SYSTEMATIC VARIATION IN ICD-10-CM OBESITY CODING BY BMI SEVERITY, SEX, AND AGE: EVIDENCE FROM THE NORSTELLALINQ LINKED CLAIMS AND EHR DATABASE
Author(s)
Spencer Friedman1, ilan behm, MPH2, Eric Mitchell, BS3, Allison Perry, PhD4.
1Norstella, Yardley, PA, USA, 2Norstella, Englewood, CO, USA, 3Norstella, Brooklyn, NY, USA, 4Norstella, New York, NY, USA.
1Norstella, Yardley, PA, USA, 2Norstella, Englewood, CO, USA, 3Norstella, Brooklyn, NY, USA, 4Norstella, New York, NY, USA.
OBJECTIVES: To characterize variation in ICD-10-CM obesity coding by BMI severity, age, and sex using exact EHR-derived BMI as the reference standard, and identify patient subgroups systematically missed by diagnosis code-based identification.
METHODS: A cross-sectional analysis was conducted using NorstellaLinQ’s linked claims and EHR database. Adults (≥18 years) with BMI ≥30 kg/m² were included. Exact BMI was derived from discrete height and weight fields; Maximum recorded BMI was used to capture the highest documented obesity severity during observation: Class I (30-<35), Class II (35-<40), or Class III (≥40). ICD-10-CM obesity coding status was assessed within ±6 months of BMI measurement across E66.xx and Z68.xx code families. Undercoding rates were stratified by obesity class, age group (18-34, 35-49, 50-64, 65+), and sex.
RESULTS: Among 7,445,857 patients with BMI ≥30 kg/m², undercoding rates decreased monotonically with increasing BMI severity across all age and sex subgroups consistently. Class I obesity showed the highest undercoding rates (67-72%), followed by Class II (55-62%), and Class III (39-47%), suggesting coding likelihood increases with clinical severity. Age was not a consistent predictor of undercoding within any BMI class. Women were more likely to be coded than men across all obesity classes and age groups, suggesting an independent sex-related difference in coding practices. The most undercoded subgroup was Class I males aged 35-49 (74.6% undercoded); the best coded was Class III females aged 35-49 (39.4% undercoded).
CONCLUSIONS: ICD-10-CM obesity coding rates vary substantially and systematically by BMI severity and sex, but not by age. Because undercoding is systematic rather than random, diagnosis code-based obesity cohorts may overrepresent patients with more severe disease and women, potentially introducing selection bias into real-world evidence studies. These findings have direct implications for the design of RWE studies, GLP-1 RA prior authorization criteria, and payer utilization management programs that rely on diagnosis codes to define eligible populations.
METHODS: A cross-sectional analysis was conducted using NorstellaLinQ’s linked claims and EHR database. Adults (≥18 years) with BMI ≥30 kg/m² were included. Exact BMI was derived from discrete height and weight fields; Maximum recorded BMI was used to capture the highest documented obesity severity during observation: Class I (30-<35), Class II (35-<40), or Class III (≥40). ICD-10-CM obesity coding status was assessed within ±6 months of BMI measurement across E66.xx and Z68.xx code families. Undercoding rates were stratified by obesity class, age group (18-34, 35-49, 50-64, 65+), and sex.
RESULTS: Among 7,445,857 patients with BMI ≥30 kg/m², undercoding rates decreased monotonically with increasing BMI severity across all age and sex subgroups consistently. Class I obesity showed the highest undercoding rates (67-72%), followed by Class II (55-62%), and Class III (39-47%), suggesting coding likelihood increases with clinical severity. Age was not a consistent predictor of undercoding within any BMI class. Women were more likely to be coded than men across all obesity classes and age groups, suggesting an independent sex-related difference in coding practices. The most undercoded subgroup was Class I males aged 35-49 (74.6% undercoded); the best coded was Class III females aged 35-49 (39.4% undercoded).
CONCLUSIONS: ICD-10-CM obesity coding rates vary substantially and systematically by BMI severity and sex, but not by age. Because undercoding is systematic rather than random, diagnosis code-based obesity cohorts may overrepresent patients with more severe disease and women, potentially introducing selection bias into real-world evidence studies. These findings have direct implications for the design of RWE studies, GLP-1 RA prior authorization criteria, and payer utilization management programs that rely on diagnosis codes to define eligible populations.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
RWD161
Topic
Clinical Outcomes, Economic Evaluation, Real World Data & Information Systems
Topic Subcategory
Data Protection, Integrity, & Quality Assurance, Distributed Data & Research Networks, Health & Insurance Records Systems, Reproducibility & Replicability
Disease
Diabetes/Endocrine/Metabolic Disorders (including obesity), No Additional Disease & Conditions/Specialized Treatment Areas