Self-reported surveys may have identified more menopause observations than structured electronic health records in the All of Us Research Program, suggesting menopause may be underdocumented in clinical records.
Investigators analyzed menopause-related survey responses, electronic health record (EHR) diagnosis codes, and genomic data from the National Institutes of Health’s All of Us Research Program from about 396,000 participants recorded as female at birth.
The investigators characterized menopause-related information using concordance between survey responses and EHR diagnoses, overlap with genomic data, and age and sociodemographic patterns. Survey measures included whether menstrual periods had stopped permanently, reasons for cessation, hysterectomy history, and ovary removal. EHR measures included structured diagnostic codes for menopause, premature menopause, and related conditions.
Among the participants, 192,655 reported that their menstrual periods had permanently stopped compared with 27,975 who had an EHR diagnosis code for menopause or premature menopause. Surveys therefore identified nearly 7 times as many menopause observations as structured EHR data. Among participants with an EHR menopause diagnosis who also completed the relevant survey, more than 99% reported that their periods had permanently stopped.
The gap in EHR documentation was similar across common menopause types among participants over 60 years. Among 53,505 participants with survey responses indicating natural menopause and 18,405 reporting surgical menopause with hysterectomy and removal of both ovaries, 18% (n = 9,565 and 3,335, respectively) had an EHR menopause diagnosis. The investigators noted that the similar proportions suggested EHR underreporting did not differ substantially by menopause type.
Among participants reporting permanently stopped menstrual periods, about 56% attributed cessation to natural menopause and 31% to surgery, while smaller proportions reported medication or other therapy, endometrial ablation, or other causes. Structured EHR data contained limited documentation of menopause-related symptoms, with 20 or less observations recorded under the code for abnormal vasomotor function.
Multimodal overlap was more limited when EHR menopause data were required. Among 250,075 female participants with genomic data, 23,615 also had menopause-related EHR data and 247,110 had menopause survey data. About 9% had both EHR and survey menopause information alongside genomic data.
Among participants reporting that their periods had permanently stopped without hormone-induced cessation, 5% were under 40 years, 42% were 40 to 60 years, and 53% were over 60 years. Natural menopause without hysterectomy or ovary removal was uncommon prior to age 40 years.
EHR diagnoses showed a different age distribution, occurring primarily between 50 and 80 years and showing a pronounced increase at 65 years. The investigators examined whether Medicare coverage might explain this pattern but could not determine its cause because just a small subset of participants insurance response data.
The investigators also observed modest sociodemographic differences in age among those with menopause information. However, similar patterns occurred in the broader program's population, suggesting that some differences may reflect the cohort’s underlying age and demographic composition rather than menopause-specific patterns. The investigators cautioned that several sociodemographic strata contained small sample sizes.
The study had several limitations. Menopause status was not clinically confirmed, and neither the survey nor EHR data provided age at natural menopause. Ages used in the analysis generally reflected the time of the survey response or EHR observation rather than the menopause event itself. Survey responses were also subject to possible response and recall bias; about 4% (n = 1,725) of the participants over 70 years reported that they had not experienced menopause. Small samples in some sociodemographic groups further limited interpretation.
The findings showed that survey data provided substantially greater ascertainment of menopause than structured EHR diagnoses in All of Us Research Program, while requiring EHR data reduced the population available for multimodal analyses. The results provided a descriptive foundation for selecting data sources, defining menopause phenotypes, and planning future studies rather than establishing which source more accurately identifies menopause. “Future studies can build on this work to investigate clinical outcomes associated with menopause,” wrote lead study author Jack W. Staples, PhD, of the Department of Biomedical Informatics at the University of Colorado Anschutz, and colleagues.
Full disclosures of the study authors can be found in the study.
Source: Menopause