OBJECTIVES : Cachexia is a common complication of cancer, characterized by extreme weight loss. However, cachexia may be under-coded in claims. This study tried to understand the extent of under-coding by comparing diagnosis-based and weight-based definitions of cachexia in cancer patients.

METHODS : IBM Watson Health linked Claims and Explorys Database was used to extract adult patients with ≥2 diagnoses of solid cancer (ICD-9 140-198; ICD-10 C00-C7B) (index date = first cancer diagnosis date) in 1/1/2014-9/30/2018. All patients had ≥12 months data without cancer pre-index (baseline) and were followed until end of data. The weight-based analysis was conducted on the subset of patients with ≥2 weight measures (including ≥1 weight in post-period). Patient were considered to have weight-based cachexia if they had 1) weight reduction >5% within 6 months or ≥10% within 1 year, 2) body mass index <20 and weight reduction >2%, or 3) sarcopenia (male <7.26 kg/m2; female <5.45 kg/m2 or diagnosis codes) and weight reduction >2%. Patients were considered as having diagnosis-based cachexia if there was ≥1 non-diagnostic claim of cachexia (ICD-9 799.4; ICD-10 R64). Prevalence and incidence rates of cachexia was estimated using both definitions.

RESULTS : Among the newly diagnosed 82,680 cancer patients, 0.53% had diagnosis-based cachexia. Among the subset of 40,301 patients with ≥2 weight measures, 18.8% had weight-based cachexia. For patients without baseline cachexia, the incidence rate of cachexia per 1000 person-years was 2.4 for diagnosis-based and 4.3 for weight-based definition. The mean (standard deviation) number of days from the first cancer diagnosis to the first evidence of cachexia was 408.4 (376.6) for diagnosis-based and 180.9 (230.8) for weight-based definitions.

CONCLUSIONS : Cachexia is under-coded in claims data. To better estimate cachexia rates or economic and clinical burden associated with cachexia in cancer patients, researchers need to use both diagnosis and weights to identify cachexia.