Data Management, Analytics, and Reporting
This domain tests whether you can trust the numbers before you act on them: governance, data quality dimensions, validation, analysis, and audience-appropriate reporting. Expect questions on data quality, dashboards, statistical process control, privacy, and analytics maturity.
Core rules and concepts
- Data governance: ownership, stewardship, data dictionary, access controls, retention, and defined change control. If definitions are not governed, dashboards drift.
- Data quality dimensions: accuracy, completeness, consistency, timeliness, validity, and uniqueness. Each should be assessed and documented.
- Sources and tradeoffs: EHR (clinical detail, workflow-dependent capture), claims (complete encounters, coding lag), registries (standardized definitions, abstraction burden), surveys (patient experience, response bias).
- Validation: logic checks, chart review against the operational definition, inter-rater reliability for abstraction, and reconciliation across sources.
- Analysis: choose run charts or control charts for time series; stratify to reveal disparities; apply risk adjustment before comparison; use confidence intervals for small denominators.
- Analytics maturity: descriptive, diagnostic, predictive, prescriptive. Predictive models require validation and monitoring for bias.
- Privacy: apply minimum necessary, de-identify using recognized methods, and use data use agreements for external sharing. Interoperability standards such as HL7 and FHIR matter for exchange, not for analytic validity.
Worked example
Two sepsis dashboards disagree: the EHR reports 78% compliance with the three-hour bundle and the abstracted registry reports 64%. Reasoning: compare the operational definitions first. The EHR counts the earliest antibiotic order time and includes patients later excluded by the registry; the registry uses administration time and excludes transfers. Neither is wrong, but they answer different questions. Resolution: adopt the registry definition for external reporting, document the EHR definition for internal operations, automate the exclusion logic, and validate quarterly with a 30-chart review.
Traps that cost marks
- Comparing two rates built from different definitions, time windows, or denominators.
- Treating missing data as zero rather than as missing.
- Presenting a single-month bar chart as evidence of improvement.
- Overloading dashboards so the signal is buried; few measures, clear targets, annotation, and action prompts work better.
- Inferring causation from correlation in observational data.
- Ignoring data latency when reporting near-real-time operational measures.