Development and validation of a premenstrual symptom screening tool for working women in relation to absenteeism.

Tables  1 and 2 present the characteristics of the study participants. Our participants are women workers aged 18–41 years (excluding pregnant and lactating women, n  = 375) in Japan. The average age was 32.6 years (sd = 5.3). In terms of occupational type, the majority of the women were clerical workers (49%), followed by service workers (18%) and professional/technical workers (15%). In terms of industry, 18% of the participants were in the medical and welfare sectors, 14% were in the manufacturing sector, 11% were in the service sector, and 10% were in the wholesale and retail sectors. In terms of working status, 77% were full-time workers, with a median of 8 daily working hours. In terms of work productivity, 393 people, or 12%, had been absent due to premenstrual symptoms, with 106 people (27%) being absent for at least 1 year, 26 people (7%) being absent for 1 week, and 216 people (55%) being absent for 1 day. Table 1 The characteristics of 3239 participants N or Mean % or SD Age, mean ± SD 32.6 5.3 Body Mass Index, mean ± SD (missing n  = 12) 20.5 3.2 Marital status  Single 2251 69.5  Married 988 30.5  Children  (+) 668 20.6  (-) 2571 79.4 Annual household income  > 8 million JPY 682 21.1  6–8 million JPY 563 17.4  4–6 million JPY 771 23.8  2–4 million JPY 980 30.3  < 2 million JPY 243 7.5 Education attainment  High/junior/elementary school 631 19.5  2 year-college 772 23.8  University/Graduated 1836 56.7 Occupation  Clerical workers 1591 49.1  Service workers 580 17.9  Professional and engineering workers 499 15.4  Sales workers 172 5.3  Manufacturing process workers 125 3.9  Administrative and managerial workers 52 1.6  Others 220 6.8 Industry  Medical, Health Care and Welfare 594 18.3  Manufacturing 463 14.3  Compound Services 350 10.8  Wholesale and Retail trade 312 9.6  Finance and Insurance 223 6.9  Construction 170 5.3  Education, Learning Support 153 4.7  Others 974 30.1 Gynecological underlying illness  Myoma Uteri 157 4.85  Endometriosis 96 2.96  Ovarian cyst 84 2.59  Others 56 1.73 Psychological underlying illness  Depression 9 0.28  Others 129 3.98 The characteristics of 3239 participants Table 2 The work characteristics of 3239 participants N or median % or IQR Working status  Full-time worker 2494 77.0  Part-time worker 548 16.9  Self-employed 197 6.1 Workplace size   1000 workers 789 24.4 Labor characteristics  Daily average hours of working, median (IQR) 8 (7–8)  Weekly average hours of working, median (IQR) 40 (30–42)  Daily average hours of standing, median (IQR) 2 (1–5)  Number of carrying heavy objects, N (%) 980 30.3  Number of night shifts in previous month, N (%) 340 10.5 Copenhagen Burnout Inventory, median (IQR)  Personal Burnout 25/100 (8.3–50)  Work-related Burnout 39.3/100 (25.0–53.6.0.6)  Client-related Burnout 37.5/100 (20.8–54.2) Work productivity  Ever been absent due to premenstrual symptoms Yes 393 12.1 No 2750 84.9 Do not remember 96 3.0  How long total absence at maximum ( n  = 393)  months Reported n  = 106 2 (1–5)  weeks Reported n  = 26 1 (1–2)  days Reported n  = 216 1 (1–1)  hours Reported n  = 45 3 (1–4) The work characteristics of 3239 participants Table 3 presents the results of the EFA and CFA [ 27 ]. The initial solution based on a scree plot with eigenvalues of 1.0 or greater suggested four factors. Using a split-half sample, we performed EFA, excluded items that were either cross-loaded or did not load enough to exceed 0.5, and repeated this process until we reached the lowest AIC in the CFA. This process eventually resulted in four domains and 27 items. We retained four factors based on the KMO measure (0.977) and Bartlett’s test below 0.05. The names of the four dimensions were determined by examining the content of items with the highest loadings in each factor and referencing theoretical frameworks and prior literature on PMS symptomatology. For instance, Factor 1 consisted mainly of physical and musculoskeletal complaints such as dizziness, palpitations, joint pain, and blurred vision, and was labeled “Somatic symptoms.” Factor 2 included items on mood depression, anxiety, irritability, and lack of concentration, and was labeled “Psychological symptoms.” Factor 3 comprised items reflecting reduced performance and increased errors in work and household tasks, and was labeled “Lack of work efficiency.” Factor 4 contained gastrointestinal symptoms such as abdominal pain, cramps, and bloating, and was labeled “Abdominal symptoms.” The final labels were confirmed through consensus in a multidisciplinary expert panel to ensure conceptual clarity and occupational health relevance. Table 3 The results of exploratory and confirmatory factor analysis Exploratory Factor Analysis ( n  = 1597; Maximum likelihood, Promax rotation) Factor1 Factor2 Factor3 Factor4 Factor 1 “Somatic symptoms”, α = 0.960, omega = 0.961  Difficulty in breathing 0.722 0.167 0.033 −0.053  Dizziness 0.536 0.156 −0.005 0.238  Stumbling or falling over 0.800 0.027 0.088 −0.034  Nausea or vomiting 0.687 −0.032 0.030 0.203  Hot flushes on face or upper body 0.591 0.098 0.043 0.149  Chest pain or squeezing in the chest 0.759 0.015 −0.059 0.128  Tinnitus 0.857 0.003 −0.023 −0.033  Heart beating or pounding 0.681 0.194 0.058 −0.019  Numbness in the limbs 0.892 −0.040 0.010 −0.047  Chilly lower back or limbs 0.528 0.068 0.078 0.165  Blurred vision or difficulty in seeing 0.799 0.054 0.048 −0.078  Pain in finger or knee joint 0.731 −0.062 0.113 0.082  Muscle pain (clumps in the legs) 0.783 −0.088 0.128 0.035 Factor 2 “Psychological symptoms”, α = 0.933, omega = 0.934  Mood Depression 0.056 0.884 −0.053 0.028  Anxiety 0.216 0.820 −0.007 −0.136  Feeling of extreme psychological instability 0.056 0.880 −0.018 −0.017  Irritability −0.075 0.719 −0.006 0.158  Lack of motivation to work −0.075 0.622 0.223 0.097  Lack of concentration −0.062 0.590 0.290 0.122 Factor 3 “Lack of work efficiency”, α = 0.947, omega = 0.948  Expending more time on routine work 0.290 0.035 0.669 −0.002  Feeling of lower accomplishment at work 0.127 0.081 0.796 −0.007  Increased careless mistakes at work 0.161 0.129 0.709 −0.029  Expending more time on finishing usual household chores 0.293 0.075 0.541 0.033  Reduced working and learning ability 0.053 0.249 0.561 0.122 Factor 4 “Abdominal symptoms”, α = 0.821, omega = 0. 827  Diarrhea or constipation 0.101 0.077 0.024 0.604  Abdominal distention 0.076 0.110 0.044 0.703  Abdominal pain, cramps 0.089 0.070 −0.021 0.673 Factor correlation ( n  = 1597)  Factor 1 1.000  Factor 2 0.553 1.000  Factor 3 0.675 0.643 1.000  Factor 4 0.576 0.585 0.594 1.000 Confirmatory Factor Analysis Result ( n  = 1642)  Root mean squared error of approximation 0.077 (0.074–0.079)  Akaike’s information criterion 79,438  Comparative fit index 0.928  Tucker-Lewis index 0.921  Standardized root mean squared 0.043  Coefficient of determination 1.000  Average Variance Extracted Factor 1 Factor 2 Factor 3 Factor 4 0.571 0.684 0.642 0.543 The results of exploratory and confirmatory factor analysis The final model was investigated with the requirement of a well-fitted model (RMSEA = 0.077, CFI = 0.928, a TLI = 0.921, SRMS = 0.043, CD = 1.000) [ 27 ]. The AVE values ranged from 0.54 to 0.68 across the four domains, and in all cases, the square root of AVE exceeded the corresponding inter-factor correlations, supporting the discriminant validity of the developed scale. The domains included “Somatic symptoms’’ (13 items, Cronbach’s α = 0.960, Macdonald’s omega = 0.961), “Psychological symptoms” (6 items, Cronbach’s α = 0.933, Macdonald’s omega = 0.934), “Lack of work efficiency” (5 items, Cronbach’s α = 0.947, Macdonald’s omega = 0.948), and “Abdominal symptoms” (3 items, Cronbach’s α = 0.821, Macdonald’s omega = 0.827) [ 28 ]. The “Somatic symptoms” had the highest factor loading, ranging between 0.528 and 0.892, followed by the “Psychological symptoms,” ranging between 0.590 and 0.884, and the “Lack of work efficiency,” ranging between 0.541 and 0.796, while the “Abdominal symptoms” had the smallest factor loadings, ranging between 0.604 and 0.703. The Cronbach’s α values for each factor exceeded 0.8, McDonald’s omega exceeded 0.7, and the item correlations between each factor were reasonably moderate, suggesting good reliability, a moderate internal consistency for each factor, but distinct constructs. Table 4 shows the Spearman’s correlation coefficients between the developed scale and the CBI scores [ 26 ]. We confirmed a moderate correlation with the three subscales of the CBI (all P s < 0.0001), suggesting a moderate concurrent validity. Table 5 presents the criterion validity between the developed scale and PMS and PMDD based on the PSST [ 7 ]. Due to the small values for PMS and PMDD, we applied the chi-squared or Fisher’s exact tests and found that women who were in the upper half for each factor and the total score were more likely to have PMS or PMDD (Table 5 , all P s < 0.0001). This indicates that each factor in the scale developed had a moderate predictive ability of PMS or PMDD. Table 4 Concurrent validity with the spearman correlation coefficient between the newly developed scale and the Copenhagen burnout score # of item median (IQR) Copenhagen Burnout Inventory Personal Burnout Work related Burnout Client related Burnout Factor 1 “Somatic symptoms" 13 15 (13–21) 0.515* 0.435* 0.409* Factor 2 “Psychological symptoms" 6 11 (8–16) 0.626* 0.564* 0.522* Factor 3 “Lack of work efficiency" 5 6 (5–10) 0.545* 0.499* 0.472* Factor 4 “Abdominal symptoms" 3 5 (3–7) 0.480* 0.404* 0.414* Total score 27 39 (31–53) 0.628* 0.548* 0.517* * p  < 0.0001 Concurrent validity with the spearman correlation coefficient between the newly developed scale and the Copenhagen burnout score Table 5 Criterion validity between the developed scale and PMS and PMDD based on the premenstrual symptom screening tool (Unadjusted analysis) PMS ( n  = 331, 10%) PMDD ( n  = 99, 3%) (+) (-) P* (+) (-) P N (%) N (%) N (%) N (%) Factor 1 “Somatic symptoms" < 0.0001 < 0.0001  Upper half ( n  = 1692) 326 98.5 1366 47 99 100 1593 50.7  Lower half ( n  = 1547) 5 1.5 1542 53 0 0 1547 49.3 Factor 2 “Psychological symptoms" < 0.0001 < 0.0001  Upper half ( n  = 1787) 330 99.7 1457 50.1 99 100 1688 53.8  Lower half ( n  = 1452) 1 0.3 1451 49.9 0 0 1452 46.2 Factor 3 “Lack of work efficiency" < 0.0001 < 0.0001  Upper half ( n  = 1699) 329 99.4 1370 47.1 98 99.0 1601 51.0  Lower half ( n  = 1540) 2 0.6 1538 52.9 1 1.0 1539 49.0 Factor 4 “Abdominal symptoms" < 0.0001 < 0.0001  Upper half ( n  = 1994) 320 96.7 1674 57.6 95 96.0 1899 60.5  Lower half ( n  = 1245) 11 3.3 1234 42.4 4 4.0 1241 39.5 Total score < 0.0001 < 0.0001  Upper half ( n  = 1627) 330 99.7 1297 44.6 99 100 1528 48.7  Lower half ( n  = 1612) 1 0.3 1611 55.4 0 0 1612 51.3 *Based on the Chi-square test or Fisher’s exact test Criterion validity between the developed scale and PMS and PMDD based on the premenstrual symptom screening tool (Unadjusted analysis) *Based on the Chi-square test or Fisher’s exact test Table 6 shows the validation of the developed scale for work absenteeism by logistic regression models. Univariate associations using chi-squared testing or logistic regression models demonstrated that women who scored in the upper half for each factor and the total score were more likely to have ever had absenteeism from work. After adjusting for age, educational attainment, marital status, working status, and the CBI scores, women with who scored in the upper half for “Somatic symptoms” were 2.13–2.59.13.59-fold more likely, those with “Psychological symptoms” were 1.81–2.15.81.15-fold more likely, those with “Lack of work efficiency” were 1.96–2.36.96.36-fold more likely, and those with " Abdominal symptoms " were 2.05–2.56.05.56-fold more likely to have ever had absenteeism. Although the OR of each factor in PBO, WBO, and CBO falls within each 95% CI, indicating no statistical difference, the OR of each factor is statistically associated with work absenteeism. Thus, the developed scale appeared to yield moderate construct validity [ 30 ]. Table 6 The results of the logistic regression model validation of the absenteeism scale (Unadjusted and adjusted analyses) Work absenteeism experience due to PMS related symptoms Logistic regression models for Work absenteeism Yes ( n  = 393) No/Do not ( n  = 2846) P* Crude odds ratio (95% CI) Copenhagen burnout inventory adjust odds ratio (95% CI)* N (%) N (%) PBO model WBO model CBO model Factor1 “Somatic symptoms" < 0.0001  Upper half ( n  = 1692) 299 76.1 1393 49.0 3.32 (2.60–4.23) 2.44 (1.88–3.17) 2.59 (2.01–3.34) 2.13 (1.67–2.73)  Lower half ( n  = 1547) 94 23.9 1453 51.1 1 1 1 1 Factor 2 “Psychological symptoms" < 0.0001  Upper half ( n  = 1787) 324 82.4 1463 51.4 4.44 (3.39–5.82) 2.09 (1.60–2.73) 2.15 (1.65–2.79) 1.81 (1.41–2.33)  Lower half ( n  = 1452) 69 17.6 1383 48.6 1 1 1 1 Factor 3 “Lack of work efficiency" < 0.0001  Upper half ( n  = 1699) 309 78.6 1390 48.8 3.85 (2.99–4.96) 2.31 (1.78–3.00.78.00) 2.36 (1.82–3.06) 1.96 (1.52–2.51)  Lower half ( n  = 1540) 84 21.4 1456 51.2 1 - - - Factor 4 “Abdominal symptoms" < 0.0001  Upper half ( n  = 1994) 337 85.8 1657 58.2 4.32 (3.22–5.79) 2.47 (1.91–3.18) 2.56 (1.99–3.29) 2.05 (1.61–2.62)  Lower half ( n  = 1245) 56 14.3 1189 41.8 1 1 1 1 Total score < 0.0001  Upper half ( n  = 1627) 319 81.2 1308 46.0 5.07 (3.90–6.60) 1.84 (1.40–2.42) 1.99 (1.52–2.59) 1.68 (1.30–2.17)  Lower half ( n  = 1612) 74 18.8 1538 54.0 1 1 1 1 *Based on the Chi-square test **Adjusted for age, educational attainment, marital status, working status, and the Copenhagen burnout binary variable divided by the median The results of the logistic regression model validation of the absenteeism scale (Unadjusted and adjusted analyses) No/Do not ( n  = 2846) *Based on the Chi-square test **Adjusted for age, educational attainment, marital status, working status, and the Copenhagen burnout binary variable divided by the median Based on the largest Youden index [ 32 ] to establish an optimal-thresholds, we estimated the total score of the disability index to be 40 points, with a sensitivity of 79%, specificity of 57%, positive predictive value of 20%, negative predictive value of 95%, positive likelihood ratio of 1.83, negative likelihood ratio of 0.36, and AUC of 0.735 (95% CI: 0.710–0.760; Fig. 1 ), which were deemed acceptable after referencing the Hosmer and Lemeshow test [ 33 ]. Fig. 1 Receiver operating characteristic curve of the scale for absenteeism. Under the optimal Youden best cut-off point 40 of total score, sensitivity yields 79%, specificity yields 57%, and area under the curve yields 0.735 Receiver operating characteristic curve of the scale for absenteeism. Under the optimal Youden best cut-off point 40 of total score, sensitivity yields 79%, specificity yields 57%, and area under the curve yields 0.735

The survey was conducted by an internet research agency, GMO Research, Inc., between September 27 and 30, 2021. GMO first recruited 23,363 working women aged 18–41 years, because PMS is most prevalent in women in their late 20 s to early 40 s [ 5 ]. Of these, 3,880 agreed to participate in the study. Participants submitted an informed consent form via the company’s website ( https://gmo-research.jp ). The inclusion criteria were as follows: (1) salaried, working women, (2) menstruating women, and (3) Japanese-speaking population, so that they were able to understand the meaning of each question. The exclusion criteria were as follows: (1) unemployed ( n = 0), (2) postmenopausal status ( n = 0), or (3) menstruation has temporarily stopped due to pregnancy, postpartum, medication, etc. ( n = 375). After the application of these eligibility criteria, 3,239 women were considered eligible for the analysis. The data set has no missing values. We asked about previous history of mental illness, including depression, and underlying gynecological illnesses, including endometriosis, myoma, and ovarian tumors. This study was approved by the Ethics Committee of Akita University (No. 2712, approval date, July 7th, 2021) and performed in accordance with the Declaration of Helsinki. One of our members is involved in the management of a pharmaceutical company; however, we are not funded by that company, and none of our members have any conflicts of interest to disclose. The initial item pool was developed based on a systematic review of existing validated instruments, including the PSST [ 7 ], Menstrual Distress Questionnaire (MDQ, Japanese version) [ 20 ], Harvard Apple Women’s Health Study, which identified 15 menstrual period-related symptoms [ 21 ], the Work Productivity and Activity Impairment Questionnaire: General Health V2.0 [ 22 ], and Work-related Physical Activity Questionnaire (WPAQ) [ 23 ]. Items were mapped to four conceptual domains: physical symptoms, psychological symptoms, work-related functioning, and abdominal symptoms. A multi-disciplinary expert panel comprising a gynecologist, a psychosomatic physician, a psychologist, occupational health specialists, and data scientists reviewed the preliminary 65-item pool. Through a series of consensus meetings, items with redundancy or poor relevance to workplace settings were removed. We selected representative items from clusters (e.g., merging overlapping expressions of reduced motivation across different life domains), and discarded ambiguous or culturally specific items that could lead to multiple interpretations to ensure face validity and clarity. Subsequently, a final version of 47 questions to identify PMS-related symptoms of women workers was established (see the appendix). For each question, we asked respondents to indicate the severity of a symptom or event that began 1–2 weeks before the onset of menstruation and disappeared within 2 or 3 days after the onset of menstruation using a 5-point Likert scale, with 0 = no symptoms, 1 = slight symptoms, 2 = moderate symptoms, 3 = severe symptoms, and 4 = very severe symptoms. No separate qualitative pretesting or pilot study was conducted prior to data collection. Data were collected using self-administered questionnaires. In addition to the aforementioned 47 items related to PMS symptoms, the following items were also collected: age; BMI (kg/m 2 ); marital status; the presence of a child; family income; educational attainment; employment status; occupation; industry of employment; company size; working hours (i.e., average daily and weekly total working hours, night shift); regularity, amount, and length of menstruation; CBI; and absenteeism from work productivity. Absenteeism was assessed using a self-reported question that asked participants whether they had ever missed work due to PMS, and, if so, the duration of absence expressed in hours per week, days per week, weeks per month, and months per year. We did not set a fixed recall period; participants were instructed to respond based on their experience since the onset of menstruation. This lifetime approach was chosen because PMS is a recurrent condition that typically occurs in each menstrual cycle, making past episodes relevant to understanding the association with current symptom burden. The International Standard Classification of Occupations was used for defining occupations and industries [ 24 ]. The CBI was selected as a validation tool based on two conceptual pathways. First, from a biopsychosocial perspective, we hypothesized that premenstrual symptoms—particularly psychological and autonomic symptoms—may elevate subjective stress, fatigue, and emotional exhaustion, which are core dimensions of burnout. This aligns with prior literature indicating that PMS can exacerbate perceived stress and mood dysregulation, thereby contributing to burnout risk [ 25 ]. Second, from a functional standpoint, burnout was used as a proxy for work-related impairment, which is conceptually linked to absenteeism and presenteeism. Given the lack of validated workplace performance scales that overlap with PMS, the CBI served as an appropriate concurrent measure of occupational health status. The CBI measures burnout and is comprised of three domains: personal burnout (PBO), work-related burnout (WBO), and client-related burnout (CBO).It has previously been shown that the higher the score on this inventory, the more likely a worker is to retire [ 26 ]. We conducted both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) to establish the factor structure of the developed PMS screening scale. We randomly split the full sample (n = 3,239) into two approximately equal subsamples using Bernoulli sampling with p = 0.50: an EFA set (n = 1,597) and a CFA set (n = 1,642), enabling independent validation of the factor structure [ 27 ].” We first performed the EFA and determined the number of factors based on a scree plot and the Kaiser criterion (eigenvalue >1) with maximum likelihood estimation and Promax oblique rotation [ 27 ]. Items were removed if they met one or more of the following criteria: (1) factor loading < 0.50, (2) substantial cross-loading on two or more factors, or (3) conceptual irrelevance or redundancy within the assigned domain (e.g., “no motivation” overlapping with “decreased interest in work, home, or social activities”). In determining the number of factors that should be retained, we used the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy over 0.5 and a significance level for the Bartlett’s test below 0.05 [ 27 ]. To determine the internal consistency of the items, we developed a final model and computed the item test, item-rest correlation, and Cronbach’s alpha and McDonald’s omega coefficients for each factor to assess whether the domains were psychometrically distinct [ 28 ]. Once the final model was determined using EFA, we performed consecutive CFAs and computed the fit indices, convergent validity, and factor loadings to confirm the best-fitting model with an root-mean-square error of approximation (RMSEA) of 0.9, Tucker–Lewis Index (TLI) of >0.95, Standardized root mean squared (SRMS) (close to 0), and Coefficient of determination (CD) (close to 1) [ 27 ]. We repeatedly created a best-fit model until the lowest number of Akaike information criterion (AIC) statistics was reached. Convergent validity was assessed using the Average Variance Extracted (AVE), calculated as the mean squared standardized loadings (acceptable if ≥ 0.50), and discriminant validity using the Fornell–Larcker criterion. We then used three types of validity to determine if the newly developed scale is valid. Concurrent validity, which assesses how well a test correlates with a criterion measured at the same time, was tested by correlation with CBI scores [ 26 ]. The CBI has 19 items and three subscales of PBO, WBO, and CBO, with established construct validity of intention to retire in women workers in Japan [ 29 ]. Criterion validity, which compares a new test with an existing test that is already considered valid or relevant to see if they produce similar results, was tested using the chi-squared or Fisher’s exact tests between existing PMS and PMDD screening tools (i.e., PSST revised for adolescents questionnaire) and the binary scores of each factor and total scores divided by the median. We selected the PSST, which consists of 19 items and is multidimensional (symptom severity and functional impairment domains), because it has been validated against DSM criteria in prior studies, including those involving Japanese women [ 6 , 7 ]. We used median splits to dichotomize the scale and subscale scores for exploratory analysis in the absence of externally validated cutoff points for each subscale. Construct validity, which measures how well a test predicts outcomes in the future (i.e., whether the developed scale could predict absenteeism) [ 30 ], we applied a logistic regression analysis and computed odds ratios with 95% confidence intervals. For exploratory purposes, we examined the association between each subscale and absenteeism by dichotomizing participants at the median value for each domain, as formal cutoff points are not previously established. Multiple logistic regression was conducted with adjustments for previously burnout-related covariates among Japanese working women [ 31 ] because burnout was considered as a proxy of absenteeism that eventually results in quitting a job: age, educational attainment, marital status, working status, and CBI binary variables divided by the median. Finally, we drew receiver operating characteristic curves for absenteeism and estimated the area under the curve (AUC) and cutoff points, with the optimal sensitivity and specificity based on the Youden Index. All analyses were performed using SAS (version 9.4, SAS Institute Inc., Cary, NC, USA) and Stata version 17 (Stata Corp., College Station, TX, USA). The significance level was two-sided and less than 0.05.

We developed a reliable and valid PMS screening tool for women workers. This scale would allow labor managers, business owners, and working women to detect PMS earlier. An appropriate intervention would resolve the health aspects of women workers, including their physical and psychological symptoms, work efficiency, and absenteeism.

In this study, we examined 3,239 women workers aged 18–41 years (excluding pregnant and lactating women) in Japan, as PMS is most prevalent in women in their late 20 s to early 40 s [ 5 , 34 ]. Of the initial 47 items we considered, the EFA identified four domains with 27 items, which were then determined to be a moderately fitted model using the CFA. The four domains included “Somatic symptoms’’ (13 items), “Psychological symptoms” (6 items), “Lack of work efficiency” (5 items), and “Abdominal symptoms” (3 items), with high Cronbach’s alphas and item correlations. The subscale scores demonstrated adequate internal consistency and moderate inter-factor correlations, supporting their use as both components of the total score and as independent domains when contextually appropriate. Content validity was ensured through a structured development process: items were generated from a comprehensive literature review and existing validated instruments, reviewed by a multidisciplinary expert panel, and refined by removing culturally specific or redundant items. The Average Variance Extracted values ranged from 0.54 to 0.68 across the four domains. In all cases, the square root of AVE exceeded the corresponding inter-factor correlations, supporting the discriminant validity using the Fornell–Larcker criterion. The developed scale had a moderate concurrent validity as a whole, which refers to a moderate correlation with the three subscales of the CBI [ 26 ], and a moderate criterion validity due to significant associations with PMS or PMDD [ 7 ]. Finally, we confirmed that the developed scale is capable of predicting work absenteeism caused by PMS-related symptoms, with an acceptable AUC. These results suggest that the scale developed in this study may be useful for helping working women to stay healthy and maximize their work productivity. Although the model fit indices were within acceptable thresholds (e.g., RMSEA = 0.077, CFI = 0.928), the RMSEA value was borderline. This suggests that while the four-factor structure is interpretable and practically useful, future refinement—such as item revision or removal—may further improve model fit. Especially, the first factor identified through factor analysis—“Somatic symptoms”—comprised the largest proportion of items in the final scale. While many of these symptoms (e.g., dizziness, palpitations, joint pain, nausea) are not specific to PMS, they have been frequently reported in studies of Japanese and East Asian populations as common premenstrual complaints [ 1 , 2 ]. These symptoms may reflect the interaction between hormonal changes and somatic symptoms during the luteal phase. Importantly, this subscale demonstrated moderate internal consistency and significant associations with burnout and absenteeism, suggesting functional relevance. Although somatic symptoms are not emphasized in DSM-5 or ICD-11 diagnostic frameworks for PMDD, their inclusion in a workplace-oriented screening tool may enhance the ecological validity and capture the broader symptomatology experienced by women workers, particularly in cultural contexts where somatic expressions of distress are prominent. The strength of this study is that we included a variety of symptoms, both directly and indirectly related to PMS, according to previous literature. Unfortunately, there have been no reference guidelines previously published for PMS. Also, PMDD has only recently been included in DSM-5 as a disease entity, with a heavy weighting on psychological symptoms [ 35 , 36 ]. Similarly, the PSST was previously developed by Steiner [ 7 ]; however, physical symptoms were treated as only one of a total of 14 symptoms, of which 10 symptoms were psychological. In contrast, our scale is balanced and evenly divided into both physical (i.e., Somatic symptoms) and psychological factors, as well as the two other domains of a lack of work efficiency and abdominal symptoms. Finally, although our developed disability index is valid for predicting absenteeism, the AUC was 0.7, which suggests a moderate acceptability [ 33 ] with the low positive predictive value (20%). This moderate AUC may partly reflect the nature of self-reported absenteeism as the predicted outcome—an event with relatively low prevalence and multifactorial determinants, many of which are unrelated to PMS (e.g., organizational policies, personal circumstances). Consequently, many individuals flagged as “at risk” may not actually experience absenteeism. This consideration underscores the importance of using the tool for supportive screening purposes only—to raise awareness and guide further evaluation—not for diagnosis or employment-related decision-making. Ethical safeguards are essential to ensure that the tool does not contribute to unintended consequences such as stigmatization, discrimination, or unwarranted anxiety among users. Ideally, as this is a screening test, it is recommended that a positive result should be followed by a visit to a gynecologist or family doctor for a detailed diagnosis of PMS/PMDD and, if necessary, medication treatment. The use of the developed scale promotes understanding of PMS within a company and might even lead to timely diagnosis of untreated underlying mental or gynecological illness (endometriosis, fibroids, etc.) that contributes to the possibility of increased work productivity. The instrument should be used in self-assessment contexts or integrated into employee health promotion programs, where positive results can lead to voluntary medical consultation or workplace support, rather than punitive outcomes. There are several study limitations that should be examined and discussed. First, only 10% of participants had PMS and 3% had PMDD, slightly lower than community-based estimates (20–30%) [ 36 ]and DSM-5 [ 37 ]–based assessments in Japanese women (5.9%) [ 6 ], suggesting our sample may have been healthier. Second, industry distribution differed slightly from national labor statistics [ 38 ]; fewer participants were in health and welfare (18.5% vs. 22%) and more in manufacturing (14.3%), with relatively light working conditions (median 8 h/day, 2 h standing/day), which may have reduced PMS severity [ 39 ]. Third, absenteeism was self-reported and attributed to menstrual symptoms, possibly causing recall bias or misclassification; objective measures such as employer records are needed. We focused on absenteeism, which is easier to quantify than presenteeism, although previous studies suggest menstrual symptoms affect presenteeism more [ 12 , 40 , 41 ]. Fourth, we did not assess psychological factors such as depression or early-life stress (e.g., emotional abuse or neglect) known to influence PMS/PMDD risk [ 42 ]. Fifth, we did not calculate inter-rater agreement in item review, which was based on expert consensus, and did not assess test–retest reliability due to the cross-sectional design. Lastly, although the present validation was conducted exclusively among Japanese working women aged 18–41 years, the symptom domains are broadly applicable to PMS experiences in other populations. Nevertheless, differences in occupational demands, healthcare systems, cultural perceptions of menstrual symptoms, and the prevalence of mental illness could influence the measurement properties of the tool. Future studies should validate the scale in older age groups, non-working women, and diverse cultural and linguistic settings, and compare its performance with additional gold-standard instruments such as validated Japanese versions of the DRSP [ 19 ] to establish broader generalizability. Longitudinal designs are also warranted to assess predictive validity over time, and repeated measurements could establish test–retest reliability. Refinement efforts may include revising or removing items with lower factor loadings, re-evaluating cut-off points to optimize sensitivity and specificity, and developing a shorter or digital version of the tool to facilitate its integration into workplace health promotion programs.

According to a Japanese government survey or relevant source, more than half of working women report that their job performance is negatively affected by menstrual symptoms [ 1 ]. In addition, a large study of [ 2 ] 19,254 women aged 15–49 years showed that the annual economic burden of menstrual symptoms in Japan amounts to be 8.6 billion United States Dollars. This estimate was derived primarily from productivity losses, including absenteeism (missed workdays) and presenteeism (reduced efficiency while at work), calculated using self-reported work impairment and national wage data. Among menstrual symptoms, premenstrual syndrome (PMS) is one of the most frequently reported in young working women and is often associated with depressive symptoms [ 3 – 5 ]. Its most severe form, premenstrual dysphoric disorder (PMDD) [ 6 ], is recognized in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision (DSM-5-TR) as a mental disorder [ 6 , 7 ] and has a prevalence of 3–8% [ 8 ]. PMS/PMDD can impair quality of life [ 9 , 10 ], contribute to absenteeism and prolonged sick leave [ 11 , 12 ], and is often exacerbated by workplace stressors such as interpersonal difficulties [ 1 , 13 ]. These symptoms create “invisible” barriers to productivity and may even lead to workplace discrimination or mistreatment. Establishing a standardized method to evaluate PMS-related workplace problems could therefore help predict productivity loss and guide strategies such as menstrual leave policies [ 14 , 15 ]. Several assessment tools have been developed to evaluate menstrual distress symptoms. Early tools included the Menstrual Distress Questionnaire [ 16 ], the Premenstrual Tension Syndrome (PMTS) [ 17 ], and the Premenstrual Assessment Form (PAF) [ 18 ]. However, these predated the recognition of Late Luteal Phase Dysphoric Disorder in the DSM and had limited psychometric properties [ 16 – 18 ]. More recently, the Premenstrual Symptoms Screening Tool (PSST) [ 7 ] was introduced to align with the DSM criteria, though it primarily emphasizes psychometric aspects. The Daily Record of Severity of Problems (DRSP) has since become the gold standard for diagnosis of PMDD, and the Japanese version became available in 2021 [ 19 ]. While international guidelines emphasize the psychological and functional dimensions of PMDs, especially PMDD, it is important to note that many working women experience a combination of psychological and physical symptoms that interfere with occupational functioning. Existing screening tools, such as the PSST and DRSP, are designed mainly for clinical diagnosis and tend to focus on mood-related symptoms, often underrepresenting physical or somatic complaints. Yet these physical symptoms can substantially affect workplace outcomes, including absenteeism and reduced productivity. Our objective, therefore, was not to redefine PMS diagnostically, but to develop a screening tool that captures a broader spectrum of symptoms—including both psychological and physical aspects rather than contradict recent international consensus trends by emphasizing functionally relevant symptom burden in occupational settings. Thus, the purpose of this study was twofold: To develop a new screening tool for PMS that can be widely applied in workplace settings, addressing both physical and psychological symptoms as well as work productivity; and. To evaluate the validity of this tool in relation to the Copenhagen Burnout Inventory (CBI), the PSST, and absenteeism. To develop a new screening tool for PMS that can be widely applied in workplace settings, addressing both physical and psychological symptoms as well as work productivity; and. To evaluate the validity of this tool in relation to the Copenhagen Burnout Inventory (CBI), the PSST, and absenteeism.