Rotger Research Foundation

Diagnostic Equity Lens (DEL)

Measure access & outcome parity across groups — offer, completion, and positive rates — with transparent fairness indicators.

Client-side only CSV ready Bootstrap 5
1) Load data
Expected columns (any case/underscores ok):
race or race_ethnicity, gender, age, zip,
test_offered, test_completed, outcome_positive
2) Configure
Groups with very small N are marked with †.
3) Scenario simulator
0%0%
Holds completion/positive conditional rates constant.
Equity Score (0–100)
—
Size-weighted average of parity ratios (capped at 1).
Groups below DI 0.80
—
Screening threshold (four-fifths heuristic).
Worst parity gap (pp)
—
Largest negative difference vs reference rate.
Group rates vs reference
Detail table
Heat-tinted cells flag DI < 0.80 (warn) or < 0.60 (bad).
Group N (denom) Rate Ref rate Parity Δ (pp) DI (ratio)

Access & outcome parity. DEL quantifies how evenly diagnostic opportunities and results are distributed across groups.

  • Offer rate: how often people are offered a test.
  • Completion rate: how often offered tests get completed.
  • Positive rate: how often completed tests are positive (or your target outcome).

Fairness indicators. We report:

  • Parity difference (Δ): group rate minus reference rate (percentage points).
  • Disparate-impact ratio (DI): group rate ÷ reference rate; values < 0.80 are flagged for review (four-fifths heuristic).
  • Equity Score (0–100): size-weighted average of parity ratios, capped at 1.0.

  1. Load data. Upload CSV or load our sample.
  2. Normalization. We auto-recognize columns (e.g., race vs race_ethnicity) and accept 0/1 or true/false indicators.
  3. Grouping. Choose how to group people (Race, Gender, Age band, ZIP3).
  4. Rates. Offer = offered ÷ total; Completion = completed ÷ offered; Positive = positive ÷ completed.
  5. Reference. Pick any group or default to the largest. We compute Δ and DI vs that group.
  6. Simulator. “Boost” offers to a selected group; the tool projects downstream effects holding observed conditional rates steady.

All analysis runs in your browser. Data never leaves your device.

CSV format. Include a header row and any of the following columns:

  • race or race_ethnicity, gender, age, zip
  • test_offered, test_completed, outcome_positive as 0/1 or true/false

Click Upload CSV, choose Metric and Group by, optionally set a Reference group, and review the KPIs, chart, and table. Use the simulator to explore targeted improvements.

  • Patients & communities: Fairer access and earlier detection by fixing where drop-offs happen.
  • Clinics & systems: Fast, visual equity QA to prioritize outreach that moves the needle.
  • Boards & funders: Simple narratives: “2 groups below 0.80 DI; Equity Score improved +9 after intervention.”
  • Data teams: Private, in-browser checks before deeper modeling.

The 0.80 DI screen is a heuristic, not a legal determination. Pair with clinical governance and context.

  • Sample size: Aggregate tiny groups; small denominators are flagged.
  • Intersectionality: Test more than one grouping (or pre-compute combined fields) if data supports it.
  • Data hygiene: Confirm logical rules (e.g., completed ⇒ offered).
  • Interpretation: Disparities point to where, not always why. Combine with qualitative insight.

  • Empty results: Ensure a header row and at least one expected column.
  • Strange rates: Verify 0/1 coding and logical consistency among offer/complete/positive fields.
  • Reference line 0: Manually set a reference group with non-zero rate.

All processing is local; refresh the page to reset.

Privacy: This tool runs entirely in your browser. No data is uploaded or stored remotely.