Presentation (CTI)

OBJECTIVES: Administrative claims data often underestimate disease prevalence by capturing only diagnosed cases. Quantifying the "undiagnosed" burden is critical for disease surveillance, population health modeling, and identifying missed intervention opportunities to enable earlier management, prevent disease progression and improve long-term outcomes. We aimed to demonstrate the utility of linked electronic medical record (EMR) and claims data for estimating the proportion of patients with clinical evidence of obesity, chronic kidney disease (CKD), and type 2 diabetes (T2D), including those without diagnosis codes.
METHODS: This retrospective study utilized the IQVIA Pharmetrics Plus linked to Ambulatory EMR and Mortality dataset. The population included nearly 4 million US adults aged ≥18 years observed from January-December 2023, with 12- and 6- month pre-index continuous enrollment for medical and pharmacy claims. Patients were identified using clinical criteria and assessed for corresponding ICD-10 codes. Criteria were: 1) BMI ≥30 kg/m², or ≥27.5 kg/m² for Asian adults (WHO Expert Consultation, 2004); 2) eGFR <60 mL/min/1.73m² on ≥2 occasions spaced ≥90 days apart. 3) HbA1c ≥6.5%.
RESULTS: Our analysis revealed significant diagnosis gaps. Obesity presented the largest "silent" cohort: 67.5% (n=301,658) of 447,228 clinically eligible patients lacked an ICD-10 code, meaning 2 in 3 patients were missed by claims alone. For CKD, 16.6% (n=4,625) of patients meeting clinical criteria were undiagnosed. T2D showed the smallest gap at 5.8% (n=2,668).
CONCLUSIONS: ICD-10 codes alone missed over two-thirds of obesity cases and nearly one-fifth of CKD cases in this network, highlighting a substantial “invisible” patient population at risk of delayed diagnosis, disease progression, and poorer outcomes. Leveraging linked EMR-claims data can more accurately characterize undiagnosed cardio-renal-metabolic burden than either source alone while identifying missed intervention opportunities. As analyses were restricted to patients with available clinical markers, results may conservatively estimate the true diagnosis gap. Further limitations include representativeness, missingness, and the single-year lookback period.