Epidemiology: Study Validity, Bias, Diagnostic Testing, and Confounding

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/31

flashcard set

Earn XP

Description and Tags

Vocabulary practice flashcards covering screening and diagnostic test validity, reliability, information bias, selection bias, confounding assessment, and standardization in epidemiology.

32 Terms

1
New cards

Validity (Measurement)

The degree to which the results of a measurement correspond to the true state of the phenomenon being measured, representing truth and accuracy in measurement.

2
New cards

Reliability

The consistency, precision, or dependability of a set of measurements, reflecting whether measurements repeatedly yield the same results under similar conditions.

3
New cards

Screening Test

A test administered to individuals who appear healthy but may be at risk, designed to establish the presence or absence of disease at an early, asymptomatic stage to reduce morbidity and mortality.

4
New cards

Diagnostic Test

A test performed to establish the presence or absence of disease in individuals presenting with abnormal signs or symptoms, often following an abnormal screening test.

5
New cards

Sensitivity

The percentage of persons with the disease who test positive, calculated as Sensitivity=[TPTP+FN]×100\text{Sensitivity} = [\frac{\text{TP}}{\text{TP} + \text{FN}}] \times 100. Highly sensitive tests are used to rule out a disease ("SNout").

6
New cards

Specificity

The percentage of persons without the disease who test negative, calculated as Specificity=[TNTN+FP]×100\text{Specificity} = [\frac{\text{TN}}{\text{TN} + \text{FP}}] \times 100. Highly specific tests are used to rule in a disease ("SPin").

7
New cards

Positive Predictive Value (PPV)

The percentage of people with positive test results who truly have the disease, calculated as PPV=[TPTP+FP]×100\text{PPV} = [\frac{\text{TP}}{\text{TP} + \text{FP}}] \times 100. PPV increases as disease prevalence increases.

8
New cards

Negative Predictive Value (NPV)

The percentage of people with negative test results who truly do not have the disease, calculated as NPV=[TNTN+FN]×100\text{NPV} = [\frac{\text{TN}}{\text{TN} + \text{FN}}] \times 100. NPV decreases as disease prevalence increases.

9
New cards

Percent Agreement

A method of measuring observer agreement calculated as the total number of concordant evaluations divided by all evaluations: Percent Agreement=a+da+b+c+d\text{Percent Agreement} = \frac{a + d}{a + b + c + d}.

10
New cards

Kappa Statistic (κ\kappa)

A reliability measure expressing the extent to which observed observer agreement exceeds that expected by chance alone relative to maximum potential improvement: κ=% observed agreement−% agreement expected by chance100%−% agreement expected by chance\kappa = \frac{\% \text{ observed agreement} - \% \text{ agreement expected by chance}}{100\% - \% \text{ agreement expected by chance}}. Values >0.75> 0.75 indicate "excellent" agreement, 0.40−0.750.40 - 0.75 indicate "intermediate" to "good" agreement, and <0.40< 0.40 indicate "poor" agreement.

11
New cards

Lead-Time Bias

A screening bias characterized by an artificial increase in measured survival time from disease detection until death without any actual lengthening of the patient's life, caused purely by detecting disease earlier in its preclinical phase.

12
New cards

Length-Biased Sampling

A screening bias occurring because slowly progressing, less aggressive forms of a disease have a longer asymptomatic detectable phase and are therefore disproportionately detected by screening compared to rapidly progressive forms.

13
New cards

Overdiagnosis

The detection by a screening program of lesions or preclinical diseases that would never have progressed to cause clinical signs, symptoms, or death during the individual's lifetime.

14
New cards

Information Bias

A distortion in the estimation of a measure of association resulting from inaccurate measurement or misclassification of exposure, disease outcome, or both.

15
New cards

Non-Differential Misclassification

Measurement error that occurs when the degree of misclassification does not differ between comparison groups (exposure is misclassified regardless of disease status, or disease is misclassified regardless of exposure status). For dichotomous variables, this biases the measure of association towards the null.

16
New cards

Differential Misclassification

Measurement error that occurs when the degree of misclassification differs between comparison groups (exposure is misclassified differentially by disease status, or disease is misclassified differentially by exposure status). This bias can be directed either towards or away from the null.

17
New cards

Misclassification Table

A 2×22 \times 2 matrix cross-tabulating classified status against the "truth" (gold standard) for either exposure or disease, used to determine sensitivity and specificity of classification and mathematically reconstruct unbiased data.

18
New cards

Selection Bias

A systematic error that occurs when the measure of association estimated in a study's analytic sample differs meaningfully from what would have been estimated if all eligible subjects from the source population had been included.

19
New cards

Internal Validity

The ability to draw true and accurate inferences about the source population based on the data obtained from the study population or analytic sample.

20
New cards

External Validity

The ability to generalize findings from the study population beyond the source population to other external populations, groups, or settings.

21
New cards

Berkson's Bias

A specific form of selection bias in hospital-based case-control studies occurring when the combination of exposure and disease increases the probability of hospital admission, typically biasing the observed odds ratio away from the null.

22
New cards

Loss to Follow-up Bias

A selection bias in longitudinal studies (cohorts or clinical trials) that occurs when participant dropout or withdrawal differs by both exposure status and disease status simultaneously.

23
New cards

Selection Probabilities (α,β,γ,δ\alpha, \beta, \gamma, \delta)

Quantitative proportions (α=aA\alpha = \frac{a}{A}, β=bB\beta = \frac{b}{B}, γ=cC\gamma = \frac{c}{C}, δ=dD\delta = \frac{d}{D}) representing the likelihood of individuals in each exposure-disease category of the source population being enrolled into the analytic sample, used in sensitivity analyses to reconstruct unbiased association measures.

24
New cards

Confounding

A systematic distortion of an exposure-disease association occurring when the measure of association is mixed with the effect of an extraneous risk factor that is unbalanced between comparison groups.

25
New cards

A Priori Criteria for Confounding

Three baseline criteria required to classify a variable as a confounder: (1) it must be an independent risk factor for the disease (not caused by the disease), (2) it must be associated with the exposure, and (3) it cannot be an intervening variable on the causal pathway between exposure and disease.

26
New cards

10% Rule (Data-Based Criterion for Confounding)

A quantitative criterion identifying confounding when the percent difference between the crude measure of association (cOR or cRR) and the adjusted measure of association (aOR or aRR) exceeds ±10%\pm 10\%, calculated as [cOR−aORaOR]×100\left[\frac{\text{cOR} - \text{aOR}}{\text{aOR}}\right] \times 100.

27
New cards

Mantel-Haenszel Method

A stratification technique that calculates an adjusted, weighted summary measure of association (RRMH\text{RR}_{MH} or ORMH\text{OR}_{MH}) across strata of a confounder; it is required when cells contain zero counts and is preferred for sparse data.

28
New cards

Residual Confounding

Confounding that persists after analytical adjustment due to unmeasured or unknown confounders, broad categorization of controlled confounders (e.g., broad age groups), or misclassification of the confounder variable.

29
New cards

Standardization

A method used to remove confounding by forcing the distribution of confounding factors (such as age) to be identical between compared groups through direct or indirect rate adjustment.

30
New cards

Direct Standardization

A rate adjustment method in which stratum-specific disease rates from two or more study populations are applied to the demographic structure of a single standard reference population to calculate expected rates and eliminate confounding.

31
New cards

Indirect Standardization

A rate adjustment method in which known stratum-specific rates from a standard reference population are applied to the demographic structure of the study population to determine the expected number of events.

32
New cards

Standardized Mortality Ratio (SMR)

The ratio or percentage of observed deaths in a study population relative to the expected deaths calculated by applying general population age-specific death rates: SMR=Observed deathsExpected deaths\text{SMR} = \frac{\text{Observed deaths}}{\text{Expected deaths}}.