Data Interpretation and Reporting
Interpretation is a translation job
Reporting turns results into decisions. Start with the audience and the decision: executives need the headline and recommended action; clinicians need patient-level context and comparability; analysts need method and definitions. State the question, the population, the measure definition, the comparison, and the limitations before presenting the numbers.
Visualization and display
Choose chart type by question - trend over time (line or run chart), comparison across categories (bar), distribution (histogram or box plot), relationship (scatter). For process data, use run charts or statistical process control: a shift of six or more consecutive points on one side of the median, or a trend of five, signals special-cause variation. Do not label normal variation as improvement. Start bar axes at zero, because truncated axes exaggerate differences.
Precision in language
Distinguish percentage points from percent change, absolute from relative risk, and incidence from prevalence. "Readmissions fell 2%" and "fell 2 percentage points, from 10% to 8%" are different claims. Round appropriately for the audience, suppress small cells to protect privacy, and always disclose the measurement period and data lag.
Traps and worked example
Look for misleading axes, cherry-picked baselines, denominators that changed mid-series, and dashboards that show activity (how many reports were run) rather than outcome. A common trap is interpreting a 20% relative reduction as a 20-point absolute drop. A clinic's diabetes control measure improves from 70% to 77%: that is +7 percentage points, a 10% relative improvement. If the denominator fell from 500 to 320 because documentation changed, the gain may be artifactual - report the denominator alongside the rate and flag the definition change before celebrating.