Kdigo Aki Staging

Final AKI Stage = MAX(creatinine stage, urine output stage, CRRT stage)

Pre-computed Tables

KDIGO Stages (Combined)

SELECT
    subject_id,
    hadm_id,
    stay_id,
    charttime,
    -- Creatinine criteria
    creat_low_past_7day,
    creat_low_past_48hr,
    creat,
    aki_stage_creat,
    -- Urine output criteria
    uo_rt_6hr,
    uo_rt_12hr,
    uo_rt_24hr,
    aki_stage_uo,
    -- CRRT
    aki_stage_crrt,
    -- Final stage
    aki_stage,
    aki_stage_smoothed  -- Smoothed over 6-hour window
FROM mimiciv_derived.kdigo_stages;

Creatinine with Baseline

SELECT
    hadm_id,
    charttime,
    creat,
    creat_low_past_7day,
    creat_low_past_48hr
FROM mimiciv_derived.kdigo_creatinine;

Urine Output Rates

SELECT
    stay_id,
    charttime,
    weight,
    uo_rt_6hr,   -- mL/kg/h over 6 hours
    uo_rt_12hr,  -- mL/kg/h over 12 hours
    uo_rt_24hr   -- mL/kg/h over 24 hours
FROM mimiciv_derived.kdigo_uo;

Critical Implementation Notes

1. Baseline Creatinine: Uses the lowest creatinine in the past 7 days as the baseline. This may underestimate AKI if patient was already in AKI on admission.

2. 48-Hour Window: The >= 0.3 mg/dL acute increase criterion uses the lowest creatinine in the past 48 hours specifically.

3. Stage 3 with Cr >= 4.0: Requires EITHER:

   • An acute increase >= 0.3 mg/dL within 48h, OR
   • An increase >= 1.5x baseline

4. Urine Output Timing: UO criteria require the patient to be in ICU for at least 6 hours before staging (KDIGO definition). Earlier times get stage 0.

5. Weight for UO Calculation: Uses documented weight from mimiciv_derived.kdigo_uo. Weight estimation methods vary.

6. CRRT as Stage 3: Any patient on CRRT is automatically Stage 3 AKI.

7. Smoothed Stage: aki_stage_smoothed carries forward the maximum stage from the past 6 hours to reduce fluctuation between creatinine/UO measurements.

8. Time Series Data: AKI is calculated at every creatinine/UO measurement time, not just once per admission.

Example: AKI Incidence

WITH max_aki AS (
    SELECT
        stay_id,
        MAX(aki_stage) AS max_aki_stage
    FROM mimiciv_derived.kdigo_stages
    GROUP BY stay_id
)
SELECT
    max_aki_stage,
    COUNT(*) AS n_stays,
    ROUND(100.0 * COUNT(*) / SUM(COUNT(*)) OVER(), 1) AS pct
FROM max_aki
GROUP BY max_aki_stage
ORDER BY max_aki_stage;

Example: AKI by Criteria Type

WITH aki_type AS (
    SELECT
        stay_id,
        MAX(aki_stage_creat) AS max_cr_stage,
        MAX(aki_stage_uo) AS max_uo_stage,
        MAX(COALESCE(aki_stage_crrt, 0)) AS max_crrt_stage
    FROM mimiciv_derived.kdigo_stages
    GROUP BY stay_id
)
SELECT
    CASE
        WHEN max_cr_stage > 0 AND max_uo_stage > 0 THEN 'Both Cr and UO'
        WHEN max_cr_stage > 0 THEN 'Cr only'
        WHEN max_uo_stage > 0 THEN 'UO only'
        WHEN max_crrt_stage > 0 THEN 'CRRT only'
        ELSE 'No AKI'
    END AS aki_type,
    COUNT(*) AS n_stays
FROM aki_type
GROUP BY 1;

Example: Time to AKI Development

WITH first_aki AS (
    SELECT
        k.stay_id,
        MIN(k.charttime) AS first_aki_time
    FROM mimiciv_derived.kdigo_stages k
    WHERE k.aki_stage >= 1
    GROUP BY k.stay_id
)
SELECT
    ROUND(
        TIMESTAMP_DIFF(f.first_aki_time, ie.intime, HOUR), 0
    ) AS hours_to_aki,
    COUNT(*) AS n_stays
FROM first_aki f
INNER JOIN mimiciv_icu.icustays ie ON f.stay_id = ie.stay_id
GROUP BY 1
HAVING hours_to_aki BETWEEN 0 AND 168  -- First week
ORDER BY 1;

Example: AKI Stage Transitions

WITH aki_trajectory AS (
    SELECT
        stay_id,
        charttime,
        aki_stage,
        LAG(aki_stage) OVER (PARTITION BY stay_id ORDER BY charttime) AS prev_stage
    FROM mimiciv_derived.kdigo_stages
    WHERE aki_stage IS NOT NULL
)
SELECT
    prev_stage AS from_stage,
    aki_stage AS to_stage,
    COUNT(*) AS n_transitions
FROM aki_trajectory
WHERE prev_stage IS NOT NULL
    AND prev_stage != aki_stage
GROUP BY 1, 2
ORDER BY 1, 2;

References

• KDIGO Clinical Practice Guideline for Acute Kidney Injury. Kidney International Supplements. 2012;2(1):1-138.
• Kellum JA, Lameire N. "Diagnosis, evaluation, and management of acute kidney injury: a KDIGO summary." Critical Care. 2013;17(1):204.