Utilization of Deep Learning for Subphenotype Identification in Sepsis-Associated Acute Kidney Injury

Abstract

Background and objectives: Sepsis-associated AKI is a heterogeneous clinical entity. We aimed to agnostically identify sepsis-associated AKI subphenotypes using deep learning on routinely collected data in electronic health records.

Design, setting, participants, & measurements: We used the Medical Information Mart for Intensive Care III database, which consists of electronic health record data from intensive care units in a tertiary care hospital in the United States. We included patients ≥18 years with sepsis who developed AKI within 48 hours of intensive care unit admission. We then used deep learning to utilize all available vital signs, laboratory measurements, and comorbidities to identify subphenotypes. Outcomes were mortality 28 days after AKI and dialysis requirement.

Results: We identified 4001 patients with sepsis-associated AKI. We utilized 2546 combined features for K-means clustering, identifying three subphenotypes. Subphenotype 1 had 1443 patients, and subphenotype 2 had 1898 patients, whereas subphenotype 3 had 660 patients. Subphenotype 1 had the lowest proportion of liver disease and lowest Simplified Acute Physiology Score II scores compared with subphenotypes 2 and 3. The proportions of patients with CKD were similar between subphenotypes 1 and 3 (15%) but highest in subphenotype 2 (21%). Subphenotype 1 had lower median bilirubin levels, aspartate aminotransferase, and alanine aminotransferase compared with subphenotypes 2 and 3. Patients in subphenotype 1 also had lower median lactate, lactate dehydrogenase, and white blood cell count than patients in subphenotypes 2 and 3. Subphenotype 1 also had lower creatinine and BUN than subphenotypes 2 and 3. Dialysis requirement was lowest in subphenotype 1 (4% versus 7% [subphenotype 2] versus 26% [subphenotype 3]). The mortality 28 days after AKI was lowest in subphenotype 1 (23% versus 35% [subphenotype 2] versus 49% [subphenotype 3]). After adjustment, the adjusted odds ratio for mortality for subphenotype 3, with subphenotype 1 as a reference, was 1.9 (95% confidence interval, 1.5 to 2.4).

Conclusions: Utilizing routinely collected laboratory variables, vital signs, and comorbidities, we were able to identify three distinct subphenotypes of sepsis-associated AKI with differing outcomes.

Keywords: AKI; acute kidney injury; acute renal failure; deep learning; dialysis; mortality; subtypes.

Graphical abstract

Figure 1.

Flow sheet for data processing and subphenotype identification. ICD9, International Classification of Diseases Ninth Revision; KDIGO, Kidney Disease Improving Global Outcomes; MIMIC III, Medical Information Mart for Intensive Care III.

Figure 2.

t-Distributed stochastic neighbor embedding (t-SNE) visual representation of subphenotype 1 in blue, subphenotype 2 in orange, and subphenotype 3 in green demonstrates separation among subphenotypes. The t-SNE plot is a method of visualizing high-dimensional data in a low-dimensional space (in this case, three dimensions). Each dot represents a patient and displays clusters within the scaled down and dimensionally reduced space of the autoencoder embeddings. As such, this space does not have any real-world units.

Figure 3.

Top 18 features found to have the largest differences between subphenotype 1 and subphenotype 3 include labs representative of kidney and liver function. Violin plots for the median value of 18 laboratory features. The 18 laboratory features with the largest log-transformed differences between subphenotypes 1 and 3 are displayed. Subphenotype 1 is represented in blue, subphenotype 2 is represented in orange, and subphenotype 3 is represented in red. pCO₂, partial pressure of carbone dioxide; PT, prothrombin time; PTT, partial thromboplastin time; RDW, red cell distribution width.

Figure 4.

Large differences on dialysis and mortality across AKI subphenotypes. Gray bars represent the proportion of patients who required dialysis, and black bars represent the proportion of patients who died within each subphenotype.