248 SEC TION II PUBLIC HEALTH SCIENCES `

PUBLIC HEALTH SCIENCES—EPIDEMIOLOGY & BIOSTATISTICS

Quantifying risk Definitions and formulas are based on the classic Disease
2 × 2 or contingency table.

or intervention
Risk factor
a b

c d

Odds ratio Typically used in case-control studies. OR a/b ad

OR = =
depicts the odds of an event (eg, disease) c/d bc
occurring giving a certain exposure (a/b) vs
the odds of an event occurring in the absence
of that exposure (c/d).
Relative risk Typically used in cohort studies. Risk of a/(a + b)
RR =
developing disease in the exposed group c/(c + d)
divided by risk in the unexposed group (eg, if
21% of smokers develop lung cancer vs 1% of
nonsmokers, RR = 21/1 = 21). For rare diseases
(low prevalence), OR approximates RR.
RR = 1 p no association between exposure and
disease.
RR > 1 p exposure associated with q disease
occurrence.
RR < 1 p exposure associated with r disease
occurrence.
Attributable risk The difference in risk between exposed and a c
AR = −
unexposed groups, or the proportion of a+b c+d
disease occurrences that are attributable to
the exposure (eg, if risk of lung cancer in
smokers is 21% and risk in nonsmokers is 1%,
then 20% of the lung cancer risk in smokers is
attributable to smoking).
Relative risk reduction The proportion of risk reduction attributable to RRR = 1 − RR
the intervention as compared to a control (eg,
if 2% of patients who receive a flu shot develop
the flu, while 8% of unvaccinated patients
develop the flu, then RR = 2/8 = 0.25, and
RRR = 0.75).
Absolute risk The difference in risk (not the proportion) c a
ARR = −
reduction attributable to the intervention as compared c+d a+b
to a control (eg, if 8% of people who receive
a placebo vaccine develop the flu vs 2%
of people who receive a flu vaccine, then
ARR = 8% − 2% = 6% = .06).
Number needed to Number of patients who need to be treated for NNT = 1/ARR
treat 1 patient to benefit. Lower number = better
treatment.
Number needed to Number of patients who need to be exposed to NNH = 1/AR
harm a risk factor for 1 patient to be harmed. Higher
number = safer exposure.
PUBLIC HEALTH SCIENCES `
PUBLIC HEALTH SCIENCES—EPIDEMIOLOGY & BIOSTATISTICS SEC TION II 249

Incidence vs Incidence = # of new cases (during a specified Incidence looks at new cases (incidents).
prevalence rate # of people at risk time period)
# of existing cases (at a point in Prevalence looks at all current cases.
Recurrence Prevalence =
Total # of people time)
Incidence
in a population
Prevalence Prevalence = Incidence rate × average duration
1 – prevalence of disease
Mortality Cure
Prevalence ≈ incidence for short duration disease Prevalence ∼ pretest probability.
(eg, common cold). q prevalence p q PPV and r NPV.
Prevalence > incidence for chronic diseases, due to
large # of existing cases (eg, diabetes).

Precision vs accuracy
Precision (reliability)
Accuracy (validity)
The consistency and reproducibility of a test.
The absence of random variation in a test.
The trueness of test measurements.
The absence of systematic error or bias in a test.
Random error r precision in a test.
q precision p r standard deviation.
q precision p q statistical power (1 − β).
Systematic error r accuracy in a test.

Accuracy Accuracy
High Low High Low

High High Low Low

Precision Precision
250 SEC TION II PUBLIC HEALTH SCIENCES
Bias and study errors
TYPE DEFINITION
Recruiting participants
Selection bias Nonrandom sampling
or treatment allocation
of subjects such that
study population is not
representative of target
population (eg, study
participants included based
on adherence or other criteria
related to outcome). Most
commonly a sampling bias.
Performing study
Recall bias Awareness of disorder alters
recall by subjects; common in
retrospective studies.
Measurement bias Information is gathered in a
systemically distorted manner. cervical cancer not observed
Procedure bias Subjects in different groups are
not treated the same.
Observer-expectancy Researcher’s belief in the
bias efficacy of a treatment changes
the outcome of that treatment
(aka Pygmalion effect; self-
fulfilling prophecy).
Interpreting results
Confounding bias When a factor is related to both Pulmonary disease is more
the exposure and outcome,
but not on the causal
pathway p factor distorts or
confuses effect of exposure on
outcome.
Lead-time bias Early detection is confused
with q survival.
`
PUBLIC HEALTH SCIENCES—EPIDEMIOLOGY & BIOSTATISTICS
EXAMPLES
Berkson bias—study population Randomization
selected from hospital is
less healthy than general
population
Healthy worker effect—study
population is healthier than
the general population
Non-response bias—
participating subjects differ
from nonrespondents in
meaningful ways
Patients with disease recall
exposure after learning of
similar cases
Association between HPV and
when using non-standardized
classifications
Hawthorne effect—participants
change their behavior in
response to their awareness of
being observed
Patients in treatment group
spend more time in highly
specialized hospital units
If observer expects treatment
group to show signs of
recovery, then he is more
likely to document positive
outcomes
common in coal workers
than the general population;
however, people who work in
coal mines also smoke more
frequently than the general
population
Early detection makes it
seem as though survival has
increased, but the natural
history of the disease has not
changed
STRATEGY TO REDUCE BIAS
Ensure the choice of the right
comparison/reference group
Decrease time from exposure
to follow-up
Use objective, standardized,
and previously tested methods
of data collection that are
planned ahead of time
Use placebo group
Blinding and use of placebo
reduce influence of
participants and researchers
on procedures and
interpretation of outcomes
as neither are aware of group
allocation
Multiple/repeated studies
Crossover studies (subjects act
as their own controls)
Matching (patients with
similar characteristics in both
treatment and control groups)
Restriction
Randomization
Measure “back-end” survival
(adjust survival according to
the severity of disease at the
time of diagnosis)
PUBLIC HEALTH SCIENCES `
PUBLIC HEALTH SCIENCES—EPIDEMIOLOGY & BIOSTATISTICS SEC TION II 251

Statistical distribution
Measures of central Mean = (sum of values)/(total number of values). Most affected by outliers (extreme values).
tendency Median = middle value of a list of data sorted If there is an even number of values, the median
from least to greatest. will be the average of the middle two values.
Mode = most common value. Least affected by outliers.
Measures of Standard deviation = how much variability σ = SD; n = sample size.
dispersion exists in a set of values, around the mean of Variance = (SD)2.
these values. SE = σ/√n.
Standard error = an estimate of how much SE r as n q.
variability exists in a (theoretical) set of sample
means around the true population mean.
Normal distribution Gaussian, also called bell-shaped.
–1σ +1σ
Mean = median = mode.
–2σ +2σ
–3σ +3σ

68%
95%
99.7%

Nonnormal distributions
Bimodal Suggests two different populations (eg,
metabolic polymorphism such as fast vs
slow acetylators; age at onset of Hodgkin
lymphoma; suicide rate by age).
Positive skew Typically, mean > median > mode. Mode
Median
Asymmetry with longer tail on right. Mean

Negative skew Typically, mean < median < mode. Median

Mode

Asymmetry with longer tail on left. Mean

Statistical hypotheses
Null (H0) Hypothesis of no difference or relationship (eg, Reality
there is no association between the disease and H1 H0
the risk factor in the population).
Alternative (H1) Hypothesis of some difference or relationship Power α
Study rejects H0
(eg, there is some association between the ( 1 – β) Type I error
disease and the risk factor in the population).

Study does not reject H0 β

Correct
Type II error
252 SEC TION II PUBLIC HEALTH SCIENCES `
PUBLIC HEALTH SCIENCES—EPIDEMIOLOGY & BIOSTATISTICS

Outcomes of statistical hypothesis testing

Correct result Stating that there is an effect or difference when
one exists (null hypothesis rejected in favor of
alternative hypothesis).
Stating that there is not an effect or difference
when none exists (null hypothesis not
rejected).
Incorrect result
Type I error (α) Stating that there is an effect or difference Also known as false-positive error.
when none exists (null hypothesis incorrectly
rejected in favor of alternative hypothesis).
α is the probability of making a type I error. p is α = you accused an innocent man.
judged against a preset α level of significance You can never “prove” the alternate hypothesis,
(usually 0.05). If p < 0.05, then there is less but you can reject the null hypothesis as being
than a 5% chance that the data will show very unlikely.
something that is not really there.
Type II error (β) Stating that there is not an effect or difference Also known as false-negative error.
when one exists (null hypothesis is not rejected
when it is in fact false).
β is the probability of making a type II error. β β = you blindly let the guilty man go free.
is related to statistical power (1 – β), which is If you q sample size, you q power. There is power
the probability of rejecting the null hypothesis in numbers.
when it is false.
q power and r β by:
ƒ q sample size
ƒ q expected effect size
ƒ q precision of measurement

Confidence interval Range of values within which the true mean If the 95% CI for a mean difference between 2
of the population is expected to fall, with a variables includes 0, then there is no significant
specified probability. difference and H0 is not rejected.
CI for population mean = x̄ ± Z(SE) If the 95% CI for odds ratio or relative risk
The 95% CI (corresponding to α = .05) is often includes 1, H0 is not rejected.
used. If the CIs between 2 groups do not overlap
For the 95% CI, Z = 1.96. p statistically significant difference exists.
For the 99% CI, Z = 2.58. If the CIs between 2 groups overlap p usually
no significant difference exists.
 PUBLIC HEALTH SCIENCES `
BEHAVIORAL SCIENCE—ETHICS SEC TION II 253

Common statistical tests

t-test Checks differences between means of 2 groups. Tea is meant for 2.
Example: comparing the mean blood pressure
between men and women.
ANOVA Checks differences between means of 3 or more 3 words: ANalysis Of VAriance.
groups. Example: comparing the mean blood pressure
between members of 3 different ethnic groups.
Chi-square (χ²) Checks differences between 2 or more Pronounce Chi-tegorical.
percentages or proportions of categorical Example: comparing the percentage of members
outcomes (not mean values). of 3 different ethnic groups who have essential
hypertension.

Pearson correlation r is always between −1 and +1. The closer the absolute value of r is to 1, the stronger the linear
coefficient correlation between the 2 variables.
Positive r value p positive correlation (as one variable q, the other variable q).
Negative r value p negative correlation (as one variable q, the other variable r).
Coefficient of determination = r 2 (amount of variance in one variable that can be explained by
variance in another variable).
r = –0.8 r = –0.4 r=0 r = +0.4 r = +0.8

Strong negative Weak negative No correlation Weak positive Strong positive

correlation correlation correlation correlation

`
BEHAVIORAL SCIENCE—ETHICS

Core ethical principles

Autonomy Obligation to respect patients as individuals (truth-telling, confidentiality), to create conditions
necessary for autonomous choice (informed consent), and to honor their preference in accepting
or not accepting medical care.
Beneficence Physicians have a special ethical (fiduciary) duty to act in the patient’s best interest. May conflict
with autonomy (an informed patient has the right to decide) or what is best for society (eg,
mandatory TB treatment). Traditionally, patient interest supersedes.
Nonmaleficence “Do no harm.” Must be balanced against beneficence; if the benefits outweigh the risks, a patient
may make an informed decision to proceed (most surgeries and medications fall into this
category).
Justice To treat persons fairly and equitably. This does not always imply equally (eg, triage).