Randomized Control Trial:

Measures Incidence and Risk Ratio

Divide a study population into Intervention and Control and measure the outcome in each
group.
Risk Ratio: incidence in outcome in exposed/incidence in outcome in unexposed
Cohort Studies:
Comparison group based on exposure
Measures Incidence and relative risk
Prospective
Case-Control:
Always retrospective
Useful when outcome is rare
Measures Odds Ratio: AD/BC, approximately equals relative risk when the outcome is
rare.
Cross-sectional:
All measurements made at one point in time
Useful for associations but does not necessitate causality.
Measures: prevalence: Number of cases/total population
Prevalence Ratio: Prevalence of disease in exposed group/prevalence of disease in
unexposed group
Useful progression of literature:
Association< Etiologic observational study<experimental study<systematic literature
synthesis<application of knowledge in decisions
Calculations:
Exposed event rate: incidence in exposed: a/a+b
Control Event Rate: incidence in unexposed: C/C+D
Relative Risk=(a/a+b)/(c/c+d)
Absolute Risk Difference= EER-CER-# of events in the exposed that are attributable to
the exposure
Relative Risk Difference= (EER-CER)/(CER)-X times more likely to have an outcome
Number Needed to Treat= 1/(EER-CER)-the number of persons you must expose to
cause 1 case.
Survival Analysis:
Used when subjects are enrolled for different lengths of time
Error:
Type I error: Accept Ha when it is false; Convict an innocent person (BAD)!=.05%
Type II error: Accept Ho when it is false: fail to convict the guilty
Beta= probability of type II error
1-beta= power=power to suggest a relationship does not exist
Rules of P’s:
Can’t compare P’s b/w 2 studies
Confidence Intervals:
If CI crosses the null, it is not significant
Mean differences: null=0
Relative Risks: null=1
Random Error:
Instrument Variability-bathroom scale vs. electron scale
Observer Variability: intra observer: fatigue; inter-obsrver; differences in opionion
Subject Variability
Systematic Error:
Instrument Bias: Calibrated Scale
Investigator/Interviewer Bias-differential interpretation of results based on knowledge of
exposure
Subject Bias
Social Desirability bias, Hawthorne
Controlling for Confounding by design:
Randomization
Selection-study only those with or without the confounder present
Matching-control is selected for each individual subject on basis of confounder
Controlling for confounding by analysis:
Stratification
Adjustment
Multivariate Analysis
Multivariate Analysis Types:
Linear Regressions: continuous outcome and continuous risk factor
Logistic Regression: binary outcome: odds ratio for each risk factor
Proportion hazard regression: time to event
Sensitivity: Probability that a person with the disease will have a positive test result
Specificity: Probability that a person without the disease will have a negative test result.
Positive Predictive Value: Probability of disease given a positive test result
Criterion for an effective screening test:
Morbidity and Mortality must be of sufficient concern
Disease must be sufficiently prevalent
Test must be sufficiently sensitive and specific
Test must be safe and acceptable
Diagnostic work up for a positive screen must be acceptable given the number of false
positives (minimal harm)
Pre-clinical intervention must improve outcomes
Test must be relatively cost-effective
ROC curve:
Tradeoff b/w sensitivity and specificity. An increase in sensitivity will be followed by a
decrease in specificity. The closer the curve to the upper left corner, the more accurate
the test.
Bias:
Spectrum Bias: study population has higher disease prevalence than typical populations
Lead time bias: detecting disease earlier appear sot increase duration of survival
fictitiously
Length time bias: screening more likely to catch cases with longer duration
Workup bias: patients with positive or negative diagnostic test results are preferentially
referred to receive verification of diagnosis by the gold standard procedure.
Evidence-based pratice:
Best Research evidence, clinical expertise, patient values.
5 A’s of evidence based practice:
Ask clinical questions-pico and know the type of information needed.
Backgroud question-general information found in textbooks
Foreground- specific to patient: pico
PICO: patient, intervention, comparison, outcome

Acquire the best evidence, appraise the evidence, Apply the evidence, Assess the
effectiveness, efficiency of the EBM process (reflection of practice)
Epidemic- sudden increase over baseline
Epidemiology: study of patterns and causes of health outcomes
Incidence approximates risk: incidents in a time period/average population in that time-
period.
Incidence approximates population risk of health events:
For rare events, for events not recurring, and over short time intervals
Prevalence:
Point: # of persons with a condition at a point in time (# living with HIV
today)/population
Period prevalence: number of persons with a condition at any time within a period
Increases with: longer survival, net in migration of cases
Decreases with: faster mortality, shorter period of illness, net migration out
Increased prevalence of HIV due to longer survival.
Useful in determining size of population experiencing a condition

 Attributable Risk : e.g., among smokers, proportion of risk of lung cancer

attributable to smoking
RiskExposed – RiskUnexposed. = RR – 1 ≈ OR - 1 *
RiskExposed RR OR
 Population Attributable Risk: e.g., among general population (smokers & non-
smokers) the proportion of risk of lung cancer attributable to smoking
Risktotal—RiskUnexposed = Proportionexp(RR—1) *
RiskTotal 1 + (Proportionexp)(RR—1)
Disease Deaths: Heart Disease, Cancer, stroke
Risk Factors: smoking
Passive Intervention- policy is always theremore effective
Active Intervention-relies on person consistently using policy
 Post-test odds =
Pre-test odds X Likelihood ratio
 Odds of disease= Probability of disease /(1 – Probability of disease)
 E.g., 60% probability is 6:4 odds