Home / Patterns / Lab-Value Reference-Range Misapplication

Lab-Value Reference-Range Misapplication

Goal Diagnosis Safety Frequency Common Category Healthcare Published View source on GitHub ↗

Issue: Agent Applies Generic Adult Reference Ranges to Lab Values Without Adjusting for Age, Sex, Pregnancy, or Assay-Specific Ranges

Frequency: Common

Symptoms

  • Pediatric or geriatric lab values flagged as “normal” or “abnormal” using standard adult reference ranges
  • Pregnancy-altered lab values (e.g., physiologic anemia, altered thyroid panels) misinterpreted as pathological
  • Sex-specific reference ranges (e.g., creatinine, hemoglobin) not applied, causing systematic over- or under-flagging in one sex
  • Cross-lab assay differences (different units or methods) ignored, causing values to be compared against the wrong range entirely

Root Cause Reference ranges are not universal constants; they vary by age, sex, pregnancy status, and even by the specific assay/instrument used by the lab. Models trained on generic adult reference tables, or that fail to parse the specific reference range printed on the lab report itself, apply a one-size-fits-all threshold that produces both false positives and false negatives across populations whose physiology deviates from the “standard” adult range.

Example

Scenario: Pregnant patient's lab report shows hemoglobin 10.8 g/dL
Standard adult reference range: 12.0-15.5 g/dL (flagged as anemic)
Pregnancy-adjusted range (2nd trimester): 9.5-15.0 g/dL (within normal physiologic range)
Model action: Flags "anemia," recommends iron supplementation work-up
Reality: Value reflects normal pregnancy-related hemodilution
Impact: Unnecessary work-up, patient anxiety, potential masking of an actually abnormal trend if baseline is reset incorrectly

Key Statistics

  • Sex-specific and age-specific reference range misapplication is a recurring source of false-positive lab flagging in automated clinical decision support audits
  • Pregnancy-related physiologic lab shifts are frequently missed by generic reference-range logic, leading to unnecessary work-ups in a non-trivial share of obstetric chart reviews
  • Cross-assay unit/method mismatches (e.g., different creatinine assay methods) account for a measurable share of lab-interpretation errors in multi-lab health systems

Contributing Factors

  • Reference ranges hardcoded rather than parsed from the lab report’s own stated range
  • No structured capture of pregnancy status, age, or sex feeding into the interpretation step
  • Multiple lab vendors with different assay methods feeding a single patient record

Mitigation Strategies

  1. Report-Sourced Reference Ranges: Always parse and use the reference range printed on the specific lab report rather than a hardcoded generic table
  2. Demographic-Conditional Range Selection: Explicitly select age/sex/pregnancy-adjusted reference ranges based on structured patient data before flagging abnormal values
  3. Assay-Method Awareness: Track which lab/assay method produced each value and avoid cross-comparing values from incompatible methods
  4. Trend-Over-Threshold: Weight within-patient trend changes alongside absolute reference-range flagging, since a “normal” value that is a sharp change from baseline can still be clinically significant

Metrics

  • False-positive abnormal-flag rate stratified by demographic adjustment applied vs. not applied
  • % of lab interpretations using report-sourced vs. hardcoded reference ranges
  • Cross-assay mismatch detection rate

Alerts

  • Pregnancy status on file but pregnancy-adjusted range not applied → P2
  • Lab value flagged abnormal using a reference range that doesn’t match the report’s stated range → P1

References

Related Patterns