National trends in inpatient endometriosis admissions: Patients, procedures and outcomes, 2006−2015

Introduction Despite guidance towards minimally invasive, outpatient procedures for endometriosis, many patients nonetheless receive inpatient care. Our objective was to assess trends in patient and hospital characteristics, surgical complications and hospital charges for women with an endometriosis-related inpatient admission in the United States. Methods We conducted a pooled cross-sectional analysis of Healthcare Cost and Utilization Project Nationwide Inpatient Sample data. Visits were stratified into three time-period-defined cohorts (2006–2007, 2010–2011, and 2014 through the first three quarters of 2015). Visits were included if the patient was aged 18–49 years and the primary diagnosis code was for endometriosis (International Classification of Diseases, 9th Revision code 617.xx). We evaluated counts of inpatient admissions and rates of patient and hospital characteristics. Results The number of inpatient admissions with a primary diagnosis code for endometriosis decreased by 72.8% from 2006 to 2015. At the same time, among those admitted for inpatient care for endometriosis, the proportions who had Medicaid insurance and multiple documented comorbidities increased. From 2006 to 2015, mean total hospital charges increased by 75% to $39,662 in 2015 US dollars, although average length of stay increased by <1 day. Conclusions The number of inpatient admissions with a primary diagnosis of endometriosis decreased over the past decade, while surgical complications and associated hospital charges increased. The share of patients with multiple comorbidities increased and an increasing proportion of inpatient endometriosis admissions were covered by Medicaid and occurred at urban teaching hospitals. These findings suggest a demographic shift in patients receiving inpatient care for endometriosis towards more complex, vulnerable patients.


Introduction
Endometriosis is a chronic inflammatory condition characterized by hormonally responsive endometrial-like tissue that exists outside of the uterus [1]. It affects 6−10% of reproductiveage women and has a considerable impact on health-related quality of life [2][3][4][5]. Endometriosis can seriously impact mental health in a variety of ways, including not only the stress of dealing with its most common symptoms (chronic pain and infertility), but also the stigma and social implications of the disease [6][7][8]. The pathogenesis of endometriosis remains enigmatic but certainly includes immunologic and hormonal factors, including oxidative stress, and genetic susceptibility and epigenetic modifications [6,[9][10][11].
First-line therapies for endometriosis are usually medical management using non-steroidal anti-inflammatory drugs and estrogen-progestin combination medications [12]. Other pharmacologic treatment options include progestins, anti-progestins, gonadotropin-receptor hormone agonists or antagonists, and aromatase inhibitors [13,14]. Surgical options most often include laparoscopic excision or ablation of lesions, hysterectomy, and/or oophorectomy [14]. Choice of surgical approach is influenced by patient preferences, family planning status, and disease severity [14]. Often a combination of medical and surgical management is necessary [14,15].
Minimally invasive surgical options have been broadly recommended to reduce the burden on patients and the healthcare system [14][15][16][17]. Slow but steady response to this recommendation is evident; for example, the share of hysterectomies performed laparoscopically increased from 10% in 1997 to 29% in 2010 and 43% in 2013 [18][19][20]. A recent study of the American College of Surgeons National Surgical Quality Improvement Program database found that 30% of recorded hysterectomies for benign indications in 2015 were abdominal, while 16% were vaginal and 55% were laparoscopic [21]. The move towards minimally invasive gynecologic surgery has led to a greater proportion of outpatient procedures, shorter hospital stays, fewer complications, and faster recovery for patients [22], though inpatient surgery remains necessary for some patients.
Despite evidence that the number of inpatient gynecological procedures performed in the United States has decreased in recent years, little is known about the patients still receiving inpatient care for endometriosis. Accordingly, we evaluated patterns of inpatient care for endometriosis using a recent, nationally representative sample of inpatient admissions in the United States. We analyzed trends in resource use, outcomes, and patient characteristics to obtain a more complete picture of inpatient care even as treatment recommendations have shifted towards outpatient management.

Data source
This pooled cross-sectional study used data from the Health Care Utilization Project National Inpatient Sample (NIS) collected by the Agency for Healthcare Research and Quality. The NIS is a publicly available data source that contains a random sample of discharges from all non- federal hospitals in the United States. It captures approximately 20% of all community-hospital discharges in the US and when weighted represents an estimated 35 million hospitalizations annually [23]. It relies on administrative billing records to collect information on patient demographic and insurance characteristics, geographic location, International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes, total charges, and hospital characteristics. NIS data were collected based on a complex sampling design. Prior to 2012, annual stratified random samples of hospitals were identified, and 100% of discharges were collected from these hospitals. In 2012, the NIS design changed to sampling at the discharge level [24]. NIS data are publicly available and completely de-identified. Thus, their use does not constitute Human Subjects Research under the United States Federal Policy for the Protection of Human Subjects (the Common Rule) 45 CFR part 46, and consideration of this study by an Institutional Review Board is not required.

Sample
We obtained data from three mutually exclusive time periods that defined our cohorts: [2006][2007][2010][2011], and 2014 through the third quarter of 2015, after which the NIS switched from ICD-9-CM to International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes. Inpatient admissions were identified with a principal diagnosis code of endometriosis (ICD-9-CM codes 617.xx). Women were excluded if they were younger than 18 years or older than 49 years old at admission, or if they had a diagnosis code for malignant neoplasms of female genital organs (ICD-9-CM codes 179 and 180.0-184.9), to increase the likelihood that observed procedures were being performed strictly for endometriosis management.

Variables
Variables of interest included patient and hospital characteristics for each time-period-defined cohort. We also assessed the prevalence of 29 Elixhauser chronic comorbid conditions, which are commonly used to measure disease burden and case mix in hospital discharge data [25]. We calculated the prevalence of seven surgical procedures commonly associated with endometriosis management across time-period-defined cohorts: hysterectomy, laparoscopy, laparotomy, oophorectomy, bladder interventions, salpingectomy, and other excisions/ablations (see S1 Table for a list of procedures and their codes). Finally, we evaluated trends in surgical complications (see S2 Table for a list of complications and their codes), length of inpatient hospital stay, and total hospital charges. Total charges were inflated to 2015 US dollars using the Personal Health Care Price Index for Hospital Care [26].

Analyses
Descriptive statistics for each time-period-defined cohort were calculated as counts and percentages for categorical variables and means for continuous variables. Unadjusted proportions and means were compared across time-period-defined cohorts using Pearson chi-squared tests or ANOVA tests. Missing values were reported as a separate category where relevant. All analyses were weighted to provide nationally representative estimates. For data prior to 2012, we used the NIS sample trend weights [27], whereas for subsequent years we used the crosssectional weights. The weighting produced a large sample size in addition to nationally representative estimates; thus, results should be interpreted in terms of clinical as well as statistical significance.

Results
For women with endometriosis as identified by an ICD-9-CM diagnosis code, the final sample consisted of 101,733 inpatient admissions for the 2006-07 cohort, 60,080 for the 2010-11 cohort, and 27,630 for the 2014-15 cohort (Fig 1).
Patient and hospital characteristics are presented by time-period-defined cohort in Table 1.
Overall, the number of inpatient admissions for endometriosis decreased over the 10-year period studied and did so at a faster rate than inpatient admissions for all conditions, which were essentially flat at between 35 and 38 million admissions (Fig 2). There were also notable changes in the composition of patient and hospital characteristics of inpatient admissions for endometriosis. There were also changes in the characteristics of hospitals at which inpatient admissions for endometriosis occurred. In 2014-15, most patients were admitted to urban-teaching hospitals (62.7% in 2014-15 vs. 40.6% in 2006-07 and 43.0% in 2010-11; p<0.001). Table 2 describes the prevalence of Elixhauser comorbidities within the three time-period-defined cohorts.  There were also changes in the frequency of endometriosis-associated surgical procedures across the decade ( Table 3). The proportion of inpatient admissions that involved a hysterectomy changed from 78.2% in 2006-07 to 67.4% in 2014-15 (p<0.001). The prevalence of   Table 4 reports trends in surgical complications by surgery type, length of inpatient stay, and total hospital charges. The prevalence of complications increased over time following hysterectomy, laparoscopy, and oophorectomy. While length of inpatient stay increased by <1

Principal findings
During 2006-2015 the number of inpatient admissions for endometriosis decreased. Although hysterectomy and oophorectomy remained the predominant inpatient surgical procedures observed, their prevalence declined slightly. The prevalence of surgical complications increased overall and by procedure type. Average length of stay increased slightly, and total hospital charges increased considerably. The composition of women being admitted for endometriosis changed over the same period. Higher proportions of women in later cohorts had multiple comorbidities, were covered by Medicaid as opposed to commercial insurance, and lived in lower-income ZIP Codes. This raises the possibility that, as inpatient admissions for endometriosis have become less common, women who are admitted for endometriosis have fewer resources and are at higher risk than women who receive care in other, presumably outpatient, settings.
Prior research suggests women with fewer resources and more comorbidities are more likely to receive a hysterectomy [28]. Further, there is compelling evidence that patients with higher income, private insurance, and white race are more likely to undergo laparoscopic hysterectomy versus open hysterectomy [19,21,29]. These trends are increasingly visible in inpatient endometriosis care. This is especially concerning as endometriosis-associated symptoms can recur following hysterectomy [30]; thus, vulnerable patients may incur higher risks for procedures that do not confer higher benefits.
Increased complications and patient comorbidities may have helped drive the increases in hospital charges. Between 2003 and 2014 the percentage of inpatient admissions with multiple chronic conditions increased from 64% to 78%, with the fastest growth among patients aged 18-44 years [31]. This suggests that patient complexity continues to be an important consideration when evaluating inpatient care, especially regarding associated resource use and costs.

Clinical implications
Our study draws attention to an important population of patients: those who continued to receive inpatient care for endometriosis even as management guidelines emphasized minimally invasive care. It is important to monitor this population of women to ensure that women who receive inpatient treatment are only those who are indicated for it, to guard against disparities based on non-clinical factors such as income, education, geographic location, or race/ethnicity. It is also important to assess the population currently requiring inpatient care to inform efforts to reduce the need for inpatient care in the future, indeed even among women for whom it may be indicated today.
Additionally, there are important economic implications to our study that are relevant to clinical care. Given that the associated mean total charges per stay increased, inpatient careparticularly surgical care-remains an important contributor to the overall economic burden of endometriosis. Total costs of surgery for endometriosis are high and include complications, medical management during recovery and follow-up, and the possibility of retreatment in addition to the cost of the procedure [32,33]. Endometriosis-related surgery is also associated with considerable indirect costs in the form of increased absenteeism and presents a significant economic burden to society [34]. Inpatient costs, and the economic burden of endometriosis overall, may continue to rise if vulnerable patients, including those with multiple comorbidities, are preferentially directed towards invasive procedures requiring inpatient care.

Research implications
Our findings add an important and objective assessment of national trends in the still sizeable number of inpatient procedures that occur for endometriosis even as guidelines recommend outpatient care settings when possible. Current outpatient medical management of endometriosis may also affect inpatient care and deserves further study. As the number of inpatient admissions related to endometriosis decreased over time, patients who received inpatient care for endometriosis were increasingly at higher risk and more disadvantaged. This may be indicative of the lack of access of vulnerable populations to timely and effective medical management or outpatient surgical care, though this would need to be confirmed with a corresponding analysis of outpatient data. Our study also provides an important characterization of the comorbidities associated with patients who are receiving inpatient care for endometriosis, which could be important information for hospitals seeking to improve inpatient endometriosis care. Finally, we observed contemporaneous increases in hospital charges and length of stay as well as surgical complications. The growth in hospital charges over 2006-2015 continues an earlier rising trend reported in an analysis of HCUP NIS data for 1993-2002; in contrast, while declining from 1993 to 2000, length of stay has not changed much since then [35]. Limitations in the currently available data preclude us from going beyond describing the national trends; future research is needed to explore potential explanations.

Strengths and limitations
Our study used the NIS database, which consists of a random sample of non-federal hospitals that, when weighted, produces a nationally representative estimate of US inpatient admissions. The NIS database has several limitations, however. It includes only individuals who received inpatient healthcare services. Also, it does not follow individual women over time or cover healthcare use outside of inpatient hospitalizations. Thus, the full scope of endometriosisrelated care, including outpatient procedures, is not captured. The study period ended in 2015, and practice patterns may have changed since then, although many of the changes observed were modest across the decade captured in this study. Only the first three quarters of 2015 were analyzed because of the change to ICD-10-CM coding and the associated risk of diagnosis and procedure misclassification [36].
Moreover, the NIS database, as is true of all large ICD-based data capture, is limited because it is based on administrative data that are collected primarily for billing and reimbursement. It offers limited detail about clinical decision-making, patient complexity, and other important aspects of the clinical encounter. The NIS data may suffer from coding inaccuracies as well as inconsistencies across hospitals. There are likely missed endometriosis-related procedures, particularly when endometriosis was present but not coded. It is also possible that patients are being incorrectly diagnosed with endometriosis, particularly if diagnoses are based on clinical symptoms. These coding patterns can vary informatively by clinician specialty and patient and hospital characteristics. In addition, the NIS reports hospital charges, which represent neither reimbursed amounts nor the costs faced by hospitals.
Of particular importance to the study of endometriosis, the NIS does not contain information about endometriosis stage nor certain details about the surgical procedures performed in terms of the intensity of dissection required. Unfortunately, given a lack of standardization of surgical documentation, it is currently not possible to abstract this information from any form of medical records across multiple practitioners within or among surgical sites [37]. All forms of endometriosis are challenging to treat; however, surgical complexity and requirement for advanced expertise is particularly high for deep endometriosis. Techniques such as superficial and deep rectal shaving and nerve sparing intestinal dissection are important considerations [38][39][40][41]. Future studies will require standardized collection of these endometriosis phenotypic presentation and surgical approach details to parse out adequately whether there has been shift over time in inpatient procedures toward women with more surgically complex endometriosis presentation that may have previously been mis-or undiagnosed.
An obvious restriction in scope of our analysis is that it examines endometriosis surgical treatment only in the inpatient setting; the inpatient setting is the focus of our review to allow characterization of this population. These findings should not be generalized to those receiving care in the outpatient surgery setting. Endometriosis treatment recommendations have focused on increasing the use of outpatient surgery (for example, laparoscopic lesion ablation/ excision, ovarian cystectomy for endometriomas, and lysis of adhesions [42]), and recent data suggest that outpatient surgery has become commonplace. For 29 states with administrative data on all inpatient admissions and ambulatory surgery visits, 78.0% of endometriosis cases in 2014 were treated in an ambulatory surgery setting [43]. Unfortunately, there is no nationally representative database of outpatient visits corresponding to the HCUP NIS that we could use to assess treatment patterns overall and compare trends in treatment patterns between care settings. There are several possible explanations for our observed trends in inpatient management, including case selection for inpatient surgery to those with more complex presentation, associated comorbidities that do not allow the patient to receive care at an outpatient surgery center (e.g., body mass index), or changes in documentation such as altered ICD code attribution. Because there is no corresponding outpatient national dataset available, we are unable to explore which factors may be contributing to our observations.

Conclusions
Surgery is an important clinical option for endometriosis, and it remains central to classifying the disease [37]. However, minimally invasive approaches should be prioritized. Because of its serious personal and societal implications, tracking clinical outcomes, economic burden and health disparities associated with endometriosis is critical to ensure that all women have access to appropriate treatment.
Supporting information S1 Table. Surgical procedures and corresponding ICD-9 diagnosis and procedure codes included in the analysis. (DOCX) S2 Table. Surgical complications and corresponding ICD-9 diagnosis and procedure codes included in the analysis. (DOCX)