Epi Methods 751 Flashcards
Occurrence & distribution of health-related events, states, & processes in specified populations; important for monitoring disease & risk factor burden, directing funding, assessing interventions/programs, identifying new health problems
Descriptive Epidemiology
Study of determinants influencing processes
Analytical Epidemiology
Sizeable aggregate of people who satisfy particular set of membership criteria; defined by person, place, & time
Population
Able to experience & measure outcome of interest; biologically susceptible & methodologically under observation
At Risk
Graphical depiction of times of onset of disease in population
Epidemic Curve
Interval from origin to time of onset of clinical illness; summarized as mean/median or graphed
Incubation Period
Cumulative incidence; proportion of study population who become cases & develop disease by certain time
Attack Rate
Persons exposed to same exposure over limited, defined period of time (1 incubation period); shape of curve rises rapidly, contains definite peak, & gradual decline; cases may appear as wave (secondary transmission)
Point Source Epidemic
Cases of disease serve as source of infection for subsequent cases, which then serve as sources for later cases; shape of curve is series of successively larger peaks until pool of susceptibles exhausted or control measures implemented
Propagated Epidemic
Substances whose presence/absence may initiate or perpetuate disease process vs. characteristics of individuals (modifiable & non-modifiable) vs. all other social, economic, & biological factors impacting health
Agent-Host-Environment Model
Examines how humans interact with each aspect of environment; identify inter-relationships for single & multi-level models
Social-Ecological Framework
Loss or attrition of subjects from follow-up; occurrence of event uncertain after this time; reasons include loss to follow-up, competing risk, administrative censoring (end of study)
Censoring
Group to whom inferences will be made; theoretical; define by person, place, & time
Target Population
Group from whom study population drawn; well-defined & enumerable; define by person, place, & time
Source Population
Group studied; must be well-defined & enumerated; define by person, place, & time
Study Population
Disease that hasn’t yet developed signs or symptoms & cannot be detected or diagnosed by test; theoretical
Preclinical Outcomes
Disease detectable by specific tests but doesn’t yet manifest signs or symptoms; surrogate for clinical disease; may be earlier in disease process (closer to causal pathway)
Subclinical Outcomes
Disease that manifests signs or symptoms; results more interpretable & applicable to clinical practice; later in disease process
Clinical Outcomes
Timely & under investigator control; quality & consistency may be improved; more representative; costly & resource intensive; difficult to sustain
Active Ascertainment
Dependent on data collected by others; less representative; less costly; easier to sustain; easier to cover large areas
Passive Ascertainment
Compare to gold standard & calculate sensitivity/specificity
Measuring Validity for Binary Variables
Compare to gold standard & create scatterplot/Bland-Altman plot or calculate correlations (Pearson’s, Spearman’s)/regression
Measuring Validity for Continuous Variables
Measure of association indicating degree to which 2 variables have linear relationship; r = -1 (perfect negative) vs. r = +1 (perfect positive)
Correlation Coefficient
Can be misleading measure of agreement; depends on range of data, doesn’t reflect slope, tests Ho of no linear relationship between variables not whether there is agreement
Pearson’s Correlation
Describes relationship between measure (X) & gold standard (Y); slope = 1 (unbiased) vs. < 1 (measure over-estimates) vs. >1 (measure under-estimates)
Regression
Compare measured values from multiple replicates to determine extent of random error; calculate Spearman’s/Pearson’s correlation, regression, % agreement, % positive agreement, kappa, CV or create scatterplot, B-A plot
Measuring Reliability
Compare replicates & calculate percent agreement, percent positive agreement, or kappa
Measuring Reliability for Binary/Categorical Variables
Compare replicates & calculate Spearman’s rank correlation coefficient
Measuring Reliability for Ordinal Discrete Factor
Sum of agreement cells/total; doesn’t take into account role of chance agreement
Percent Agreement
[(Observed agreement)-(Agreement Expected by Chance)]/[100-(Agreement Expected by Chance)]; takes into account role of chance agreement
Kappa Statistic
Compare replicates & create scatterplot/Bland-Altman plot or calculate correlation coefficient/regression/coefficient of variation
Measuring Reliability for Continuous Variable
Shows agreement between measurements; plot difference against mean; mean of difference not equal to zero line indicates bias vs. points outside limits of agreement indicates random error
Bland-Altman Plot
SD replicates/mean replicates; typically want <10%
Coefficient of Variation
Compare measured values to gold standard to determine extent of bias; calculate Spearman’s/Pearson’s correlation, regression, % agreement, % positive agreement, kappa, sensitivity/specificity or create scatterplot, B-A plot
Measuring Validity
Less costly, easier to design & carry out, monitoring trends over time vs. low sensitivity, limited availability of data, not representative
Passive Surveillance
Highly sensitive, detailed information, representative vs. costly/resource intensive, difficult to sustain
Active Surveillance
Population tracking; choose entire population or representative sample to monitor for condition
Universal Surveillance
Warning signs; choose key location (most susceptible to change) to monitor for condition
Sentinel Surveillance
A/A+C vs. D/B+D
Sensitivity vs. Specificity
A/A+B vs. D/C+D
Positive vs. Negative Predictive Value
Relationship to cumulative incidence when IR constant over time interval & IR*t<0.1
R(t)=IRt & R(t)=1-e^(-Sum(IRt))
Relationship between prevalence & IR in steady state, constant IR, & disease rare (P<0.1)
P=IR*D; D=duration of disease
Number of fatal cases/total number of cases, over a given time interval
Case Fatality
Number of deaths from specific cause/number of deaths from all causes, over a given time interval
Proportional Mortality
(IR exposed - IR unexposed)/IR exposed or (IRR-1)/IRR; fraction of exposed cases for which exposure responsible for disease
Excess Fraction
Percentage of cases that would be avoided if exposure eliminated from population; highly dependent on prevalence of exposure in population
Population Attributable Fraction
Case definition that is more encompassing/less stringent –> more individuals who have outcome counted as cases
Higher Sensitivity, Lower Specificity
Case definition that is less encompassing/more stringent –> more individuals who don’t have outcome counted as non-cases
Lower Sensitivity, Higher Specificity
Estimate of time for IR in steady state
Midpoint population*time between enumerations
The study of the occurrence & distribution of health-related events in specified populations, including study of determinants influencing health states, & application of knowledge to control the health problems
Epidemiology