ICD-Coded Versus Undiagnosed Acute Kidney Injury in the Atherosclerosis Risk in Communities (ARIC) Study.

Rationale & Objective: Acute kidney injury (AKI) is common among older adults and associated with adverse outcomes, yet it often goes undiagnosed. We aimed to compare the prevalence of undiagnosed versus Study Design: Prospective cohort study. Setting & Participants: A total of 5,692 adults with proteomic and metabolomic profiling from the Atherosclerosis Risk in Communities study, excluding individuals with end-stage kidney disease at baseline (2011-2013). Exposures: AKI, defined as either ICD-coded or undiagnosed based on 3 abstracted levels of serum creatinine measured during hospitalizations. Outcomes: Incident coronary heart disease, heart failure, stroke, and all-cause mortality. Analytical Approach: ICD-coded AKI was compared with undiagnosed AKI across all hospitalizations. Risk factors for ICD-coded and undiagnosed AKI were assessed separately using Cox proportional hazards models from study visit to the first AKI encounter. Time-varying Cox models were used to evaluate the associations of ICD-coded and undiagnosed AKI with adverse outcomes. Results: ICD codes identified 51.4% of AKI defined based on changes in serum creatinine, with higher sensitivity for more severe stages and a specificity of 87.0%. Undiagnosed AKI was more common in women; otherwise, clinical risk factors for undiagnosed AKI were similar to those for ICD-coded AKI, as were proteomic and metabolomic risk factors. Both AKI groups were strongly associated with coronary heart disease, heart failure, stroke, and mortality. Limitations: AKI was estimated using 3 values of serum creatinine only and not urinary output. Diagnosed AKI was identified using billing codes, which may underestimate clinical diagnosis of AKI. Conclusions: Undiagnosed AKI is prevalent, especially among women, and is biologically and prognostically similar to ICD-coded AKI, motivating efforts to improve AKI recognition and diagnosis.