Kidney Disease: Improving Global Outcomes (KDIGO) acute kidney injury (AKI) definitions were evaluated for cases detected and their respective outcomes using expanded time windows to 168 h. AKI incidence and outcomes with expanded time intervals were identified in the electronic health records (EHRs) from 126,367 unique adult hospital admissions (2012-2014) and evaluated using multivariable logistic regression with bootstrap sampling. The incidence of AKI detected was 7.4% (n = 9357) using a 24-h time window for both serum creatinine (SCr) criterion 1a (≥0.30 mg/dL) and 1b (≥50%) increases from index SCr, with additional cases of AKI identified: 6963 from 24-48 h.; 2509 for criterion 1b from 48 h to 7 days; 3004 cases (expansion of criterion 1a and 1b from 48 to 168 h). Compared to patients without AKI, adjusted hospital days increased if AKI (criterion 1a and 1b) was observed using a 24-h observation window (5.5 days), 48-h expansion (3.4 days), 48-h to 7-day expansion (6.5 days), and 168-h expansion (3.9 days); all are p < 0.001. Similarly, the adjusted risk of in-hospital death increased if AKI was detected using a 24-h observation window (odds ratio (OR) = 16.9), 48-h expansion (OR = 5.5), 48-h to 7-day expansion (OR = 4.2), and 168-h expansion (OR = 1.6); all are p ≤ 0.01. Expanding the time windows for both AKI SCr criteria 1a and 1b standardizes and facilitates EHR AKI detection, while identifying additional clinically relevant cases of in-hospital AKI.

Original languageEnglish (US)
Article number3304
Journaljournal of clinical medicine
Issue number15
StatePublished - Jul 27 2021


  • Acute kidney injury
  • AKI timing
  • Computer-assisted
  • Detection limit
  • Diagnosis
  • Epidemiology

ASJC Scopus subject areas

  • Medicine(all)


Dive into the research topics of 'Optimization of Acute Kidney Injury (AKI) Time Definitions Using the Electronic Health Record: A First Step in Automating In-Hospital AKI Detection'. Together they form a unique fingerprint.

Cite this