Epidemiology, biostatistics, and informatics concepts in your own words

Age-adjusted mortality rate. A mortality rate statistically modified to eliminate the effect of different age distributions in the different populations.

Age-specific mortality rate. A mortality rate limited to a particular age group. The numerator is the number of deaths in that age group; the denominator is the number of persons in that age group in the population.

Agent. A factor, such as a microorganism, chemical substance, or form of radiation, whose presence, excessive presence, or (in deficiency diseases) relative absence is essential for the occurrence of a disease.

Alternative hypothesis. See "Hypothesis, alternative."

Analysis of covariance (ANCOVA). An analysis used when the independent variables are a mixture of continuous and nominal variables. It can be viewed as a multiple regression in which nominal independent variables are included in the regression equation.

Analysis of variance (ANOVA), factorial. An analysis in which dependent-variable values are treated as categories of more than one factor. It allows additional null hypotheses about the differences among means to be tested, including main effects and interactions.

Analysis of variance (ANOVA), one-way. An analysis in which dependent-variable values are treated as categories of a single factor. The only hypothesis tested is that all group means are equal.

Analytic epidemiology. The aspect of epidemiology concerned with the search for health-related causes and effects. Uses comparison groups, which provide baseline data, to quantify the association between exposures and outcomes, and to test hypotheses about causal relationships.

Analytic study. A comparative study intended to identify and quantify associations, test hypotheses, and identify causes. Two common types are the cohort study and the case-control study.

Antigenic drift. Minor change in influenza subtype, usually resulting from RNA point mutation in the coding region of hemagglutinin.

Antigenic shift. Major change in subtype that occurs with recombination of an RNA genome segment.

Application programming interface (API). A defined set of rules that lets one software system request data or services from another. Modern health data standards such as FHIR expose data through web-based APIs.

Artifactual changes (or differences in disease incidence).

• diagnostic inaccuracy;
• new or more sensitive diagnostic tests;
• more complete ascertainment of cases;
• differences in quality of medical care or access to care can affect accuracy of diagnosis and completeness of case ascertainment;
• but inadequate access to care or quality of care can also be a real risk factor for development of disease;
• various types of respondent error in interview surveys: inadequate knowledge; inaccurate recall; intentionally inaccurate responses; interviewer error or bias; data entry error; non-response; changes in definition or classification of a disease; errors in the enumeration of the population; in special cases, failure to remove appropriate persons from the denominator because they are no longer at risk of developing the disease; and sampling error.

Ascertainment bias. Selection biases related to ascertainment of disease status. Due to:

• differential medical surveillance of exposed or unexposed, and differential tendency to diagnose a condition among exposed or unexposed;
• differential hospital referral patterns among exposed or unexposed cases; or
• selecting controls from individuals with a disease that might be caused by the exposure of interest.

(See "Artifactual changes.")

Association. Statistical relationship between two or more events, characteristics, or other variables.

Attack rate. A variant of an incidence rate, applied to a narrowly defined population observed for a limited period of time, such as during an epidemic.

The attack rate (AR) is the proportion of people in an exposed or unexposed group who became ill.

Attributable risk among the exposed. The proportion of disease among those exposed to the factor that can be attributed to that exposure.

Bar chart. A visual display of the size of the different categories of a variable. Each category or value of the variable is represented by a bar.

Bias. Deviation of results or inferences from the truth, or processes leading to such systematic deviation. Any trend in the collection, analysis, interpretation, publication, or review of data that can lead to conclusions that are systematically different from the truth.

Binary data (dichotomous data). A variable with only two possible categories, a special case of nominal data. Examples: disease status (yes or no) and sex.

Biologic transmission. The indirect vector-borne transmission of an infectious agent in which the agent undergoes biologic changes within the vector before being transmitted to a new host.

Blinding. Reduces bias from:

• differing psychological effect of one treatment versus the other on the occurrence of symptoms and side effects;
• subjective perception or reporting of symptoms and side effects by study subjects;
• difference in concomitant or compensatory treatment by either the subject or the investigator.

Box plot. A visual display that summarizes data using a "box and whiskers" format to show the minimum and maximum values (ends of the whiskers), interquartile range (length of the box), and median (line through the box).

Carrier. A person or animal without apparent disease who harbors a specific infectious agent and is capable of transmitting the agent to others. Carrier state may occur in an individual with an infection that is inapparent throughout its course (known as an asymptomatic carrier), or during the incubation period, convalescence, and postconvalescence of an individual with a clinically recognizable disease. Carrier state may be of short or long duration (transient carrier or chronic carrier).

Case. In epidemiology, a countable instance in the population or study group of a particular disease, health disorder, or condition under investigation. Sometimes, an individual with the particular disease.

Case-control study. A type of observational analytic study. Enrollment into the study is based on presence ("case") or absence ("control") of disease. Characteristics such as previous exposure are then compared between cases and controls.

Case definition. A set of standard criteria for deciding whether a person has a particular disease or health-related condition, by specifying clinical criteria and limitations on time, place, and person.

Case-fatality rate. The proportion of persons with a particular condition (cases) who die from that condition. The denominator is the number of incident cases; the numerator is the number of cause-specific deaths among those cases.

Causality, guidelines for (guidelines for assessing causality).

Major criteria:

• temporal relationship;
• biological plausibility;
• consistency;
• alternative explanations.

Other considerations:

• dose-response relationship;
• strength of the association;
• cessation effects.

Source: Adapted from Hill AB. The environment and disease: association or causation? Proceedings of the Royal Society of Medicine. 1965;58(5):295–300.

Cause of disease. A factor (characteristic, behavior, event, etc.) that directly influences the occurrence of disease. A reduction of the factor in the population should lead to a reduction in the occurrence of disease.

Cause-specific mortality rate. The mortality rate from a specified cause for a population. The numerator is the number of deaths attributed to a specific cause during a specified time interval; the denominator is the size of the population at the midpoint of the time interval.

Central limit theorem. The result that the sampling distribution of an estimate such as the mean approaches a normal (Gaussian) shape as the sample size increases, whatever the shape of the underlying data. It affects the sampling distribution, not the distribution of the data.

Class interval. A span of values of a continuous variable which are grouped into a single category for a frequency distribution of that variable.

Cluster. An aggregation of cases of a disease or other health-related condition, particularly cancer and birth defects, which are closely grouped in time and place. The number of cases may or may not exceed the expected number; frequently the expected number is not known.

Cohort. A well-defined group of people who have had a common experience or exposure, who are then followed up for the incidence of new diseases or events, as in a cohort or prospective study. A group of people born during a particular period or year is called a birth cohort.

Cohort definition. A set of computable criteria that specify how individuals enter and leave a study cohort. In observational informatics, a cohort definition is often written so it can be executed against a common data model.

Cohort study. A type of observational analytic study. Enrollment into the study is based on exposure characteristics or membership in a group. Disease, death, or other health-related outcomes are then ascertained and compared.

Collectively exhaustive events. A set of events that names every possibility, so that at least one of them must occur.

Common data model (CDM). A specification that defines a standard structure, and usually standard vocabularies, for organizing data from multiple sources, so that the same analysis can run against each source without being rewritten.

Common source outbreak. An outbreak that results from a group of persons being exposed to a common noxious influence, such as an infectious agent or toxin. If the group is exposed over a relatively brief period of time, so that all cases occur within one incubation period, then the common source outbreak is further classified as a point source outbreak. In some common source outbreaks, persons may be exposed over a period of days, weeks, or longer, with the exposure being either intermittent or continuous.

Complement (complementary event). The complement of an event is the probability that the event does not happen.

Computable phenotype. An algorithm, expressed in code or logic, that identifies the set of individuals with a particular condition or characteristic from health data, using defined inclusion and exclusion criteria. (Compare "Case definition.")

Concept. In a standardized vocabulary, such as those used by the OMOP common data model, a concept is a distinct clinical idea with a unique identifier (a concept ID) to which source codes are mapped.

Confidence interval. A range of values for a quantity of interest, for example a rate, constructed so that the range has a specified probability of including the true value. The specified probability is the confidence level, and the end points are the confidence limits. In plain terms, it gives an interval within which we have a stated level of confidence that the population value lies.

Confidence limit. The minimum or maximum value of a confidence interval.

Confounding. Occurs when exposed and unexposed groups are not comparable with regard to the distribution of another factor which is itself a risk factor for the disease under study. The presence of the confounding factor distorts the risk estimate relating exposure and disease. If Risk Factor #2 meets two criteria, it will confound the relationship between Risk Factor #1 and the outcome:

1. Is Risk Factor #2 associated with the outcome in the study population?
2. Is Risk Factor #2 associated with Risk Factor #1?

If both are answered yes, then Risk Factor #2 is a confounder. The degree of confounding depends on how strongly the confounder is associated with both the outcome and the other risk factor. Confounding can be controlled through study design or in analysis, for example by stratified analysis or by including the confounder in a regression model.

Consolidated Clinical Document Architecture (C-CDA). A Health Level Seven (HL7) standard that specifies a library of templates for structured clinical documents, such as discharge summaries and care summaries, so they can be exchanged and understood across systems.

Contact. Exposure to a source of an infection, or a person so exposed.

Contingency table. A two-variable table with cross-tabulated data.

Continuous data. A variable with a large number of possible values that are evenly spaced. Examples: age, weight, blood pressure, and pH.

Control. In a case-control study, the comparison group of persons without disease.

Controlled vocabulary. A curated set of terms with defined meanings, used to record information consistently and to reduce ambiguity across systems and analyses. Also called a terminology.

Correlation coefficient. A measure that reflects the strength and direction of the association between two or more variables. A sample correlation can be tested against a population value using the same Gaussian approach used for means, slopes, and intercepts, under the null hypothesis that the population correlation coefficient equals zero.

Crude mortality rate. The mortality rate from all causes of death for a population.

Cumulative frequency. In a frequency distribution, the number or proportion of cases or events with a particular value or in a particular class interval, plus the total number or proportion of cases or events with smaller values of the variable.

Cumulative frequency curve. A plot of the cumulative frequency rather than the actual frequency for each class interval of a variable. This type of graph is useful for identifying medians and other percentiles.

Data dictionary. A structured reference that documents the variables in a data set, including their names, definitions, formats, and permissible values.

Data governance. The policies, roles, and processes that ensure data are managed as an asset, covering quality, access, security, stewardship, and appropriate use.

Data model. A description of how data are structured and related, including entities, fields, data types, and the relationships among them.

Data provenance. A record of the origin of data and the transformations applied to them over time, used to understand and trust how a data set was produced.

Data quality. The degree to which data are fit for their intended use, assessed across dimensions such as completeness, conformance to expected formats and vocabularies, and plausibility.

De-identification. The process of removing or obscuring information that could identify an individual from a data set, so the data can be used or shared with reduced privacy risk.

Demographic information. The person characteristics (age, sex, race, and occupation) of descriptive epidemiology, used to characterize the populations at risk.

Denominator. The lower portion of a fraction used to calculate a rate or ratio. In a rate, the denominator is usually the population (or population experience, as in person-years) at risk.

Dependent variable. The outcome variable, or the variable whose values are a function of other variables (the independent variables). It is the data about which we want to draw a conclusion.

Descriptive epidemiology. The aspect of epidemiology concerned with organizing and summarizing health-related data according to time, place, and person.

Dichotomous data. See "Binary data."

Differential misclassification. When classification errors differ between groups. Risk estimates can be biased in either direction.

Direct transmission. The immediate transfer of an agent from a reservoir to a susceptible host by direct contact or droplet spread.

Distribution. In epidemiology, the frequency and pattern of health-related characteristics and events in a population. In statistics, the observed or theoretical frequency of values of a variable.

Distribution of data. See "Sampling distribution" for the contrast between the distribution of data and the sampling distribution.

Dose-response relationship. As the dose of exposure increases, the risk of disease also increases.

Electronic health record (EHR). A digital record of a person's health information generated across one or more care encounters, designed to be maintained and shared across care settings.

Endemic disease. The constant presence of a disease or infectious agent within a given geographic area or population group; may also refer to the usual prevalence of a given disease within such area or group.

Environmental factor. An extrinsic factor (geology, climate, insects, sanitation, health services, etc.) which affects the agent and the opportunity for exposure.

Epidemic. The occurrence of more cases of disease than expected in a given area or among a specific group of people over a particular period of time.

Epidemic curve. A histogram that shows the course of a disease outbreak or epidemic by plotting the number of cases by time of onset.

Epidemiologic triad. The traditional model of infectious disease causation. Includes three components: an external agent, a susceptible host, and an environment that brings the host and agent together, so that disease occurs.

Epidemiology. The study of the distribution and determinants of health-related states or events in specified populations, and the application of this study to the control of health problems.

Source: Porta M, ed. A Dictionary of Epidemiology. 6th ed. Oxford University Press; 2014. This definition is also the one used by CDC in Principles of Epidemiology in Public Health Practice, 3rd ed.

Epidemiology, objectives of.

1. To identify the cause and risk factors of a disease.
2. To determine the extent of disease found in the community.
3. To study the natural history and prognosis of disease.
4. To evaluate existing and new preventive and therapeutic measures and modes of health care delivery.
5. To provide the foundation for developing public policy and regulatory decisions relating to environmental problems.

Ethics guidelines for epidemiologists. The American College of Epidemiology (ACE) Ethics Guidelines set out eleven key duties and obligations. Several carry subheadings, shown nested below.

1. Professional role of epidemiologists.
2. Minimizing risks and protecting the welfare of research participants.
3. Providing benefits.
4. Ensuring an equitable distribution of risks and benefits.
5. Protecting confidentiality and privacy.
6. Obtaining informed consent.
   • Elements of informed consent. (See "Informed consent, elements of.")
   • Avoidance of manipulation or coercion.
   • Conditions under which informed consent requirements may be waived.
7. Submitting proposed studies for ethical review.
8. Maintaining public trust.
   • Adhering to the highest scientific standards.
   • Involving community representatives in research.
9. Avoiding conflicts of interest and partiality.
10. Communicating ethical requirements and confronting unacceptable conduct.
11. Obligations to communities.
    • Reporting results.
    • Public health advocacy.
    • Respecting cultural diversity.

Source: American College of Epidemiology. Ethics Guidelines. Annals of Epidemiology. 2000;10(8):487–497. Structure summarized in McKeown RE, Weed DL, Kahn JP, Stoto MA. American College of Epidemiology Ethics Guidelines: Foundations and Dissemination. Science and Engineering Ethics. 2003;9(2):207–214.

Etiology. Cause of a disease.

Experimental study. A study in which the investigator specifies the exposure category for each individual (clinical trial) or community (community trial), then follows the individuals or community to detect the effects of the exposure.

Exposed (group). A group whose members have been exposed to a supposed cause of disease or health state of interest, or possess a characteristic that is a determinant of the health outcome of interest.

External validity. External validity, or generalizability, refers to the extent to which the population we have studied is representative, and whether we can extrapolate our findings to other populations.

Extract, transform, load (ETL). The process of extracting data from source systems, transforming them to fit a target structure and vocabulary, and loading them into a target database, for example when mapping source data to a common data model.

Factorial analysis of variance. See "Analysis of variance (ANOVA), factorial."

Fast Healthcare Interoperability Resources (FHIR). A Health Level Seven (HL7) standard for exchanging healthcare information electronically. FHIR organizes information into modular components called resources and exchanges them through web-based APIs.

Frequency distribution. A complete summary of the frequencies of the values or categories of a variable; often displayed in a two-column table: the left column lists the individual values or categories, and the right column indicates the number of observations in each category.

Frequency polygon. A graph of a frequency distribution with values of the variable on the x-axis and the number of observations on the y-axis; data points are plotted at the midpoints of the intervals and are connected with a straight line.

Generation time. Interval from infection to time of maximal infectivity for others.

Graph. A way to show quantitative data visually, using a system of coordinates.

Health Level Seven (HL7). A standards development organization, and the family of standards it produces, for the exchange, integration, sharing, and retrieval of electronic health information. C-CDA and FHIR are HL7 standards.

Herd immunity. The resistance of a group to invasion and spread of an infectious agent, based on the resistance to infection of a high proportion of individual members of the group. The resistance is a product of the number susceptible and the probability that those who are susceptible will come into contact with an infected person.

Histogram. A graphic representation of the frequency distribution of a continuous variable. Rectangles are drawn in such a way that their bases lie on a linear scale representing different intervals, and their heights are proportional to the frequencies of the values within each of the intervals.

Homoscedasticity. The assumption that the variances are equal across different groups of dependent-variable values. It is assumed so that a single pooled estimate of the variance can be used, drawing on information from all groups.

Host. A person or other living organism that can be infected by an infectious agent under natural conditions.

Host factor. An intrinsic factor (age, race, sex, behaviors, etc.) which influences an individual's exposure, susceptibility, or response to a causative agent.

Hypothesis. A supposition, arrived at from observation or reflection, that leads to refutable predictions. Any conjecture cast in a form that will allow it to be tested and refuted.

Hypothesis, alternative. The hypothesis adopted if the null hypothesis proves implausible, stating that exposure is associated with disease, or more generally that there is an effect.

• A two-sided alternative hypothesis includes values on both sides of the null value; the p-value is calculated from both tails of the sampling distribution.
• A one-sided alternative hypothesis includes values on only one side of the null value; the p-value is calculated from the single tail on that side.

Almost all alternative hypotheses should be two-sided. The only exception is when it is impossible for the population value to fall on the other side of the null value.

Hypothesis, null. The hypothesis tested in a test of statistical significance, which assumes that the exposure is not related to disease, or more generally that there is no effect. Null hypotheses are usually specific statements about a population: that a difference equals zero, that a ratio equals one, or that nothing changes in relation to anything else. A workable null hypothesis satisfies both the statistician, who needs a specific statement about a population, and the researcher, who needs something biologically interesting.

Hypothesis testing, statistical. The steps are:

1. State the hypothesis to be tested. It should describe a condition worth testing and make a specific statement about the population. Because such statements usually say that differences equal zero or ratios equal one, the hypothesis tested is the null hypothesis.
2. Collect data.
3. Calculate the probability of getting a sample estimate at least as far from the hypothesized value as the observed estimate, if the null hypothesis were true. This probability is the p-value.
4. Reject or fail to reject the null hypothesis. A p-value of 0.05 or less is generally considered small enough to reject it. Rejecting the null hypothesis provides support for the alternative hypothesis.

Immunity, active. Resistance developed in response to stimulus by an antigen (infecting agent or vaccine) and usually characterized by the presence of antibody produced by the host.

Immunity, herd. See "Herd immunity."

Immunity, passive. Immunity conferred by an antibody produced in another host and acquired naturally by an infant from its mother, or artificially by administration of an antibody-containing preparation (antiserum or immune globulin).

Immunogenicity. Ability of an agent to produce an immune response in a host.

Incidence rate. A measure of the frequency with which an event, such as a new case of illness, occurs in a population over a period of time. The denominator is the population at risk; the numerator is the number of new cases occurring during a given time period.

Real changes or differences in disease incidence:

• increase or decrease in environmental or behavioral risk factors for disease;
• differences between populations in relevant demographic characteristics that are associated with environmental or behavioral risk factors;
• differences in genetic factors between populations.

Incubation period. Interval from infection to onset of clinical illness.

Independent variable. An exposure, risk factor, or other characteristic being observed or measured that is hypothesized to influence an event or manifestation (the dependent variable). It is the condition under which we examine the dependent variable.

Indicator variable. A variable that distinguishes categories, for example two treatment groups. Its regression coefficient gives the difference between the group means of the dependent variable. When an interaction is present, the indicator variable's coefficient gives the difference between the group intercepts.

Infectivity. Ability of an agent to invade and multiply in a host.

Inference, statistical. In statistics, the development of generalizations from sample data, usually with calculated degrees of uncertainty.

Information bias. Inaccurate classification of study subjects with respect to disease or exposure status. Can result from: inaccurate or incomplete recall of prior exposures, symptoms, etc. by study subjects; inaccurate reporting of disease or exposure information; systematic error due to interviewers' gathering of selective data; inaccurate measurement of exposure or disease-related parameters; or inaccurate diagnosis of disease.

Informed consent, elements of.

• purpose of study;
• sponsors;
• investigators;
• scientific methods and procedures;
• anticipated risks and benefits;
• anticipated inconveniences or discomfort;
• the individual's right to refuse participation or to withdraw from the research at any time without repercussions.

Source: American College of Epidemiology. Ethics Guidelines. Annals of Epidemiology. 2000;10(8):487–497. (Subheading under "Obtaining informed consent"; see "Ethics guidelines for epidemiologists.")

Interaction. The effect of one risk factor (or independent variable) on the outcome differs depending on the level of another. When interaction is present, the effects of the two factors cannot be separated; they are linked. In factorial analysis of variance, an interaction indicates that the main effects alone do not explain the differences among the group means, so the means must be examined individually.

Interaction variable. A variable created by multiplying an indicator variable by another independent variable (either a continuous variable or another indicator variable). It is included in a regression so that the regression lines can have different slopes; its coefficient gives the difference between the slopes for the groups.

Intercept. The value of the dependent variable where the regression line crosses the y-axis; the value of y when x = 0.

Internal validity. Refers to whether a particular study was well done and, therefore, whether the findings are valid for the population that was studied.

International Classification of Diseases (ICD). A classification system maintained by the World Health Organization for coding diagnoses and causes of death. National clinical modifications, such as ICD-10-CM in the United States, add detail for recording illness.

Interoperability. The ability of different information systems and software to exchange data and to use the information that has been exchanged.

Latency period. A period of subclinical or inapparent pathologic changes following exposure, ending with the onset of symptoms of chronic disease.

Least squares. An estimation approach that chooses parameter values to minimize the squared differences between observed and predicted values. It is preferred when possible because it is computationally simpler than maximum likelihood.

Logical Observation Identifiers Names and Codes (LOINC). A standard set of codes for identifying laboratory tests, measurements, and clinical observations.

Logistic regression. A regression model for a binary outcome that uses the logit transformation so that predicted probabilities stay between 0 and 1.

Logit transformation. A transformation used in logistic regression to keep predicted probabilities between 0 and 1.

Main effect. The effect of a single factor, tested as the null hypothesis that the population mean is the same across all categories of that factor.

Maximum likelihood. An iterative estimation approach that searches for the parameter values that make the observed data most probable. It is more computationally intensive than least squares.

Mean, arithmetic. The measure of central location commonly called the average, calculated by adding all the individual values in a group of measurements and dividing by the number of values. Intuitively, the center of gravity of the data.

Measure of association. A quantified relationship between exposure and disease; includes relative risk, rate ratio, and odds ratio.

Metadata. Data that describe other data, such as their meaning, structure, source, format, or quality.

Missing data. Values that are absent for some observations in a data set. Patterns of missingness can bias analyses if they are not understood and handled appropriately.

Mixed epidemic. This type of epidemic begins with a common source exposure to an infectious agent, with subsequent propagative spread (for instance, a foodborne outbreak, followed by secondary person-to-person spread from those who were initially ill).

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

Mortality rate. A measure of the frequency of occurrence of death in a defined population during a specified interval of time.

Mortality rate, infant. A ratio expressing the number of deaths among children under one year of age reported during a given time period, divided by the number of births reported during the same time period. The infant mortality rate is usually expressed per 1,000 live births.

Mortality rate, neonatal. A ratio expressing the number of deaths among children from birth up to but not including 28 days of age, divided by the number of live births reported during the same time period. The neonatal mortality rate is usually expressed per 1,000 live births.

Multicollinearity. A condition in which two or more independent variables in a model are highly correlated with one another. Multicollinearity makes it difficult to separate the individual effects of those variables and inflates the standard errors of their estimated coefficients.

Multiple comparison problem. The inflation of the overall chance of a Type I error that occurs when more than one statistical test is performed, so that the combined chance of a false positive exceeds the nominal 5%. Procedures such as Student-Newman-Keuls and Tukey address the problem. The Bonferroni correction also controls the error rate but is more conservative, which makes it harder to reject the null hypothesis.

Mutually exclusive events. Events such that only one of them can occur; if one happens, the others cannot. Mutually exclusive events cannot be statistically independent. Mutual exclusion is often mistaken for statistical independence.

Natural history of disease. The temporal course of disease from onset (inception) to resolution.

Natural language processing (NLP). Computational methods for extracting structured information from unstructured text, such as clinical notes.

Naturalistic sample. A sample obtained by randomly sampling values of the independent variable, so that their distribution in the sample reflects their distribution in the population.

Necessary cause. A causal factor whose presence is required for the occurrence of the effect (of disease).

Negative predictive value (NPV). Percent of people who do not actually have the disease among those who test negative for the disease.

Net sensitivity. People who correctly screen positive on both tests as a percent of all people with the illness.

Net specificity. People who screen negative on the first test, plus people who correctly screen negative on the second test, as a percent of all people without an illness. With serial screening, there is a loss in net sensitivity and a gain in net specificity.

Nominal data. A variable whose values cannot be placed in a meaningful order. Examples: sex, race, disease status, and country of origin. A variable with only two categories is a special case called binary or dichotomous data.

Non-differential classification. Classification errors of the same magnitude in both groups; risk estimates biased towards the null (that is, towards 1.0). This reduces power, reducing the apparent sample size of the study.

Nonparametric test. A test that does not rely on assumptions about the shape of the underlying distribution, such as a Gaussian distribution. Nonparametric tests are used for ordinal dependent variables and when distributional assumptions are in doubt, for example when a sample is too small for the central limit theorem to apply.

Normal approximation. Approximating a sampling distribution that is not exactly Gaussian with a Gaussian distribution. A common guideline is that at least 5 events and 5 non-events are needed to use the normal approximation to the binomial distribution.

Nosocomial infection. Infection acquired in a hospital setting.

Null hypothesis. See "Hypothesis, null."

Numerator. The upper portion of a fraction.

Observational Health Data Sciences and Informatics (OHDSI). An open, international, multi-stakeholder collaborative that develops open-source tools and methods for observational health research, and that maintains the OMOP common data model and its standardized vocabularies.

Source: The Book of OHDSI. Observational Health Data Sciences and Informatics; 2021. ohdsi.github.io/TheBookOfOhdsi

Observational study. Epidemiological study in situations where nature is allowed to take its course. Changes or differences in one characteristic are studied in relation to changes or differences in others, without the intervention of the investigator.

Odds ratio. A measure of association which quantifies the relationship between an exposure and health outcome from a comparative study; also known as the cross-product ratio.

OMOP common data model. A common data model, developed and maintained by the OHDSI community, that standardizes the structure and content of observational health data, including a set of standardized vocabularies, so that analyses can be shared and run across different databases. OMOP stands for Observational Medical Outcomes Partnership.

Source: The Book of OHDSI. Observational Health Data Sciences and Informatics; 2021. ohdsi.github.io/TheBookOfOhdsi

One-sided alternative hypothesis. See "Hypothesis, alternative."

One-way analysis of variance. See "Analysis of variance (ANOVA), one-way."

Ontology. A formal, structured representation of the concepts within a domain and the relationships among them. Biomedical examples include SNOMED CT and disease ontologies such as Mondo.

Ordinal data. A variable with a limited number of possible values that can be placed in a meaningful order. Examples: patient satisfaction scores and stage of disease.

Outbreak. Synonymous with epidemic. Sometimes the preferred word, as it may escape the sensationalism associated with the word epidemic. Alternatively, a localized as opposed to generalized epidemic.

Outbreak investigation, steps.

1. Establish the existence of the epidemic.
2. Confirm the diagnosis.
3. Define a case.
4. Identify and count cases to determine who is at risk.
5. Describe the cases by time, place, and personal characteristics.
6. Develop one or more hypotheses about what caused the epidemic, and evaluate.
7. Conduct an analytic study; that is, a study with a comparison group.
8. Implement control and prevention efforts.
9. Prepare a written report and communicate findings.

Source: Centers for Disease Control and Prevention. Principles of Epidemiology in Public Health Practice. 3rd ed. Lesson 6: Investigating an Outbreak.

Pandemic. An epidemic occurring over a very wide area (several countries or continents) and usually affecting a large proportion of the population.

Parallel screening test. When more than one test is administered at the same time.

Pathogenicity. Ability of an agent to produce clinically apparent illness in a host.

Period prevalence. The amount of a particular disease present in a population over a period of time.

Person-time rate. A measure of the incidence rate of an event, for example a disease or death, in a population at risk over an observed period of time, that directly incorporates time into the denominator.

Phenopackets. An open standard, developed by the Global Alliance for Genomics and Health, for representing an individual's phenotypic and disease information in a structured, shareable form.

Source: Jacobsen JOB, Baudis M, Baynam GS, et al. The GA4GH Phenopacket schema defines a computable representation of clinical data. Nature Biotechnology. 2022;40(6):817–820.

Point prevalence. The amount of a particular disease present in a population at a single point in time.

Population. The total number of inhabitants of a given area or country. In sampling, the population may refer to the units from which the sample is drawn, not necessarily the total population of people.

Population attributable risk. The proportion of disease in a total population (of both exposed and non-exposed) that can be attributed to a particular exposure. Important from a public health policy standpoint; can help set prevention priorities.

Population attributable risk (PAR) in a cohort study:

where OR is the odds ratio.

Population attributable risk from a case-control study:

Positive predictive value (PPV). Percent of people who actually have the disease among those who test positive for the disease.

Predictive value positive. A measure of the predictive value of a reported case or epidemic; the proportion of cases reported by a surveillance system or classified by a case definition which are true cases.

Prevalence. The number or proportion of cases or events or conditions in a given population.

Duration can be shortened by (1) cure and (2) death.

Primary prevention. Actions taken to prevent the development of a disease among persons who are well and do not have the disease in question.

Probability. The number of times an event occurs compared to the number of observations, or to the number of times it could occur.

Probability difference and probability ratio. Two ways to compare two probabilities. The probability difference (the absolute difference between them) reflects the underlying frequency of the event. The probability ratio (the relative comparison) reflects how many times more or less likely the event is in one group than in another.

Propagated outbreak. An outbreak that does not have a common source, but instead spreads from person to person.

Proportion. Tells us the fraction of the population affected.

Proportionate mortality. The proportion of deaths in a specified population over a period of time attributable to different causes. Each cause is expressed as a percentage of all deaths, and the sum of the causes must add to 100%. These proportions are not mortality rates, since the denominator is all deaths, not the population in which the deaths occurred.

Protected health information (PHI). Individually identifiable health information that is protected under privacy regulations, such as HIPAA in the United States.

Public health surveillance. See "Surveillance, public health."

Purposive sample. A sample obtained by deliberately selecting values of the independent variable so that their distribution in the sample differs from their distribution in the population.

P-value. The probability of observing a result (for example, a relative risk, odds ratio, or sample estimate) as far from the null value as the one observed, or farther, if the null hypothesis were true. It is a conditional probability, written P(data or more extreme | H0 true). A p-value does not give the probability that the null hypothesis is true; it gives the probability of the observed data, assuming the null hypothesis is true. A value of 0.05 or lower is commonly treated as small enough to reject the null hypothesis.

Randomization. Randomization reduces selection bias:

• removes self-selection in treatment allocation;
• removes investigator bias in the allocation of subjects to treatment group;
• the goal is to produce study groups comparable with regard to both known and unknown risk factors.

Rate. The number of cases that occur, divided by the population at risk. A rate indicates the risk of an event occurring, or how fast events happen.

Rate ratio. A comparison of two groups in terms of incidence rates, person-time rates, or mortality rates.

Ratio. The value obtained by dividing one quantity by another.

Real reasons for differences in evidence of disease.

1. Increase or decrease in environmental or behavioral risk factors for disease.
2. Differences between populations in relevant demographic characteristics that are associated with environmental or behavioral risk factors.
3. Differences in genetic factors between populations.
4. Improvement in treatment or survivorship of a disease can affect mortality while incidence remains unchanged.

Real-world data (RWD). Data relating to health status or the delivery of care that are collected outside of conventional randomized trials, for example from electronic health records, insurance claims, registries, or patient-generated sources.

Source: U.S. Food and Drug Administration. Framework for FDA's Real-World Evidence Program. 2018.

Real-world evidence (RWE). Clinical evidence about the use, benefits, or risks of a medical product, derived from the analysis of real-world data.

Source: U.S. Food and Drug Administration. Framework for FDA's Real-World Evidence Program. 2018.

Record linkage. Methods for identifying and joining records that refer to the same individual across or within data sources. In a clinical setting, this is often called patient matching.

Regression analysis. A modeling method that relates a dependent variable to one or more independent variables. Confounders can be controlled by including them as additional independent variables. (See also "Logistic regression" and "Analysis of covariance (ANCOVA).")

Regression coefficient. A slope in a model with more than one independent variable.

Registry. An organized system that collects uniform data to evaluate specified outcomes for a population defined by a particular disease, condition, or exposure. Also called a patient registry.

Source: Gliklich RE, Leavy MB, Dreyer NA, eds. Registries for Evaluating Patient Outcomes: A User's Guide. 4th ed. Agency for Healthcare Research and Quality; 2020.

Relative risk (RR). The ratio of the risk of a disease in exposed individuals to the risk of disease in non-exposed individuals.

Relative risk of a disease, interpreting.

• If RR = 1: risk in exposed is equal to risk in non-exposed (no association).
• If RR > 1: risk in exposed is greater than risk in non-exposed (positive association, possibly causal).
• If RR < 1: risk in exposed is less than risk in non-exposed (negative association, possibly protective).

Reliability. The ability of a test to produce the same results if the test is repeated.

Representative sample. A sample whose characteristics correspond to those of the original population or reference population.

Reservoir. The habitat in which an infectious agent normally lives, grows, and multiplies; reservoirs include human reservoirs, animal reservoirs, and environmental reservoirs.

Risk. The probability that an event will occur, for example that an individual will become ill or die within a stated period of time or age.

Risk factor. An aspect of personal behavior or lifestyle, an environmental exposure, or an inborn or inherited characteristic that is associated with an increased occurrence of disease or other health-related event or condition.

RxNorm. A standardized naming system for clinical drugs, maintained by the US National Library of Medicine, that links the drug vocabularies used by different systems.

Sample. A selected subset of a population. A sample may be random or non-random, and it may be representative or non-representative.

Sampling distribution. The distribution of an estimate (such as the mean) across repeated samples, also called the distribution of estimates. It is distinguished from the distribution of the data: the spread of the data distribution is measured by the standard deviation, while the spread of the sampling distribution is measured by the standard error. The central limit theorem affects the sampling distribution, not the distribution of the data. The distribution of data concerns the selection of individuals; the sampling distribution concerns the selection of samples.

Secondary attack rate. A measure of the frequency of new cases of a disease among the contacts of known cases.

Secondary prevention. Actions taken among people who have already developed a disease to improve the prognosis of disease, through early detection (screening) and early intervention (at an early stage in the disease's natural history).

Selection bias. Error due to systematic differences in characteristics between those who are included in a study and those who are not. Can result from non-random sampling methods, differential participation, the healthy worker effect, or selecting prevalent cases rather than incident cases of disease.

Sensitivity. The ability of a system to detect epidemics and other changes in disease occurrence. The proportion of persons with disease who are correctly identified by a screening test or case definition as having disease.

Sentinel surveillance. A surveillance system in which a pre-arranged sample of reporting sources agrees to report all cases of one or more notifiable conditions.

Serial screening test. A test in which an initial screening test is conducted; then the positives from that test undergo additional screening tests.

Sex-specific mortality rate. A mortality rate among either males or females.

Significance, statistical. Does not indicate a causal relationship.

Slope. How much the dependent variable changes as the independent variable increases by one unit.

SNOMED CT. A comprehensive, multilingual clinical terminology used to record clinical findings, procedures, and other clinical concepts in a computable form.

Specificity. The proportion of persons without disease who are correctly identified by a screening test or case definition as not having disease.

Sporadic. A disease that occurs infrequently and irregularly.

Spot map. A map that indicates the location of each case of a rare disease or outbreak by a place that is potentially relevant to the health event being investigated, such as where each case lived or worked.

Standard deviation. A measure of the spread of a distribution of data, equal to the square root of the variance and expressed in the same units as the data.

Standard error. The standard deviation of a sampling distribution; a measure of the spread of an estimate across repeated samples.

Standardization. See the "Templates and Worksheets" chapter for direct and indirect standardization templates.

Statistical independence. Two events are statistically independent when the probability of one is not affected by whether the other occurs. It does not mean the events are separate or that one prevents the other; independent events must be able to occur at the same time. If exposure and disease are independent, the probability of disease is the same whether or not a person was exposed.

Stratified analysis. A method for controlling confounding by dividing the data into strata within which the confounder takes a single value (or a narrow range), then analyzing the association within each stratum.

Strength of association. The magnitude of the association between an exposure and an outcome, commonly measured by the relative risk or odds ratio. A stronger association provides more support for a causal relationship, although strength alone does not establish causation.

Structured data and unstructured data. Structured data are organized in predefined fields, such as coded diagnoses or laboratory values. Unstructured data, such as free-text notes, are not organized in a predefined model and often require natural language processing to analyze.

Sufficient cause. A causal factor or collection of factors whose presence is always followed by the occurrence of the effect (of disease).

Surveillance, public health. The systematic collection, analysis, interpretation, and dissemination of health data on an ongoing basis, to gain knowledge of the pattern of disease occurrence and potential in a community, in order to control and prevent disease in the community.

Survival curve. A curve that starts at 100% of the study population and shows the percentage of the population still surviving at successive times for as long as information is available. May be applied not only to survival as such, but also to the persistence of freedom from a disease, or a complication, or some other endpoint.

Table. A set of data arranged in rows and columns.

Table shell. A table that is complete except for the data.

Terminology. A curated set of terms with defined meanings, used to record information consistently. See also "Controlled vocabulary."

Transmission of infection. Any mode or mechanism by which an infectious agent is spread through the environment or to another person.

Two-sided alternative hypothesis. See "Hypothesis, alternative."

Type I error. Rejecting the null hypothesis when it is in fact true (a false positive). Its probability is denoted alpha.

Type II error. Failing to reject the null hypothesis when it is in fact false (a false negative). Its probability is denoted beta. See the table under "Type I error."

Unstructured data. See "Structured data and unstructured data."

Vaccine coverage. The proportion of a target population that has been vaccinated.

Vaccine efficacy. The proportional reduction in disease attack rate between unvaccinated and vaccinated groups.

Validation. The process of checking that data, or a definition such as a computable phenotype, perform as intended, often by comparison against a reference standard. Sensitivity and specificity are commonly used to report how well a phenotype performs.

Validity. The degree to which a measurement actually measures or detects what it is supposed to measure.

Variability. Four sources of variation contribute to screening test reliability:

• Biological variation. Many biological measurements are based on variable characteristics.
• Test measurement variability. The accuracy of the measuring device may vary from time to time.
• Intraobserver variability. A single observer might interpret the same test result differently from time to time.
• Interobserver variability. Differences in test result interpretation from observer to observer.

Variable. Any characteristic or attribute that can be measured.

Variance. A measure of dispersion equal to the average squared difference between the data values and the mean.

Vector. An animate intermediary in the indirect transmission of an agent that carries the agent from a reservoir to a susceptible host.

Vehicle. An inanimate intermediary in the indirect transmission of an agent that carries the agent from a reservoir to a susceptible host.

Virulence. Ability of an agent to produce severe disease in a host.

Vital statistics. Systematically tabulated information about births, marriages, divorces, and deaths, based on registration of these vital events.

Vocabulary mapping. The process of linking source codes from one vocabulary to concepts in a standard vocabulary, for example mapping local codes to SNOMED CT or to OMOP standard concepts. Also called a crosswalk.

Years of potential life lost. A measure of the impact of premature mortality on a population, calculated as the sum of the differences between some predetermined minimum or desired life span and the age of death for individuals who died earlier than that predetermined age.

Z-score (standard score). The number of standard deviations a value lies from the mean of its distribution, calculated as (value − mean) / standard deviation. Z-scores let values be compared against the standard normal distribution.

Z-test. A hypothesis test based on the standard normal (Z) distribution. It is appropriate when the sampling distribution is approximately normal and the standard error is known or well estimated, for example with large samples.