypothesis

Concept of 95% Confidence Interval

• If you repeated this study 100 times,
• In 95 out of 100 times,
• You might get an RR value between 2.1 and 4.2
• Most likely 2.5

Question for the class - What does RR of 2.1 mean?

• (A) E is high risk for O
• (B) E has no association with O

How do we rule out the play of chance?

• Select a large enough sample suitable for the effect you want to study
• Perform sample size estimation and power calculation ahead of the study
• Deal with it during the planning of the study

What is bias?

• Bias = Systematic errors in observation or conduct of the study
• Selection Bias: Where the groups are not comparable the way they were identified
• Response Bias: When the participants of the study provide erroneous information

How can you eliminate biases in the study design?

• In the study design phase,
• The investigator should be careful about selection of the sample
• In experimental studies, use randomisation and blinding
• Train the data collectors in the study

Tests of association: control for confounding variable

What is a confounding variable?

Example of a confounding variable

Explanation why Age is a confounding variable

• In the study on the association between smoking and lung cancer,
• Age is a confounding variable
• Old people tend to smoke more than younger people, AND
• Old age is also a risk factor for ANY cancer!
• Smoking CANNOT CAUSE Aging!

How do we control for confounding variable?

• Matching
• Multivariable analysis
• Allocation to groups being compared using Random Numbers Table

How do we find causal linkage using Bradford-Hill Criteria

Break for a couple of minutes

Putting these ideas together: study designs

Epidemiological study designs

• Single Case studies
• Case series (used in Epidemic Surveillance)
• Cross-sectional surveys
• Case control studies (Most widely used study designs)
• Cohort studies

Case control studies

Cohort studies

Rounding up everything we learned with Snow's Cholera investigation

Cholera Epidemic

https://www.youtube.com/watch?v=KvHL0dHj3RM

The ghost map

Snow's table

Summary