Lab Result Interpretation Guide

This guide helps users understand how to use the lab module to interpret laboratory data.

Foundation: LOINC codes

Before interpreting anything, the most precise approach is to work from LOINC codes, because the same test (glucose, say) can have different specimen sources (whole blood, plasma, urine) or measurement methods. This system supports LOINC first and foremost.

Example: interpreting a health check-up report

Suppose you receive a check-up report with several flagged values: 1. Fasting glucose: 110 mg/dL 2. Triglyceride: 180 mg/dL 3. Total cholesterol: 240 mg/dL

The patient is a 55-year-old man.

Step 1: Determine the LOINC codes (if the report does not supply them)

Tool call:

search_loinc(mode="code", keyword="空腹血糖")
# Returns: 1558-6 (Glucose [Mass/volume] in Serum or Plasma --Fasting)

search_loinc(mode="code", keyword="三酸甘油酯")
# Returns: 2571-8 (Triglyceride [Mass/volume] in Serum or Plasma)

Step 2: Interpret a single value

Analyse each of these values.

Tool call:

interpret_lab_result(loinc_code="1558-6", value=110, age=55, gender="M")

Result: The tool picks the reference interval that applies to the age / sex (fasting glucose is 70–100 mg/dL in data/loinc/lab_reference_ranges.csv) and returns result.flag = "H", status = "偏高 (High)", the interval and its source, and a generic sentence ("above the normal reference value; further evaluation advised"). The tool does not diagnose: a clinical reading such as impaired fasting glucose (pre-diabetes) is for the caller (the LLM) or a clinician to make from the result.

Step 3: Batch interpretation

To analyse the whole report at once.

Tool call:

batch_interpret_lab_results(
    results_json='[
        {"loinc_code": "1558-6", "value": 110},
        {"loinc_code": "2571-8", "value": 180},
        {"loinc_code": "2093-3", "value": 240}
    ]',
    age=55,
    gender="M"
)

Result: The tool returns total_tests, abnormal_count, normal_count and the per-item interpretations (all three are H). Tests without a reference range are skipped and not counted. An overall risk assessment (for example a metabolic-syndrome concern) is likewise left to the caller or a clinician.