Epidemiology Lectures for UG

Epidemiology
Causation and Causal Inference
Dr. Amita Kashyap
Sr. Prof. Community Medicine
S.M.S. Med. College, Jaipur

Concept of Sufficient Cause and
Component Causes
• Need to define “cause” – if we define cause as
an antecedent event, condition or
characteristic that was necessary for the
occurrence of the disease at the moment it
occurred, given that other conditions are
fixed……
• This definition provides only a component of a
complete causal mechanism

General Model of Causation
• We begin life as a pragmatic philosopher,
developing a general causal theory; that some
events or states of Nature are causes having
specific effects.
• Though rudimentary it suggests that we are
equipped with curiosity to understand, by logic,
doubt, speculation, and developing methods to
prove (experiments)
• ? - a working knowledge of the essential
system of causal relations that enables each of
us to navigate our complex world

A “Sufficient Cause”
• A “Sufficient Cause”, a complete causal mechanism –
a set of minimal conditions and events that
inevitably produce disease.
• Minimal – implies that all conditions are necessary
Ex. – Tobacco smoking and Lung cancer
• Not all smokers get lung cancer- there are certain
individuals primed by certain unknown conditions
and just adding smoking causes lung cancer. Eg
asbestos exposure
• ? Heavy smokers have approx. a 10% lifetime risk
developing lug cancer

U
U U
A
A B B E
E
Three Sufficient Causes of a Disease – each constellation
(I, II, and III) of component causes is sufficient to produce
the disease
Strength of Effect
II
I I III
The condition under which ‘E’ acts as “necessary and sufficient cause”
= “presence of A or B but not both”

Exposure to
component causes
Response Frequency of Exposure
(combination) 1000 each
A B E Outcome Pop. 1 Pop. 2
1 1 1 1 100 900
1 1 0 1 100 900
1 0 1 1 900 100
1 0 0 0 900 100
0 1 1 1 900 100
0 1 0 0 900 100
0 0 1 0 100 900
0 0 0 0 100 900
Exposure frequencies for three component causes
in two hypothetical populations 1 and 2
B=1, E = 1
B=1, E=0
B=0, E=1
B=0, E=0
Assumption: disease is a non recurrent event, such as death or first occurrence of disease
1 = present; 0 = absent for exposure and Response
The Proportion getting Disease = Numbers getting exposure pattern X response

B = 1, E = 1 B = 1, E = 0 B = 0, E = 1 B = 0, E = 0
CASES 1000 100 900 0
TOTAL 1000 1000 1000 1000
Proportion 1.00 0.10 0.9 0.0
Incidence proportion for combo of “B and E” in Population 1
Incidence proportion for combo of “B and E” in Population 2
B = 1, E = 1 B = 1, E = 0 B = 0, E = 1 B = 0, E = 0
CASES 1000 900 100 0
TOTAL 1000 1000 1000 1000
Proportion 1.00 0.9 0.10 0.00
Why “E” is much stronger determinant in Population 1 ?

Interaction among Causes
• Two component causes acting in the same
‘sufficient cause’ may be thought of as
interacting biologically to produce disease
• This need not to be ‘simultaneous’ – e.g. head
injury leading to Hip fracture??
• The extent or apparent strength of biologic
interaction between two factors is dependent
on the prevalence of some other factors
A
B
C D
E
A
B
F G
H
I II III
C
A
F I
J

Proportion of Disease due to
sufficient cause
• What fraction of disease is caused by ‘U’ if these are
the only sufficient causes to cause a specific disease ?
• The answer is all of it, bcz without ‘U’ there is no
disease, it’s a ‘necessary cause’.
U U
A B E
E
U
A B
I II III

Induction Period – specific cause-effect
pair; not just the effect
• If in ‘Sufficient Cause’ I, the sequence of action of the causes is – A,B,C,D,
and E and we want to study the effect of B (which acts at some narrowly
defined time)
• Disease occurs only after the sequence is completed
• The interval btw the action of B and the disease occurrence is the
induction period for the effect of B
A
B
C D
E
‘Sufficient Cause’ I

Factor B
Disease
Initiation
Disease
occurrence
Disease
Detection
Induction Period
Latent Period Period
We can reduce ‘Latent Period’ by improved methods of
disease detection BUT not the induction period as it ends
with disease occurrence.
!! – Role of Biomarkers (attempt to focus on causes
more proximal to the Disease occurrence)

Philosophy of Scientific Inference
• Inductivism : making generalizations;
observations induce the formulation of a natural
law in scientist’s mind e.g. observation of lack of
smallpox in milkmaids induced in Jenner’s mind
the theory that cowpox confers immunity
against smallpox.
• Based on ‘assumption’; there is no logic or force
of necessity
• Logical Fallacy; (after this therefore on account
of this!!!)

Philosophy of Scientific Inference
• Refutationism :
process of elimination- conjecture and
refutation (no matter how many times we boil
water in an open flask and get boiling point as
100C; we cannot prove that water’s boiling
point is 100C. but one attempt to boil water
in a closed flask or at higher altitude will
refute the preposition that water always boil
at 100C )

• Since it is possible for any observation to be
consistent with many hypotheses that
themselves may be mutually in-consistent;
consistency btw an observation and
hypothesis is no proof of hypothesis
• In contrast, a valid observation that is
inconsistent with an hypothesis refutes it- if
you wring the neck of the rooster before
sunrise; you have disproved that rooster’s
crowing is a necessary cause for sunrise

Bayesianism
• In classic logic, premises of the deductive
argument need to be 100% truth e.g. “if A
implies B, and B is false, then A must be false”
• The conclusion from this argument “A must be
false” will be valid only when assumptions - ‘A
implies B’, and ‘B is false’ are true statements
• All observation about physical world are
subject to some error
• If we can assign some prior probability to our
statements we can use laws of probability to
derive certainty to the conclusion

• Bayesian philosophy provides a methodology
for sound reasoning and, in particular,
provides many warnings against being overly
certain about ones conclusions
• Such warnings are echoed in Refutationist
philosophy : the intensity of the conviction
that a hypothesis is true has no bearing on
whether it is true or not
• Most epidemiologists prefer Interval
Estimates

Consensus
• The ultimate goal of scientific inference is to
capture some objective truths about the
material world
• We may know a theory is false bcz it
consistently fails the tests we put it through
BUT we cannot know that it is true, even if it
passes every test we can devise, for it may fail
a test yet un-devised
• Hence any theory of inference should ideally
be evaluated by how well it leads us to detect
errors in our hypothesis and observation

Consensus
• When confronted with a refuting observation,
a scientist faces the choice of rejecting either
the validity of the theory or its scientific
infrastructure
• Observations that are falsifying instances of
theories may be treated as “anomalies”,
tolerated without falsifying the theory in hope
of future explanations, e.g. observation that
shallow-inhaling smokers had higher lung
cancer rates than deep-inhaling smokers!

Causal inference in Epidemiology
• Epidemiologic hypothesis usually are based on
vague assumptions lacking biologic knowledge
e.g. “smoking causes CVDs”
• To cope with this vagueness, epidemiologists
usually focus on negation of causal
hypothesis, the Null Hypothesis that exposure
does not have causal relation
• Then, any observed association can potentially
refute the hypothesis (ensuring no biases
present)
• Testing Competing Epidemiologic Theories

Causal Criteria
• Since there can not be a set of sufficient
criteria, a list of causal criteria (proposed by
Bradfor Hill 1965) provide road map through
complicated territory
• He suggested some aspects of association to
be considered in attempting to distinguish
causal from non-causal association……

Causal Criteria
1. Strength
2. Consistency
3. Specificity
4. Temporality
5. Biologic gradient
6. Plausibility
7. Coherence
8. Experimental evidence
9. Analogy
Hill’s verdict – none of my nine viewpoints can bring
indisputable evidence for or against the cause and
effect hypothesis.

Conclusion
• Science advances by a process of
elimination; “conjecture and refutation”
• Alternative hypothesis
• Evaluating competent causal theories
using crucial observations

A Patient’S Profile:
• A 60 year old previously healthy female, research
chemist recently developed shortness of breadth and
nosebleeds.
• Pale, pulse 110/ min, low (20%) hematocrit, elevated
(20000/l) leukocyte counts, low platelet (15000/l)
with PBF showing atypical myeloblasts
• Hospitalized for Suspected acute myelogenous
leukemia; confirmed by bone marrow aspirate and
biopsy.
• Chemotherapy started, about 3 weeks later, her temp.
abruptly rose to 39C and neutrophil count dropped to
100 /l.
• No source of apparent infection;

Patient Profile…ctd:
• Importance of Risk assessment!!
• How likely is it that patient has a bacterial
infection?
• Her blood and urine cultures were taken, and
broad spectrum antibiotics administered (empiric
treatment)
• Potential Risk of complications from delayed
antibiotic outweighed empiric use of antibiotic
• Cultures confirmed staphylococcus aureus in blood

Measures of Disease Occurrence
Epidemiologic measures - to assess
outcomes and thereby guide decisions
• Risk (the likelihood that a person will
contract a disease)
• Prevalence (Load; the amount of disease
already present in the population)
• Incidence Rate (how fast is the new
occurrence of disease)

Defined
Population
Have
Disease
Do not
have
disease
Do not have
disease at
baseline
PAR
Prevalent
cases
1. Identify
Population
3. Follow
only those
who did not
have the dis.
2. Determine
who has the Dis.
& who doesn’t
Do not have
disease at
baseline
Develop Dis.
Do not have
disease
Follow up for 1 year
incident
cases

Risk (cumulative incidence)
• It is a measure of the occurrence of new cases
• i.e. Proportion of unaffected persons (PAR) in
the population who, will contract the disease
over a specified period of time
New cases
Person at Risk
R =
• Has no unit;
• lies between 0 and 1

onset end
A
B
C
D
E
F
Hypothetical study of group of six subjects
1995 96 97 98 99 00 01 02 03 04
Dx …………………………………………Death
97 02
99
97
99 02
Dx,,,,,,,,,,,,,,,,,,,,,,,
97
02
What is the Risk of Dis. development within 2 years of enrolment
New cases
R =
Person at risk
= 1/6 = 0.17 OR 17%

Example of HAI in cancer patients
• Estimate a cancer patient’s risk of getting HAI
from a study of 5031 patients admitted in
comprehensive cancer centre.
• If 596 patients met criteria for Hosp. Acquired
infection
• Risk period? - Starts 48 hrs after hospitalization
and ends 48 hrs after discharge.
New cases
R =
Person at risk
= 596/5031 = 0.12 OR 12%

• Can we apply this risk to our patient?
• More likelihood of infection for our patient
can come from studies on similar
subjects…having fever, and low granulocyte
count….
• Now if 1022 such cancer patients were studied
and 530 had HAI then the Risk is 530/1022 =
0.52 i.e. 52%

Measures of Disease Occurrence ctd…
• Prevalence (Burden of Disease)–
indicates number of existing cases of a disease in a
population at a time.
• E.g. An important question in deciding antibiotic
use to the patient is the type and magnitude of
infection anticipated!!
• We know that individuals with low neutrophil
count are susceptible to wide variety of infections…
– S.aureus was cultured from 62 out of 96 patient’s
specimens
• Prev. of S.aureus infection = 62/ 96 = 0.65 i.e. 65%

onset end
A
B
C
D
E
F
1995 96 97 98 99 00 01 02 03 04
Dx …………………………………………Death
97 02
99
97
99 02
Dx,,,,,,,,,,,,,,,,,,,,,,,
97
02
What is the Prevalence of Disease in 2001
Total cases
p =
Total population
= 1/4 = 0.25 OR 25%
B
….left

Measures of Disease Occurrence
ctd…
• Incidence Rate – measures the rapidity with
which new cases of the disease develop.
• Estimated by observing a population and
counting the number of new cases over Net
Time (person years) i.e.
• Incidence Rate = New cases/ person time
• A subject at risk of disease followed for 1 yr, or
5 yrs contributes 1 or 5 person-years of
observation respectively.

onset end
A
B
C
D
E
F
0 1 2 3 4 5 6 7 8 9
Dx …………………………………………Death
97 02
99
97
99 02
Dx,,,,,,,,,,,,,,,,,,,,,,,,
,,,,,,
97
02
How many person years are contributed by A, B, C, D E and F?
04
Total new cases
IR=
Total person years
= 2/22 = 0.09 cases /person years
i.e. 9 cases/ 100 person-yrs
04
04
98
Observation years
95

Example of HAI ctd…
• Those 5031 remained under observation for a
total of 127859 patient days
• What is the average length of stay?
• Since 596 patients developed HAI the IR would
be – 596/ 127263= 0.00468 cases/ patient days
• Can be expressed for better readability as 4.7
cases/ 1000 patient days
• Interpretation: among patients similar to those
studied, on average, about 0.47% patient/day
would be expected to develop a HAI
127859/5031
= 25.41

Calculation of IR for a Large Pop.
• Calculating person-years (PT) for each individual
would be too cumbersome! Alternatively
• PT = (Av. Size of PAR) X (Length of observation)
• In many instances, relatively few people develop
the disease and there is no other demographic
shift hence whole Pop. Can be taken as At
Risk…i.e. not excluding patients
• PT = (Size of entire Pop.) X (Length of observation)
• 596/127859=0.00466 while 596/127263=0.00468 !!!

Calculation of IR for a Large Pop.
• If there are an estimated 1,91,85,836 women in
an area btw 1996 and 2000 (5 yrs period) and
2957 women were newly diagnosed with Acute
myelocytic leukimia (AML)
• What is the annual incidence rate of AML ?
• 1,91,85,836 women x 5 Yrs = 9,59,29,180 WY
• IR = 2957 new cases/ 9,59,29,180 Wyrs =
3.1cases /1,00,000 WY

Characteristic Risk Prevalence Incidence Rate
What is
measured
Probability of
Disease
Proportion of
Pop. With
disease
Rapidity of
Disease
Occurrence
Units None None Cases/ person-
time
Time of disease
Dx
Newly
diagnosed
Existing cases Newly
diagnosed
Synonyms Cumulative
Incidence
- Incidence
Density
Characteristics of Risk, Prevalence & Incidence Rate
In our Hypothetical Ex. In 2001 Prev. was 25%,
2 Yr. Risk was 17% and the IR was 9cases/ 100 yrs

Problems with Incidence and
Prevalence Measurements
• Problems with Enumerator
– The first problem is defining who has the disease.
– The next issue is Method of data collection –
interview, self reporting , survey… associated biases!!
• Problems with Denominators
– everyone in the group represented by the
denominator must have the potential to enter the
group that is represented by the numerator…
PAR concept
• Problems with Hospital Data

Relationship Between Incidence
and Prevalence
• There is an important relationship between
incidence and prevalence: in a steady-state
situation, in which the rates are not changing
and in-migration equals out-migration, the
following equation applies:
• Prevalence = Incidence × Duration of disease

Example
• 2,000 persons are screened for tuberculosis,
Using chest x-rays: 1,000 are upper-income
(HIG) individuals and 1,000 are lower-income
(LIG) individuals.
• X-ray findings are positive in 100 of the HIG
and in 60 of the LIG.
• Can we therefore conclude that the risk of
tuberculosis is higher in HIG people than in
LIG people?

Screened
Population
Point Prev./
1000
Incidence
(Occurrences
/yr)
Duration
(yrs)
Hitown 100 4 25
Lotown 60 20 3
Prevalence = Incidence × Duration

20 30 40 50 60 70 80
0
100
200
300
400
20%
15%
10%
5%
0%
Annual
Rate/
100000
Percent
of
total
cases
Breast cancer incidence rates and distribution of cases by age
Age in yrs
The incidence is increasing so dramatically with
age, why are only fewer than 5% of the cases
occurring in the oldest age group of women?

Incidence increasing but prevalence
decreasing – How?
46
0
5
10
15
20
25
30
35
40
1
9
9
0
1
9
9
3
1
9
9
6
1
9
9
9
Prevalence
Incidence
Fatal, Or short duration

Incidence stable but prevalence
increasing indicates:-
47
0
5
10
15
20
25
30
35
40
45
1
9
9
0
1
9
9
3
1
9
9
6
1
9
9
9
Prevalence
Incidence
New Program or
Better Dx Test !!!
•Death is prevented
and Dis is not cured
• Diagnosed more
•Immigration of cases

Incidence maintained but prevalence
declining means:-
48
0
5
10
15
20
25
30
35
1
9
9
0
1
9
9
2
1
9
9
4
1
9
9
6
1
9
9
8
incidence
prevalence
New effective drug!
Or Dis. Became more
Virulent/ fatal,
Emigration of cases

Incidence Rate:Expressed as-
Morbidity rate-
New cases total population at risk
Mortality rate-
No. Of deaths due to a disease/ total population
Case fatality rate-
No. Of deaths due to a disease/ total no. Of cases of that disease
Attack rate-
No. Of new cases of a disease, during a specified time/ total
population at risk for the same time
Secondary Attack Rate- No. of exposed persons developing disease
within the Range of “IP” following exposure to a Primary Case.

Survival
• Probability of being alive for a specific length of
time
• For a Ch. Dis. Like cancer, 1 and 5 Yr survival
rates are often used as indicator of the severity
of the disease and the prognosis.
• E.g. if 5-Yr survival for AML is 0.19, it means that
only 19% of patients with AML survive at least 5-
Yrs after diagnosis
• Survival = Newly Dx Pts. – Deaths/ Newly Dx Pts.
For a specified time

Dx onset end
A
B
C
D
E
F
0 1 2 3 4 5
Observation years
Patients
Censored
Death
Censored
Death
What is the 2 year survival rate?
2 year survival rate = 5/6 = 0.83 i.e. 83%
What is the 2 year Risk of Death?
2 year Risk of Death = 1/6 = 0.17 i.e. 17%
5 yr S If we assume B & E survive all 5 yrs = 4/6= 0.67=67% !
5 yr S If we assume B & E didn’t survive all 5 yrs = 2/6= 0.33=33% ! !

Methods to account for censored cases
• Life Table analysis
• Kaplan-Meier analysis AND Graphs
0 1 2 3 4 5
20
40
60
80
100
0
Survivors
Percent
Years since Dx
47%
68%
58%
? Median Survival Time
50

Case Fatality
• The propensity of a disease to cause Death
• If N = 15 and 5 of whom develop disease of
concern , then Risk = 5/15= 0.33 = 33%
• If only 2 of the affected die CF = 2/5 = 0.40 = 40%
• Survival = incident cases – death /total affected
• = 5-3 / 5 = 3/5 = 0.6 = 60% i.e. 100 – CF = Survival
Number of Deaths
CF =
No. of Dx Cases
New cases
R =
Person at risk

onset end
A
B
C
D
E
F
0 1 2 3 4 5 6 7 8 9
Dx …………………………………………Death
97 02
99
97
99 02
Dx,,,,,,,,,,,,,,,,,,,,,,,,
,,,,,,
97
02
How many person years are contributed by A, B, C, D E and F?
04
Total new cases
IR=
Total person years
= 2/22 = 0.09 cases /person years
i.e. 9 cases/ 100 person-yrs
04
04
98
Observation years
95
2 person yrs
2 person yrs
2 person yrs
3person yrs
7 person yrs
6 person yrs

Comparing disease occurrence
(in groups with different exposures )
• To calculate the Risk that a health effect will
result from an exposure
• Risk Difference (Excess Risk)- expressed as:-
Incidence in exposed - Incidence in un-exposed
Smoking category Stroke cases Person yrs of
observation
Stroke IR/ 100000
Person yrs
Never smoked 70 3,95,594 17.7
Ex-smoker 65 2,32,712 27.9
Smoker 139 2,80,141 49.6
total 274 9,08,447 30.2
= 49.6 – 17.7 = 39.1/ 100,000 person yrs

Comparing disease occurrence
(in groups with different exposures )
• Attributable Fraction (exposed) – proportion
of cases that can be attributed to exposure
Incidence in exposed - Incidence in un-exposed
/ Incidence in exposed = (49.6 – 17.7/ 49.6) X 100
= 64%
Indicating 64% Risk Reduction if exposure is
removed

• Population Attributable Risk –
determine relative importance of
exposure for entire population
= incidence in total population –
incidence among un-exposed / incidence
in total population

• Relative Risk –
ratio of the risk of occurrence of
disease among exposed people to that
among un-exposed people (baseline
level of exposure) e.g.
(in our Ex. = 49.6/17.7 = 2.8)
• Good indicator of strength of
association because it is expressed
relative to baseline level of exposure

Measures of Mortality:
• Mortality rate
–Crude death rate
–Cause specific death rate
–Age specific death rate
• Case-fatality rate
• Proportionate mortality rate
• Standardized Mortality Rates

Adjusted Rates: Standardization
• Standardization:
– The process to derive a summary figure to
compare health outcomes of groups
–The process can be used for mortality,
natality, or morbidity data, race
• Standardization Methods
–Direct
–Indirect

Example: Age-Adjustment
A. Direct Method: requires –
1. Age-specific rates in the sample population
a) The age of each case
b)The population-at-risk for each age group
in the sample
2. Age structure of a standard population
Summary figure is an Age-adjusted rate

Direct Age Adjustment
Population 1 Population 2
Population No. of
Deaths
Death
rate/
100000
Population No. of
Deaths
Death
rate/
100000
900000 862 96 900000 1130 126
Standard Population can be taken from outside or
both population can be clubbed to get Standard Population

Direct Age Adjustment:
Comparison of Age specific death rates
Population 1 Population 2
Age
Gr.
popula
tion
No. of
Deaths
Death
Rate/
100000
popula
tion
No. of
Deaths
Death
Rate/
100000
All
ages
900000 862 96 900000 1130 126
30-49 500000 60 12 300000 30 10
50-69 300000 396 132 400000 400 100
70+ 100000 406 406 200000 700 350

Direct Age Adjustment: using total of
two pop. As standard Population
Age
Group
Standard
Population
1996-2000
Age
specific
mortality
rates
Expected
no. of
deaths /
100000
2001-2005
Age
specific
mortality
rates
Expected
no. of
deaths /
100000
All Ages 1800000
30-49 800000 12 96
(8 x 12)
10 80
50-69 700000 132 924
(7 x 132)
100 700
70+ 300000 406 1218 350 1050
Total 2238 1830
2238 1830
Age adjusted Rate = ---------- X 100000 = 124.3, --------- X 100000 = 101.7
1800000 1800000

B. Indirect method: requires
1. Age structure of the sample population
2. Total deaths in the sample population
3. Age-specific rates for the standard
population
4. No need for stratum-specific rates of the
sample
Summary figure is a Standardized
Mortality ratio (SMR)

Indirect Standardization
• Stratum specific Death rates of standard population are
applied to each stratum of the sample population to get
Expected Deaths
• Overall DR of sample population from records gives
Observed Deaths
Observed
SMR = ----------------- X 100
Expected
SMR of 100 means no difference between the
number of outcomes in the sample population
and that which would be expected in the
standard population

Indirect Standardization (cont.)
Total expected deaths per year: 2,083
Total observed deaths per year: 1,464 (from Records)
SMR = 1,464 / 2,083 x 100 = 70.3% (30% less than expected)
Age
Group
Number people
(Census, 2001)
Standard Death
Rates per 1,000,000
(All Causes of Death)
Expected Number of
Deaths per 1,000,000
(1) (2) (3) = (1) X (2)/ 1,000,000
20-24 7,989 1,383 11
25-34 37,030 1,594 59
35-44 60,838 2,868 174
45-54 68,687 8,212 564
55-64 55,565 22,953 1,275
2,083

Patterns of occurrence
• Distribution Patterns (TPP analysis)of a disease
within a population
– Who develops the disease? (Person)
– Where does the disease occur? (Place)
– When does the disease occur ? (Time)
• Level (rate of occurrence)- Endemic or Epidemic
• Causal Role - Genetic or environmental

Patient profile
• A 30 yr old female domestic worker; recently
migrated from India to USA presented with 6 weeks
h/o cough, fever, night sweats, weakness, fatigue
and shortness of breath.
• h/o two normal deliveries followed by Tubal ligation
• Chest X-ray shows cavity lesions, sputum is AFB +ve
and mycobacterium grew on culture which was
sensitive to all drugs
• Administered 4 drugs under DOTS

• After 2 months put on 2 drugs 3 times a week
as she was asymptomatic with no bacilli in
sputum.
• She resided in a low town apartment building,
tuberculin test was done on her husband and
two children
• Results were +ve for her husband and 3 yr old
• Although no active disease was found yet
prophylactic Tt was given to all three of them
• Out of 54 neighbors; 1 was infected without any
evidence of clinical disease and received PT
• None of the work place contacts were +ve

Environment Infectious Individual Susceptible
Individual
Close contacts of
infected, susceptible
people in close
spaces
Pulmonary or
Laryngeal disease
with bacilli in sputum
Compromised
immune system
Poor ventilation Forceful cough with
uncovered mouth
Predisposing disease
or condition e.g.
silicosis, cancer
Recirculation 0f
contaminated air
Less than 2-3 weeks
of appropriate anti-
microbial therapy
Lack of adequate
Nutrition
Injectable drug use
or heavy alcohol
intake
Factors that increase the probability of T.B. transmission

Epidemiology
Observational
Epidemiology
Experimental
Epidemiology
Descriptive
Epidemiology
Analytical
Epidemiology
Hypothesis about disease causation
Leads to
Help explain descriptive patterns (possible causes)
And to improve Dis. Surveillance
Verified by Analytical/ Experimental Studies

Person
(disease do not occur at random!)
• Variation of occurrence in relation to
personal characteristics reflects:
– differences in level of exposure to causal
factors,
–susceptibility to causal factors,
–or both.
• Personal characteristics include…….

946
1499
5286
4191
3147
0
1000
2000
3000
4000
5000
6000
up to 14 15-24 25-44 45-64 65+
cases
Age in Years
Number of T.B. cases by Age in a year...
? From where does such Data Comes
Notification of diseases!! List is updated as per
changing scenario/ needs. And Surveillance data.
Interpretation? Highest Risk in 25-44 yrs??

Surveillance
• “Ongoing systematic collection, analysis,
and interpretation of data essential for-
– planning, implementation, and evaluation of
public health practice closely integrated with
the timely Feedback.”
• Types - Passive or Active
• Help to Know- Changes in either disease
rates or levels of environmental risk factors

Surveillance goals
• Identification of patterns of disease
occurrence
• Detecting disease outbreaks at nascent age
• Development of clues for possible Risk
Factors
• Anticipation of health service needs
• Finding cases for further investigation

1.5
3.7
6.2 6.3
8.8
0
1
2
3
4
5
6
7
8
9
10
0-14 15-24 25-44 45-64 65+
incidence
per
100000
person
years
Age in Years
Incidence Rates for Reported T.B. Cases
Incidence among persons in oldest age group is
over 40% higher than that for 25-44 years group
Possible factors
Contributing
•Long latent period
•Elderly lived through
times when T.B. was
rampant (Birth Cohort
Effect)
•Other illness like DM
and Cancers more in
elderly
•Declined immunity in old
age
•More chances of living
in Closed settings
How to interpret ?
Higher incidence rate in
certain minority group!!
Gender Differences !!!

Place (spot maps!)
• International
• National
• State and/ or
• Local comparisons gives insight to probable
reasons
– Estimated 8 million people develop T.B. each year
worldwide
– 95% of these comes from developing world
– Most rapid rise of T.B. IR is in sub Saharan Africa!

• Usual rate of occurrence – endemic rate
• A rapid and dramatic increase over the
endemic rate is - epidemic rate
• Epidemic can develops in a matter of days
or weeks (few hrs for staphylococcal food
poisoning) but for chronic condition like
cancer it takes years to decades
• Establishing linkages between RF and
Disease Occurrence become difficult if
there is greater time lag (latent period)

Endemic vs. Epidemic
Endemic Epidemic
No.
of
Cases
of
a
Disease
Time
Usually plot weekly

No rapid rise in incidence of T.B. but departure from decline!!

Correlation with Disease Occurrence
• To develop hypothesis about possible
causes of disease,
– Presence of a suspected RF is measured
in different populations and compared
with incidence of disease (Ecologic Study)
–Examine extent to which two
characteristics are related e.g. (RF and
disease occurrence)

Incidence rates of TB and AIDS in 15 States of USA
r = 0.91, coefficient of determination (r2) = 0.98
Regression equation = a + b*x
Tuberculosis IR = - 0.8 + 0.57 X AIDs IR
Ecologic fallacy!!

Migration studies
• To clarify whether a disease of unknown
cause is determined principally by genetic
inheritance or environmental exposure
• For diseases with long latent periods, it
may take years for the reduced rate of
occurrence
• If environmental exposure early in life is
critical, then effects may be visible in
offspring's only!

Ways to interpret data and generate Hypothesis

Risk
Ratio
No Racial
Difference

Investigating An Outbreak
Dr. Amita kashyap
96

What is a Disease Outbreak?
Outbreak vs Epidemic
What does it Require?
A pathogen in sufficient quantities,
A mode of transmission,
And a pool of susceptible people
97

98
A Scenario!
A 23 yr old male student; presented at 10:30 pm
on 17th Jan 2014, at the emergency complaining
of a sudden onset of abdominal cramping,
nausea and diarrhea. He was weak, not
severely distressed, had no fever or vomiting.
A No. of other students, all with the same
symptoms, visited emergency over next 20 Hrs
All treated with Fluid replacement recovered
fully within 24 hrs. of the onset of illness.

Does it warrant an
investigation?
• Why?
99

When should we Investigate?
• Number and severity of persons
affected!
• Uncertainty about cause!
• Level of Public Concern/ Political
pressure!!
• Potential for contributing to medical
knowledge! 100

101
Reasons for Outbreak Investigation
Quantifying the epidemic (Descriptive
epidemiology)
Getting at the source and reasons (Analytic
epidemiology)
for
Preventing others from becoming affected

Investigation in our scenario!
Quick information revealed 47 students out of
1164 college enrollment got affected by 8 PM
on 18th Jan (20 Hrs)
 What is the quantitative measure of the extent
of an outbreak?
No. of New Cases
AR = Persons at Risk
What is the AR for this period?
= 47/ 1164 X 100 = 4% 103

Hostel wise distribution of 47 known cases, AR,
population and sex of the occupants of each hostel
Hostel Sex PAR No. of
Cases
AR
1 F 80 19 23.8
2 F 62 2 3.2
3 F 89 0 0
4 F 61 1 1.6
5 F 53 5 9.4
6 M 35 0 0
7 M 63 0 0
8 F 103 4 3.9
9 M 35 1 2.9
10 M 37 0 0
11 F 34 1 2.9
12 M 62 13 21.0
13 M 32 1 3.1
14 M 10 0 0
Total - 756 47 6.2
Attack Rate (all students)
= 47/ 1164 X 100= 4%
Attack Rate (hostellers)
= 47/ 756X100= 6.2%
Attack Rate (hostel 1, 12)
= 19+13/ 80+62 = 22.5%
Attack Rate (other hostels)
= 15/ 614 = 2.4%
Risk Ratio = AR hostel (1, 12)
/ AR (Other hostels) X 100
= 22.5/ 2.4 = 9.4
? Sex difference in AR =

Further :
Visit to hostels revealed that not all
students who became ill reported to
emergency.
Needed un-baised data- hence…
Seven hostels were randomly selected for
information collection on desired areas! 105

Response to the questionnaire survey by hostels
Questionnaire returned
Selected
Hostel
Population Number Percent No. of ill St.
5 53 49 92.5 13
6 35 26 74.3 13
7 63 28 44.4 15
8 103 65 63.1 21
9 35 19 54.3 5
12 62 44 71.0 22
Nurses’ hostel 60 60 100 17
Unidentified - 13 - 4
Total 411 304 74.0 110
106
AR = 110/304 X100 = 36.2%
Note: initial hostel wise AR for Hostel 6, and 12 were 0% and 21%
As per survey data ! - AR (H6) =13/26X100=50% and AR (H12) =22/44X100 = 50%

• AR of hostel 6 and 12 were 0% and 21% by
emergency data but by survey data both are
50% - Approach for data collection!
• Was emergency data useless?
• Is 36.2% the true AR of AGE on campus ?
• Explain factors why AR estimated from
emergency records were low?
• Why more cases from hostel 1 and 12 at
emergency?
107

Additional information…..
• No large gathering of students..... hence inquiries were
made about meals eaten on 16th and 17th Jan
• Most students ate at college cafeteria
• How will you zero down to source of infection?
108
St. who ate specific meal St. who did not eat specific meal
Ill Well Total AR(%) Ill Well Total AR(%)
Jan 16
Breakfast 52 100 152 34.2 51 94 145 35.2
Lunch 89 150 239 37.2 20 44 64 31.3
Dinner 87 150 237 36.7 23 44 67 34.3
Jan 17
Breakfast 56 105 161 34.8 42 89 131 32.1
Lunch 106 145 251 42.2 3 49 52 5.8 RR!
Dinner 78 130 208 37.5 31 64 95 32.6
42.2/ 5.8=7.3

Can we now calculate IP?
• Having identified the meal at which the
students most probably were exposed to the
causal pathogen and
• Knowing each student’s time of food
consumption and onset of symptoms; we can!!
109
IP(hrs) No. of Students Cumulative No. of St.
8 22 22
9 11 33
10 18 51
11 8 59
12 42 101

0
10
20
30
40
22
8
11
9
18
10
08
11
42
12
Time (hours)
Number
of
cases
0
20
40
60
80
22
8
33
9
51
10
59
11
100
12
Time (hours)
Cumulative
frequency
%
100
Median I P = 10 hrs

What next?
A follow up survey to obtain information about
particular foods that 251 students ate at lunch
on Jan 17!
If students were uncertain about whether they
ate or not the meal in question, they were not
included in the analysis of the particular food.
 As a result total of those who ate or did not eat
each specific item did not equal 251 for all items
111

Food specific histories of students who ate lunch at
the college cafeteria on Jan 17th
Food/ beverage St. who ate Sp. Food /
Beverage
St. who did not eat Sp. Food /
Beverage
Ill Well Total AR (%) Ill Well Total AR (%)
Fish Curry 16 36 52 30.8 87 103 190 45.8
Lamb Gravy (RR = 8) 95 56 151 62.9 7 82 89 7.9
Chicken noodle 12 57 69 17.4 92 80 172 53.5
Dal Fry 58 54 112 51.8 39 69 108 36.1
Fruit salad 32 39 71 45.1 63 82 145 43.4
Cabbage salad 4 5 9 44.4 95 126 221 43.0
Plain vanilla Ice cream 19 29 48 39.6 80 102 182 44.0
Rabri 62 77 139 44.6 39 56 95 41.1
Milk 91 127 218 41.7 12 13 25 48.0
Coffee 10 31 41 24.4 89 103 192 46.4
tea 23 19 42 54.8 78 114 192 40.6

Further investigation -
• About preparation of Lamb Gravy revealed
that it was cooked on 16th Jan, refrigerated
and warmed on the morning of 17th Jan
• Now, even without Lab investigation we can
speculate the etiologic agent? Cl. perfringens
• Suggesting features:-
–Gastrointestinal symptoms without fever
and vomiting
–Median I P is 10 Hrs
–Meat Gravy Dish is the most likely food
113

115
• This is the most common form of transmission in food-
borne disease, in which a large population is exposed
for a short period of time.
Point Source Transmission

116
• In this case, there are several peaks, and the
incubation period cannot be identified.
Continuing Common Source or
Intermittent Exposure

Warning Signals of an impending outbreak
• Clustering of cases/ deaths in Time/Place
• Unusual increase in cases/ deaths
• Even a single case of measles , AFP, Cholera, Plague,
Dengue, or JE
• Ac. febrile illness of unknown etiology
• Two or more epidemiologically linked cases of
outbreak potential
• Unusual isolates
• Shifting in age
• High or sudden increase in vector density

Unusual
Health Event
No
Yes
Is this an
outbreak
Etiology, Source
& Transmission
known?
No
Yes
Institute control
measures
Further Investigation
Describe outbreak
in terms of TPP
Continued….

Develop Hypothesis regarding
Source, Transmission, Etiology & PAR
yes No
Does the
Hypothesis
Fit with facts
Institute control
measures
Special studies
Remember that outbreak is usually
a sudden & unexpected event!
There is need to act quickly.
A systematic Approach Helps

Epidemic preparedness
• Formation & Training of RRT
• Regular review of data
• Alertness during known ‘outbreak season’
• Identifying outbreak prone areas
• Ensuring that these areas have necessary drugs
and materials (including transport media)
• Identifying & strengthening the labs
• Designating vehicles
• Ensuring communication channels

Steps of Outbreak Investigation
• Verification of the outbreak
• Sending the RRT
• Monitoring the situation
• Response to an outbreak
• Interim report by RRT within one week
• Declaring the outbreak to be over
• Final report & its Review within 10 days of the
outbreak declared to be over

Nullification of source
Minimizing transmission Protecting
the host
Response to outbreak

Analysis
• Analyze and interpret - within 24 hours
• Identify EWS
• Frequency count by reporting unit helps in identifying
outbreaks or potential outbreaks
• Analysis in terms of person, time and place will be able to
focus the intervention;.
• During an outbreak, analysis of the data identifies the most
appropriate and timely control measures.
• Analysis of routine data provides information for predicting
changes of disease rates over time and enables appropriate
action.
Data compilation/analysis and response should be at
all levels.

Feedback
Essential to maintain know-how, moral and
support the peripheral staff.
Monthly Feed back Report should be sent
regularly even when there are no
epidemics
Feed back report should also be provided
on the quality of data submitted to the
district surveillance officer

Epidemiology Lectures for UG

More Related Content

What's hot (20)

Similar to Epidemiology Lectures for UG (20)

More from amitakashyap1 (20)

Recently uploaded (20)

Epidemiology Lectures for UG