Biostatistics — USMLE Step 2 CK Notes

Case (disease) Control (no disease)

A——> Has disease

O——> No disease

Retrospective: What happened?

Risk factor frequency

Odds Ratio ——> (Odd Case): Exposure Odds Ratio -> 4.0 v 1.0 -> First has 400% chance cf second

Odds Ratio = ad/bc

Cohort: Incidence of disease. RR = a/(a+b)

————

O —-> No disease. c/(c+d)

O —-> No disease

Retro and Pro Risk factor for eventually getting disease etc

Relative Risk

Determines RFs or outcomes

Keep your last name

B—-> A: Decreases FN (eliminates) and Increases FP

—>>> Increases Sensitivity and decreased Specificity

TN

TNT

FP

TP

ITP

FN

Ata

A

B

Specificity: Rules IN

Sensitivity: Rules OUT

SNAP and SPIN

Number needed to treat: NNT = 1/ARR

ARR is number who had placebo and got disease - those who had drug and got disease

RRR is ARR/Placebo group

Significance: p<0.05. (Alpha)

Two sample t test: 1 categorical and 1 numerical

Comparison of 2 means

ANOVA: 3 or more means

Chi square: Categorical data and proportion of ppl (not mean!): 2 categorical!

Attributable risk per cent (ARP) -> (RR - 1)/RR

Factorial Design

Use of 2 or more experimental interventions

—->> each with 2 or more variables studied independently eg BP ranges

Matching used to control confounding

Cross sectional study: Prevalence Odds Ratio +/- risk factor —->> disease prevalence

Data gathered at one point in time

Failure Mode and Effects Analysis (hazard analysis) —-> Find solution to potential problems

Prospective process performed before problems arise

Failure modes: All things that could go wrong

Effect of things going wrong

Causes of failure modes

Root Cause Analysis

Used once problem occurred

Rigorous approach used in reaction to an event to identify underlying cause

Control Chart Analysis

Used once problem occurred

Plot data of a variable over time with control limits

—-> if variable falls outside these limits —> signals potential issue —> evaluate

Primary Prevention: lifestyle

Secondary prevention: Early detection and slow progression: Screening

Tertiary Prevention: Disease present —- measures decrease mortality/morbidity

SDs -> distance from MEAN

1 —> 68% within

2 —> 95%

3 —> 99.7 %

Anchoring: Fixation on initial impressions to make dx

Related to confirmation bias

Eg: Burning throat pain dx as GERD despite weight loss (true dx is cancer)

Availability: Dx swayed by recently seen or memorable ‘high stakes’ cases

Eg SOB dx as flu during flu season when true dx is PE

CONFIRMATION (like OCD)

Emphasis on evidence supporting presumed dx

Overlooking of info supporting alternative dx; related to anchoring bias

Framing

Dx approach influenced by context and presentation of info

Eg Abdo pain dx as opiate withdrawal in pt described as drug seeking ——-> true dx is SBO

Susceptibility bias: Type of selection bias

Pygmalion: Self fulfilling Prophecy

Ascertainment Bias

Results from atypical population extrapolated to entire population

Type of selection bias

Observer bias

Observer unconsciously influences outcome of study by knowing exposure status of participants

Hawthorne Effect (Measurement Bias)

Change in behaviour if study participants know they are being observed

Lead time Bias: Early detection -> increased survival

Confidence Interval

95% CI of mean = mean +/- 1.96 x SEM

A larger CI indicates that there is a wider range of possible effects (small sample) -> less precise

If CIs overlap eg group A (83-85) and group B (84-86) ==> not statistically significant

A 95% CI means 95% chance true value lies between interval points

To check closest risk etc remember Hazard Ratio -> estimates RR

Medical Emancipation: Sex, drugs and rock n roll

STIs, pregnancy care or contraception

Mental health and addiction services

Emergency care

RR = incidence in exposed/ incidence in unexposed

RR < 1 —> Reduced Risk

Hypotheses

Null Hypothesis (H0)

Alternative Hypothesis (H1)

Hypothesis of difference —> There is a link between disease and risk factor

Type I error (False positive) —> happened by chance

Stating there is an association when none exits (incorrectly rejecting null hypothesis)

α = probability of type I error ——-> WORSE

General rule of thumb is that statistical significance is reached if p < 0.05

Type II error (False negative)

Stating there is no effect when an effect exists —->. Incorrectly accepting null hypothesis

β = probability of type II error

Power (True Positive)

Probability of correctly rejecting null hypothesis

Power = 1 - β

Power depends on Sample size ——> Increasing sample size increases power

Size of expected effect ——> Increasing effect size increases power

True Negative: Probability of correctly accepting null hypothesis

Precision: Repeatedly same; increased power -> increased precision

Accuracy: absence of systematic error (validity); trueness

Clinical Trials

Phase I: small number of healthy subjects ——>>>> safety and dosage

Phase II: small number of diseased patients ——>>> efficacy and side effects

Phase III: large number of diseased patients ——>>> efficacy and side effects

-> compare to standard of care

Phase IV: post-marketing surveillance —->>> detects rare or long-term adverse effects

Attributable Risk (risk difference)

—-> risk difference between exposed and unexposed groups

Likelihood ratio (extent to which test result correct)

LR + = Sensitivity/ (1-specificity). False + ratio = 1-specificity

LR - = Specificity/ (1-sensitivity). False - ratio = 1 -sensitivity

Skew: alphabet positive

prevalence = incidence x duration

Matching prevents confounding