RESULTS

Study Population

This study identified 2 218 115 eligible patients with a BMI measurement in the HIRD during the intake period, with 966 427 (44%) having a BMI ≥30 kg/m2. Of these patients with obesity, 523 165 (54%) had a medical claim with an ICD-10 diagnosis code for obesity within 6 months before or after index date. Patients having a medical claim code with an ICD-10-CM code for obesity demonstrated distinct differences in demographic and clinical characteristics compared with those without a code (Table 1; Supplementary Table S1). On average, patients with a code were older at index date (53.6 vs 51.3 years) and were more likely to be insured through Medicare Advantage/Supplemental as opposed to commercial insurance (22% vs 16%) compared with those without a code. They also had a higher mean BMI (37.6 vs 34.6). Comorbidity burden, measured using the mean Quan Charlson Comorbidity Index, was also higher in this group (1.6 vs 1.2). Coded patients also had a greater proportion of individuals with each comorbidity of interest over the full pre-index period and each medication and healthcare service of interest over the 6-month pre-index period. Across all eligible patients, the most common comorbidities included hypertension, dyslipidemia, and lower back pain; the most frequently prescribed medications included antihypertensives, antihyperlipidemics, and antidepressants; and the most common healthcare encounters were visits to a cardiologist, physical therapy, and orthopedic encounters.

 | Medical Claims Code (N = 523 165) | No Medical Claims Code (N = 443 262)
Demographics |  | 
Age in years, mean (SD) | 53.6 (14.6) | 49.7 (16.0)
Female, % | 61 | 55
Insurance type, % |  | 
Commercial | 76 | 82
Medicare Advantage | 22 | 16
Unknown/other | 2 | 2
Race, % |  | 
American Indian or Alaska Native, NH | 1 | <1
Asian, NH | 1 | 2
Black, NH | 15 | 12
Hispanic | 6 | 6
Native Hawaiian or other Pacific Islander, NH | <1 | <1
White, NH | 75 | 77
Unknown/other | 2 | 3
BMI, kg/m2, % |  | 
Mean (SD) | 37.6 (6.7) | 34.6 (5.1)
Obesity class I (30-34.9) | 44 | 66
Obesity class II (35-39.9) | 28 | 22
Obesity class III (≥40) | 28 | 12
Healthcare encounters, % |  | 
Endocrinologist visit | 6 | 3
Cardiologist visit | 18 | 11
Bariatric surgery | 2 | <1
Orthopedic encounter | 13 | 8
PT visit | 13 | 10
Comorbidities, % |  | 
QCI, mean (SD) | 1.6 (2.1) | 1.2 (1.9)
Anxiety disorder | 41 | 36
Dyslipidemia | 66 | 55
Hypertension | 67 | 53
Lower back pain | 49 | 42
OSA | 29 | 17
Osteoarthritis of knee | 23 | 16
T2D | 33 | 22
Medication use, % |  | 
Antihyperlipidemic medication | 35 | 27
Antihypertensive medication | 49 | 36
Chronic weight management medication | 1 | <1
GLP-1 RA for T2D | 6 | 3

Validity of Codes

In the evaluation of the validity of obesity diagnosis codes, a total of 380 606 TP, 17 422 FP, 1 234 266 TN, and 585 821 FN were identified (Table 2). This distribution yielded a PPV of 95.6%, which implies that when a code for obesity is identified, there is a 95.6% probability that the patient’s BMI is ≥30 kg/m2. The NPV was recorded as 67.8%, indicating that if there is no observed code for obesity, the patient has a 67.8% chance of having a BMI <30 kg/m2. Specificity was 98.6%, suggesting that when a patient’s BMI is <30 kg/m2, there is a 98.6% likelihood that there will be no code for obesity. However, the sensitivity was found to be 39.4%, meaning that when a patient’s BMI is ≥30 kg/m2, there is only a 39.4% chance that an obesity diagnosis code will be present within 60 days before or after BMI documentation in the EHR. This relatively lower sensitivity points towards potential missed or underdocumentation of obesity in patients with a BMI ≥30 kg/m2.

 | ICD-10-CM Obesity Code Present | ICD-10-CM Obesity Code Absent | Total
BMI ≥30 kg/m2 (obesity) | True positive: 380 606 | False negative: 585 821 | 966 427
BMI <30 kg/m2 | False positive: 17 422 | True negative: 1 234 266 | 1 251 688
Total | 398 028 | 1 820 087 | 2 218 115

Prevalence and Predictors of Coding

Overall, the prevalence for obesity coding in medical claims within 6 months before or after BMI documentation of obesity in the EHR was found to be 54%. The prevalence of obesity coding exhibited a wide range, from 39% to 88%, depending on demographic and clinical characteristics (Table 3). Coding prevalence was lowest among patients in the age groups 18-24 years (39%) and 25-34 years (46%), as well as those identified as obesity class I (44%). The highest prevalence of obesity coding was found among patients receiving chronic weight management medication (88%), those who underwent bariatric surgery (86%), and those receiving diet counseling and surveillance (84%). However, these groups represented less than 1%, 1%, and 3% respectively of patients with a BMI ≥30 kg/m2. Further examination revealed that among patients exhibiting the most common clinical characteristics described above, obesity coding prevalence were slightly higher but did not vary appreciably from the mean (58%-65%).

 | Prevalence, % (95% CI)
Age, years | 
18-24 | 38.7 (38.3-39.2)
25-34 | 46.3 (46.0-46.6)
35-44 | 52.5 (52.2-52.7)
45-54 | 55.8 (55.6-56.0)
55-64 | 56.3 (56.1-56.4)
65-74 | 60.1 (59.8-60.3)
≥75 | 52.9 (52.6-53.3)
BMI kg/m2 | 
Obesity class I (30-34.9) | 44.0 (43.9-44.2)
Obesity class II (35-39.9) | 59.8 (59.6-60.0)
Obesity class III (≥40) | 73.9 (73.7-74.1)
Clinical characteristics | 
Dyslipidemia | Present: 58.8 (58.6-58.9); absent: 46.9 (46.7-47.1)
Hypertension | Present: 59.9 (59.8-60.0); absent: 45.2 (45.1-45.4)
Lower back pain | Present: 57.8 (57.7-58.0); absent: 51.1 (50.9-51.2)
Antidepressant medication | Present: 59.9 (59.7-60.1); absent: 52.1 (52.0-52.2)
Antihyperlipidemic medication | Present: 60.5 (60.3-60.7); absent: 51.2 (51.1-51.4)
Antihypertensive medication | Present: 61.3 (61.2-61.5); absent: 48.7 (48.6-48.3)
Chronic weight management medication | Present: 87.6 (86.6-88.6); absent: 54.0 (53.9-54.1)
Bariatric surgery | Present: 85.8 (85.2-86.5); absent: 53.8 (53.7-53.9)
Diet counseling and surveillance | Present: 65.3 (65.0-65.5); absent: 52.2 (52.1-52.3)
Orthopedic encounter | Present: 63.2 (62.9-63.4); absent: 53.0 (52.9-53.1)
PT visit | Present: 60.8 (60.5-61.1); absent: 53.3 (53.1-53.4)

The study found that certain clinical characteristics increased the likelihood of having a code for obesity (Supplementary Table S2). These included being female, being older than 24 years (up to age 74), identifying as non-White, being insured by Medicare Advantage, and having lower socioeconomic status, higher BMIs, and the presence of most comorbidities, medications, and encounters included in the model (Figure 1). Nevertheless, the impact of most of these clinical characteristics was modest with odds ratios (OR) close to 1, implying the increase in coding odds was not meaningful. However, receiving weight management interventions and having a BMI ≥35 kg/m2 considerably increased the odds of obesity coding, consistent with prevalence estimates. Specifically, patients who received chronic weight management interventions had over three times the odds of being coded for obesity, including bariatric surgery (OR, 3.01; 95% CI, 2.83-3.20), diet counseling and surveillance (OR, 3.47; 95% CI, 3.35-3.60), and receiving weight management medication (OR, 4.63; 95% CI, 4.19-5.13). Additionally, markedly higher BMI, specifically within obesity class III, compared with class I, tripled the odds of coding for obesity (OR, 3.11; 95% CI, 3.07-3.15).

Figure 1.: Logistic Regression Model Exploring Predictors of Medical Claims Coding for Obesity Among Patients with BMI ≥30 kg/m2 in the EHRAbbreviations: BMI, body mass index; CKD, chronic kidney disease; EHR, electronic health record; ESKD, end stage kidney disease; GERD, gastroesophageal reflux disease; GLP-1 RA, glucagon-like peptide-1 receptor agonist; HF, heart failure; MAFLD, metabolic dysfunction-associated fatty liver disease; MASH, metabolic dysfunction-associated steatohepatitis; MDD, major depressive disorder; NH, Non-Hispanic; NSAID, nonsteroidal anti-inflammatory drug; OSA, obstructive sleep apnea; PPI, proton pump inhibitor; SGLT2, sodium-glucose cotransporter-2; SES, socioeconomic status; T2D, type 2 diabetes).Reference groups: Sex, male; age, 18-24; race/ethnicity, White, NH; health plan product, commercial; region, South; SES category, 4; index provider specialty, primary care provider; BMI, 30-34.9; comorbidities, absent; medications, no use.Note: SES category 1 is lowest quartile and 4 is highest quartile.