Data Analytics, Reporting, and Visualization
From question to decision
Analytics in healthcare spans descriptive (what happened), diagnostic (why), predictive (what will happen), and prescriptive (what to do). Start with the decision the analysis supports, then choose measures. Know the difference between a measure's numerator, denominator, exclusions, and the specification that makes it reproducible.
Concepts to master
- Measure specification: numerator, denominator, exclusions, and the data elements required.
- Rates vs. counts; crude vs. risk-adjusted or stratified comparisons.
- Confounding, selection bias, and why a raw comparison between hospitals can mislead.
- Visualization choices: bar for comparison, line for trend over time, scatter for relationship, and the danger of dual axes and truncated y-axes.
- Small numbers and suppression rules to protect privacy and avoid unstable rates.
- Dashboards: audience, refresh cadence, definitions on the page, and drill-down.
- Benchmarking and attribution; the difference between process and outcome measures.
Traps
Questions often present a rate comparison without adjustment and expect you to spot confounding. Another trap: choosing a pie chart for many categories or a truncated axis that exaggerates differences. Also watch for denominator errors-excluding patients who should be in the population, or including those who should be excluded-which silently change the measure.
Rule of thumb
State the decision, then the measure, then the denominator, then the visualization. If the audience cannot tell what the denominator is, the chart is not finished.
Worked example
Hospital A has a higher mortality rate than B, but A treats older, sicker patients. Without risk adjustment, the comparison is confounded; the defensible report stratifies or adjusts and discloses the method.