Impact of Endometriosis Diagnostic Delays on Healthcare Resource Utilization and Costs

This retrospective study used the Optum Research Database, a geographically diverse US database representing approximately 67 million individuals from 1993 to present. Medical and pharmacy claims and enrollment information were obtained from 1 January 1999 to 31 July 2017 (study period). Medical claims consisted of International Classification of Disease, Ninth and Tenth Revisions, Clinical Modification (ICD-9-CM and ICD-10-CM) diagnosis and procedure codes, Healthcare Common Procedure Coding System (HCPCS) codes, and revenue codes. Pharmacy claims included National Drug Codes for filled prescriptions and days and quantity of drug supplied. The Optum Research Database is fully de-identified and HIPAA compliant and did not require Institutional Review Board approval or waiver of authorization.

Patients were required to be aged 18–49 years and have ≥ 1 medical claim for endometriosis in any position (ICD-9-CM/ICD-10-CM diagnosis code 617.x/N80.x) from 1 January 2004 to 31 July 2016 (identification period). The date of the first medical claim with an endometriosis diagnosis code was considered the index date. Patients were required to have continuous health plan enrollment with medical and pharmacy benefits for ≥ 60 months (1825 days) prior to the index date (pre-diagnosis period) and ≥ 12 months (365 days) following the index date and a medical claim for ≥ 1 endometriosis symptom in any position (dyspareunia, generalized pelvic pain, abdominal pain, dysmenorrhea, or infertility) during the pre-diagnosis period. A pre-diagnosis period of 5 years was selected based on recent evidence from Soliman et al. who reported the average delay from first symptom to endometriosis diagnosis was approximately 4.4 years [13]. Endometriosis symptoms were selected based on published literature [4, 5] and guidance from the clinician author. Patients with a medical claim for endometriosis or malignancy prior to the index date were excluded from the study. To rule out patients with conditions that have symptoms similar to endometriosis, patients with an ICD-9/ICD-10 diagnosis code for genitourinary or intra-abdominal infection (e.g., chlamydia, gonorrhea), inflammatory bowel disease, diverticulitis, appendicitis, peritonitis, other genitourinary conditions (cystitis, urethritis), or kidney stones any time prior to the index date were also excluded.

Patients were assigned to delay cohorts based on the length of time from the date of the first medical claim for a non-diagnostic service for an endometriosis symptom to the index date categorized as: short delay (≤ 1 year), intermediate delay (1–3 years), and long delay (3–5 years). These cutoffs are conservative estimates of disease burden and were chosen to balance the clinical burden and sample size requirements for the study.

All study variables were analyzed descriptively, and comparisons between delay cohorts were made. Mean healthcare utilization and costs over the 60-month pre-diagnosis period were calculated and presented separately for each of the three diagnostic delay cohorts. Statistical tests of significance for differences across the three cohorts were conducted using chi-square tests for categorical variables and ANOVA and t test for continuous variables. For endometriosis-specific healthcare resource utilization and costs, the hypothesis tested was whether patients who have had endometriosis symptoms for a longer period of time would have mean utilization and costs equal to patients who have had endometriosis symptoms for a shorter period of time. Calculated p-values were adjusted for multiple comparisons (Bonferroni correction) with a threshold of statistical significance of p < 0.017.

After applying inclusion and exclusion criteria, 11,793 patients were included in the study, of which 37.7% (n = 4446) had a short delay, 27.0% (n = 3179) had an intermediate delay, and 35.3% (n = 4168) had a long delay (Fig. 1). Patients with a short delay were slightly older (39.8 ± 7.0) than patients who had intermediate (38.9 ± 7.8) or long delays (38.9 ± 7.6) (p < 0.001 for both comparisons) (Table 1). Approximately half of all patients resided in the South, and 66% had a point of service healthcare plan (p value not significant).

Patients in this study had a mean diagnostic delay of 763.9 ± 631.0 days (2.09 ± 1.77 years) (Table 1). Patients with a short, intermediate, or long delay averaged 90.2 days, 733.4 days, and 1505.9 days, respectively, from the onset of endometriosis symptoms until diagnosis. Common symptoms identified were abdominal pain (67.3%), dysmenorrhea (52.0%), and dyspareunia (13.0%) (Table 1). Patients with a short delay were least likely to have abdominal pain and infertility, but were most likely to have dysmenorrhea compared with patients who had intermediate and long delays. The proportion of patients with abdominal pain increased significantly with increasing diagnostic delay (p < 0.001 for all comparisons). Using the proxy for disease severity, patients with a long delay had a less concentrated presence of endometriosis symptoms than those with shorter delays ranging from a symptom severity of 0.3–1.0 (p < 0.001 for all comparisons; Table 1).

Almost all patients (95.8%) had ≥ 1 comorbid condition (Table 1), with the most common being fatigue/neurasthenia (49.2%), headache and migraine (42.8%), ovarian cysts (40.6%), urinary tract infections (38.8%), depression and anxiety (37.6%), and uterine fibroids (34.2%). Comorbidities tended to be the highest among patients with longer delays.

There are several limitations to this study. Healthcare claims are collected for the purpose of payment, not research, which leads to several inherent limitations. The presence of an endometriosis diagnosis code on a medical claim is not proof of the presence of disease. The diagnosis code may be incorrectly coded or included as rule-out criteria rather than actual disease. It is possible that patients may have had a diagnosis of endometriosis prior to the baseline period, which may explain the older age of onset found in this study. The baseline period was extended to 5 years to minimize this risk. Additionally, endometriosis symptoms may not have been fully captured in the claims database. Endometriosis-related healthcare utilization and costs in the pre-diagnosis period were based on the presence of claims for five common endometriosis symptoms and endometriosis-related surgical and pharmacologic treatment. While this was done to provide conservative estimates, it is possible that the true costs and utilization due to endometriosis were higher in this population. This study included a managed care population and may not be generalizable to other populations. Additionally, due to the coverage of the health plan underlying the claims database, about half of the study patients were from the South. Since racial and ethnic data were not collected, we cannot know if this may have skewed study results. Lastly, due to the observational nature of this study and the descriptive analyses performed, it is possible that confounding factors not accounted for could contribute to the difference in costs and utilization between the diagnostic delay cohorts.