Development and Validation of a Nomogram for Predicting Nutritional Risk Based on Frailty Scores in Older Stroke Patients

Abstract

Background: In older stroke patients with frailty, nutritional deficiencies can amplify their susceptibility, delay recovery, and deteriorate prognosis. A precise predictive model is crucial to assess their nutritional risk, enabling targeted interventions for improved clinical outcomes.

Objective: To develop and externally validate a nutritional risk prediction model integrating general demographics, physical parameters, psychological indicators, and biochemical markers. The aim is to facilitate the early identification of older stroke patients requiring nutritional intervention.

Methods: This was a multicenter cross-sectional study. A total of 570 stroke patients were included, 434 as the modeling set and 136 as the external validation set. The least absolute shrinkage selection operator (LASSO) regression analysis was used to select the predictor variables. Internal validation was performed using Bootstrap resampling (1000 iterations). The nomogram was constructed based on the results of logistic regression. The performance assessment relied on the receiver operating characteristic curve (ROC), Hosmer--Lemeshow test, calibration curves, Brier score, and decision curve analysis (DCA).

Results: The predictive nomogram encompassed seven pivotal variables: Activities of Daily Living (ADL), NIHSS score, diabetes, Body Mass Index (BMI), grip strength, serum albumin levels, and depression. Together, these variables comprehensively evaluate the overall health and nutritional status of elderly stroke patients, facilitating accurate assessment of their nutritional risk. The model exhibited excellent accuracy in both the development and external validation sets, evidenced by AUC values of 0.934 and 0.887, respectively. Such performance highlights its efficacy in pinpointing elderly stroke patients who require nutritional intervention. Moreover, the model showed robust goodness of fit and practical applicability, providing essential clinical insights to improve recovery and prognosis for patients prone to malnutrition.

Conclusions: Elderly individuals recovering from stroke often experience significant nutritional deficiencies. The nomogram we devised accurately assesses this risk by combining physiological, psychological, and biochemical metrics. It equips healthcare providers with the means to actively screen for and manage the nutritional care of these patients. This tool is instrumental in swiftly identifying those in urgent need of targeted nutritional support, which is essential for optimizing their recovery and managing their nutrition more effectively.

Keywords: Elderly people; Frailty; Nomogram; Nutritional risk; Primary care; Stroke.

Fig. 1

LASSO regression screened 7 potential out of 38 candidate variables. Note: (a) Displays the distribution of LASSO regression coefficients for 38 characteristics, producing a logarithm (lambda) sequence coefficient profile graph; (b) Two vertical dashed lines from left to right represent lambda.min and lambda.1se, respectively. Lambda.min corresponds to the λ value that results in the smallest estimated model error, while lambda.1se corresponds to the λ value where the internal cross-validation error is at its maximum within one standard deviation. By determining the optimal λ, seven non-zero coefficients are derived

Fig. 2

Nomogram for predicting nutritional risk in frail older stroke patients

Fig. 3

Roc curve of nutrition risk model in training and validating cohorts

Fig. 4

Calibration curve of nutrition risk model in training and validating cohorts

Fig. 5

DCA curve of nutrition risk predictive nomogram in training and validating cohorts