Handgrip strength is a simple and reliable marker to check the overall health of older adults. It can help to predict problems like disease onset, mortality, and weakness. It is influenced by various factors such as age-related sarcopenia, cognitive function, nutrition, chronic disease, and lifestyle. Many studies have investigated individual risk factors. However, healthcare providers still have limited ways to use this knowledge in real-life care. Comprehensive tools combine all these multiple factors into a single comprehensive assessment. A solution for this challenge is provided by nomograms. It transforms complex statistical models into easy-to-use clinical tools. These tools give personalized risk estimations that help to guide clinical decisions.
A recent study published in Scientific Reports developed and validated logistics regression-based nomograms to predict low handgrip strength using key predictors. This study utilized data from 2011 of the China Health and Retirement Longitudinal Study (CHARLS). This included the participants aged ≥45 years in China. Individuals with missing key characteristics and lost to follow-up were excluded.
The Bayesian Information Criterion (BIC) logistic regression method was used to identify factors associated with low handgrip strength. The model performance was evaluated by using different curves and tests, such as receiver operating characteristic (ROC) curves, decision curve analysis (DCA), and calibration curves. Internal validation was assessed by using 10-fold cross-validation and bootstrapping.
A total of 1,138 individuals (mean age = 64.3 ± 8.7 years, male = 582, female = 556) were included in this study. These are divided into the low handgrip strength group (n = 422) and the normal handgrip strength group (n = 716).
Significant risk factors for low handgrip strength included age >65 years (odds ratio [OR] = 2.35, 95% confidence interval [CI]: 2.14–2.58, p < 0.001), cancer history (OR = 1.61, 95% CI: 1.10–2.33, p = 0.01), stroke history (OR = 2.19, 95% CI: 1.79–2.68, p < 0.001), limitations in daily activities (OR = 1.25, 95% CI: 1.13–1.38, p < 0.001), and elevated glycated hemoglobin level >5.7% (OR = 1.11, 95% CI: 1.02–1.22, p = 0.02). Protective factors were alcohol consumption (OR = 0.82, 95% CI: 0.75–0.91, p < 0.001), regular exercise (OR = 0.87, 95% CI: 0.78–0.97, p = 0.01), married (OR = 0.73, 95% CI: 0.66–0.80, p < 0.001), moderate physical activity (OR = 0.83, 95% CI: 0.74–0.92, p < 0.01), and higher education (OR = 0.76, 95% CI: 0.66–0.87, p < 0.01) had significantly lower risk of low handgrip strength group participants. High body mass index (BMI) was protective, especially in the third quartile, with an OR of 0.62 (95% CI: 0.55–0.71) and p < 0.001. A higher index also reduced risk at the highest risk with an OR of 0.79 (95% CI: 0.71–0.88) and p < 0.01. The overall test showed the model was highly statistically significant with χ² of 278.5, degrees of freedom of 16, and p < 0.001. This indicates that the 12 chosen variables work well together to predict the risk of the low handgrip strength group.
Despite reducing the number of variables by 64%, the simplified model performed similarly to the full model (Area under curve [AUC]: 0.78 vs. 0.77; Brier score: 0.179 vs. 0.183). The DeLong test confirmed no significant AUC difference (p = 0.631) among the two models, and the lower BIC (1245.3 vs. 1389.6) showed improved model simplicity.
The simplified model demonstrated good discrimination ability (AUC = 0.78, 95% CI: 0.77–0.79), significantly better than random (AUC = 0.5, DeLong test, p < 0.001) by using a non-parametric method. The calibration slope stayed consistent across the validation samples, ranging from 0.89 to 0.95 based on 1000 bootstraps and 10-fold cross-validation, indicating that performance was stable and generalizes well.
The threshold of 0.40 was the optimal, offering good sensitivity (72.5%) and specificity (69.8%). DCA showed that this model has provided greater clinical net benefit across most thresholds, especially at 0.40, where the net benefit (0.19) exceeds both treat all (0.11) and treat none (0), supporting its potential in clinical decision-making.
In conclusion, this nomogram serves as a practical tool for clinicians to identify individuals at risk of reduced handgrip strength and as a guide for targeted interventions. By integrating multiple risk factors into a user-friendly, simple format, it translates complex data into actionable insights. It supports early detection and facilitates targeted intervention, and aims to maintain functional independence and enhance quality of life in older people.
Reference: Wu T, Li B. A grip strength prediction tool for older adults based on logistic regression: construction, validation, and clinical application value. Sci Rep. 2025;15:15283. doi:10.1038/s41598-025-00291-0
