Epidemiology, Biostatistics, and Evidence-Based Medicine

Comprehensive practice flashcards covering epidemiology, biostatistics, study designs, diagnostic testing, literature evaluation, and evidence-based medicine.

Causal Inference

The systematic process of determining whether an observed empirical association between an exposure and an outcome is likely to be causal rather than spurious.

The Epidemiologic Approach

A multi-step methodology progressing from initial observation and descriptive studies to formulating specific hypotheses, collecting and analyzing data, assessing internal validity, judging causality, and refining hypotheses.

Association

A statistical dependence or relationship between two or more events, characteristics, or variables; it does not inherently prove or imply causation.

Sir Austin Bradford Hill

A world-renowned medical statistician ($1897$-$1991$) who in $1965$ published a landmark set of nine criteria to evaluate whether observed associations reflect true cause-and-effect relationships.

U.S. Surgeon General's 1964 Report on Smoking and Health

The foundational public health report that described preliminary causal guidelines that were later expanded by Sir Austin Bradford Hill in $1965$.

Strength of the Association (Hill's Criteria)

A causal guideline assessing the numerical magnitude of the measure of effect, such as relative risk ($RR$) or odds ratio ($OR$), by comparing disease rates in exposed versus unexposed groups.

No Association (Effect Size Threshold)

An odds ratio, risk ratio, or rate ratio between $1.0$ and $< 1.2$ for increased risk, or between $> 0.83$ and $1.0$ for decreased/protective risk.

Weak Association (Effect Size Threshold)

An odds ratio, risk ratio, or rate ratio between $1.2$ and $< 1.5$ for increased risk, or between $> 0.67$ and $0.83$ for decreased/protective risk.

Moderate Association (Effect Size Threshold)

An odds ratio, risk ratio, or rate ratio between $1.5$ and $< 3.0$ for increased risk, or between $> 0.33$ and $0.67$ for decreased/protective risk.

Strong Association (Effect Size Threshold)

An odds ratio, risk ratio, or rate ratio between $3.0$ and $10.0$ for increased risk, or between $0.10$ and $0.33$ for decreased/protective risk.

Approaching Infinite Association (Effect Size Threshold)

An odds ratio, risk ratio, or rate ratio of $> 10.0$ for increased risk, or $< 0.10$ for decreased/protective risk.

Consistency with Other Investigations (Hill's Criteria)

The repeated observation of the same findings across diverse study populations, geographic settings, methodological designs, and time periods.

British Doctors Study (1956 Lung Cancer RR)

A landmark investigation documenting a relative risk of $14.0$ for cancer of the lung and bronchus among current cigarette-only smokers compared to non-smokers.

U.S. Veterans Study (1966 Lung Cancer RR)

A major epidemiological investigation that identified a relative risk of $12.14$ for cancer of the lung and bronchus among current cigarette-only smokers.

Temporality (Hill's Criteria)

The indispensable causal requirement that the putative exposure of interest must unequivocally precede the onset of the outcome or disease.

20-Year Lag Time

The observed chronological delay between the peak in population cigarette consumption per person per year and the subsequent surge in lung cancer mortality per $100{,}000$ men.

Specificity of the Association (Hill's Criteria)

The condition wherein a single causal factor produces a single distinct outcome; in modern epidemiology, this guideline is recommended almost exclusively for infectious diseases.

Biological Gradient (Dose-Response Relationship)

A causal criterion demonstrated when progressive increments in the level or duration of an exposure correspond to predictable increases in the frequency or severity of the outcome.

Biological Plausibility (Hill's Criteria)

The degree to which an observed empirical association is consistent with existing biological, physiological, pharmacological, or pathological knowledge.

Coherence and Analogy (Hill's Criteria Guidance)

Two of Hill's original criteria that modern epidemiologists advise not relying upon due to their weak explanatory and discriminative power.

Screening Test

A medical test or procedure administered to asymptomatic members of a designated population to detect disease at an early, unrecognized stage.

Medical Surveillance

Another term for a screening test or systematic monitoring protocol applied to asymptomatic populations to assess disease likelihood.

Pre-Clinical Phase

The window in the natural history of disease that occurs between biological onset and the emergence of clinical symptoms, during which screening can detect pathology.

Goal of Screening

To detect disease during the pre-clinical detectable phase so that early intervention or treatment can successfully reduce patient morbidity and mortality.

WHO Screening Principles

Criteria establishing that a screened condition must be an important public health problem with high prevalence/serious outcome, an identifiable latent phase, understood natural history, agreed treatment, and cost less than subsequent medical care.

The Perfect Diagnostic Test

A theoretical diagnostic test wherein all diseased individuals produce one uniform value, all non-diseased produce another distinct uniform value, and zero ambiguous results occur.

Gold Standard Test

The definitive clinical test (such as tissue biopsy, autopsy, or DSM-$5$ criteria) assumed to be $100\%$ accurate in establishing true disease status.

Sensitivity

The probability that a screening or diagnostic test will correctly return a positive result among individuals who truly have the disease ($\frac{a}{a+c}$).

Specificity

The probability that a screening or diagnostic test will correctly return a negative result among individuals who truly do not have the disease ($\frac{d}{d+b}$).

Positive Predictive Value (PV+)

The probability that a patient who tests positive on a screening test actually possesses the target disease ($\frac{a}{a+b}$).

Negative Predictive Value (PV-)

The probability that a patient who tests negative on a screening test is truly free of the target disease ($\frac{d}{c+d}$).

Sensitivity-Specificity Trade-Off

The inherent diagnostic compromise where adjusting cutoffs to increase sensitivity inherently reduces specificity, and vice versa.

High Sensitivity / Low Specificity Example

Mammography, which is calibrated to minimize false-negative results so that true malignancy is rarely missed, even at the cost of false positives.

High Specificity / Low Sensitivity Example

The urine nitrate dipstick test, used in hospitalized patients to reliably rule in urinary tract infections with minimal false positives.

Phenylketonuria (PKU)

A rare inborn metabolic disorder wherein the body cannot metabolize phenylalanine due to enzyme deficiency, causing toxic neurological accumulation, seizures, and severe intellectual disability.

Phenylalanine Hydroxylase (PAH)

The metabolic enzyme that normally hydroxylates dietary phenylalanine into tyrosine, the impairment of which causes classic phenylketonuria.

PKU Heel Stick Test

An inexpensive, mandated blood screening test performed on newborns to detect abnormal phenylalanine accumulation before brain injury occurs.

PKU Newborn Screening Sensitivity (1993-1994 Data)

Calculated at $100.00\%$ ($\frac{245}{245}$), ensuring that zero true PKU cases yielded a false-negative result in the screened population.

PKU Newborn Screening Specificity (1993-1994 Data)

Calculated at $99.76\%$ ($\frac{3{,}751{,}508}{3{,}760{,}505}$) based on national neonatal screening data.

PKU Positive Predictive Value Paradox

Despite a near-perfect specificity of $99.76\%$, the positive predictive value was only $2.65\%$ ($\frac{245}{9{,}232}$) due to the extremely low disease prevalence in the population.

Inferential Statistics

Mathematical and statistical procedures that allow researchers to analyze sample data to make generalized inferences about the larger population of interest.

Population (Statistics)

The complete, entire collection of individuals, objects, or measurements about which a scientific inquiry is made.

Sample (Statistics)

A selected subset or portion drawn from an overarching population to represent that population in an investigation.

Population Parameter

A numerical value that describes a characteristic of an entire population, symbolized by Greek letters such as $\mu$ (mean) and $\sigma$ (standard deviation).

Sample Statistic

A numerical value calculated from sample observations that estimates a population parameter, symbolized by Latin letters such as $\bar{X}$ (mean) and $s$ (standard deviation).

Statistical Estimation

The inferential process of approximating unknown population parameters from computed sample statistics.

Role of Chance (Inferential Statistics)

The natural random variation inherent in sample selection that can cause sample findings to deviate from true population parameters.

Null Hypothesis (H0)

A statistical statement specifying that there is no change, no difference, no effect, or no association between variables in the population.

Alternative Hypothesis (H1)

A statistical statement asserting that there is a true difference, effect, or relationship between study variables in the target population.

Directional (One-Tailed) Test

A statistical hypothesis test wherein the researcher specifies the expected direction of the effect in advance, concentrating the rejection region into a single tail of the distribution.

Non-Directional (Two-Tailed) Test

The conventional, more conservative statistical test that splits the significance region across both tails, detecting differences regardless of direction.

Physicians' Health Study (1982)

A double-blind randomized clinical trial of $22{,}071$ male physicians stopped early by its DSMB because low-dose aspirin ($325\,\text{mg}$ every other day) demonstrated a major protective effect on first MI.

Alpha Level (Significance Level)

The maximum acceptable probability of making a Type I error, established a priori and standardly set at $\alpha = 0.05$.

Type I Error (Alpha Error)

The error of incorrectly rejecting a true null hypothesis, thereby concluding an effect or association exists when it does not (analogous to convicting an innocent defendant).

Type II Error (Beta Error)

The error of failing to reject a false null hypothesis, thereby missing a genuine treatment effect or association (analogous to failing to convict a guilty defendant).

Statistical Power

The probability of correctly rejecting a false null hypothesis ($\text{Power} = 1 - \beta$), indicating the study's ability to detect a true effect.

p-value

The calculated probability, obtained a posteriori, of observing results as extreme as or more extreme than the study data assuming the null hypothesis is true.

Statistical Significance

A determination achieved when the calculated $p$-value is less than the predetermined $\alpha$ level ($p < 0.05$), indicating that results are unlikely due to chance alone.

Clinical vs. Statistical Significance

The distinction where an observed outcome may achieve statistical significance ($p < 0.05$) but represent a difference so small (e.g., $1.9\%$ or $0.04\,\text{L}$ in $\text{FEV}_1$) that it provides no practical clinical benefit to the patient.

Confidence Interval (CI)

An interval estimate constructed around a sample point estimate ($\text{point estimate} \pm \text{margin of error}$) that has a specified likelihood (e.g., $95\%$) of containing the true population parameter.

Null Value for Difference in Means

A value of $0$; if a $95\%$ confidence interval for the difference between two means includes $0$, the finding is not statistically significant ($p \ge 0.05$).

Null Value for Difference in Proportions (Ratios)

A value of $1.0$; if a $95\%$ confidence interval for a relative risk or odds ratio includes $1.0$, the finding is not statistically significant ($p \ge 0.05$).

Confidence Interval Width

An index of the precision of an estimate; narrower intervals reflect higher precision and are driven by larger sample sizes and lower outcome variance.

Relative Risk (RR)

A comparative ratio measuring the incidence of disease in an exposed cohort relative to the incidence of disease in an unexposed cohort ($\frac{\text{Incidence in exposed}}{\text{Incidence in unexposed}}$).

2x2 Contingency Table

A cross-tabulation table classifying individuals according to exposure status (rows) and disease or outcome status (columns).

Relative Risk Formula (2x2 Table)

The algebraic ratio computed from a contingency table as $RR = \frac{a / (a+b)}{c / (c+d)}$.

SGLT2 Inhibitor UTI Cohort Example

A study evaluating $2{,}390$ women where $27$ of $482$ flozin users ($5.6\%$) and $77$ of $1{,}908$ non-users ($4.0\%$) developed urinary tract infections, yielding an $RR$ of $1.4$.

Relative Risk Interpretation (RR > 1)

Expressed as exposed individuals being $RR$ times as likely to develop the outcome, or having a percentage increase calculated by $(RR - 1) \times 100\%$.

Relative Risk of 14.0

An effect size indicating a $14$-fold increase in risk, which corresponds to a $1{,}300\%$ increase in the outcome among exposed individuals compared to unexposed individuals.

Relative Risk of 0.5

An effect size indicating a $0.5$-times risk, which corresponds to a $2$-fold or $100\%$ relative reduction in risk ($\frac{1}{0.5} = 2.0$).

Relative Risk of 0.1

An effect size indicating a $10$-fold relative risk reduction, or a $900\%$ reduction in risk ($\frac{1}{0.1} = 10.0$).

ACIP HepB Infant Dose Recommendation

Advisory Committee on Immunization Practices guidance recommending $\ge 3$ doses of hepatitis B vaccine for infants weighing $\ge 2{,}000\,\text{g}$, and $\ge 4$ doses for infants weighing $< 2{,}000\,\text{g}$.

Infant Hepatitis B Vaccine Cohort Study

An investigation of infants of infected mothers showing disease incidence of $0.0667$ ($3/45$) in those with $< 3$ doses versus $0.0105$ ($97/9{,}207$) in those with $\ge 3$ doses, yielding an $RR$ of $6.35$.

Units of Relative Risk

Relative risk has no physical units of measurement because it is a direct dimensionless ratio of two incidence proportions.

Forest Plot

A graphical representation utilized in meta-analyses displaying the point estimates and confidence intervals of individual studies alongside an overall pooled effect estimate.

Logarithmic Scale for Relative Risk

A plotting transformation that renders an $RR$ of $0.5$ (two-fold decrease) and an $RR$ of $2.0$ (two-fold increase) equidistant from the baseline value of $1.0$.

Morbidity

Any departure, subjective or objective, from a state of physiological or psychological well-being.

Prevalence

The proportion of individuals in a specified population who have a particular disease or condition at a designated point or period in time, encompassing both new and pre-existing cases.

Incidence

A measure capturing the rate of occurrence of new cases of a disease arising within a population at risk during a specified timeframe.

Population at Risk

Individuals who are biologically capable of developing the disease of interest and do not already have it at the beginning of the observation interval.

Point Prevalence

The number of individuals with a disease or attribute at a single moment in time divided by the total number of individuals in the population at that specific moment.

Adult Lifetime Asthma Prevalence (NM 2021)

Recorded at $14.0\%$ in New Mexico adults aged $18+$, which is comparable to the United States national rate of $14.6\%$.

Adult Current Asthma Prevalence (NM 2021)

Recorded at $10.6\%$ in New Mexico adults aged $18+$, higher than the United States national rate of $9.7\%$.

Childhood Lifetime Asthma Prevalence (NM 2021)

Recorded at $8.9\%$ among New Mexico children aged $0$-$17$, lower than the national United States rate of $10.2\%$.

Incidence Rate for Aggregate Data

A vital statistics metric defined as the number of new disease cases during a time period divided by the estimated mid-period population, usually reported per $100{,}000$ people.

Mid-Period Population

The population estimate determined as of July $1$, standardly used as the denominator for calculating annual aggregate incidence and mortality rates.

New Mexico Cancer Incidence Rate (2019)

Calculated as $443.7$ new cases per $100{,}000$ population, based on $9{,}330$ new cases among a mid-year population of $2{,}102{,}656$.

Cumulative Incidence (CI)

The proportion of a closed population initially at risk that becomes diseased during a specified time interval ($\frac{\text{New cases in period } T}{\text{Population at risk at start of } T}$), expressed as a percentage.

Tuberculosis in Lung Cancer Patients Example

A study following $904$ Japanese lung cancer patients for $2$ years in which $9$ developed TB, yielding a cumulative incidence of $0.996\%$ ($\frac{9}{904} \times 100\%$).

Prevalence Formula (Steady-State)

The mathematical relationship where prevalence roughly equals incidence multiplied by the average duration of the disease ($P = I \times D$).

Incidence-Prevalence Reservoir Model

A conceptual model wherein new cases (incidence) flow into a pool (prevalence), which is depleted through recovery (cures) or death.

Crude Mortality Rate

The total number of deaths occurring from all causes in a population during a specified period divided by the population size, standardly expressed per $1{,}000$ persons.

Cause-Specific Mortality Rate

The number of deaths attributed to a specific disease during a defined period divided by the mid-period population, typically expressed per $100{,}000$ persons.

Case-Fatality Rate

The proportion of individuals diagnosed with a specific medical condition who die from that disease, serving as a direct measure of disease severity.

Infant Mortality Rate

The number of deaths in children under $1$ year of age occurring during a given year per $1{,}000$ live births.

High School Smoking Peak (CDC YRBS)

The peak cigarette smoking prevalence among United States high school students, which reached $36.4\%$ in $1997$ before falling to $10.8\%$ in $2015$.

Descriptive Statistics

Procedures used to summarize, organize, and describe sample data through tables, figures, and numerical metrics without testing population hypotheses.

Table 1 (Biomedical Publications)

The standardized first table in a clinical trial or study paper displaying the baseline descriptive statistics and demographics of the study sample.

Categorical Data

Qualitative data comprised of mutually exclusive groupings or classes into which observations are placed.

Continuous Data

Quantitative metric data that can assume any numeric value along a continuous scale or continuum within a specified range.