Validity of Self-reported Endometriosis

A Comparison Across Four Cohorts

A.L. Shafrir; L.A. Wise; J.R. Palmer; Z.O. Shuaib; L.M. Katuska; P. Vinayak; M. Kvaskoff; K.L. Terry; S.A. Missmer


Hum Reprod. 2021;36(5):1268-1278. 

In This Article


Although large sample sizes of women with endometriosis are needed in order to fully investigate the heterogeneity of endometriosis in relation to risk factors and response to treatments, the majority of large studies rely on self-report of endometriosis diagnosis from self-administered questionnaires, the accuracy of which has not been fully investigated. Our results indicated that women report endometriosis fairly well, with an overall confirmation proportion of 84% (ranging from 72% to 95% among cohorts), indicating that if a participant reported being told by a physician that she had endometriosis, she reliably reported that information. Women who reported that their endometriosis diagnosis was confirmed by laparoscopic surgery had an exceptionally high overall confirmation proportion of over 94%. Confirmation of endometriosis diagnosis did not substantially vary by infertility status at the time of endometriosis self-report. Information on pathology results, AFS or rASRM stage, and the presence or absence of deep endometriosis or endometrioma(s) were often absent from the medical records reviewed.

For the purposes of the present study, the overarching goal was to determine validity of endometriosis case status via self-administered questionnaire in large, diverse populations. Thus, although there is debate about the requirement of surgical visualization with and without pathology confirmation for 'true' endometriosis diagnosis (Agarwal et al., 2019), we deemed that if a physician had noted in the clinical record that this patient likely had endometriosis, it is reasonable to believe that, when asked, that woman would indeed report that she had been told by a physician that she had endometriosis.

Previous studies on the validity of self-reported endometriosis have observed varying confirmation proportions, ranging from 32% to 89% (Treloar et al., 2000; Missmer et al., 2004; Saha et al., 2015, 2017). These variations in confirmation proportions could be due to: differences in the definition of a confirmed case (e.g. surgical with and without pathologic confirmation versus clinical); different population groups (e.g. health professionals versus general population); differences in methods for confirming endometriosis (e.g. registries versus medical records); and differences in how questions about self-reported endometriosis were asked (e.g. confirmation by laparoscopic surgery). Saha et al., (2015) reported a confirmation proportion for endometriosis of 82%, which included surgical, histologic and/or clinical medical record evidence within two large Swedish twin cohorts. Within the same twin cohorts, Saha et al., (2017) observed a confirmation proportion of only 32% when comparing self-reported endometriosis to Swedish inpatient registry data. This discrepancy highlights that the choice of the gold standard confirming endometriosis and the definition of a confirmed endometriosis case can dramatically alter findings.

Within our study, we observed a fairly wide range of overall confirmation proportions, potentially due to differences in the populations of the four cohort studies. BWHS, a cohort of women from the general US population (aged 21–69 years at enrollment), had the lowest overall confirmation proportion (72%), whereas GUTS, a cohort of children of registered nurses (aged 9–17 years at enrollment), had the highest overall confirmation proportion (95%). The overall confirmation proportion in the E3N, a cohort of women from the general French population (aged 39–66 years at enrollment), was fairly high (82%). This may be due to differences in access to medical records between the US and French populations, as French participants possessed and provided their own medical records, whereas almost all US participant records had to be obtained from a healthcare provider. These results suggest that consideration of the accuracy of self-reported endometriosis should depend on the context/population within which the study is conducted, and therefore should inform the level of detail and format for the wording of questions to maximize validity.

We additionally aimed to capture information on histologic confirmation. Often reviewers and readers of endometriosis scientific papers assume that histologic information is or should be readily available. This study reinforces that that is not the case in the USA. Less than half of the GUTS and NHSII 1st and second wave participants had a relevant pathology report from which we could assess histologic confirmation. For those who did have a pathology report, additional challenges arose in abstracting information on pathology findings including: most often only a single index lesion was sent to pathology; and lack of evidence of assessment for endometriosis by the pathologist, particularly during a hysterectomy, for an indication other than endometriosis. For example, one NHSII second wave participant had surgery with a suspected endometrioma sent to pathology, which was determined to be an inclusion cyst by the pathologist. However, the surgical report documented visualization of additional lesions believed to be superficial peritoneal endometriotic implants—none of which were sent to pathology. Thus, the participant must remain—perhaps erroneously—classified as surgically but not histologically confirmed. With the paucity of pathology reports and the lack of physicians sending appropriate lesion biopsies to pathologists, caution should be used when abstracting histologic confirmation from medical records and by overemphasis of diagnostic criteria that are not widely standard of clinical care.

Although we found that women self-reported their endometriosis diagnosis fairly well, there are ways in which self-report could be improved. Among some of the cohorts studied, approximately 10–17% of participants contacted denied their initial self-report upon re-contact. It may have been that at the time of completing a given questionnaire, they believed they had endometriosis and later learned this was not the case, or it could simply be that an incorrect box was checked on that questionnaire. Therefore, re-contacting the participant to confirm self-report could help to improve the accuracy of initial self-reports. Additionally, the overall confirmation proportion was higher among women reporting laparoscopic confirmation of their endometriosis diagnosis compared with those not reporting laparoscopic confirmation. This may be driven by the number and clarity of discussions between a patient and practitioner when choosing to undergo laparoscopy and subsequently reviewing the results. The addition of a question about whether an endometriosis diagnosis was confirmed by laparoscopic surgery would help to subset participants with the most accurate reporting of endometriosis.

Information on AFS or rASRM stage was highly variable in medical records from NHSII 1st and second waves and GUTS. Although 44% of NHSII first wave medical records included information on stage, only 13% of NHSII second wave and 24% of GUTS included stage information. The majority of reports indicated stage either numerically (e.g. Stage III) or qualitatively (e.g. mild-to-moderate stage) with very few reports listing out the actual score. The presence or absence of deep endometriosis was also rarely documented. Interestingly, among surgically confirmed cases, a higher proportion of GUTS (6%) as compared to NHSII second wave (3%) medical records reported deep endometriosis; the opposite of what we would expect, given the age distribution of the two cohorts. However, the GUTS cases were evaluated for endometriosis during a time period (1996–2016) when surgeons and radiologists or other imaging specialists have been more thoroughly assessing for deep endometriosis as compared to the time period when NHSII second wave cases (1972-2008) were diagnosed. These results highlight the existing limitations of using medical records to abstract information on surgical information relevant for subgrouping endometriosis cases.

The presence of an endometrioma was reported in almost 30% of NHSII second wave medical records; aligning with the expected 20–40% of endometriosis patients who will have at least one endometrioma. However, the presence of an endometrioma was reported in only 8% of GUTS medical records, potentially due to the younger age at diagnosis for GUTS compared to NHSII participants. Accurate reporting of endometriomas in medical records is crucial for appropriate subgrouping of endometriosis patients as the presence of an endometrioma contributes to the rASRM categorization of stage III or IV disease, which have been more strongly associated with genomic loci compared to stage I and II disease (Sapkota et al., 2015). Additionally, endometriomas may be an important subgroup potentially contributing to the risk of ovarian cancer among women with endometriosis (Saavalainen et al., 2018), and lack of information on the endometriosis subtypes will make it impossible to rigorously explore these associations.

The lack of information on histologic confirmation, AFS or rASRM stage, and endometriosis macro-phenotypes calls into question the feasibility of using medical records for abstracting subtyping information and may yield biased data when analyses are restricted to only those records that do include this information. Furthermore, the lack of consistency in the information documented additionally hampers discovery. Without required, standardized documentation tools, medical records are subject to recording bias due to: time pressures placed on physicians limiting the amount of information they record; variations in how well physicians document the full medical history of their patients; and correlation with the intensity of specialization or personal interest of the physician (Luck et al., 2000). Luck et al., (2000) used patient 'actors' with four different medical conditions to assess how well the quality of care received by the patients was reflected in the medical records. Using standardized checklists filled out by the patients (gold standard) and chart abstraction of visits with the physicians, the authors found a sensitivity of only 70% and specificity of 81% for reporting of absolutely necessary care within the medical records. The authors concluded that medical records are neither sensitive nor specific. Therefore, for information to be reliably abstracted from medical records, particularly about histologic findings, rASRM stage and endometriosis lesion characteristics, adoption of harmonized surgical and pathology reporting is imperative. For endometriosis, surgical documentation tools such as those developed by the World Endometriosis Research Foundation (WERF) Endometriosis Phenome and Biobanking Harmonization Project (EPHect) are essential (Becker et al., 2014). For cancer discovery, consistent accuracy and detail, including tumor and patient characteristics, has been critical to informative sub-phenotyping and staging (Reis-Filho and Pusztai, 2011; Vaughan et al., 2011). This must occur for women with endometriosis.

The present study is the largest to evaluate the validity of self-reported endometriosis; it includes four large cohort studies with a diverse range of age and population groups including both general population and healthcare providers. Additionally, for GUTS, NHSII and BWHS, we had detailed information on confirmation by laparoscopic surgery, and within GUTS and NHSII, we had information on reporting of AFS or rASRM stage and the presence or absence of endometriomas or deep endometriosis. The medical record abstraction was conducted separately among the four cohort studies, potentially leading to differences in classification of participants. However, we standardized the information abstracted from the medical records as much as possible to minimize this effect. Additionally, we did not have complete information for all the cohorts. We were missing information on histologic confirmation for E3N and BWHS and on reporting of AFS/rASRM stage and endometriosis macro phenotype for BWHS and E3N. However, we do not expect to have seen large differences in the availability of stage and endometriosis macro phenotype information in the BWHS and E3N given that these data were absent from the majority of the records from NHSII and GUTS. The response for the BWHS validation study was low (20%), probably because the BWHS sought records from all women who reported endometriosis instead of restricting to a random sample of women who had recently responded to questionnaires, as was done in the other cohorts.