Biostatistics &

Epidemiology
0

Notes by MedSN
 BIOSTATISTICS & EPIDEMIOLOGY

CANCER SCREENING .................................................................................................................................. 1

CANCERS WHICH YOU DO NOT SCREEN FOR .................................................................................................... 1
SCREENING FOR MEDICAL CONDITIONS ......................................................................................................... 2
VACCINATION .......................................................................................................................................... 3
TYPES OF STUDIES .................................................................................................................................... 4
DIAGNOSTIC TESTS ................................................................................................................................... 5
CUT-OFF................................................................................................................................................. 6
INCIDENCE .............................................................................................................................................. 7
PREVALENCE ........................................................................................................................................... 7
SPECIFIC MEASURE OF DISEASE OCCURRENCE................................................................................................. 8
ATTRIBUTABLE RISK .................................................................................................................................. 9
POPULATION ATTRIBUTABLE RISK ................................................................................................................ 9
CHOOSING CONTROL IN CASE-CONTROL STUDIES ............................................................................................ 9
BIAS .................................................................................................................................................... 10
FUNNEL PLOT ........................................................................................................................................ 11
SYSTEMIC ERRORS .................................................................................................................................. 12
CONFOUNDING ...................................................................................................................................... 13
CONFOUNDING VS RISK MODIFICATION ...................................................................................................... 13
POWER OF A STUDY ................................................................................................................................ 14
HAZARD RATIO ...................................................................................................................................... 14
STUDY TYPES......................................................................................................................................... 15
TEMPORALITY OF DIFFERENT STUDY DESIGNS ............................................................................................... 16
FACTORIAL DESIGN STUDY ....................................................................................................................... 17
CLUSTER ANALYSIS ................................................................................................................................. 17
POST-HOC ANALYSIS .............................................................................................................................. 17
NON-INFERIORITY AND SUPERIORITY STUDY................................................................................................. 18
CASE-CONTROL VS RETROSPECTIVE COHORT STUDY....................................................................................... 19
ROC (RECEIVER OPERATING CHARACTERISTIC CURVES) .................................................................................. 20
CHOOSING TEST ..................................................................................................................................... 21
MCNEMAR TEST .................................................................................................................................... 21
MEAN, MEDIAN, MODE .......................................................................................................................... 22
EXTERNAL VALIDITY ................................................................................................................................ 22
INTERNAL VALIDITY ................................................................................................................................ 22

1
 BIOSTATISTICS & EPIDEMIOLOGY

CANCER SCREENING
CANCER WHEN TO START HOW TO DO WHEN TO STOP
50 years of age Colonoscopy ever 10 years or 75 years of age
10 years before diagnosis of
a parent Flexible sigmoidoscopy every 5
Colon years + FOBT every 3 years or

FOBT every year

Breast 50 years of age Mammography every 2 years 75 years of age
21 years of age Pap smear every 3 years or 65 years of age + 3
consecutive normal pap
Cervical Pap smear every 5 years + HPV smears
testing every 5 years
Total abdominal
hysterectomy
55-80 years of age + ≥30 Low dose CT every year 80 years of age
Lung pack years + Quit smoking
<15 years Quit smoking >15 years

CANCERS WHICH YOU DO NOT SCREEN FOR

1. Prostate cancer
a. Because of lead time bias, screening for it is useless
2. Ovarian cancer:
a. Normally we do not screen for it
b. However, screening should be done if the patient is BRCA positive

1
 BIOSTATISTICS & EPIDEMIOLOGY

SCREENING FOR MEDICAL CONDITIONS

CONDITION WHO TO SCREEN HOW TO SCREEN
Abdominal aortic Male + ≥65 years + ever smoked 1-time ultrasound of abdomen
aneurysm in their lifetime CT scan of abdomen done for
some other condition
Osteoporosis Female + ≥65 DEXA scan (start giving
bisphosphonates if positive)
Hepatitis C Baby boomers (1945-1965) Hepatitis C antibodies
HIV Everyone once or anyone who ELISA
asks for it
HTN Everyone at every visit Ambulatory monitoring
Diabetes Mellitus Hypertensive patients HbA1c
Male ≥35 years
Female ≥45 years

Hyperlipidemic Anyone ≥20 years with risk Lipid panel

diseases factors for coronary artery
disease

Depression Geriatric patients PHQ-9

Mobility Geriatric patients Get up and go

2
 BIOSTATISTICS & EPIDEMIOLOGY

VACCINATION
VACCINE WHEN TO GIVE CONTRAINDICATIONS
Every 10 years (3 lifetime doses)
>3 <10 years since Clean
lifetime last vaccination the
doses wound

Tdap >10 years since Give Td Anaphylaxis

last vaccination booster
<3 <5 years since last Give Td
lifetime vaccination booster
doses >5 years since last Give Td
vaccination + IVIG
Pneumococcal One 13-valent at age <60 Do not give 13 and 23-valent at the
One 23-valent at age >65 same time
Do not give in
Zoster Age >60 immunocompromised patients
If there is egg allergy, instead of
Flu Give to everybody, every year using IM form, use the intranasal
(live attenuated) form
HPV Everyone age 9-26 years (once) -
Meningococcal Age: 11-18 (preferably 11-12) -
Give booster at age 16-21 if primary
vaccination was done at age <16
Give to adolescents in college
dormitories or in army

3
 BIOSTATISTICS & EPIDEMIOLOGY

TYPES OF STUDIES

4
 BIOSTATISTICS & EPIDEMIOLOGY

DIAGNOSTIC TESTS
1. Sensitive tests:
a. Have many FP
b. SN-N-OUT (sensitive tests, if negative, rule out the disease)
2. Specific tests:
a. Have very few FP
b. SP-P-IN (specific tests, if positive, rule in the disease)
3. Positive predictive value:
a. With increasing prevalence of a disease, the PPV increases
4. Precision:
a. It means that you can reproduce your results and they are consistent
5. Accuracy:
a. It means how much the test results agree with the gold standard test
6. While writing the formula, make sure that TRUE (positive or negative) is the numerator while
sum is the denominator

5
 BIOSTATISTICS & EPIDEMIOLOGY

CUT-OFF

6
 BIOSTATISTICS & EPIDEMIOLOGY

INCIDENCE
1. Incidence is the measure of new cases that develop in a population over a certain period of
time
2. Measuring incidence:
a. Make sure you define the TIME PERIOD during which new cases is counted
b. Make sure you include only the population at risk of acquiring the disease

PREVALENCE
1. It is a measure of the total number of cases (new and old) measured at a particular point in
time
a. Conceptualize it as a ‘snapshot’ of the number of diseased individuals at a given time

7
 BIOSTATISTICS & EPIDEMIOLOGY

SPECIFIC MEASURE OF DISEASE OCCURRENCE

MEASURE FORMULA
Crude Mortality Rate
𝑁𝑁𝑁𝑁.𝑜𝑜𝑜𝑜 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑ℎ𝑠𝑠
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠

Cause-specific Mortality Rate

𝑁𝑁𝑁𝑁.𝑜𝑜𝑜𝑜 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑ℎ𝑠𝑠 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑎𝑎 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠

Case-fatality rate
𝑁𝑁𝑁𝑁.𝑜𝑜𝑜𝑜 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑ℎ𝑠𝑠 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
𝑁𝑁𝑁𝑁.𝑜𝑜𝑜𝑜 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑏𝑏𝑏𝑏 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑

Standardized Mortality Ratio (SMR)

𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂𝑂 𝑛𝑛𝑛𝑛.𝑜𝑜𝑜𝑜 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑ℎ𝑠𝑠
𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸 𝑛𝑛𝑛𝑛.𝑜𝑜𝑜𝑜 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑ℎ𝑠𝑠

Attack Rate
𝑁𝑁𝑁𝑁.𝑜𝑜𝑜𝑜 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑤𝑤𝑤𝑤𝑤𝑤ℎ 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑎𝑎𝑎𝑎 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟

Maternal Mortality Rate

𝑁𝑁𝑁𝑁.𝑜𝑜𝑜𝑜 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑ℎ𝑠𝑠
𝑁𝑁𝑁𝑁.𝑜𝑜𝑜𝑜 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏ℎ𝑠𝑠

Crude Birth Rate

𝑁𝑁𝑁𝑁.𝑜𝑜𝑜𝑜 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏ℎ𝑠𝑠
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠

8
 BIOSTATISTICS & EPIDEMIOLOGY

ATTRIBUTABLE RISK
1. You can think of it as “risk difference”
2. It is a measure of the excess incidence of a disease due to a particular factor (ie, exposure)
a. Measure it by taking the difference between risk of exposed vs risk of unexposed people
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒−𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢
3. Attributable risk % is calculated by:
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑖𝑖𝑖𝑖 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒
𝑅𝑅𝑅𝑅−1
a. It can also be calculated by:
𝑅𝑅𝑅𝑅

POPULATION ATTRIBUTABLE RISK

1. Population attributable risk is the impact of exposure on the entire study population
2. Population attributable risk = Risk of developing disease in exposed – risk of
developing disease in whole population
𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅
3. Population attributable risk % =
𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 𝑜𝑜𝑜𝑜 𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑𝑑 𝑖𝑖𝑖𝑖 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝

CHOOSING CONTROL IN CASE-CONTROL STUDIES

1. The control group is chosen from the same population
2. The control should not have the disease under study
3. The control should not be chosen on the basis of exposure because comparing the exposure
status in cases and controls underlies the analysis

9
 BIOSTATISTICS & EPIDEMIOLOGY

BIAS
BIAS DETAIL
Selection Bias It results from selection of study subjects that are not
representative of the study population

Referral Bias It results when patients are sampled from specialized

medical centers and therefore, they do not represent the
general population
Non-response Bias It results when study design allows subjects to decide
whether or not to participate in the study
Prevalence Bias (Neyman It results when incidence of a disease is estimated based
Bias) on prevalence, and data become skewed selective
survival
Measurement Bias It results from inaccurate estimation of exposure and/or
outcome
Recall Bias It results mainly in case-control studies where effect of
exposure may be overestimated
Observer Bias
(Ascertainment Bias, It results when the investigator’s decision is adversely
Detection Bias, Assessment affected by knowledge of the exposure status
Bias)
Attrition Bias It is a type of selection bias in which there is a loss to
follow-up which occurs disproportionately between
exposed and non-exposed subjects
Surveillance Bias It occurs when the exposed group undergoes increased
monitoring relative to the general population. This tends
to increase disease diagnoses compared to the general
population
Ecologic Fallacy This is a bias which occurs when population-level
information is applied to an individual level

Volunteer Bias This is a bias which occurs when a sample of volunteers

misrepresents its target population and threatens
generalizability (ie, external validity)

10
 BIOSTATISTICS & EPIDEMIOLOGY

FUNNEL PLOT
1. It is a plot which is helpful in assessment of publication bias
2. Each study’s treatment effect is plotted against a measure of that study’s size or precision
(usually using the standard error of the treatment effect)
3. The triangle is centered on a summary estimate of the treatment effect and with its sides
corresponding to standard errors
4. If there is no bias, 95% of studies should lie within the triangle
a. A larger sample size increases precision so small studies will be scattered widely at the
base of the triangle whereas larger studies (more powerful) will be at the top and have a
narrow spread
5. Funnel plots should be symmetric in the absence of study heterogeneity and publication bias

11
 BIOSTATISTICS & EPIDEMIOLOGY

SYSTEMIC ERRORS

12
 BIOSTATISTICS & EPIDEMIOLOGY

CONFOUNDING

CONFOUNDING VS RISK MODIFICATION

13
 BIOSTATISTICS & EPIDEMIOLOGY

POWER OF A STUDY
1. It depends on following:
a. Alpha level (statistical significance level):
i. Lowering the alpha level decreases the power of the study
ii. Decreasing alpha level increases type II error (beta error)
b. The magnitude of difference in outcome between the study groups:
i. Subtle difference is more difficult to detect than a big difference
c. Sample size:
i. Increasing sample size increases the power of the study

HAZARD RATIO
1. Hazard ratio is the ratio of an event rate occurring in the treatment group compared to an event
rate occurring in the non-treatment group
a. <1 = treatment group has lower event rate than control group
b. >1 = treatment group has higher event rate than control group
c. Hazard ratio equal to or close to 1 implies there is little or no difference between the
event rate in both groups

14
 BIOSTATISTICS & EPIDEMIOLOGY

STUDY TYPES

15
 BIOSTATISTICS & EPIDEMIOLOGY

TEMPORALITY OF DIFFERENT STUDY DESIGNS

16
 BIOSTATISTICS & EPIDEMIOLOGY

FACTORIAL DESIGN STUDY

1. A factorial design involves 2 or more experimental interventions, each with 2 or more variables
that are studied independently.

CLUSTER ANALYSIS
1. Cluster analysis is the grouping of different data point into similar categories
2. It involves randomization at the level of groups rather that at the level of individuals

POST-HOC ANALYSIS
1. It refers to performing unplanned statistical tests on patterns that were identified after the fact
in data from a completed study

17
 BIOSTATISTICS & EPIDEMIOLOGY

NON-INFERIORITY AND SUPERIORITY STUDY

1. Non-inferiority Study:
a. In this study, the goal is to prove that a new dug is not unacceptably worse than a
comparator. This means that the new drug should not be worse by more than an
acceptable margin
i. The non-inferiority margin is used for this purpose and any effect estimate that
lies entirely to the left of the non-inferiority margin is “not non-inferior”.
ii. If the effect estimate lies to the right of the margin, it is non-inferior
iii. If the confidence interval crosses the margin, this implies that non-inferiority is
not proven
2. Superiority Study:
a. In this study the goal is to prove that a new drug is better than the comparator
i. A vertical line centered at 0 is used for this
ii. All effect estimates to the right of the line imply the drug is superior
iii. All effect estimates to the left of the line imply the drug is not superior
iv. If confidence interval crosses the line, superiority cannot be proven

18
 BIOSTATISTICS & EPIDEMIOLOGY

CASE-CONTROL VS RETROSPECTIVE COHORT STUDY

1. In retrospective cohort study, risk factor exposure is determined BEFORE the outcome is known,
thus allowing calculation of the relative risk
2. In simple words, always look for the order in which outcomes and risk factors are assessed:
a. In case-control studies, outcome is determined first and then we look for risk factors in
the past
b. In retrospective cohort study, we determine the risk factor exposure first and then
determine if they are statistically significantly associated with the outcome
3. Case-control is helpful in finding odds ratio while cohort study is helpful in finding the relative
risk and incidence of disease.

19
 BIOSTATISTICS & EPIDEMIOLOGY

ROC (RECEIVER OPERATING CHARACTERISTIC CURVES)

1. The closer the plotted curve approaches the left and top borders of the graph, the more
accurate the test

20
 BIOSTATISTICS & EPIDEMIOLOGY

CHOOSING TEST

MCNEMAR TEST
1. McNemar test compares the difference between 2 paired proportions
2. Patients serve as their own control (eg, success/failure before and after treatment in the same
subjects)

21
 BIOSTATISTICS & EPIDEMIOLOGY

MEAN, MEDIAN, MODE

EXTERNAL VALIDITY
1. External validity (generalizability) is defines as the applicability of a study’s results beyond the
group that was initially assessed
2. It answers the question, “How generalizable are the results of a study to other populations?”

INTERNAL VALIDITY
1. Internal validity relates to conclusions regarding cause and effect in a study
2. It answers the question, “Are we observing/measuring what we think we are
observing/measuring?”
3. The major threat to internal validity is confounding

22