Answers are grounded in this paper and cite their sources.
BACKGROUND: Identification of malnutrition relies heavily on dietitian-led screening and assessment. Structured healthcare data allows for automated identification of suspected malnutrition. This study aimed to develop and validate a computable phenotype for case-finding suspected malnutrition in geriatric outpatients. MATERIALS AND METHODS: A cross-sectional analysis was performed using a clinical data repository. This study included 8,872 older adults aged ≥65 years with at least two encounters at a Midwestern outpatient clinic. Using the Global Leadership Initiative on Malnutrition (GLIM) criteria as a framework, a computable phenotype was created and compared with ICD-10 malnutrition codes and a registered dietitian's chart-review classification. Performance was evaluated using sensitivity, specificity, positive predictive value, negative predictive value, balanced accuracy, and Cohen's kappa. Predictive validity was assessed using a Cox proportional hazards model and visualized with Kaplan-Meier survival curves. RESULTS: The computable phenotype displayed good specificity (96%) but poor sensitivity (69%) compared with ICD-10 malnutrition codes. The two methods had fair agreement (κ = 0.27). The computable phenotype had good sensitivity (95%) and specificity (82%) and showed substantial agreement (κ = 0.76) with a dietitian's malnutrition classification. ICD-10 malnutrition codes (HR = 3.22, CI 2.15-4.82, p < 0.001) and the computable phenotype (HR = 2.85, CI 2.25-3.59, p < 0.001) identified suspected malnutrition subjects were associated with an increased all-cause mortality risk. The computable phenotype (4.78%) identified more older adults as suspected of being malnourished than the ICD-10 codes (1.27%). CONCLUSION: These findings suggest that a GLIM-structured computable phenotype performs with high sensitivity and agrees substantially with a dietitian's classification of malnutrition.
No takes yet. Share an insight, caveat, or question.
Sullivan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: