PUBH 3131 Study Guide - Final Guide: Case Fatality Rate, Automated Theorem Proving, Null Hypothesis
Epidemiology: study humans:observation:GROUP:chroni
Endemic: specific to area
Epidemic: outbreak, more than expected
Pandemic: worldwide
Prevalence: total # of cases % proportion
● # of people ill/total # in group
Incidence: number of new cases, higher than prevalence
● # of new cases during time t/# at risk
Cumulative incidence: all cases throughout the time
Relative Risk: incidence rate of disease in
exposed/incidence rate of disease in nonexposed
(A/A+ B) / (C/C+D)
RR < 1: beneficial RR>1: harmful
Risk Difference:Attributable risk:describes absolute
impact of exposure on disease occurrence
● R exposed-R unexposed
● A/A+B - C/C+D
Odds Ratio: AD/BC how high the odds of exposure are
among cases compared with controls
SMR: # observed/#expected deaths
Proportionate Mortality: mortality due to specific causes/
mortality due to all causes x 100
Case Fatality: # deaths due to disease/# cases of
disease x 100
PPV: probability of a true positive, given you test positive
:more prevalent=higher PPV TP/TP+FP
NPV: probability of a true neg, given test neg TN/TN+FN
Sensitivity: probability testing POS, given disease
present 100% : NO FN
● Rules OUT disease: TP/TP+FN
● Too sensitive: waste resources,health conse
Specificity: probability testing NEG, given disease
absent 100% :NO FP
● Rules IN disease: TN/TN+FP
● Too specific: missed diagnosis, undercount
Truth table PRESENT ABSENT
POSITIVE
TP
FP
NEGATIVE
FN
TN
Population Attributable risk:PAR
● incidencepop-incidenceunexp
Observational:things already happened,uncontrollable
factors, difficult to determine causation
Experimental: trial
Analytical:test hypotheses, comparison groups
Descriptive: describe populations,disease,distributions
Exposed diseased non-diseased
Yes
No
Study Designs:
1. Ecological: compares measures of outcome and risk
factor by GROUPS (STAT)
● S: uses info that is available:fast+cheap
● W: Ecological fallacy:attributing associations on
group level to individuals, confounding
2.Cross Sectional:exposure+distributions determined at
one point: each person assessed once: SELF REPORT:
no INCIDENCE,good preval (OR) PREVALENCE RATIO
● W: no temporality,self report, no predict futur
3.Case control:(analytical) retrospective:begins
w/outcome and then determine exposure status (OR)
● Need cases and controls
● S:efficient for rare +long latency disease
● W: No temporality, selection bias, no inciden
4.Cohort: prospective:comparison by exposure, share
time related characteristic (RD,RR,)
● W: bad for rare/long latency disease
● S:temporality
5. Trials: experimental:randomize:eliminate cofound
● RCT:gold standard:highest level of validity:
● Double blind: temporality,controls confounding,
minimizes bias
Bias:systematic error, not random
Random error: due to chance, accounted for statisti
1.Selection bias:distortions due to
procedures/choice/participation of subjects (no EV)
2. Information bias:mismeasures of outcome:poor
validity, missing/wrong info, subjects misidentified
3. Confirmation bias:tendency to search,interpret info as
confirmation of one’s existing beliefs
4.Recall bias: reduction in reliability of info(case control)
5.Self selection bias:participants can choose not to
participate: differential effect
Differential bias:effects comparison group unequal
Non-differential:info equally wrong in comparison
groups, groups seem more alike than they are
Confounding:groups differ by a factor that changes
outcome: artifactual (location,ses, age:standardize)
Effect modifier:variable changes effect of study variable:
cannot remove/control (sex,race,age)
Threats to Internal (study population:variables)
● Maturation:time,selection bias, instrument
Threat to External validity EV (general population
● Selection bias, replicability
Randomization: taking ppl+randomly assigning
Eliminate cofounders
Random sampling: randomly select people
External validity
Causal Criteria
1.Strength of Association:check artifactual
2.Consistency:different conditions same result
3. Repeatability
4.Specificity: risk factor produces 1 particular effect
5.Temporality:NECE,exposure precede disease
6.Biological Gradient:Dose Response:greater
exposure:greater harm
7.Plausibility:makes biological sense
8.Coherence:fits w/ other research finding
9.Analog:other disease models have similar mech
10.Reversibility:remove exposure:remove disease
Multiplestudies:plausib,coherence,analog,consistency
Necessary:must have a to get b
Sufficient: a alone results in b
Temporality is necessary but not sufficient: coming first
doesn’t make something casual
Psychiatric Epidemiology:
● Unipolar major depression:leading
● Diagnosis by observation:detection is late
● Etiology unknown,lack of standard assess
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
# of people ill/total # in group. Incidence: number of new cases, higher than prevalence. # of new cases during time t/# at risk. Relative risk: incidence rate of disease in exposed/incidence rate of disease in nonexposed (a/a+ b) / (c/c+d) Risk difference:attributable risk:describes absolute impact of exposure on disease occurrence. Odds ratio: ad/bc how high the odds of exposure are among cases compared with controls. Proportionate mortality: mortality due to specific causes/ mortality due to all causes x 100. Case fatality: # deaths due to disease/# cases of disease x 100. Ppv: probability of a true positive, given you test positive. Npv: probability of a true neg, given test neg tn/tn+fn. Sensitivity: probability testing pos, given disease present 100% : no fn. Specificity: probability testing neg, given disease absent 100% :no fp. Observational:things already happened,uncontrollable factors, difficult to determine causation. Study designs: ecological: compares measures of outcome and risk factor by groups (stat)