Epidemiology and Public Health Study Designs, Measures, and Causal Inference

Comprehensive practice vocabulary flashcards covering epidemiologic study designs (descriptive, cross-sectional, ecological, case-control, cohort, and experimental), sampling strategies, morbidity/mortality measures, validity, bias, confounding, and causal inference.

Descriptive Epidemiology

An epidemiologic approach whose goal is to characterize the amount and distribution of health and disease within a population by person, place, and time to alert researchers to emerging problems, indicate the extent and severity of disease, and suggest potential solutions.

Analytic Epidemiology

An epidemiologic approach whose goal is to test specific etiologic hypotheses regarding causal associations between exposures and health outcomes to identify what interventions work or do not work.

Case Report

A detailed descriptive epidemiologic report documenting the clinical profile and medical history of a single individual patient ($n = 1$).

Case Series

A descriptive epidemiologic study that tracks and describes clinical characteristics and outcomes across a small group of patients presenting with the same diagnosis or condition.

Cross-Sectional Study

An observational study design that assesses exposure and disease status simultaneously in a defined sample or population at one specific point in time, providing a snapshot to assess prevalence.

Ecological Study

An observational study in which the unit of analysis is a group or population (typically defined geographically by countries, states, or counties) rather than individuals, using aggregate measures for both exposures and outcomes.

Ecological Fallacy

A logical bias that occurs when an association observed between aggregate-level variables at the group or population level is incorrectly assumed to hold true for individuals within those groups.

Pearson's Correlation Coefficient ($r$)

A statistical measure for two continuous, quantitative variables assessing direction ($r > 0$ for positive, $r = 0$ for none, and $r < 0$ for negative) and magnitude on a scale ranging from $-1.0$ to $+1.0$, where values between $0.7$ and $1.0$ (or $-0.7$ and $-1.0$) indicate strong association.

Simple Random Sampling

A probability sampling strategy where every element in the target population has an equal probability of selection, identified through a random number table or computer generator.

Systematic Sampling

A probability sampling method in which sampling elements are chosen at regular intervals through a calculated sampling interval ($k = \frac{N}{n}$), beginning at a randomly selected start point.

Stratified Sampling

A probability sampling method where the target population is first divided into mutually exclusive subgroups (strata) based on specific characteristics (e.g., sex, race, urbanicity), followed by random sampling within each stratum to ensure proportional representation.

Cluster Sampling

A probability sampling technique where naturally occurring units containing multiple sampling elements (such as city blocks, schools, or clinics) are selected as the sampling elements, useful for dispersed populations lacking an exhaustive sampling frame.

Random Digit Dialing (RDD)

A telephone survey probability sampling technique in which researchers identify regional area codes and telephone exchanges, randomly generating the remaining four digits to recruit participants without relying on a published directory.

Prevalence

The proportion of a population that has a specific disease or condition at a given time, calculated as $\text{Prevalence} = \frac{\text{\# existing cases (including new cases)}}{\text{Total population}} \times \text{rate base}$, useful for measuring disease burden and allocating healthcare resources.

Point Prevalence

The proportion of a population that has a disease or health condition assessed at one single, specific point in calendar time.

Period Prevalence

The proportion of a population that exhibits a disease or health condition at any point over a specified span or window of time.

Incidence

The rate of transition from a non-diseased state to a diseased state over a specified follow-up period, calculated as $\text{Incidence} = \frac{\text{\# new cases}}{\text{Population at risk over a time period}} \times \text{rate base}$, and useful for identifying risk factors.

Bathtub Model of Disease

A steady-state conceptual framework expressing the dynamic relationship $\text{Prevalence} = \text{Incidence} \times \text{Duration}$ ($P = I \times D$), where new incident cases flow in like water from a tap, pool as existing prevalence, and exit through cure or mortality.

Age-Adjusted Rate

A summary morbidity or mortality rate derived by statistical standardization procedures to remove the confounding effect of differences in age composition when comparing two or more distinct populations.

Age-Specific Rate

A disease frequency rate calculated exclusively within a designated age bracket by dividing the number of cases in that age group by the total population within that same age group.

Secular Trends

Gradual changes in the frequency and occurrence of disease or mortality that unfold consistently over extensive time periods spanning multiple decades or centuries.

Cyclic (Seasonal) Trends

Recurrent increases and decreases in the incidence or prevalence of an illness over recurring periods within a calendar year or across intervals of several years.

Point Epidemic

An outbreak or epidemic curve indicating that an exposed group was affected almost simultaneously by a common, circumscribed source of infection, contamination, or etiologic agent.

Clustering

An unusual aggregation of health events or cases of disease that are closely grouped together in space, in time, or in both space and time.

Case-Control Study

An observational analytic study design in which participants are selected based on whether they have the disease (cases) or are free of the disease (controls), and their previous exposure histories are retrospectively gathered and compared.

"WOULD" Criterion

The fundamental guideline for control selection in case-control research: if an individual selected as a control had developed the outcome of interest during the study interval, they would have been eligible and identified as a case in the study.

Matched Case-Control Study

A case-control study design in which controls are paired with cases based on shared confounding characteristics, performed either on an individual level (e.g., matching a case to a control within $\pm 2$ years of age) or frequency level (matching strata distributions).

Odds Ratio ($OR$)

The measure of association used in case-control studies defined as the odds of disease among exposed individuals divided by the odds of disease among unexposed individuals, computed from a standard $2 \times 2$ table as $OR = \frac{a \times d}{b \times c}$.

Cohort Study

An observational study design where a disease-free group of individuals is classified based on exposure status and tracked over time to determine and compare the incidence of one or more health outcomes.

Prospective Cohort Study

A cohort design in which exposure status is evaluated at the start of the study, and participants are followed forward in time into the future to identify incident disease cases.

Retrospective Cohort Study

A cohort design in which both exposure and outcome have already occurred prior to the initiation of the study, reconstructed historically using existing baseline and medical records.

Ambidirectional Cohort Study

A hybrid cohort design that uses historical records to capture past exposures and outcomes while simultaneously following the established cohort forward in time into the future.

Framingham Heart Study

A landmark population-based prospective cohort study initiated in 1948 by the NHLBI that enrolled 5,209 adult residents of Framingham, Massachusetts to track cardiovascular disease risk factors across generations.

Air Force Ranch Hand Study

A special occupational cohort study evaluating long-term adverse health outcomes among 2,483 military service personnel who flew or supported C-123 aircraft spraying Agent Orange during the Vietnam conflict from 1962 to 1971.

Counterfactual

The theoretical ideal state representing what the disease risk would have been among exposed individuals had they simultaneously not been exposed, approximated in observational epidemiology through unexposed comparison groups.

Relative Risk ($RR$)

The ratio of the cumulative incidence of disease in an exposed cohort group to the cumulative incidence of disease in an unexposed comparison group: $RR = \frac{CI_E}{CI_U} = \frac{a / (a + b)}{c / (c + d)}$.

Attributable Risk ($AR$)

The measure of absolute excess disease risk among exposed individuals that is directly attributable to the exposure: $AR = CI_E - CI_U = \frac{a}{a + b} - \frac{c}{c + d}$.

Attributable Risk Percent ($AR\%$)

The proportion of disease incidence in the exposed cohort that could be eliminated if the exposure were completely removed: $AR\% = \left(\frac{CI_E - CI_U}{CI_E}\right) \times 100$.

Risk Difference ($RD$)

A measure of public health impact reflecting the absolute excess cases in the exposed group relative to the unexposed group per unit population: $RD = \left(\frac{a}{a + b} - \frac{c}{c + d}\right) \times \text{rate base}$.

Randomized Controlled Trial (RCT)

An experimental investigation in which the researcher randomly allocates eligible human participants to either an active intervention arm or a comparison arm (placebo, standard of care, or no treatment) to test causal efficacy.

Quasi-Experimental Study

An experimental study design where the researcher manipulates the exposure or intervention but does not randomly assign individual subjects to study arms, often relying on naturally or opportunistically formed groups.

Cross-Over Design

An experimental trial in which each subject receives multiple consecutive interventions in a planned sequence, switching between active treatment and control arms partway through the trial.

Cluster-Randomized Trial

An experimental trial in which intact social or administrative groups (such as clinics, schools, or geographic areas) are randomized to intervention or control arms rather than individual participants.

Community Trial

A subtype of quasi-experimental study designed to evaluate the real-world impact of a health promotion policy, educational campaign, or environmental intervention delivered across an entire defined community.

Natural Experiment

An observational condition resembling an experiment where exposure levels vary across population subsets due to external forces outside the researcher's control, such as state-level policy variations or environmental events.

Randomization

The process of allocating enrolled trial participants to treatment or control arms by chance, ensuring that each participant has an equal probability of assignment and balancing both measured and unmeasured confounders at baseline.

Placebo

An inert substance, procedure, or sham intervention designed to resemble the active treatment in appearance and administration without possessing specific pharmacological or therapeutic properties.

Placebo Effect

A measurable physical or psychological improvement in health status observed in control group participants attributable to expectations, beliefs, and the power of suggestion rather than the active treatment.

Nocebo Effect

The development of adverse side effects or worsening of symptoms in a control patient receiving an inactive substance, caused by negative expectations and anticipatory anxiety.

Double-Blind Design

A masking methodology employed in clinical trials where neither the study participants nor the investigators interacting with them know which treatment group the participants have been assigned to.

Compliance

The degree to which study participants adhere strictly to the prescribed intervention protocol and schedule throughout the duration of a clinical trial.

Equipoise

A state of genuine clinical uncertainty among investigators regarding whether one intervention is superior to another, accompanied by the belief that withholding the experimental treatment will not harm participants.

Belmont Report

A 1978 foundational ethical report issued by the National Commission establishing three core principles governing research involving human subjects: Respect for Persons, Beneficence, and Justice.

Respect for Persons

An ethical principle outlined in the Belmont Report requiring that individual autonomy be acknowledged and honored through voluntary informed consent, with additional protections for vulnerable populations.

Beneficence

An ethical principle from the Belmont Report obligating investigators to maximize tangible research benefits while minimizing physical, psychological, and social harms.

Justice

An ethical principle from the Belmont Report requiring the fair and equitable distribution of research burdens and benefits, ensuring groups selected for study are positioned to benefit from the findings.

Epidemiologic Study Design Algorithm

A hierarchical decision-tree framework used to identify epidemiologic study designs by sequentially assessing the unit of analysis, exposure assignment, sample identification method, and temporal direction.

Koch's Postulates

A classical four-point microbiological causal framework requiring that an agent be present in all disease cases, isolated in pure culture, capable of inducing disease when inoculated into a healthy host, and recoverable from that host.

Temporality

The sole non-negotiable criterion among Hill's criteria of causation, dictating that the exposure must precede the onset of the disease in chronological sequence.

Biological Gradient

A Hill criterion of causation demonstrated when increasing or decreasing levels of an exposure lead to a proportional, monotonic change in the risk or severity of the outcome (dose-response curve).

Specificity of Association

A Hill criterion positing that causation is strengthened when an exposure produces a single distinct outcome rather than multiple disparate conditions.

Consistency

A Hill criterion of causation met when the observed association is repeatedly demonstrated across different research investigators, geographic locations, target populations, and study methodologies.

Plausibility

A Hill criterion met when an observed epidemiological relationship is supported by an understood biological or physiological mechanism grounded in existing scientific knowledge.

Coherence

A Hill criterion satisfied when the cause-and-effect narrative does not conflict with the known natural history, biology, and laboratory findings of the disease.

Deterministic Causality

A causal model stating that a given cause is invariably followed by a specific effect, categorized according to whether the cause is necessary, sufficient, both, or neither.

Probabilistic (Stochastic) Causality

A modern epidemiological causal paradigm that incorporates randomness and variation, estimating how an exposure shifts the statistical probability of a health event rather than guaranteeing its occurrence.

Necessary Cause

A component causal factor whose presence is mandatory for the disease event to occur; without it, the outcome cannot develop.

Sufficient Cause

A minimal constellation of component causes and conditions that inevitably triggers the development of disease.

Rothman's Causal Pie Model

A multicausal conceptual model illustrating disease etiology as complete pie circles representing sufficient causes made of individual pie slices (component causes), where a component cause present in all pies is deemed a necessary cause.

Web of Causation

An ecological framework modeling chronic non-communicable disease etiology as an intricate network of interconnected factors spanning social, regional, community, worksite, and individual biological determinants.

Internal Validity

The degree to which the design, conduct, and statistical analysis of an epidemiologic investigation accurately measures the association between exposure and outcome without distortion from bias or confounding.

External Validity

The degree to which the empirical results and inferences of a study can be generalized and applied to external populations, alternative settings, and different time periods.

Random Error

The non-systematic divergence of an observed study result from the true underlying population parameter caused by poor measurement precision, sampling error, or biological variability, assessed using $p$-values.

$p$-Value

The probability of observing a result as extreme as, or more extreme than, the study findings purely by random chance under the null hypothesis, conventionally deemed statistically significant when $p < 0.05$.

Systematic Error (Bias)

Any non-random deviation of study results, measurements, or inferences from the truth arising from flaws in participant selection, data collection, or analytic procedures.

Selection Bias

A systematic distortion of an association resulting from the manner in which subjects are recruited, enrolled, or retained in a study, such that the exposure-outcome relationship differs between participants and non-participants.

Berkson's Bias

A form of selection bias arising in hospital-based case-control studies when differing hospital admission rates between cases and controls with different exposure statuses distort the exposure-disease odds ratio.

Healthy Worker Effect

A type of selection bias seen in occupational cohort studies where workers exhibit lower overall morbidity and mortality rates than the general population because severely ill individuals are excluded from employment.

Information Bias

A systematic error in estimating an effect measure caused by misclassification or inaccurate measurement of participant exposure status, confounders, or health outcomes.

Recall Bias

A type of information bias common in retrospective case-control studies where cases report or recall past exposure histories more or less accurately than controls.

Interviewer Bias

A systematic error occurring when study interviewers or abstractors consciously or unconsciously probe more thoroughly for past exposures among cases than among controls.

Observer Bias

A type of information bias where an investigator's prior knowledge of a participant's exposure status influences how they diagnose, measure, or classify the study outcome.

Respondent Bias

A form of information bias that occurs when a study participant systematically misreports their own outcome or disease status during data collection.

Confounder

An extraneous variable that distorts an association because it is independently associated with the exposure in the source population, causally associated with the outcome, and not an intermediate step on the causal pathway.

Restriction

A design-phase technique used to control confounding by limiting study enrollment exclusively to individuals who fall within a specific range or category of the confounding factor.

Stratification

An analytic method to control for confounding where the study sample is separated into mutually exclusive strata of the confounding factor to assess the exposure-disease association separately within each stratum.

Multivariable Modeling

A mathematical regression modeling approach used in the analytic phase to estimate the independent effect of an exposure on an outcome while simultaneously controlling for multiple categorical and continuous confounders.