Documente Academic
Documente Profesional
Documente Cultură
Epi= Upon
Demos= People
Logos= Study of (Body of Knowledge)
Descriptive
Analytic
Final
Descriptive Epidemiology
Characteristics of Time
Characteristics of Person
Age
Gender
Ethnicity / Race
Marital status
Socio-economic status (SES)
Occupation
Final
Characteristics of Place
Descriptive Epidemiology
Types of study:
Case Reports
Case Series
Correlation studies
Cross sectional studies = Community health survey
Final
Descriptive vs Analytic
DESCRIPTIVE ANALYTIC
EPIDEMIOLOGY EPIDEMIOLOGY
Analytic Epidemiology
Host Factors
Agent Environment
Agents
Nutrients
Allergens
Radiation
Physical trauma
Microbes
Psychological experiences
Host Factors
Genetic endowment
Immune System (e.g. vaccinated)
Nutritional Status
Environment
Temperature
Sanitation
Pollution of water / air
Population density
Disease Transmission
Modes of communication
Direct Indirect
transmission transmission
Trans-placental (vertical)
Modes of transmission
Modes of transmission
Vector
An animate intermediary in disease transmission.
Most vectors are arthropods such as mosquitoes,
fleas, or ticks.
Vehicle
Inanimate objects such as food, water, biologic
products (e.g. blood), and fomites that may
indirectly transmit an infectious agent from a
reservoir to a host.
Frequency Measurements
Incidence & Prevalence
This Lecture
When?
Annual occurrence, seasonal variation,
Passive Surveillance
Reporting of cases by health care providers on a periodic and consistent
way. Usually thru legislatively mandated reporting of certain conditions; not
actively seeking new cases
Sentinel Surveillance
Monitoring rate of occurrence of specific conditions to assess stability or
change in health levels of a population; by specific orgs to know how
disease patterns have changed
Count and Ratio
Count, Proportion,
Ratio and Rates
Count
Numerator
Denominator
In Epidemiology:
Prevalence
= # of existing cases at or over a point in time
aka total
Incidence rate
The rate of:
new events or incidents
(disease, injury or death)
in population at risk
during a period of time
=Often an estimate
Example:
x = developed disease
-- = time followed
Incidence Density = 2 / 8
= 25 per 100 person-years
A Prospective Study of post-menopausal hormones and coronary heart
disease
NEJM 313:1044, 1985
Can be measured at
a specific point in time (point prevalence) or
over a specified period of time (period prevalence)
Point vs. Period Prevalence
a) Correct
b) Incorrect, the comparison is not based on rates
c) incorrect, because no control or comparison group is used
d) incorrect, because prevalence is used instead of incidence
Incidence and Prevalence
P I x D !!!
Duration is shorter ie. Cold, prevalence of it will be lower too
Paradox between
Incidence and Prevalence
Example
30% of all deaths from myocardial infarction
occur within 24 hours of the onset of symptoms
in people having no prior evidence of disease
Incidence and Prevalence
Disease of long duration are well represented in a
prevalence study, even when there incidence is
low.
Attack Rate
Morbidity Rate
Mortality Rate
Case Fatality Rate
Relative Risk
Attributable Risk
Attack Rates
Number of events among a population at risk in a period of
time
10 attack rate =
20 attack rate in the families =
Question: Calculate primary and secondary attack rates for the
following epidemic of hepatitis A
Incidence Rate
Attack Rate
Secondary Attack Rate
Point Prevalence
Period Prevalence
Mortality Frequency Measures
But usually:
denominator is midpoint pop so that makes it an
estimate & not an exact figure
Crude Mortality Rates
28 to 47 750
> 48 1000
Total 1800
Example : Age Specific Mortality Rate
Proportionate Mortality
Final
Case-Fatality Rate Final
Case Fatality is
18 / 20 = 0.9 or 90%
Example: Calculation of
Infant Mortality Rates
Yes a b a+b
Exposure
No
c d c+d
AR = Ie Io (subtract!)
Ie is incidence in the exposed
Io is incidence in un-exposed
RR: How much more likely; How much the risk for a
patient who smokes increased compared to a non smoker.
Descriptive Studies
Case Report and Case series
Cross-sectional Studies
Correlational Studies
Case series
Cross-Sectional Surveys
May describe:
Previously described disorder involving a new
population or subgroup
Known disorder with unusual clinical
presentation or clinical course
Condition in which novel tx methods are used
Examples:
Obesity and TV viewing
Alcohol and CHD
Hypertension and physical inactivity
Final
Cross-sectional survey of coronary heart disease
Not 89 14 157.2/1000
physically
active
Physically 90 3 33.3/1000
active
Depression
3.5
20 2.5
3 10 1.5
2.5
2
55 3.8
1.5 60 3.7
1
0.5
50 3.1
0
0 10 20 30 40 50 60 70
35 2.8
Is there a linear relationship between BMI and
Body weight?
3.5
2.5
1.5
0.5
0
0 10 20 30 40 50 60 70
Linear Regression or
Least Squares
Using this method, we choose a line such that the sum of
squares of vertical distances of all points from the line is
minimized.
Final
Correlation Coefficient
Final
Coefficient of Determination = r2
Another important measure of linear
association between x and y (0 r2 1)
r =1 r =0.9
r =0.5 r=0
Correlation and linear regression
Cost effective
Time
B 0.22
C +0.72
D 0.85
E0
The postnatal weight gain (Y) in kilograms over a specified
period of time was related to varied amounts of a formula
supplement (X), in unit doses, taken during the same
period for a sample of 50 infants. The following results
were obtained:
An effective
Descriptive Analytic Intervention
Epidemiology Epidemiology =
Prevention
An injury story
A B C D
Biologic Disease detectable Symptoms Recovery
onset by screening develop Disability
or
Death
Detectable
pre-symptomatic phase
Primary Prevention:
Modifying risk factors or eliminating causes
Phase:
Before disease is present
Avoids the development of disease
Strategy:
Reduce or remove risk factors
Educate on health promotion
Good nutrition
Vaccinations
Regular exercise Final
Methods of Primary Prevention:
Health promotion
Promoting Breastfeeding
Final
Methods of Primary Prevention:
Health promotion
Environment:
Final
Methods of Primary Prevention:
Specific protection
Final
Secondary Prevention
Phase:
Disease present
Strategy:
Diagnose the disease early
Prompt treatment
Arrest diseases process Final
Prevent disability
Methods of Secondary Prevention
Early diagnosis and treatment
Screening tests
- Pap smear
- Colonoscopy
- mammography
Regular checkups
Final
Tertiary Prevention
Phase:
Disease present; clinical course after occurrence of
disease
Strategy:
Restore normal / near normal functioning
Reduce fatalities and complications
Final
Methods of Tertiary Prevention
Disability limitation
Rehabilitation
Medical Treatment / Therapy
Prosthetics
Physiotherapy
Final
The Prevention Plan
Smoking
Obesity
Institute a Rational Health Care System
Eliminate Health Disparities
Clean up and protect the Environment
Prevention and Epidemiology
Descriptive epidemiology measures
The extent of certain disease (frequencies)
Who (where and when) is at risk (TPP)
The consequences of this disease (Incidence, Case Fatality,
Prevalence=Disease Load)
Analytic epidemiology
What are the risk factors involved
(= determinants of disease)
Prevent, detect, reduce
Cohort Studies Case control studies
- Starts w/group of exposed ppl & non- - Starts w/group of ppl already
exposed ppl then follow them to see how w/disease & group w/out disease
many ppl develop disease & how many did then look at their histories to see how
not many were* exposed to risk factor & how
- Prospective: only go fwd from date of study many were not
- Retrospective: events have already occurred; - Calculate odds ratio b/c no known
go back, even when you do this however, still incidence rate (for relative risk)
start study w/exposure present or not; & then
see who developed disease & who didnt (as AKA CCO
very similar to case control) - Adv: easy to study rare disease, cheap,
- calculate relative risk b/c known incidence short duration (less time consuming)
- Disadv: difficult to establish temporal
rate AKA CSR association, biased due to confounders
- Adv: able to study temporal association
(exposures precedes disease can be clearly
established), study rare exposures, can KNOW WHAT TYPE OF
control confounders (preventing bias)
- Disadv: time consuming & expensive, high STUDY IT IS; MORE QUES
drop out rate ON THAT THAN CALCS!!!
- Nested case control study: start w/exposed
& unexposed, eventually get ppl w/disease &
no disease & then go back into history to Final
study other factors of interest, once these
grouped so develops into case control
Clinical trials: ppl w/disease used as reference Cross-over study: pts in group A & B receives
pop expal group divided into diff drugs; then wait (wash out period) 1 yr
- Drug group switch drugs given to each
- Placebo group - adv: each n can compare themselves (as
subjects become their own controls), need
Placed by randomization to reduce selection less ppl
bias, reduce effect of known & unknown - 2 types of clinical analysis in clinical trials
confounders - Intention to treat: drug & placebo
Blinding groups aka ALL participants will
Single-blind: participants blinded, be analyzed, even if non-
investigators know compliant, taken into analysis
Double-blind: both dont know
produces/reveals effectiveness of
Triple-blind: both + analysts dont
know who got drug & who got placebo drug!**More VALUE
Phases of clinical trial: - Explanatory analysis: drug &
1. Preclinical phase: animal studies placebo groups but ONLY
2. Phase I: healthy inds compliant pts analyzed; reveals
3. Phase II: diseased inds, ie. drug for HTN efficacy of drug! More reptd by
is it really reducing the disease (HTN) in drug companies
pt? Is it valid for that particular disease?
4. Phase III: randomized control trial
5. Phase IV: after drug released into market to
see any other efx
STUDY conducted to know association of alcoholism & cirrhosis of liver;
total of 300 agreed to participate in study; 100 pts w/cirrhosis selected &
matched w/200 ppl w/out cirrhosis; interviewed & asked about their
alcohol consumption in past. Of 100 pts who had cirrhosis, 80 were
alcoholics. In remaining 200 w/out cirrhosis, 40 were alcoholics. What is
the type of study conducted? Calculate measurement of risk.
= [80*160] / [40*20]
100 200
Disease No disease
Exposed 80 40
Unexposed 20 160
In 1945, study conducted to know efx of exposure of Dutch famine to
fetus & its association w/childhood & adult illnesses. 500 pregnant women
who were exposed to famine & IDd & 500 women who werent exposed to
famine were also studied. Children born to both groups were followed. It
was observed that more # of kids that were born to famine exposed
women devd early CV & respiratory diseases.
Also called:
Longitudinal studies
Follow-up studies
Incidence Studies
Cohort Studies
Examples:
Group of smokers
Follow up Follow up
Disease Disease Does Totals Incidence
develops Not Develop
Exposed A B A+B A/A+B
Not exposed C D C+D C/ C+D
PROSPECTIVE (FUTURE
STUDIES)
RETROSPECTIVE
Prospective Cohort Study
Exposure Disease
?
Final
= Concurrent Cohort
= Longitudinal Study
Cohort Study: Lead level and Affective Disorders
not exposed 5 95
Exposure Disease
?
= Non-concurrent
Cohort
= Historical cohort
Study
FAMOUS COHORT STUDIES
The Framingham Study
Time
Money expensive
Hospital records
Bias
Loss to follow up
Non participation
Bias in cohort study
Observational
Case-control
Cohort
Experimental
Randomized control trial
Case-control Study
Final
Case-control Study Design - benefits
Time & money issues: cost efficient & less time to complete
Final
Case-Control Design - issues
Final
Design of Case-control Studies
Final
I. Definition and Selection of Cases
Neighborhood controls
Best friend control (same habits + limiting
confounders)
Spouse or sibling control
Selection of Hospitalized Controls
Disease:
Hospital records, case-registries, pathology log books
etc
Exposure:
Interview, mail questionnaires, medical records etc
IV. Stratified Analysis
Disadvantage:
It is extremely cumbersome
Disease (n = 50)
LC (n = 50) No LC (n = 150)
Smokers 45 60
Exposure
Non-smokers 5 90
ODDS RATIO
Cases Controls
with LC without LC
smokers A=45 B=60
nonsmokers C=5 D=90
Odds Ratio =
ratio of odds of exposure among cases
ratio of odds of exposure among controls
Case Control
Yes
a b
Exposure
No
c d
Odds Ratio = ad
bc
Odds Ratio (OR)
Final
Selection Bias
= sampling bias
Sample selected differs in properties in cases
and controls
Final
Recall Bias - example
Pygmalion effect
Confounder
Exposure Disease
Final
Confounders
Final
Controlling for Confounding
Study design:
Matching
Restriction
Analysis:
Stratified analysis
Multivariate analysis
Final
Matching
Cases and controls are matched by usual
confounders (e.g. age, sex, SES, smoking, alcohol
etc.) so that these factors are equally distributed in
both groups and will not confound the association
between the variables
Volunteers Volunteers
RANDOMIZE
Experimental population
Participants
Final
Conducting a Clinical Trial
Do necessary exclusions
Random assignment
Outcome measurement
Final
Selection of Study Population
Final
Blinding
Final
Blinding
Final
Uncontrolled trials
Sample size
Analysis of data
Post-randomization Changes in
Comparison Groups
Migration bias:
Study participants may drop out, switch
Tx groups, become non-compliant
Compliance bias:
Inds in 1 tx arm may drop out at higher
rates due to factors such as side efx
Analysis of Data
Study population
Drug A Placebo
1000 1000
250 250
cured cured
Drug A Placebo
1000 1000
250 250
cured cured
Explanatory analysis means that the cure rate for Drug A is calculated as
250/800 = 31%
The cure rate for placebo arm is 250/1000 = 25%
Informed consent
Protecting the interests of the patient
Withholding treatment known to be effective
Monitoring for toxicity and adverse effects
Stopping rules
When to withdraw a patient from study
Informed Consent
Final
Stopping Rules
Reliability
Validity
Reliability aka Reproducibility
Present Absent
+ +
Results of
Positive a b
True +ve False +ve
Screening test
- -
Negative
c d
False ve True ve
Final
Screening Tests
Disease status
Present Absent
True False
Positive
Results of positive positive
Screening Test
Negative False True
negative negative
Sensitivity
The proportion of persons w/disease whore correctly
identified by test
Positive TP or a FP or b
Test
Negative
Test FN or c TN or d
Positive TP or a FP or b
Test
Negative
Test FN or c TN or d
Final
Population distribution of intraocular
pressures in those with and without
Low specificity Glaucoma
High specificity
High sensitivity Low sensitivity
IDd out
Hence less false +ves
Final
Positive Predictive Value = PPV
Final
Positive Predictive Value
Disease
Present Absent
Positive TP or a FP or b
Test
Negative
Test FN or c TN or d
Final
TP/(TP+FP)
Measures only the distribution of persons with a positive test
Uses data from the top row
Negative Predictive Value
Positive TP or a FP or b
Test
Negative
Test FN or c TN or d
Final
TN/(TN+FN)
Measures only the distribution of persons with a negative test
Uses data from the bottom row
Two examples
Disease
(as determined by Final
"Gold standard")
Present Absent
Sensitivity Specificity
FOB screen test is used in 203 people to look for bowel cancer:
Patients with bowel cancer
(as confirmed on endoscopy) Final
Present Absent
= TP / (TP + FP)
Pos TP = 2 FP = 18 = 2 / (2 + 18)
FOB = 2 / 20 10%
test = TN / (TN + FN)
Neg FN = 1 TN = 182 182 / (1 + 182)
= 182 / 183 99.5%
= TN / (FP + TN)
= TP / (TP + FN)
= 182 / (18 + 182)
= 2 / (2 + 1)
= 182 / 200
= 2 / 3 66.67%
PPV=10%: Positive test is poor at confirming 91%
cancer
Sensitivity: It will pickup 66.7% of all cancers
Specificity: As initial screen it correctly identifies 91% of those who do not have cancer
NPV=99.5%: As a screening, a negative result is very good at reassuring a patient
does not have cancer
Breast Cancer Detection and Implications for
Periodicity of Screening. Am J. Epi 100: 357-366,1974
Breast Cancer
Present Not Present
Negative 47 62,295
Positive
132 985
Screening Test
Negative 47 62,295
(a+c)/ (a+b+c+d)
Predictive Value,
specificity & prevalence
The Positive Predictive Value
the probability that if the test is positive, the patient truly has the
disease
depends on:
the Specificity and
even more on Prevalence of the disease
Positive
900 4950
Screening Test
Prevalence 1%
Negative 100 94,050
Positive
900 1980
Screening Test
Prevalence 1%
Negative 100 97,020
Positive
4,500 4,750
Screening Test
Prevalence 5%
Negative 500 90,250
0.1 1.8 90 95
1.0 15.4 90 95
5.0 48.6 90 95
50 94.7 90 95
Key Points
If the prevalence of a disease low, then PV+ will be
low even if you have a test w/high sensitivity &
specificity
AGE 35 40 41 43 45
I I I I I
Biologic Disease Patient A Symptoms A&B
onset of detectable diagnosed develop: both
disease by screen at screening B diagnosed die
A. 7% Whats sensitivity?
B. 17% 30% 30/ [30+70] = 30/100 =
T+ F+
C. 18% F- T-
N = 1000
D. 30%
True +ve = 30 Total true = 100
E. 83% True ve = 750
30 150
70 750
False ve = 70 Total false = 900
False +ve = 150
Cut point of a screening test intended for
detecting Diabetes Mellitus was lowered from
140 mg to 130 mg of glucose per dL of blood.
Change in the cut point for this screening test
would
A. Increase specificity
B. Increase sensitivity
C. Decrease true positive rate
D. No change in sensitivity or specificity
since the number of people with DM will
remain the same
Confidence Interval
for Relative Risk and Odds ratio
Confidence Interval
(in terms of RR or OR)
Provides an interval range around the odds ratio or
the relative risk and represents the range within
which the true magnitude of effect lies
Final
8 Hills Criteria:
1. Study design
2. Strength of Association
3. Consistency
4. Correct Temporal Relationship clearly established or not?
5. Dose Response Relationship
6. Plausibility is there a known scientific explanation?
7. Specificity
8. Analogy
Final
1. Study Design
For example:
Many studies of different designs (case control, cohort, case
series etc.), using different subjects, found an association
between smoking and lung cancer
4. Correct Temporal Relationship
Final
5. Dose Response Relationship
Final
6. Biologic Plausibility
Do Not
Develop Develop
Disease Disease
Cases Subgroup
Selected as
Controls
CASE-CONTROL STUDY
Evidence based medicine
EBM not only identifies which txs are effective but also those which are ineffective and may do
more harm than good, and identifies areas where more investigation is needed and where there may
be gaps in knowledge
Steps for practicing EBM
Step 1: Formulating a well built question
Clinicians seek to assemble the best and most up-to-date evidence with
which to answer that question.
Objectives
To examine the effects of DMT for depression with or without standard care, compared to no treatment or standard care alone, psychological
therapies, drug treatment, or other physical interventions. Also, to compare the effectiveness of different DMT approaches.
Search methods
The Cochrane Depression, Anxiety and Neurosis Review Group's Specialised Register (CCDANCTR-Studies and CCDANCTR-References) and
CINAHL were searched (to 2 Oct 2014) together with the World Health Organization's International Clinical Trials Registry Platform (WHO ICTRP)
and ClinicalTrials.gov. The review authors also searched the Allied and Complementary Medicine Database (AMED), the Education Resources
Information Center (ERIC) and Dissertation Abstracts (to August 2013), handsearched bibliographies, contacted professional associations, educational
programmes and dance therapy experts worldwide.
Selection criteria
Inclusion criteria were: randomised controlled trials (RCTs) studying outcomes for people of any age with depression as defined by the trialist, with at
least one group being DMT. DMT was defined as: participatory dance movement with clear psychotherapeutic intent, facilitated by an individual with a
level of training that could be reasonably expected within the country in which the trial was conducted. For example, in the USA this would either be a
trainee, or qualified and credentialed by the American Dance Therapy Association (ADTA). In the UK, the therapist would either be in training with, or
accredited by, the Association for Dance Movement Psychotherapy (ADMP, UK). Similar professional bodies exist in Europe, but in some countries
(e.g. China) where the profession is in development, a lower level of qualification would mirror the situation some decades previously in the USA or
UK. Hence, the review authors accepted a relevant professional qualification (e.g. nursing or psychodynamic therapies) plus a clear description of the
treatment that would indicate its adherence to published guidelines including Levy 1992, ADMP UK 2015, Meekums 2002, and Karkou 2006.
Main results
Three studies totalling 147 participants (107 adults and 40 adolescents) met the inclusion criteria. Seventy-four participants took
part in DMT treatment, while 73 comprised the control groups. Two studies included male and female adults with depression. One
of these studies included outpatient participants; the other study was conducted with inpatients at an urban hospital. The third
study reported findings with female adolescents in a middle-school setting. All included studies collected continuous data using
two different depression measures: the clinician-completed Hamilton Depression Rating Scale (HAM-D); and the Symptom
Checklist-90-R (SCL-90-R) (self-rating scale).
Statistical heterogeneity was identified between the three studies. There was no reliable effect of DMT on depression (SMD -0.67
95% CI -1.40 to 0.05; very low quality evidence). A planned subgroup analysis indicated a positive effect in adults, across two
studies, 107 participants, but this failed to meet clinical significance (SMD -7.33 95% CI -9.92 to -4.73).
One adult study reported drop-out rates, found to be non-significant with an odds ratio of 1.82 [95% CI 0.35 to 9.45]; low quality
evidence. One study measured social functioning, demonstrating a large positive effect (MD -6.80 95 % CI -11.44 to -2.16; very
low quality evidence), but this result was imprecise. One study showed no effect in either direction for quality of life (0.30 95% CI
-0.60 to 1.20; low quality evidence) or self esteem (1.70 95% CI -2.36 to 5.76; low quality evidence).
Authors' conclusions
The low-quality evidence from three small trials with 147 participants does not allow any firm conclusions to be drawn regarding
the effectiveness of DMT for depression. Larger trials of high methodological quality are needed to assess DMT for depression,
with economic analyses and acceptability measures and for all age groups.
Meta-analysis
Frequency Distribution
Measures of Central Location
Measures of Variance
Why Study Statistics?
Researchers
Consumers of medical research
Statistics in Medical Research
The goal:
To design the process and extent of sampling
in order to form valid and accurate inferences
Real differences?
Due to chance?
Frequency Distribution
- Numerical
- Pictorial
Always: values of the variables on horizontal axis and
their frequencies on the vertical axis
Frequency Distribution;
Graphical Display: Pie Chart
Frequency Distribution;
Graphical Display
Frequency Polygon:
The midpoints of the top
of each bar of the
histogram are plotted
and connected with
straight lines.
This makes it easier to
put two or more sets of
data on same graph
Shape of distributions
Properties of frequency
distribution:
- Variation or Dispersion
Shape of distribution
Symmetric
- Normal distribution (Gaussian Curve)
Skewed
- Tail to the right: positively skewed
- Tail to the left: negatively skewed
Symmetric Distribution
15
RBC cholinesterase
mmol/min/ml
10 Frequency
Freq
5
0
5.95-7.95 7.95-9.95 9.95-11.95 11.95- 13.95-
13.95 15.95
Cholinesterase levels
RBC Cholinesterase
5.
0
5
10
15
95
-7
.9
7. 5
95
-9
.9
9.
95 5
-1
11 1.
95
.9
5 -1
13 3.
95
.9
5 -1
15 5.
95
.9
5 -1
7.
95
Negatively skewed
Frequency
RBC Cholinesterase
5.
95
0
2
4
6
8
10
12
14
16
-7
.9
7. 5
95
-9
9. .9
95 5
-1
11 1.
.9 95
5-
13
13 .9
.9 5
5-
15
15 .9
.9 5
5-
17
.9
5
Positively skewed
Frequency
Normal distribution
Symmetric (bell-shaped) curve
Measures of central tendency
1, 4, 6, 3, 2, 7, 9, 11, 5, 10, 8?
Question:
Calculate mean, median and mode
c) Mode = 2
Measures of central tendency
Variation or Dispersion
Properties of frequency
distribution:
- Variation or Dispersion
Measures of Spread or Variation
RANGE
VARIANCE
STANDARD DEVIATION
Range:
-Arrange the data in ascending order
-Find out the maximum and minimum values
- Maximum value minimum value
Measures of Variation
Standard deviation
Measure of absolute variation in a given data
set, and a supplement to the mean
Measures of Variation
Standard Deviation
s2 = (xi - x)2
n-1
Measures of Variation
BIRTHWEIGHT
2000
1000
Frequency
0 N = 9747.00
25 75 12 17 22 27 32 37 42 47 52 57 62
0. 0. 50 50 50 50 50 50 50 50 50 50 50
0 0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0 .0
BIRTHWEIGHT
Statistics
BIRTHWEIGHT
N Valid 9747
Mis sing 0
Mean 3367.19
Median 3405.00
Mode 3430
Std. Deviation 623.36
Variance 388574.54
Minim um 312
Maxim um 6605
Percentiles 25 3061.00
50 3405.00
75 3749.00
Birth weight
Birth weight is approximately normally
distributed (bell-shaped curve)
Normal Distribution
Blood pressure
Why do we need to know this?
If a particular data shows a normal distribution, we
can apply the specific characteristics of normal
distribution to it
Normal Distribution
What is normal distribution?
It is a theoretically perfect frequency polygon which:
2. Is symmetric
Normal Distribution
Biostatistics II
The Normal Distribution
The Normal Distribution
Questions
Test B 60 40 10
MCQ
The distribution of factor X in a population of
men ages 20-40 follows a multi-modal
distribution with mean of 20 mg/dl and
standard deviation 2 mg/dl
95% of men will have Factor X levels between
ranges
A. 16-24
B. 18-22
C. 15-25
D. Can not be calculated
Bimodal distribution of height
= 2 modes
Statistical Inference
And SEM = SD / n
Example
length of stay in a Patient Length of
hospital for 5 patients Stay
1 3
we want to calculate 2 5
the 95% CI 3 2
4 3
5 2
Patient Length of Deviation (xi - x)2
Stay (xi - x)
1 3 3-3=0 0
Mean=3
2 5 5-3=2 4
3 2 2-3=-1 1
4 3 3-3=0 0
5 2 2-3=-1 1
Applying Z=(x x) / s
= 112-120 /6 = -1.33
Samples and Populations
Normal Range :
Mean +/- 2SD
Statistics I
Both SEM and confidence interval indicate how
precise (or imprecise) our estimate is.
The standard error of the mean, is based on variability
in data (the standard deviation) and the size of the
sample
SEM = SD / n
SAMPLING
STATISTICAL SIGNIFICANCE
ERRORS
T TEST
VARIABLES
Confidence interval
Statistical Significance
Confidence intervals and p-values are
used to demonstrate statistical significance
P-value< 0.05 and 95% C.I. Both state
that a result as extreme as the one
obtained is likely to have occurred by
chance only 5% of the time
i.e. you can assume that the result is
unlikely to have occurred by chance
How to use CI around Relative Risk
or Odds Ratio as a measure of
statistical significance?
Testing for statistical difference
using CI
Look at the CI carefully
Types of hypothesis:
-Null hypothesis (Ho)
--Alternative hypothesis (Ha)
Hypothesis Testing
The way we compute the test statistic, depends on sample size, type
of variables , & sometimes shape of population distribution.
Type of variable
t-test
p-value t-value
0.00001 15.890287
Smoking and Birth weight
Mean birth weights:
born to non-smokers is 3417 grams (~ 7 lbs. 8 oz.)
Chi-Square test
Scenario: As clinicians, knowing that LBW is
associated with increased morbidity and mortality, we
are concerned that pregnant patients who smoke are
more likely to deliver a low birth weight (LBW) infant.
Chi-Squares P-values
----------- --------
Uncorrected: 61.45 0.000001
Mantel-Haenszel: 61.44 0.000001
Yates corrected: 60.71 0.000001
Smoking and LBW
Definition:
The occurrence of cases of an illness
in excess of expectancy
Identify the existence of
the outbreak
No. of cases
10
9
8
7
6
5
4
3
2
1
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
Date
Propagated or
Person-person Outbreak
35
30
25
# CASES 20
15
10
5
0
1 4 7 10 13 16 19 22 25 28
Continuous Source Outbreak
No. of cases
10
9
8
7
6
5
4
3
2
1
9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
Date
Initial assessment / action
Data collection
Time
incl. epidemic curve
Place Remember:
incl. place of residence, use standard format for
work, travel etc. data collection
Case-control study
Starting point: disease status (ill - not ill)
Find odds of exposure of cases and non-cases
Calculate odds ratios: ad/bc
Test for statistical significance
Calculate confidence interval of odds ratio
Look for possible confounding
Environmental inspection
Environmental inspection
Get info on usual practices
Inspect premises and practices
Take samples
Get info on food storage and handling (cold chain,
hot chain)
Get info on personnel (disease history?)
Ask for maps and plans
Need for specialist health inspector
Have your eyes open for clues from unexpected
sources
Control Measures
Remove source
Isolate / treat cases
Destroy food, recall products
Stop production, close premises
Ensure good practice procedures
Protect persons at risk
General hygiene
Vaccination
Other prophylaxis
Initial assessment/action
Make recommendations
Produce guidelines
Make proposals for change in law
Communication
4. Define cases
Establish a case definition
5. Identify cases
Identify and count cases by Line listing
6. Describe and orient the data in terms of
time, place and person
Outbreak curve
Map
Identify demographic and other characteristics of
persons at risk
12 Steps of Outbreak Investigation
Analyze
7. Develop hypotheses
Open-ended and wide-ranging interviews with a few
people
8. Evaluate hypotheses
Comparison: hypotheses with established facts
Analytic epidemiology
Cohort studies (RR; 95% CI)
NNT = 1/ARR
Answer:
NNT = 13
Number needed to harm (NNH)
= # of PPL, who must be exposed to something in order for one of them to experience
an adverse effect
55 out of 75 people died due to usage of an experimental drug. Among 75 people who
took placebo only 35 of the them died. What is the number needed to harm?
HIGHER VALUE = BETTER; TAKES LONGER TO GET ADVERSE EFX ON PT
NNH = 4
CLINICAL PROBABILITY
Example:
Chance of having a brown hair is 0.3
Chance of getting a cold is 0.2
What is the chance of meeting brown haired person
with a cold?
0.30.2 = 0.06
Addition rule
addition rule of probability states that probability of
any one of several particular events occurring is
equal to the sum of their individual probabilities,
provided the events are mutually exclusive (i.e. they
cannot happen at one time)
0.25+0.25 = 0.5
Chi-square value and statistical
significance
Suppose the calculated value was 4.25
Degree of freedom was 1
P < 0.05
Chi-square value
If the calculated value was 2.5
Degree of freedom was 1