Presentation (CTI)

OBJECTIVES: To characterize sarcopenia diagnoses in the real-world setting by comparing Appendicular Skeletal Muscle Mass Index (ASMI) and the Sarcopenia Index (SI) against diagnostic thresholds using structured and unstructured electronic health record (EHR) data.
METHODS: This retrospective study used EHRs from the OMNY Health real-world data platform (2017-2025). Patient encounters with a sarcopenia diagnosis code were selected. Encounters with clinical notes containing the term “sarcopenia” or having procedure codes related to body composition measurement or functional performance testing were flagged to assess support for the sarcopenia diagnosis. ASMI and SI were estimated at each encounter using an anthropometric model and the ratio of serum creatinine to cystatin C, respectively. Diagnostic threshold values for sarcopenia (males and females) were defined as 7.0 kg/m² and 5.4 kg/m² for ASMI and 89.9 mg/dL and 79.0 mg/dL for SI.
RESULTS: A total of 18,980 encounters (10,404 unique patients) with a sarcopenia diagnosis code were selected. Patients were predominantly White (68.6%) and female (51.4%) with mean age of 68.7 (males) and 71.5 (females) years. Among diagnoses, 17.8% had support from clinical notes only, 4.8% from procedure codes only, and 2% from both. Mean (standard deviation) ASMI was 7.8 (1.1) and 6.3 (1.4) kg/m² for males and females, respectively (n=8,630). Analogous values were 119 (63.7) and 82.2 (40.3) mg/dL for SI (n=413). Notably, many diagnoses did not meet the diagnostic threshold for ASMI (73.6%) and SI (49.2%).
CONCLUSIONS: Sarcopenia is increasingly recognized as a predictor for delayed recovery from disease and mortality, yet diagnosis codes often lack supporting evidence in EHRs as reported in clinical notes and procedure codes. These findings underline the importance of unstructured clinical notes and suggest that diagnosis coding alone may be insufficient for robust phenotyping of sarcopenia in the real-world setting.