NPCNet: Navigator-Driven Pseudo Text for Deep Clustering of Early Sepsis Phenotyping

TL;DR

Sepsis presents a heterogeneous clinical syndrome requiring phenotype-informed therapy. The authors introduce NPCNet, a deep clustering framework that converts temporal EHRs into pseudo-text embeddings and uses a target navigator to encode clinical relevance, identifying four phenotypes $\alpha$, $\beta$, $\gamma$, $\delta$. These phenotypes exhibit divergent SOFA trajectories and differential responses to early vasopressor administration, with $\alpha$, $\beta$, and $\delta$ showing potential treatment benefits. NPCNet outperforms baselines on internal clustering metrics, reveals clinically meaningful trajectories, and generalizes to external datasets (eICU) with some treatment-effect signals affected by data limitations. The work advances precision medicine in sepsis by linking computable phenotypes to actionable clinical outcomes and treatment strategies.

Abstract

Sepsis is a heterogeneous syndrome. Identifying clinically distinct phenotypes may enable more precise treatment strategies. In recent years, many researchers have applied clustering algorithms to sepsis patients. However, the clustering process rarely incorporates clinical relevance, potentially limiting to reflect clinically distinct phenotypes. We propose NPCNet, a novel deep clustering network with a target navigator that integrates temporal Electronic Health Records (EHRs) to better align sepsis phenotypes with clinical significance. We identify four sepsis phenotypes ($α$, $β$, $γ$, and $δ$) with divergence in SOFA trajectories. Notably, while $α$ and $δ$ phenotypes both show severe conditions in the early stage, NPCNet effectively differentiates patients who are likely to improve ($α$) from those at risk of deterioration ($δ$). Furthermore, through the treatment effect analysis, we discover that $α$, $β$, and $δ$ phenotypes may benefit from early vasopressor administration. The results show that NPCNet enhances precision treatment strategies by uncovering clinically distinct phenotypes.

Figures (4)

• Figure 1: The overview of the study pipeline.a Data extraction. We use static variables, including demographics and comorbidities, together with time-varying variables, such as laboratory test results and vital signs, as the model input. b NPCNet architecture. Through the text embedding generator, we first bin the value of time-varying variables into bin indices according to the distribution of the training set. We then transform the series of time-varying variables into pseudo texts, combining static information. NPCNet trains with an objective function that consists of $\mathcal{L_\text{rec}}$, $\mathcal{L_\text{clustering}}$, and $\mathcal{L_\text{navigator}}$. Through the clustering network in the clustering operator and clinical relevance through the target navigator, we can derive the patient embeddings and the centroids of computable phenotypes during the training stage. c Phenotype derivation. We only use the information within the first six hours of ICU admission, without the target navigator, to get the patient embeddings. At last, we compute the distance between the patient embeddings and the centroids of phenotypes in the embedding space. Finally, we assign the patients to the nearest phenotypes for further evaluations. d Evaluation process. We assess the clustering results from three perspectives: (1) internal evaluation, (2) clinical significance, and (3) treatment effect analysis.
• Figure 2: Abnormal clinical variables, grouped into eight organ systems, among the sepsis computable phenotypes. The ribbon connects from a phenotype to an organ system if the group median is more abnormal than the overall median (Supplementary Table 2). The more clinical variables are abnormal for the phenotype, the broader the ribbon.
• Figure 3: Multivariable logistic regression on in-hospital mortality with phenotypes.
• Figure 4: The clinical outcomes across phenotypes in the testing set.a SOFA trajectories during the 18 hours following phenotype derivation by NPCNet, stratified by the SOFA score at six hours after ICU admission. b Pairwise comparisons of SOFA trajectories between four phenotypes at each hour within each stratum for NPCNet and benchmarks. The cell in the heatmap represents the number of significantly different phenotype pairs, out of six total possible pairwise combinations, at each time point within each SOFA stratum. The values range from 0 to 6, with darker colors indicating more pairwise differences, thereby reflecting how distinguishable the phenotypes are over time. The Trajectory Divergence Index (TDI) was then derived by normalizing the number of statistically significant pairs by the total number of pairwise comparisons, yielding a metric between 0 and 1 that quantifies the overall performance of models in clinical significance. c Kaplan-Meier curves of one-year mortality for the phenotypes derived by NPCNet and benchmark models.