Innovative nomogram for predictive risk stratification of aspiration pneumonia in post-stroke dysphagia patients

TL;DR

This study created a new tool to predict aspiration pneumonia risk in stroke patients with swallowing difficulties, using common clinical data for better early identification.

Contribution

The first dedicated nomogram for AP risk prediction in PSD patients with multi-center validation and novel predictor combinations.

Findings

The nomogram model achieved excellent discrimination with a C-index of 0.885.

It showed good agreement between predicted and observed AP probabilities.

The model provided net benefit across various threshold probabilities.

Abstract

Post-stroke dysphagia (PSD) affects up to 76% of stroke patients and increases aspiration pneumonia (AP) risk, leading to higher mortality among older survivors. Current risk assessment tools for AP in PSD patients lack precision.

We conducted a retrospective study of 7,134 stroke patients admitted to Jinshan Hospital from 2019 to 2023. We used multivariable logistic regression to identify AP predictors and constructed a nomogram model using these predictors. Model performance was evaluated using bootstrap resampling, calibration, and decision curve analysis. Internal validation was conducted on 30% of cases, and external validation was performed on 500 PSD patients from community health centers.

Among 2,663 PSD patients, 578 (21.7%) developed AP. Independent predictors included age, stroke severity, hyperlipidemia, hyperhomocysteinemia, heart failure, CRP, WBC, neutrophil ratio, Hb, FBG, prealbumin, BNP, and serum sodium. The nomogram model showed excellent discrimination (C-index: 0.885) and good agreement between predicted and observed AP probabilities. It provided net benefit across various threshold probabilities.

Our study developed the first dedicated nomogram for AP risk prediction in PSD patients, incorporating novel predictor combinations and demonstrating robust validation across multi-center cohorts. This fills an important clinical need under community conditions by enabling early identification of high-risk PSD patients using routinely available clinical variables.

Figures