Reference

Glossary

Terms are defined where they first appear in the chapters; this master glossary gathers them in one place.

TermDefinition
Absolute referenceA cell reference locked with dollar signs so it does not move when copied, such as $B$2.
ActionA decision or change made in response to insight, such as adjusting reminder timing.
Adjusted R-squaredA version of R-squared that penalizes added predictors, used to compare models of different sizes.
Administrative code setA code system that identifies services, products, or billing items, such as CPT or HCPCS Level II.
Analysis ToolPakAn add-in included with Excel that provides statistical procedures, including linear regression.
Analytic sampleThe records finally used in an analysis, which may be smaller than the respondent sample after missing-data, eligibility, or analysis-specific exclusions.
Anecdotal evidenceA single vivid case offered in place of a sample; it cannot support a general claim.
AppendStacking rows from files that share the same structure.
AssumptionsThe conditions a recommendation depends on, disclosed so the decision-maker can weigh the risk.
AUROCThe area under the ROC curve; the probability the model ranks a random event case above a random non-case.
BaselineA metric’s typical level over a longer period, used as the reference for judging change.
BenchmarkA reference point, such as a past baseline, a target, or a matched peer, used to judge whether a number is good.
Binary outcomeAn outcome with two possible values, such as showed up or did not.
BlindingKeeping participants or assessors unaware of group assignment so expectations do not color the result.
BlockingGrouping similar units before randomizing within each block, to reduce variation attributable to the blocking factor.
Bottom line up frontThe practice of stating the conclusion first and the supporting detail afterward.
Business associateA person or organization that handles PHI on behalf of a covered entity, and is bound by HIPAA.
Calculation traceabilityThe ability to follow any result back to the rows and formulas that produced it; note this is not a formal change-history audit trail, which records who changed what and when.
Care settingThe place and mode in which care happens, such as an ambulatory clinic, an emergency department, a hospital, or a care-management program.
ClassificationA code system that sorts clinical detail into a fixed set of reporting categories, such as ICD-10-CM.
Clinical analyticsAnalysis focused on care processes and outcomes.
Clinical terminologyA code system that captures fine-grained clinical meaning and relationships, such as SNOMED CT.
Cluster sampleA sample of whole natural groups, such as clinics, with everyone in the chosen groups included.
Code setA standardized list of categories used to record one part of care, such as ICD-10-CM for diagnoses.
CodebookA record of each column’s name, meaning, type, and allowed values; also called a data dictionary.
CoefficientThe change in the outcome for a one-unit increase in a predictor, holding the other predictors constant.
Cohort studyAn observational study that follows a defined group over time, prospectively or retrospectively, to see who develops an outcome.
Common-cause variationThe natural, expected wobble of a stable process from one period to the next.
Confidence intervalAn estimate reported as a range, the point estimate plus and minus a margin of error, with a stated level of confidence.
ConfounderA variable related to both the exposure and the outcome and not on the causal pathway between them, which can explain an apparent link; it need not be hidden.
ConfoundingWhen a predictor’s apparent effect is partly explained by another variable related to both it and the outcome.
Construct validityThe degree to which what a measure counts actually stands in for the thing the question is about.
Continuous variableA quantitative variable measured on a scale that can take any value in a range, such as cost or minutes.
Continuum of careThe sequence of settings a patient moves through over time, from prevention and primary care through post-acute care; most continuity risk sits at the handoffs between them.
Control chartA run chart with calculated limits that formally distinguish ordinary variation from signals.
Control groupThe units in an experiment that do not receive the treatment under study, instead getting usual care, a placebo, or an alternative, used as a comparison.
Convenience sampleA sample of whoever is easiest to reach; cheap but prone to bias.
Correlation coefficientA number from -1 to 1 measuring how tightly two numerical variables move together; written r.
Coverage errorThe mismatch between the target population and the sampling frame: eligible people the frame omits, and ineligible people it wrongly includes.
Covered entityA health plan, healthcare clearinghouse, or healthcare provider that transmits health information electronically in connection with HIPAA-standard transactions.
CPTThe American Medical Association’s licensed code set for outpatient and professional services.
Cross-sectional studyAn observational study that measures a population at a single point in time.
Cumulative gainsA view of how much of an outcome is captured by acting on the top-ranked share of a population.
DashboardA small, curated set of views that supports recurring decisions.
DataRaw observations, such as a single appointment date, diagnosis code, or survey response.
Data cleaningThe work of making a dataset trustworthy and ready for analysis.
Data dictionaryA document recording what each field means, its allowed values, and how it is calculated.
Data governanceThe rules and roles that decide how data is defined, produced, and used.
Data lineageThe record of where a field came from and what was done to it along the way.
Data stewardA person accountable for a dataset who can say authoritatively what its fields mean.
Data use agreementA contract under which a covered entity may share a limited data set for research, public health, or operations.
Data validationA rule that restricts what a cell will accept when a value is typed in.
De-identificationThe process of reducing identifiers until data no longer reasonably identifies anyone.
Decision briefA short document that leads with a recommendation, then gives impact, a visual, assumptions, and limits.
Decision supportAnalysis aimed at a specific choice someone must make, judged by whether it changes the action.
DenominatorThe population the numerator events are drawn from.
Descriptive analyticsAnalysis that describes what happened.
Descriptive statisticsNumbers that condense a column into a short summary of its center, spread, and shape.
Diagnostic analyticsAnalysis that explores why something happened.
Dimension tableA table that describes the stable actors in the system, with one row per actor and no events.
Discrete variableA quantitative variable that counts and takes separate, whole values, such as a number of visits.
DiscriminationA model’s ability to separate cases from non-cases, summarized by AUROC.
DriftThe decay of a model’s accuracy as the world it learned from changes over time.
Dummy variableA variable that codes a category as 1 or 0 so a regression can use it; also called an indicator variable.
Dynamic arrayA formula result that automatically spills across neighboring cells, available in Microsoft 365 and Excel 2021 or later.
ETLExtract, transform, load; the general pattern of moving data from sources into an analysis-ready table.
Excel TableA structured, named object Excel manages as a unit, expanding automatically when rows are added and supporting references by column name.
ExperimentA study that deliberately assigns a treatment to measure its effect, which can support a causal claim.
Expert DeterminationThe HIPAA de-identification method in which a qualified expert determines and documents that re-identification risk is very small.
Fact tableA table that records events, with one row per event and a time attached.
Financial analyticsAnalysis focused on reimbursement, denials, payer mix, and margin.
Foreign keyA column in one table that refers to the primary key of another.
FormulaA cell entry beginning with an equals sign that computes a result from other cells.
FunctionA named calculation that takes inputs in parentheses, such as AVERAGE(range).
HCPCS Level IIA code set for supplies, drugs, and non-physician services, maintained by CMS.
Health information exchangeThe electronic infrastructure that lets separate organizations share a patient’s records when care crosses organizational lines.
Healthcare systemThe loose network of settings, staff, and software through which care is delivered.
HIPAAThe Health Insurance Portability and Accountability Act, which sets US rules for using and disclosing PHI.
ICD-10-CMThe system of alphanumeric codes used in the United States to record diagnoses.
ICD-10-PCSThe code system used to record inpatient procedures.
ImputationFilling a true missing value with an estimated one, as a documented and deliberate choice.
Inclusion and exclusion criteriaThe rules that decide which records belong in the denominator.
InformationData organized so that a pattern can be seen, such as a monthly no-show rate.
InsightAn answer to a question that matters, such as learning that no-shows are more common at some clinics than others.
InterceptThe predicted outcome when all predictors are zero; where the line crosses the axis.
Interquartile range (IQR)The distance between the 25th and 75th percentiles, covering the middle half of the data.
JoinCombining two tables by matching a foreign key to a primary key, producing a wider table whose rows carry columns from both.
k-anonymityA disclosure-risk property in which every record shares its quasi-identifier values with at least k minus 1 other records.
Least squaresThe method that fits the line by minimizing the sum of squared residuals.
LevelsThe distinct categories a qualitative variable can take.
LimitationsWhat an analysis cannot show or does not know, stated plainly.
Limited data setPHI with direct identifiers removed but some dates and partial geography retained, shareable only under a data use agreement.
Linear regressionA model of a numerical outcome as a straight-line function of one or more predictors.
Log-oddsThe logarithm of the odds of an event; logistic regression is linear in this quantity.
Logistic regressionA model that predicts the probability of a binary outcome, bounded between 0 and 1.
LOINCThe standard for identifying laboratory tests and clinical observations across systems; it does not encode the measured result value.
Master dataThe authoritative record of core business entities, such as the patient, provider, and facility lists.
MeanThe arithmetic average of a set of values.
MeasureA precise, repeatable definition of something you can count or calculate.
MedianThe middle value when the data is sorted; half of records fall below it.
MergeAdding columns from a related table by matching on a shared key.
MetricThe number a measure produces when it is run on data.
Minimum necessaryThe HIPAA standard of making a reasonable effort to limit PHI to what a task requires.
MS-DRG (Medicare Severity Diagnosis-Related Group)A three-digit payment group that represents an inpatient stay for Medicare payment.
Multiple linear regressionA regression with more than one predictor, where each coefficient holds the others constant.
Multistage sampleA sample drawn in stages, such as selecting whole clusters first and then sampling within each.
Named rangeA cell or range given a readable name so a constant or threshold can be defined once and reused.
NDC (National Drug Code)A drug-product identifier assigned by the FDA as 10 digits in three segments, commonly normalized to 11 digits for billing; the FDA moves to a 12-digit format in 2033.
Neighborhood deprivation quintileOne of five equal-sized groups of neighborhoods ranked by a combined measure of socioeconomic disadvantage, from least to most deprived.
Nominal variableA qualitative variable whose categories have no inherent order, such as visit type or payer.
Nonresponse biasDistortion that occurs when those who do not respond differ from those who do.
Normal distributionThe idealized symmetric, bell-shaped distribution in which about 68 percent of values fall within one standard deviation of the mean and 95 percent within two.
NumeratorThe events of interest being counted in a rate.
ObservationOne row of a table; a single unit the data describes.
Observational studyA study that records what happens without intervening; it can show associations but cannot by itself establish causation.
Odds ratioThe exponentiated logistic coefficient; the factor by which a predictor multiplies the odds of the event.
Operational analyticsAnalysis focused on the flow of work, such as access, staffing, and throughput.
Operational definitionA measure written in terms of the actual columns and values in the data, so that anyone can reproduce it.
Ordinal variableA qualitative variable whose categories have a meaningful order but not necessarily equal gaps, such as a 1-to-5 satisfaction rating.
Outcome variableThe quantity a regression tries to explain; also called the dependent variable.
OutlierA value far from the rest of the data.
p-valueThe probability that chance alone would produce a difference at least as large as the one observed, if there were truly no difference.
PatientThe person who receives care; the party whose health the whole system exists to serve.
PayerA party that finances care and administers coverage and payment, such as a private insurer, a self-funded employer, Medicare, or Medicaid.
PivotTableA tool for rapidly summarizing tabular data by dragging fields into rows, columns, filters, and values.
Population health analyticsAnalysis that looks across groups rather than single visits.
Power QueryExcel’s data preparation layer, which records refreshable import and cleaning steps.
Predictive analyticsAnalysis that estimates what is likely to happen next.
PredictorA variable used to explain the outcome; also called an independent variable.
Prescriptive analyticsAnalysis that recommends what action to take.
Primary keyA column, or a combination of columns, whose value uniquely identifies each row of a table and is never missing.
Primary useThe operational job a piece of data was built to do in its source system.
Principal diagnosisThe condition found after study to be chiefly responsible for a hospital admission.
Projected impactAn honest estimate of what a recommended action will produce.
Protected health information (PHI)Health data that can be tied to an identifiable individual, held by a covered entity or business associate.
ProviderA party that delivers and documents care, such as a hospital, clinician, or pharmacy.
Qualitative variableA variable that holds a label placing a record in a group; also called categorical. It is either nominal or ordinal.
Quantitative variableA variable that holds a numeric quantity you can do arithmetic on; also called numeric. It is either discrete or continuous.
Quasi-experimental studyA study that compares a treatment and a comparison group without full randomization; it can support causal claims under stated, defensible assumptions.
Quasi-identifierA field that is not a direct identifier but can single out a person in combination, such as birth date, sex, and ZIP code.
R-squaredThe share of the outcome’s variation that the model explains, from 0 to 1.
RandomizationAssigning units to treatment or control by chance so the groups are alike on average beforehand.
Range checkA validation that confirms a field’s values fall within an allowed set or interval.
RateA measure formed by dividing a numerator by a denominator over a fixed time window.
Raw dataSource data kept exactly as received, so it stays recoverable and auditable.
Re-identificationLinking apparently anonymous data back to a person, often by combining it with an outside source.
Reasonableness checkA quick, independent test that a result makes sense, such as confirming a filtered row count matches a formula’s denominator.
Reference dataA small, shared lookup list of valid values, such as facility types or payer categories.
Referential integrityThe property that every foreign key value exists in the table it points to.
RefreshRe-running a query’s steps so new or updated source data flows through automatically.
RegulatorA stakeholder, such as CMS or NCQA, that defines how a measure must be calculated and can tie payment or a public rating to the result.
Relative referenceA cell reference that shifts when the formula is copied, such as B2.
ReplicationIncluding enough units in each group that a chance difference is unlikely to be mistaken for a real effect.
ReproducibilityThe property that a cleaning process can be rerun to produce the same result.
ResidualThe difference between an actual value and the model’s prediction for it.
Respondent sampleThe subset of the selected sample that actually responds or yields data.
Revenue codeA four-digit code on an institutional bill that marks where and what type of service was provided.
Risk adjustmentThe use of diagnosis codes to account for how sick a population is, which affects payment.
Risk stratificationSorting a population into tiers by predicted risk so resources can be targeted.
Run chartA plot of a metric over time, read against its own baseline to separate signal from noise.
RxNormA terminology of normalized names for clinical drugs, maintained by the National Library of Medicine.
Safe HarborThe HIPAA de-identification method that removes or reduces eighteen specified categories of identifier.
SampleThe records you observe. It narrows in two steps: the selected sample is what you draw from the frame, and the respondent sample is the subset that actually provides data.
Sampling frameThe list of records you can actually reach and sample from.
Secondary useAny later use of the data to answer a question it was not originally designed for, which includes most analytics.
Selected sampleThe records drawn from the sampling frame before any nonresponse.
Selection biasDistortion that occurs when the way units enter the sample is related to what is being measured and is not corrected for, so the sample does not represent the population.
SignalA change in a metric that reflects a real shift in the process, not ordinary variation.
Simple random sampleA sample in which every unit on the frame has an equal chance of selection.
SkewAsymmetry in a distribution, where one tail is longer than the other.
SlopeThe change in the outcome for a one-unit increase in a predictor.
SNOMED CTA detailed clinical terminology used in electronic health records to capture findings and problems.
SolverAn Excel add-in for optimization, used later to fit a logistic regression by maximizing likelihood.
Source systemThe software where a piece of data is first created, such as a scheduling system, electronic health record, or billing platform.
Spreadsheet-ready dataData laid out tidily, with one row per unit, one variable per column, one value per cell, and consistent codes.
StakeholderAnyone with a stake in what the data says, such as a clinician, operations manager, finance leader, public health team, patient, or regulator.
Standard deviationA measure of how far values typically fall from the mean, calculated in Excel with STDEV.
Standard errorA measure of how much a sample estimate would vary from sample to sample; it shrinks as the sample grows.
Stratified sampleA sample drawn separately within defined subgroups so each subgroup can be represented according to the allocation plan.
Structural missingA blank that is correct because the value does not apply to that record.
Structured referenceA formula reference that uses a Table name and column name, such as Appointments[attendance_outcome], instead of cell coordinates.
Study designThe overall plan for how evidence is collected, such as a trial, cohort, or cross-sectional study.
Systematic reviewA study that does not enroll new participants but systematically synthesizes evidence from existing studies; a meta-analysis pools their results statistically.
Systematic sampleA sample that takes every kth record from a random start in an ordered frame.
TableA rectangle of data in which rows are records and columns are variables; also called a data matrix.
Target populationEveryone the question is about.
ThresholdThe cutoff probability above which a predicted case is acted upon.
Tidy dataData in which each row is one observation at a single grain and each column is one variable.
TransformationReshaping data into the form analysis needs without changing its meaning.
True missingA blank where a value should exist but was not recorded.
Unit of analysisWhat one row of a table represents; also called the grain. It decides what the table can count.
ValidationThe set of checks that confirm the data means what the analyst thinks it means.
VariableA column of a table, one attribute measured for every record.
Version controlManaging successive versions of files through clear naming and separation, so milestones are not lost.
Working dataThe transformed copy of the data that analysis is actually built on.
XLOOKUPA lookup function in Microsoft 365 and Excel 2021 or later that returns a value from one column matched on another; INDEX with MATCH is the older equivalent.