Data Governance, Analytics, and Information Retrieval
Governance and data quality
Data governance defines who can make decisions about data: ownership, stewardship, definitions, access, quality, and lifecycle. It is a business and clinical accountability structure, not an IT function alone. Know data-quality dimensions: completeness, accuracy, timeliness, consistency, validity, and uniqueness. Value metadata, the master patient index, and standardized definitions. Analytics moves from source systems to an enterprise data warehouse or lakehouse, then to reports, dashboards, and models.
Analytics and retrieval
Candidates should understand cohort definition, denominators, confounding, and visualization choices, plus basic SQL logic and information retrieval metrics. Precision = retrieved relevant / all retrieved; recall = retrieved relevant / all relevant; F1 balances them. Natural language processing can extract concepts from notes but needs validation against a gold standard. Dashboards should show provenance, refresh date, and denominator.
Traps
Common errors include comparing rates with different denominators, treating a dashboard as truth without validation, and confusing precision with recall. Another trap is assuming governance means locking data down; appropriate access enables use while protecting privacy. A statistically significant difference may not be operationally meaningful.
Worked example / rule of thumb
A search for diabetic patients returns 100 charts; 80 are truly diabetic. There are 200 true diabetics in the population. Precision = 80/100 = 80%; recall = 80/200 = 40%; F1 is about 53%. Rule of thumb: define the numerator, denominator, population, and time window before building any metric; document the data source and owner.