CHDA cheat sheet

Analytics hierarchy: descriptive (what happened) → diagnostic (why it happened) → predictive (what will happen) → prescriptive (what should be done).

HITECH Act (2009) funded EHR adoption through incentive payments tied to meaningful use criteria and expanded HIPAA enforcement.

HIPAA de-identification: Safe Harbor removes 18 specified identifiers; expert determination is the statistical alternative.

Core code sets: ICD-10-CM (diagnoses), ICD-10-PCS (inpatient procedures), CPT/HCPCS (outpatient services and supplies), NDC/RxNorm (drugs), LOINC (labs), SNOMED CT (clinical findings).

Claims data reflect billing rather than clinical truth; diagnosis codes are subject to upcoding and to 'at least once per year' capture conventions.

HEDIS (NCQA) measures health plan performance on preventive care and chronic disease; CMS Star Ratings score Medicare Advantage plans on similar quality domains.

HIMSS Analytics Adoption Model (AMAM) stages 0–7 describe escalating EHR capability and data-use maturity.

PMPM = total cost ÷ member months; a normalized rate required to compare per-member cost across populations and contracts.

Value-based arrangements - shared savings, capitation, bundled payments - shift financial risk from payer to provider and drive demand for analytics.

SMART criteria: Specific, Measurable, Achievable, Relevant, Time-bound - the standard test for a usable business objective.

Gap analysis = future-state requirement minus current-state capability; the residual defines the project scope.

MoSCoW (Must / Should / Could / Won't have) forces explicit priority when requested scope exceeds delivery capacity.

ROI = (benefit − cost) ÷ cost × 100%; payback period = initial investment ÷ annual net benefit.

NPV discounts future cash flows at the cost of capital; a positive NPV indicates the initiative creates value.

Stakeholder power–interest grid: high power/high interest = manage closely; high power/low interest = keep satisfied; low power/high interest = keep informed; low/low = monitor.

A requirements traceability matrix links each requirement to design, build, and test evidence, preventing silent scope creep.

Root-cause tools: 5 Whys (iterative questioning), fishbone/Ishikawa (cause categories), Pareto analysis (80/20 ranking of causes).

A user story records role, need, and value ('As a ___, I need ___, so that ___') and is verified by written acceptance criteria.

ETL extracts, transforms, then loads into a target store; ELT loads raw data first and transforms it inside the target platform.

HL7 v2 is a pipe-delimited messaging standard for ADT, orders, and results; ADT^A01 = admit, ADT^A03 = discharge, ADT^A08 = update.

FHIR is a RESTful API standard exchanging modular resources (Patient, Encounter, Observation); it underpins current CMS interoperability requirements.

C-CDA/CCD documents are XML clinical summaries used for care transitions and exchanged as consolidated patient records.

Five core data-quality dimensions: completeness, accuracy/validity, consistency, timeliness, uniqueness.

MPI matching: deterministic requires exact agreement on defined identifiers (high precision); probabilistic scores field agreement against a threshold tuned between false links and missed links.

Terminology binding maps local codes to standard value sets; unmapped local codes surface as 'unknown' or 'other' in downstream reporting.

Batch loads run on a schedule (e.g., nightly); streaming or real-time feeds support alerting and near-real-time monitoring.

Sample size for a proportion: n = z²p(1−p) ÷ e²; margin of error shrinks with the square root of sample size.

Sensitivity = TP/(TP+FN); specificity = TN/(TN+FP); PPV = TP/(TP+FP); NPV = TN/(TN+FN).

Sensitivity and specificity are properties of the test; PPV and NPV change with prevalence.

95% CI ≈ estimate ± 1.96 × standard error; a two-sided CI excluding the null corresponds to p < 0.05.

Type I error = rejecting a true null (alpha); Type II error = failing to reject a false null (beta); power = 1 − beta.

Test selection: t-test for two means, chi-square for two categorical variables, ANOVA for three or more group means, logistic regression for binary outcomes.

Odds ratio is used in case-control and logistic models; relative risk is used in cohort designs.

Risk adjustment: Charlson and Elixhauser comorbidity indices for outcomes; CMS HCC models for payment and benchmark comparison.

Direct standardization applies the study population's age-sex-specific rates to a standard population; indirect standardization applies standard rates to the study population.

Control charts use ±3 sigma limits; one point beyond a limit, or a run of seven or more points on one side of the centerline, signals special-cause variation.

Chart selection: bar = category comparison, line = trend over time, scatter = correlation, histogram = distribution, box plot = spread and outliers.

Suppress cells with counts below a threshold (commonly <11 in CMS reporting) to limit re-identification risk in aggregate reports.

Counts describe volume; rates normalize by a denominator and are required to compare populations of different sizes.

Statistical significance is not clinical significance: report effect size and confidence intervals alongside p-values.

Benchmark types answer different questions: percentile rank (relative standing), external peer comparison (market context), internal historical baseline (own trend).

Operational dashboards track daily or real-time tactical metrics; strategic dashboards track periodic KPIs mapped to organizational goals.

Standard report structure: question → method → key finding → caveat → recommended action.

Every metric should be reported with its denominator, time window, population definition, and rounding rule.

Run charts (median line) and control charts (control limits) are the standard tools for displaying variation over time.

Governance defines decision rights and accountability for data; stewardship executes quality, definition, and appropriate-use duties day to day.

Metadata types: business (definitions and rules), technical (schemas, models), operational (lineage, job logs); a data dictionary documents business metadata.

HIPAA Security Rule safeguards are administrative, physical, and technical - e.g., access control, audit logging, encryption, contingency planning.

Minimum necessary standard limits PHI use and disclosure to what the purpose requires; role-based access control enforces it technically.

Breach Notification Rule: notify affected individuals without unreasonable delay and no later than 60 days after discovery; HHS notification thresholds depend on breach size.

Business associate agreements are required with vendors that create, receive, maintain, or transmit PHI on behalf of a covered entity.

Data lifecycle: create/acquire → validate → store → use/share → archive → destroy, with retention and destruction rules at each stage.

Data lineage documents source-to-report transformation paths and enables impact analysis when a source system or mapping changes.

Data-quality monitoring KPIs: completeness %, duplicate rate, validity % against value sets, and timeliness measured as source-to-warehouse lag.