(P-121) Clinical Notes versus Diagnosis Codes in Identification and Characterization of Women with Endometriosis

Background: Identification of women with endometriosis (EM) via ICD diagnosis codes in administrative databases is challenging due to diagnostic delay marked by non-specific symptom documentation, surgical diagnosis, and non-specific treatment options, leading to selection bias in quantification of disease burden. Electronic health record (EHR) databases include data from clinical management notes by the provider, which may capture different patients than those identified using data entered for administrative purposes.

Objectives: To identify and characterize women with EM using different data sources

Methods: Women aged 18-50 years were identified within Optum’s Market Clarity EHR and Clinformatics DataMart (CDM) databases from 01/01/07 to 03/31/2020. Index for the “clinical cohort” was the first date of EM in the free-text clinical notes (N = 2449). Index for the “administrative EHR cohort” (N = 501) and “claims cohort” (N = 99,204) was the date of the first ICD diagnosis code for EM (ICD-9 617.x or ICD-10 N80.x). Patients were required to be active (EHR) or have continuous enrollment (claims) for 1 year prior to index. Characteristics were defined by ICD code in the 365 days prior to index.

Results: Median [IQR] age was 38 [30-44] years in the clinical, 36 [30-44] years in the administrative EHR, and 37 [32, 43] years in the claims cohorts. Only 25% of women in the administrative EHR cohort also had EM in the clinical notes at baseline; 4% of women in the clinical cohort had an ICD code for EM during follow-up. The baseline prevalence of uterine fibroids ( <1% v 11% v 22%), migraines (8% v 19% v 19%), low back pain (8% v 15% v 20%), pelvic pain (9% v 55% v 46%), dysmenorrhea (2% v 18% v 24%), dyspareunia (1% v 5% v 7%), and depression (9% v 19% v 16%) were lower in the clinical v administrative EHR and claims cohorts, respectively. After index, the incidence of hysterectomy per 100 person-years was 0.5 in the clinical, 4.8 in the administrative EHR, and 9.3 in the claims cohorts. Median [IQR] time to hysterectomy was 394 [45, 1160] days in the clinical, 79 [6, 362] days in the administrative EHR, and 55 [10, 274] days in the claims cohorts.

Conclusions: Characteristics of women with EM vary depending on the data source and criteria used to identify the patient population. Further research is needed to assess if the disease burden differs or if patients identified from diagnoses recorded in claims or EHR administrative files are more advanced in their disease course. These factors may modify conclusions concerning clinical outcomes, healthcare utilization, and disease management.