Endometriosis treatment pathways in the largest private health insurance in Brazil: A real-world data study

Endometriosis is a gynecological condition characterized by the ectopic growth of endometrial-like tissue outside the uterine cavity, including the peritoneum, mesentery, ovaries, and fallopian tubes [ 1 ]. In some cases, the disease deeply infiltrates pelvic organs such as the bowel, bladder, and ureters, characterizing deep endometriosis, which represents the most severe form of the disease. Deep infiltrating endometriosis (DIE) accounts for approximately 20% of all cases and is considered the most severe form of the disease, with peritoneal involvement reported in up to 44% of patients [ 2 ]. Affecting approximately 6% to 10% of women of reproductive age, endometriosis ranks among the most prevalent chronic gynecological diseases [ 3 ]. The hallmark symptom is cyclic pelvic pain, often accompanied by dysmenorrhea, deep dyspareunia, dysuria, and dyschezia, depending on the organs involved [ 4 ]. Although these symptoms are often inappropriately normalized and overlooked due to their chronic nature, they significantly impair quality of life and frequently contribute to psychological conditions, including anxiety and depression [ 5 ]. Furthermore, up to 50% of infertile women are diagnosed with endometriosis, underlining its profound impact on hormonal function and fertility [ 6 ]. Despite its histological similarities to ectopic endometrium, ectopic lesions exhibit invasive behavior and angiogenic activity, complicating diagnosis and treatment [ 7 ]. Surgical treatment is indicated in patients with symptomatic endometriosis refractory to medical therapy, in cases of organ involvement or damage, and in selected patients with persistent infertility [ 8 ]. Although surgery may significantly improve symptoms and fertility outcomes, it does not eliminate the underlying disease process, and recurrence remains a relevant clinical concern.[ 9 ]. Real-world evidence (RWE) has emerged as a complementary paradigm to traditional randomized controlled trials (RCTs), offering insights derived from real-world data (RWD) collected in clinical practice [ 10 ]. Unlike RCTs, which are designed for high internal validity under controlled conditions, RWE captures a broader spectrum of patient experiences, allowing for the examination of outcomes in more diverse populations and real-world settings [ 11 ]. This integrative approach not only enhances the applicability of clinical findings but also enables the study of rare diseases and long-term outcomes. By leveraging RWE, healthcare systems can build learning environments where research is seamlessly integrated with clinical care, addressing gaps left by traditional methodologies and expanding the evidence base for conditions like endometriosis [ 12 ]. Studying the patient journey in endometriosis is particularly crucial due to the gaps in understanding key timelines, such as the interval from symptom onset to diagnosis, treatment, and surgical intervention. Notably, the time required for a definitive diagnosis is estimated to take an average of 7 years, highlighting the challenges patients face in receiving timely and accurate care [ 13 ]. These delays often exacerbate disease progression and psychological burden, yet data on these critical periods remain scarce in the literature. It is also important to highlight the substantial costs incurred by patients throughout this journey, from the prolonged diagnostic process to the initiation of appropriate treatment. These costs include both direct expenses, such as consultations, diagnostic exams, and surgical interventions, and indirect costs, such as lost productivity, absenteeism, and the overall impact on quality of life [ 14 ]. Despite advances in imaging techniques such as transvaginal ultrasound with bowel preparation and magnetic resonance imaging, the diagnosis of endometriosis—particularly deep endometriosis—remains challenging. Variability in imaging expertise, limited access to specialized centers, and the absence of universally adopted classification systems contribute to persistent diagnostic delays. This study aims to address these knowledge gaps by exploring real-world patient experiences with endometriosis, providing valuable insights into diagnostic and therapeutic timelines. By shedding light on these underexplored aspects, we hope to inform clinical practices and improve outcomes for women suffering from this complex condition.

In our cohort, we identified 5740 women that received surgical treatment for endometriosis from 2006 to 2024. Most patients are from the Northeast Region of Brazil ( Fig 1 ), but there are patients in all five regions in Brazil (North, Northeast, Center-West, Southeast and South). Median age for surgery was 37 (32–42) years-old, with range from 14 to 59 years-old. There were 10 deaths in the cohort period (0,17%). Only one patient had the cause of death endometriosis related but died 67 months after index event. There were 19,714 people-year before index event and 9,574 people-year after index event. The median time of each patient on the insurance company with an active plan was 3.78 (2.28–5.28) years before index event and 1.58 (0.49–2.87) after surgery (p < 0.001). Infertility and pelvic/abdominal pain were the first assessed symptoms to manifest. Abdominal pain, with infertility reported on outpatient consultations about 19(9–36) months prior to surgery and pelvic or abdominal pain being reported about 18(8–37) months prior to surgery ( Fig 2 ). Dyspareunia was a symptom that appeared significantly later than the others (p < 0.001), with its onset about 13 (6–27) months prior to surgery ( Fig 2 ). Patients with deep endometriosis took more time to undergo surgery from endometriosis than patients with superficial endometriosis ( Fig 3 ). a. Comparing the onset of pelvic/abdominal pain in our cohort. Median time for superficial endometriosis was 12 months, and 20 months for deep endometriosis, HR 1.441 (IC 95%, 1,303 − 1,595, p < 0.001). b. Comparing the onset of infertility in our cohort. Median time for superficial endometriosis was 15.21 months, and 20.38 months for deep endometriosis, HR 1.206 (IC 95%, 1,033 to 1,408, p < 0.001). c. Comparing the onset of dyspareunia in our cohort. Median time for superficial endometriosis was 8 months, and 14 months for deep endometriosis, HR 1.455 (IC 95%, 1,231 to 1,718, p < 0.001). After surgery all symptoms had a significant improvement (p < 0.001, Table 1 ). Outpatient consultations were discriminated by medical specialty on Table 2 -a. The incidence of outpatient consultations in gynecology diminished from 1.64 to 1.42 consultations/people-year, meanwhile the number of outpatient consultations in obstetrics increased from 2.16 to 2.64 consultations/people-year. This behavior was found in deep endometriosis, where gynecology outpatient consultations decreased from 1.97 to 1.70. Obstetrics outpatient consultations increased from 2.50 to 3.01 consultations/people-year. In superficial endometriosis we found the same increase in outpatient consultations in obstetrics, but we also observed a modest numerical increase in gynecology outpatient consultation rates. Regarding physiotherapy outpatient consultations there were only 208 consultations (1055 consultations/ 100.000 people-year) before index event and 64 consultations (668 consultations/ 100.000 people-year) after index event. Psychology consultations were 6013 (30,501/ 100,000 people-year) before index event and 3,520 (3676/ 100,000 people-year) after index event. The incidence of ER consultations ( Table 2 -b) in gynecology decrease from 0.246 to 0.170 consultations/people-year and increase in obstetrics from 0.258 to 0.811 consultations/people-year. The incidence of hospital admissions ( Table 2 -c) for labor increased from 0.27 to 0.44 admissions/people-year. This trend was found in deep and superficial endometriosis. Regarding the index event, median hospital stay was 1 (1–1) day, with the minimal hospital stay of 1 day AND maximum of 27 days. Sixty-six (1.1%) patients were sent to ICU after surgery, with a median ICU stay of 1.62 (0.62–2.62 days), minimal ICU stay of 1 day and maximum stay of 13 days. Before index event, patients had a median number of 1(0.5–1.5) procedures. Most common procedures ( Table 3 ) were hysteroscopy, Implantation of Hormonal Intra-Uterine Device and Laparoscopic Cholecystectomy. The rate of Cesarean Section was 557.97/100.000 people-year. Normal Delivery rate was 116.66/100.000 people-year. After index event, patients had a median number of 1(0.5–1.5) procedures. Most common procedures ( Table 4 ) were hysteroscopy, cesarean section and Laparoscopic surgical treatment of endometriosis. The rate of Cesarean Section was 1984.541/100.000 people-year. Delivery rate was 470.023/100.000 people-year. Two hundred and thirty-nine (4.1%) patients underwent one additional surgery for endometriosis and 3 patients underwent two additional surgeries for endometriosis in our cohort. Findings at index event and operation for recurrence were similar (p = 0.901, Table 5 ). Five cases (0.08%) had elective reoperations within 30 days for findings in the operating room where more complex than preoperative findings and surgical team did not have specialists available for intestinal or ureteral resections. From the 5740 patients, 93 (1.6%) had urgent reoperations within 30 days from index event. Procedures are discriminated in Table 6 . There were 48 cases of reoperation for abdominal obstruction by adherences, 22 cases of persistent urine leakage, 14 fistulas, 6 rectal perforations and 3 ileal perforations. Deep endometriosis had more urgent reoperations and the most severe causes for reoperations in out cohort (p < 0.001). Regarding ER visits ( Table 7 ), the most frequent diagnosis was abdominal and pelvic pain, accounting for 28.34% of consultations (917 cases), followed by encounter for other postprocedural aftercare, representing 21.72% (703 cases), mainly for drain and stitches removal. Other notable diagnoses included endometriosis (5.07%, 164 cases), other disorders of the urinary system (3.34%, 108 cases), and complications of procedures, not elsewhere classified (3.03%, 98 cases). Less common diagnoses included headache (1.92%, 62 cases), dorsalgia (1.89%, 61 cases), other abnormal uterine and vaginal bleeding (1.85%, 60 cases), and cystitis (1.73%, 56 cases). Additionally, 24.65% of consultations (798 cases) fell under the category “others,” while 6.46% (209 cases) had no specific information available ( Table 8 ). The MRI findings show a significant reduction in cases of deep endometriosis (from 609 to 124) and superficial endometriosis (from 505 to 133) in the post-index period, alongside a proportional increase in cases classified as no endometriosis (from 709 to 472) (p < 0.001, Table 9 ). When compared to surgical findings, MRI demonstrated a sensitivity of 63.7%, correctly detecting endometriosis cases, and a specificity of 100%, accurately identifying all cases without endometriosis without generating false positives. The median cost per endometriosis surgery was 22.8% in hospital admission costs, 47.2% in medical team costs and 30% in medications and materials ( Table 10 ). Cases with deep endometriosis had significant more expensive costs in all parameters (p  0.999). The median cost on exams, including laboratory tests and imaging exams per patient was equivalent to 22% of the overall (deep + superficial) endometriosis surgery cost. When stratified by severity of endometriosis, it accounts for 19.3% of the costs in superficial cases, versus 23.9% in deep endometriosis cases (p < 0.001). The median cost per ER visit was divided in 34.2% in unit admission costs, 40.02% in medical team costs and 25% in medications and materials. Patients’ with deep endometriosis had more expensive hospital admission costs in ER (p < 0.001), but similar costs with medical team, medications and materials. The total ER visit was equivalent to 1.6% of a endometriosis surgery cost ( Table 11 ).

Our study underscores the significant impact of surgical interventions on improving patient symptoms and reducing the economic burden associated endometriosis. Our findings advocate for the implementation of more effective diagnostic and management strategies to enhance patient care. Moreover, the insights gained from this large-scale study contribute to a better understanding of the real-world dynamics of endometriosis treatment, supporting the development of policies aimed at improving the quality of life for affected women and optimizing healthcare resources.

This is a population-based, retrospective cohort study, considering the Hapvida Notredame portfolio, which consists of over 8.8 million lives, which accounts for about 4.7% of Brazil population. Hence, this is the largest sample of patients with detailed information in the country. The study received approval from the Institutional Ethics Committee and adheres to the highest standards of research practice as outlined in the Declaration of Helsinki. This project was assessed and approved by the Brazilian National Ethics on Research Comittee with approval number 7.101.673. Approval in within the submission. We designed our cohort to encompass all patients that were submitted to surgical treatment for endometriosis. The first surgical procedure for endometriosis was our index event. All information of these patients was extracted from 5 years before de index event to 3 years after the index event. The first surgical procedure for endometriosis was identified through the authorizations database. There is a national unified table of procedures that identifies all procedures from public and private health systems (TUSS table). We selected all the surgical procedures codes from this table regarding simple endometriosis, deep endometriosis (including bladder and intestinal) and peritoneal endometriosis. Open and laparoscopic approaches were considered. Robotic-assisted approach is unavailable. After these patients were identified in our datalake by their respective registration number. Our next step was to identify the period each patient was active in the health insurance company. Hence, we identified all activations and deactivations of each patient insurance policies to determine for how long each patient was active on the insurance company. The access in the datalake was performed from October 1 st 2024 to December 31st 2024. The extracted data was anonymized to prevent individual participants identification. We also collected data regarding the age of each patient in the index event, geographical distribution, MRI for endometriosis, number of outpatient consultations, number of emergency room consultations, hospital admissions, intensive care unit admissions and surgical procedures. All these events were extracted with their respective ICD-codes. Data regarding mortality and cause of death was also extracted. The analyzed surgical procedure reports detailed endometriosis findings as deep endometriosis (characterized by resection of the intestine, colon, or bladder, or explicitly described in the surgical report as deep endometriosis), superficial endometriosis (when findings did not meet the criteria for deep endometriosis), absence of endometriosis, and unavailable information. We considered surgery-related complications as ER consultations and hospital readmissions through ER up to 30 days after endometriosis surgery. Furthermore, no standardized classification system (such as rASRM or ENZIAN) was systematically applied in the surgical reports, as data were derived from routine clinical documentation within a real-world administrative database. Ethnical data was not explicitly collected for Brazil; however, as a highly multiracial country with data gathered from various regions, the study inherently reflects a diverse representation of ethnic groups. To analyze patient symptoms, we developed three AI models to extract information related to infertility, dyspareunia, and abdominal pain from patient records. For each symptom, patient records were labeled as “absent,” “present,” or “unavailable,” and a BERT-based algorithm (BioBERTpt-all) was fine-tuned to classify the information into these categories. The infertility model achieved recall performances of 94.1%, 94.4%, and 93.2% for the “absent,” “present,” and “unavailable” classes, respectively. The dyspareunia model attained recall rates of 98.5%, 89.7%, and 97.1% for the “absent,” “present,” and “unavailable” classes, respectively. Lastly, the abdominal pain model achieved recall performances of 97.1%, 96.9%, and 92.8% for the “absent,” “present,” and “unavailable” classes, respectively. Data regarding costs of each patient in the Insurance company was also extracted. Only data from 2019 to 2024 was available, hence median and interquartile costs for each treatment step were calculated for estimation. Monetary correction for inflation was performed by the IPCA index (in English, “Broad Consumer Price Index”). The analysis was performed for values of 2024. Values are presented in relative value for preserving company’s strategical information. Statistical analyses were performed using SPSS software version 29.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism version 10.0.0 for Windows (GraphPad Software, Boston, Massachusetts, USA) and a significance level of 5% was considered, meaning results with a p-value of 5% or less (p ≤ 0.05) were considered statistically significant.