Introduction
Endometriosis is a chronic gynecological disorder characterized by the presence of endometrial-like tissue (glands and stroma) outside the uterus [
1]. It is estimated that close to 200 million women worldwide will experience endometriosis [
2]. Over 10 million women in the USA have endometriosis with the highest prevalence among women aged between 30 and 40 years [
3,
4].
Current therapeutic options for managing endometriosis include pharmacological and surgical treatments [
5,
6]. While pharmacological agents (including oral contraceptives, nonsteroidal anti-inflammatory drugs, progestogens, gestrinone, and gonadotropin-releasing hormone agonists) are often used as first-line therapy for reducing pain, surgical interventions (including laparoscopic resection, laparotomy, bilateral salpingo-oophorectomy, and hysterectomy) are the mainstay of treatment for patients with advanced endometriosis, or who remain refractory to pharmacotherapy, or those with medication-related side effects [
7,
8].
Treatment of endometriosis through pharmacotherapy and/or surgery imposes a potentially significant economic burden on the US healthcare system [
9] and on society in general through work loss and reductions in the quality of life of the sufferers [
10‐
12]. The annual healthcare cost burden in the USA associated with endometriosis was estimated to be $22 billion in 2002, of which $17.3 billion was due to direct medical costs (outpatient and hospitalization) and $4.7 billion was due to indirect costs (loss of productivity) [
13]. Later, in a multicenter, prospective study, the annual societal burden of endometriosis in the USA was estimated to be €49.6 billion (equivalent to $69.4 billion; US$ October 2010) [
14]. The same study reported that approximately two-thirds of the costs were due to lost productivity and one-third was due to direct healthcare costs. Given the prevalence of endometriosis, there is a need to estimate and update the contemporary cost burden of endometriosis in a real-world setting. The purpose of the current retrospective study was to quantify incremental direct and indirect healthcare costs in the USA among newly diagnosed endometriosis patients in the year following diagnosis compared to those without endometriosis.
Methods
Data Source
This study used a similar study design and analytic approach of a previously published study that estimated the direct and indirect cost burden associated with undergoing an endometriosis-related surgery [
15]. This retrospective database analysis used healthcare claims data from the Truven Health MarketScan
® Commercial Claims and Encounters (Commercial) and the MarketScan
® Health and Productivity Management (HPM) databases for the period January 1, 2009 through June 30, 2015. The commercial database contains pooled inpatient and outpatient healthcare experience, including pharmacy claims, of approximately 40 million employees and their dependents covered under a variety of health plan types in the USA. Claims-based datasets contain relatively comprehensive real-world data on healthcare with the ability to track large numbers of patients. An important limitation of such data is that it is created from a system designed for reimbursement purposes rather than for research so that, for example, there is an absence of information on medical expenses not subject to insurance reimbursement such as for over-the-counter medications. More specific information on the limitations of the dataset are detailed in the “
Discussion”.
The HPM database includes details on workplace absenteeism, short-term disability, and workers’ compensation data for a subset of individuals in the commercial database. The data is provided from employer payroll systems and disability case records supplied by data contributors to the Commercial database. Information on specific causes of absences or disability are not directly coded in the database.
All database records are de-identified and were accessed with protocols compliant with US patient confidentiality requirements, including the Health Insurance Portability and Accountability Act (HIPAA) of 1996, and thus were exempted from institutional review board approval. This article does not contain any studies with human participants or animals performed by any of the authors.
Study Population
Endometriosis Cohort
Female patients (aged 18–49 years) with a diagnosis of endometriosis (International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) 617.x) on a non-diagnostic medical claim between January 1, 2010 and June 30, 2014 during the study time frame were initially selected for analysis. The date of first diagnosis was defined as the index date. Additionally, patients were considered eligible for inclusion if they had continuous health plan coverage for 12 months before (pre-index period) and 12 months after (follow-up period) the index date. Patients with a diagnosis of endometriosis (ICD-9-CM 617.x) in any position on a non-diagnostic medical claim preceding the index date, or diagnosis of malignant neoplasm of female genitourinary organs (ICD-9-CM: 179.x-184.x) or radical hysterectomy during the 12-month pre-index period were excluded. ICD-9-CM codes are used here and throughout for identifying diagnostic information because it was the required coding system during the relevant time period on insurance reimbursement claims used in the commercial database.
Non-Endometriosis Control Cohort
Patients in the non-endometriosis cohort were identified using selection criteria similar to the endometriosis cohort. Female patients with no diagnosis of endometriosis during the entire study period (January 1, 2009 through June 30, 2015) were randomly selected and directly matched 10:1 to endometriosis patients on the basis of data availability in the index year and age at index). Index date for the control cohort was randomly assigned during this time period with a date distribution matching the date distribution found among endometriosis patients.
Outcome Measures
Healthcare resource utilization and costs (all-cause and endometriosis-related) were evaluated for the 12-month pre- and post-index periods. The direct healthcare costs incurred were assessed for healthcare services across various care settings: inpatient, emergency room (ER), outpatient services including obstetrician–gynecologist (OB/GYN) visits, and outpatient prescriptions. Endometriosis-related utilization and expenditures were defined as (a) medical claims with a principal diagnosis of endometriosis, (b) pharmacy claims for drugs used to treat and manage endometriosis (leuprolide, danazol, depot medroxyprogesterone, oral medroxyprogesterone acetate, levonorgestrel implants and intrauterine devices, megestrol acetate, progestin-only oral contraceptive pills, histrelin, goserelin, nafarelin, triptorelin, ganirelix, degarelix, cetrorelix, abarelix, and other oral contraceptive pills), and (c) endometriosis-related surgeries (number and proportion of patients with at least one endometriosis-related surgery overall, and stratified by laparotomy, laparoscopy, hysterectomy, oophorectomy, and other excision/ablation procedures). All expenditures were inflation adjusted to 2014 US$ using the Medical Care component of the Bureau of Labor Statistics Consumer Price Index [
16].
Work losses associated with absenteeism (the number of days absent from work), short-term disability (STD), and long-term disability (LTD) in the 12-month post-index period were assessed for both endometriosis and non-endometriosis cohorts in terms of the proportion of patients with each category of work loss as well as average number of days lost. To assess indirect costs due to work loss, a daily wage was calculated using the age-, gender-, and geographic region-adjusted wage rate from the US Bureau of Labor Statistics from the year 2010. An estimated daily wage rate ($240 per day) was considered for calculating costs from lost productivity due to STD/LTD. As STD/LTD benefit typically does not replace full wages, 70% of wages, which approximates the proportion of total wages and benefits paid by employers in the HPM database, was used.
Patient Characteristics
Patient characteristics including age, geographic region (US Census division), urbanicity, health plan type, and index year were measured on the index date. In addition, the Deyo Charlson Comorbidity Index (DCI), comorbid conditions (based on the presence of ICD-9-CM diagnosis), and medication use (based on Healthcare Common Procedure Coding System (HCPCS) and National Drug Code (NDC) codes) were measured during the 12-month pre-index period.
Statistical Analysis
Patient characteristics, healthcare utilization, and costs for each study cohort were analyzed descriptively. Continuous measures were summarized as means and standard deviations, and categorical measures were summarized as counts and percentages. Testing for statistical differences in the study outcomes between cohorts was performed using Chi square test for categorical measures, and ANOVA and t tests for continuous measures.
The “real-world” costs as determined from the Commercial and HPM databases are presented first. However, the reported cost differences for the cohorts may be influenced by differences in their baseline characteristic, so adjusted cost estimates were also computed. Multivariate analysis was used to compare both direct and indirect (absence, STD, LTD, and total) cost outcomes in patients with and without endometriosis. Generalized linear models (GLMs) were used to estimate incremental costs for patients with endometriosis versus those without. The models used generalized estimating equations to account for within-patient correlation and used a gamma distributed error and log link to account for the distributional characteristics of cost data. Estimation of the standard error of the marginal cost was conducted by using the delta method. Two-part models were used for potentially infrequent outcomes and associated costs such as total STD costs [
17]. All models controlled for the relevant patient characteristics (age, gender, geographic region, and health plan type, DCI, and baseline comorbidities) between the two cohorts. The recycled prediction method was used to estimate the marginal or incremental costs associated with endometriosis. A
p value of less than 0.05 was considered statistically significant.
Discussion
This retrospective study estimated the incremental direct and indirect costs associated with endometriosis among commercially insured patients in the USA and showed that the direct healthcare costs for endometriosis patients measured both 12 months prior to and after endometriosis diagnosis were significantly higher than those in matched controls. Additionally, costs due to work loss through absenteeism, STD, and LTD were also higher in patients with endometriosis. To our knowledge, this is the first major study that compared both direct and indirect healthcare costs between patients with and without endometriosis.
Analogous to the findings of previous studies [
9,
13,
14] this study showed that endometriosis patients incurred significantly higher healthcare costs compared to controls. The incremental costs per patient in the first year post-diagnosis (over $10,000 with multivariable adjustment) are substantial compared to estimates for other common gynecological ailments such as uterine fibroids ($6873 in 2014 US$ [
18]) or menorrhagia ($2878 in 2014 US$ [
19]). Notably, over 60% of the total annual healthcare costs for endometriosis patients were accrued within 3 months of the index date, thereby adding to the existing body of literature which has reported highest healthcare resource utilization and costs in the first year after endometriosis diagnosis [
20‐
22].
Indirect costs associated with endometriosis are substantial as well, with average incremental indirect costs estimated at US $2132 per patient (adjusted estimate) in the 12 months following the index diagnosis. A point to notice here is that claims data can only capture indirect losses due to the absenteeism component; data on presenteeism (reduced productivity due to illness while on the job) was not available. Estimates from a study of European Union countries suggest that indirect costs due to presenteeism can substantially exceed the costs of absenteeism, by a factor of 2: €6298 due to presenteeism vs. €3280 due to absenteeism [
14]. A study of 10 countries across multiple continents estimated that endometriosis resulted in the equivalent work loss of 6.4 h/week due to presenteeism and 4.4 h/week due to absenteeism [
23] and provided indirect cost estimates for the USA of about $3200/year for absenteeism due to endometriosis symptoms and about $14,800 for presenteeism (values converted into 2014 US$). Instead of using records of absence and disability claims, that study based indirect cost estimates on responses by patients to versions of the Work Productivity and Activity Impairment (WPAI:GH) and Short-Form–36 version 2 (SF36v2), and estimated the work loss in terms of lost wages without adjustments for reduced pay for short- and long-term disability.
A previous study characterized the healthcare costs and utilization of endometriosis patients [
22]. Although the prior study followed costs and utilization patterns over a longer period of time (up to 5 years for a small percentage of patients), the current analysis expands upon this previous work in several ways: by reporting endometriosis-related costs separately, examining trends within the first year post-diagnosis, and by providing multivariate models to provide cost estimates adjusted for variations in baseline covariates. It also contains a somewhat broader sample of patients, including women up to 49 years of age (as opposed to 45 years in Fuldeore et al.’s analysis [
22]). Importantly, this study provides an update in an evolving field of treatment, and confirms changes in approaches to gynecological treatment.
A recent analysis shows that hysterectomies and oophorectomies (as treatments across all indications and not just endometriosis) have shown modest decreases in prevalence since the year 2000 with a dramatic shift towards outpatient instead of inpatient procedures [
21]. In an analysis spanning the years 2000–2010, Fuldeore et al. [
22] reported that 40% of endometriosis patients received an inpatient admission in their first year. In contrast, in the current analysis spanning the years 2010–2014 only 29% received an inpatient admission. These numbers include admissions for all reasons, not just endometriosis, but because surgical treatment is very common within the first year of diagnosis [
9,
21], this is very likely related to a drop in inpatient treatment for endometriosis.
Despite changes in the treatment landscape, surgery remains a mainstay of endometriosis therapy in our analysis, with 38% of the patients undergoing a hysterectomy in the first year post diagnosis, and two-thirds undergoing a surgical procedure used to treat endometriosis. This high rate of hysterectomy in the presence of less invasive surgical interventions and pharmacotherapy might represent a diagnostic delay necessitating a more invasive treatment approach; future research should explore this further.
Most newly diagnosed endometriosis patients have incurred pharmacy claims for pain management, and an exceedingly high proportion of our sample of endometriosis patients, nearly 90%, had at least one prescription for opioids. The use of opioids for pain management for endometriosis warrants further investigation.
A notable strength of our study is that it included a large cohort of endometriosis and matched non-endometriosis patients to evaluate the direct and indirect healthcare costs. Additionally, this study provides a current cost trend for endometriosis by using data from a more contemporary time frame than previously available.
This study has several limitations inherent to administrative claims data. As patients were identified and variables were captured through administrative claims data, endometriosis-related symptoms, which are not recorded as a diagnosis on a medical claim, may not have been captured. The diagnosis of endometriosis and clinical characteristics were determined using ICD-9-CM diagnosis or procedure codes, which are subject to data coding limitations and data entry error, and it is possible that women with endometriosis may have inadvertently been included in the control group including those with symptoms but for whom a diagnosis had not been made yet or those for whom a diagnosis had not been recorded. The use of over-the-counter medications and other self-management techniques is not usually captured in the claims data, so the use of over-the-counter NSAIDs or other pain management medications is not included in our analysis. The indirect costs included in this study are limited to absenteeism and disability claims by the employee and do not include reduced presenteeism or potential caregiver burden, as well as any other humanistic burden associated with this chronic condition. The absence costs themselves may be conservative, as the disruption of missing workers may result in effectively more work loss than is reflected in each patient’s missing hours [
24]. Endometriosis may have important effects, such as on the psychological well-being of patients [
25‐
27], that fall outside of the scope of the data used in this analysis. Finally, findings from this observational study may be prone to bias from non-random selection into the treatment group and was limited to endometriosis patients covered by Commercial health plans. Therefore, the results are not generalizable to those covered under other types of insurance or who lack coverage.