Mammography screening reduced lifetime loss of employment duration and productivity for women with breast cancer: Real‐world evidence of societal impacts

Abstract

The objective of this study is to evaluate the personal and societal impacts of mammography screening by estimating the life expectancy (LE), LE loss, lifetime employment duration (LED), LED loss, lifetime productivity (LP), and LP loss of women with breast cancer (BC). We recruited 113,169 women with BC diagnosis (2002–2017) from the Taiwan Cancer Registry, following survival until 2018. The monthly employment‐population ratio (EMRATIO) and insured salary were abstracted from the National Health Insurance (NHI) until 2017. Lifetime survivals were extrapolated using relative survival with age‐, sex‐, and calendar‐year‐matched referents from vital statistics through a rolling‐over algorithm, which were used in hazard models estimating LED and LP, with losses calculated by comparing the BC cohort to matched referents and relative loss percentages. We summarized the overall losses of LED and LP weighted by stage shifts between those who received screening versus those who did non‐screening. We found no LED or LP losses for stages 0–I. LED losses for stages II–IV were 0.8, 2.3, and 5.1 years, respectively, with relative losses of LP being 9.4%, 30.2%, and 69.0%, respectively. Younger women faced more significant losses, with those aged 18–39 experiencing at least triple LED loss in stages III and IV compared to stage II. Mammography screening led to overall reductions of 2.9 years in LE, 0.6 years in LED, and 7.1% in LP, equivalent to a savings of US$ 7169. The mammography screening program in Taiwan improved the health and productivity of the population, especially among younger women.

What's New?

Despite abundant evidence showing cost‐effectiveness, the impact of mammography screening on lifetime employment duration and productivity in women with breast cancer remains unclear. Using real‐world data, here the authors estimated the reductions allowed by mammography screening in terms of employment duration and productivity by weighting the lifetime estimates by the stage proportions detected by mammography versus no‐mammography in Taiwan. Mammography reduced the losses in life expectancy by 2.9 years, in lifetime employment duration by 0.6 years, and in lifetime productivity by $7169. The mammography screening program in Taiwan improved the health and productivity of the population, especially among younger women.

Abbreviations

BC

breast cancer

CI

confidence interval

EMRATIO

employment‐population ratio

FCM

friction‐cost method

HCM

human capital method

LE

life expectancy

LED

lifetime employment duration

LP

lifetime productivity

NHI

National Health Insurance

RL

relative loss

UHC

universal health coverage

1. INTRODUCTION

Mammography screening has been implemented for several decades, with abundant evidence showing cost‐effectiveness. ¹ , ² , ³ In addition to reducing medical expenditures, early detection by screening also improved the quality of life of women. ³ , ⁴ Taiwan also launched a government‐subsidized mammography screening for women aged 45–69 and those aged 40–44 with a family history since 2004 and expanded to 40–74 in 2025, ⁵ following the latest recommendation from the US Preventive Task Force Recommendation. ⁶ However, confronting an aging society in the post‐pandemic era, one had better conduct the economic evaluation more accurately by integrating the products of survival probability and productivity at each time point (say, monthly) throughout the lifetime for a more efficient allocation of limited resources. ⁷ Valuing and quantifying the potential savings of employment duration and productivity loss ⁸ would be one of the significant steps, especially for women who have contributed to paid and unpaid work. Furthermore, these savings may contribute to improving the quality of life and reducing the financial burden on women and their families. Existing studies on people with cancer have considered absenteeism, early retirement, or premature mortality to estimate the indirect costs of cancer. ⁹ , ¹⁰ Few incorporated the lifetime employment duration (LED) and productivity to assess human capital losses. ¹¹

Researchers have recently begun to pay attention to the productivity loss of women with breast cancer (BC). ¹² , ¹³ , ¹⁴ , ¹⁵ The loss in working years of a woman after a BC diagnosis in Sweden has been estimated since she aged 50, 55, or 60 years, stratified by stages and event type (death, disability, and old‐age retirement) using flexible parametric survival modeling. ¹² Some researchers explored the issue collectively over an observable follow‐up period. The total labor productivity losses and years of potential work‐life loss due to the death of BC in Spain were estimated yearly from 2005 to 2014 using a simulation model based on the conventional human capital method (HCM). ¹³ Researchers from Poland applied the HCM to estimate the productivity loss and public finance burden attributable to breast cancer from 2010 to 2014. ¹⁴ In Taiwan, 88.8% of BC women were found to be employed continually within 1 year after diagnosis. ¹⁵ However, the lifetime estimation for an individual under different conditions would be more informative in personal decision‐making and societal perspectives.

An ongoing debate exists on valuing productivity losses. ¹⁶ , ¹⁷ The friction‐cost method (FCM) assesses the productivity loss due to temporary absenteeism from an employer's perspective. The HCM assesses the loss of a worker's lifetime earnings due to morbidity and/or premature death from both an individual's and societal perspectives. The data were usually collected by self‐report or interview instruments; only some are suitable for HCM or FCM. ¹⁸ Additionally, although many cost‐effectiveness studies have incorporated productivity loss/gains in the cost‐effectiveness studies, the impact and methods may not concern and extrapolate the whole lifetime. ¹⁹

Using claim data, we adopted a novel approach: multiplying the lifetime survival probability with a second function each month, either employment‐population ratio (EMRATIO) or insured salary, and summing it to a lifetime to quantify LED or productivity. We quantify the stage‐specific life expectancy (LE), LE loss, LED, LED loss, lifetime productivity (LP), and LP loss and their relative loss percentages compared to the age‐, sex‐, and calendar‐year matched referents for women with BC. As a result, we estimated the savings of mammography screening on LED and LP by weighting the lifetime estimates by the stage proportions detected by mammography versus none from 2010 to 2018 in Taiwan.

2. METHODS

2.1. Cohort

We collected all women with a valid diagnosis of breast cancer aged 18–64 years from 2002 to 2017 from the Taiwan Cancer Registry to estimate their employment status after the mandatory education in Taiwan. We abstracted their monthly employment status and insured salary directly from the National Health Insurance (NHI) dataset during the same period. We followed their survival until the end of 2018 through linkage with the National Mortality Registry. Anyone who had no death records in the registry was deemed censored. We stratified the cohort into stage 0 to stage IV according to American Joint Committee on Cancer (AJCC) records in the Cancer Registry. We adopted the pathology stage prior to the clinical stage. Those without stage information were classified as having the unknown stage. We included all women with valid BC diagnoses within the above age range and classified those without stage information as the stage unknown, or about 2.6% (2967/113,169), as in Table 1. We stratified the BC cohort by 5‐year age strata, which resulted in 36 subgroups.

TABLE 1.

Demographic characteristics at the diagnosis of women with breast cancer during 2002–2017, stratified by stages and ages.

			Classification of employment, n (%)			Monthly insured salary, US$
Stages, ages	No.	EMRATIO a	Employer	Government sector	Private sector	Unemployed	Including unemployed, median (IQR)	Only employed, mean
Stage 0
18–39	1584	0.79/0.74	18 (1.1)	155 (9.8)	1078 (68.1)	333 (21.0)	865 (689)	1277
40–44	2108	0.82/0.76	38 (1.8)	304 (14.4)	1389 (65.9)	377 (17.9)	903 (928)	1431
45–49	3889	0.81/0.77	109 (2.8)	495 (12.7)	2565 (66.0)	720 (18.5)	865 (777)	1414
50–54	3351	0.75/0.74	118 (3.5)	339 (10.1)	2065 (61.6)	829 (24.7)	865 (718)	1425
55–59	2617	0.63/0.62	64 (2.4)	155 (5.9)	1425 (54.5)	973 (37.2)	790 (630)	1519
60–64	2054	0.42/0.43	54 (2.6)	85 (4.1)	721 (35.1)	1194 (58.1)	715 (1218)	2038
Stage I
18–39	4333	0.78/0.75	62 (1.4)	401 (9.3)	2897 (66.9)	973 (22.5)	752 (715)	1253
40–44	5272	0.79/0.76	101 (1.9)	601 (11.4)	3469 (65.8)	1101 (20.9)	752 (772)	1350
45–49	7777	0.80/0.77	220 (2.8)	910 (11.7)	5058 (65.0)	1589 (20.4)	752 (746)	1367
50–54	7102	0.76/0.74	210 (3.0)	644 (9.1)	4513 (63.5)	1735 (24.4)	828 (658)	1359
55–59	6003	0.63/0.62	194 (3.2)	369 (6.1)	3209 (53.5)	2231 (37.2)	828 (689)	1496
60–64	4760	0.42/0.44	103 (2.2)	160 (3.4)	1739 (36.5)	2758 (57.9)	715 (1158)	1967
Stage II
18–39	5055	0.75/0.75	68 (1.3)	424 (8.4)	3289 (65.1)	1274 (25.2)	715 (596)	1238
40–44	5701	0.76/0.75	100 (1.8)	532 (9.3)	3727 (65.4)	1342 (23.5)	715 (655)	1278
45–49	7953	0.78/0.77	189 (2.4)	747 (9.4)	5251 (66.0)	1766 (22.2)	715 (596)	1246
50–54	7380	0.73/0.74	206 (2.8)	519 (7.0)	4648 (63.0)	2007 (27.2)	715 (536)	1268
55–59	6325	0.61/0.62	169 (2.7)	284 (4.5)	3411 (53.9)	2461 (38.9)	752 (596)	1426
60–64	4782	0.41/0.44	79 (1.7)	113 (2.4)	1774 (37.1)	2816 (58.9)	687 (1050)	1898
Stage III
18–39	2048	0.70/0.75	29 (1.4)	168 (8.2)	1246 (60.8)	605 (29.5)	715 (609)	1254
40–44	2250	0.74/0.75	35 (1.6)	215 (9.6)	1418 (63.0)	582 (25.9)	687 (596)	1242
45–49	3227	0.74/0.77	73 (2.3)	242 (7.5)	2072 (64.2)	840 (26.0)	687 (489)	1224
50–54	3431	0.70/0.74	77 (2.2)	212 (6.2)	2099 (61.2)	1043 (30.4)	715 (489)	1225
55–59	2961	0.61/0.63	74 (2.5)	147 (5.0)	1580 (53.4)	1160 (39.2)	715 (536)	1400
60–64	2060	0.39/0.44	28 (1.4)	53 (2.6)	733 (35.6)	1246 (60.5)	687 (1052)	1927
Stage IV
18–39	618	0.69/0.75	12 (1.9)	41 (6.6)	371 (60.0)	194 (31.4)	687 (498)	1152
40–44	748	0.68/0.76	10 (1.3)	64 (8.6)	438 (58.6)	236 (31.6)	687 (490)	1164
45–49	1190	0.67/0.77	17 (1.4)	77 (6.5)	702 (59.0)	394 (33.1)	687 (451)	1226
50–54	1371	0.61/0.74	29 (2.1)	70 (5.1)	742 (54.1)	530 (38.7)	715 (502)	1306
55–59	1307	0.53/0.62	35 (2.7)	53 (4.1)	604 (46.2)	615 (47.1)	715 (1036)	1499
60–64	945	0.34/0.43	15 (1.6)	34 (3.6)	273 (28.9)	623 (65.9)	687 (1052)	2192
Stage unknown
18–39	362	0.65/0.73	11 (3.0)	20 (5.5)	204 (56.4)	127 (35.1)	687 (502)	1265
40–44	439	0.66/0.75	8 (1.8)	31 (7.1)	249 (56.7)	151 (34.4)	687 (600)	1439
45–49	631	0.69/0.76	13 (2.1)	57 (9.0)	365 (57.8)	196 (31.1)	687 (502)	1317
50–54	643	0.63/0.74	9 (1.4)	35 (5.4)	361 (56.1)	238 (37.0)	687 (451)	1409
50–59	511	0.49/0.63	10 (2.0)	16 (3.1)	224 (43.8)	261 (51.1)	687 (1052)	1692
60–64	381	0.38/0.45	12 (3.1)	8 (2.1)	127 (33.3)	234 (61.4)	658 (1028)	1955

Note: The employment status was retrieved from the month of diagnosis. US$1 = NT$31.9 which is the average exchange rate of Central Bank of the Republic of China (Taiwan) during 2002–2017.

Abbreviations: EMRATIO, employment‐population ratio; IQR, interquartile range.

^a Women with BC/age, sex, and calendar‐year matched referents.

2.2. The theoretical framework of lifetime estimation

The conventional monthly employment rate is calculated by dividing the number of people employed by the labor force. It indicates what percentage of people who are able and looking for work are currently employed. This study applies the EMRATIO indicator in Organization for Economic Co‐operation and Development (OECD) countries. ²⁰ EMRATIO is calculated as the number of people employed divided by the total population at a specific age, sex, and calendar month. Thus, the lifetime measure of employment duration or productivity of a study cohort x can be estimated using the formula: ${\sum }_{t=0}^{L}S\left(t|x\right)\times E\left(t|x\right),$ where $S\left(t|x\right)$ denotes the survival function of the study cohort, $E\left(t|x\right)$ denotes the employed of the index population/the index population (whole study cohort x) (EMRATIO) or sum of insured salary of the employed/the index population (approximated productivity) of women with BC alive at time $t,$ and $L$ denotes the time when the survival probability reaches zero. ²⁰ , ²¹ , ²² , ²³ The lifetime survival function can be estimated using the rolling‐over algorithm: We applied logit transformation for the relative survival between BC subcohorts divided by age‐, sex‐, and calendar year‐matched referents simulated from general population mortality. Then, we constructed restricted cubic spline models from diagnosis to the end of follow‐up and extrapolated them month by month by repeatedly treating the newly extrapolated month as the “really observed one” until the survival probability <0.1%. ²³ , ²⁴

We quantified the loss of LE, LED, or LP by subtracting the estimates of the index cohort from the average of the general population (namely, simulated from national life table for LE or all enrollees of NHI for LED or LP) matched with every index woman on age, sex, and calendar year of the first diagnosis. For example, the loss of LEDs in the index cohort is the difference between the LEDs of the index cohort and the matched referents. These estimates, standard errors, and 95% confidence intervals (CIs) were basically calculated using the R package iSQoL2, which is accessible (http://sites.stat.sinica.edu.tw/isqol/).

2.3. Estimating the LED and LED loss

We used the NHI databases to perform the analysis, had been composed of over 98% of the civilians in Taiwan since 2004. ²⁵ We estimated the employment status until 65 years old because it is the suggested legal age for retirement currently applied in Taiwan. ²⁶ We used the monthly EMRATIO to construct a model for estimating LED for study cohort i, denoted as $E\left(t|i\right)$ ²⁰ , ²² In the registry for beneficiaries of Taiwan NHI dataset, if the insured woman's premium is paid through her salary account and her insured unit is not one of the followings: obligatory/alternative military, military dependent/student, low‐income household, district office, resident in social welfare organizations, monk, nun, or international student, she is classified as employed. Otherwise, she is assumed to be unemployed. The numerator of EMRATIO of the BC cohort at time t is the employed number that month and the denominator is the number of women who survived BC that month. At the same time, the EMRATIO of the matched referent is derived by matching the age, gender, and calendar year of diagnosis to the EMRATIO of the whole population in Taiwan, which is also calculated by all the beneficiaries in the same dataset (2002–2017).

The EMRATIO function $E\left(t|i\right)$ of the BC cohort during the follow‐up period is calculated by the monthly $E\left(t|i\right)$ and using the bidirectional seasonal moving average to mitigate the fluctuations, which were assumed to be zero after 65 years old (conditional). We followed the concepts developed by Wang et al. ²² to extrapolate it to lifetime. In general, EMRATIO decreases when the disease progresses to near the end of life, which is likely associated with an increased mortality hazard. We used the complete EMRATIO function $E\left(t|r\right)$ and hazard function $h\left(t|r\right)$ of the matched referents (r) and the association between EMRATIO and hazard to construct the following regression model:

$$\text{logit}\left(E\left(t|i\right)\right)-\text{logit}\left(E\left(t|r\right)\right)={\beta }_0+{\beta }_1\left(\log \left(h\left(t|i\right)\right)-\log \left(h\left(t|r\right)\right)\right)+{\epsilon }_t.$$

With estimated ${\beta }_0$ and ${\beta }_1$, we obtained lifetime function of $E\left(t|i\right)$. We then estimated the LED loss by subtracting the LED of BC from that of the matched referents. As the formal retirement age in Taiwan is 65, we only estimated the lifetime employment up to this age for both the index and reference cohorts. We also calculated the proportion of relative loss of employment duration by dividing the loss of LED of the BC cohort by the LED of the referents.

2.4. Estimating the LP and LP loss

We applied the same concepts to estimate LP function, LP, and productivity loss using the insured salary information from the NHI dataset. According to Taiwan's NHI Act, premiums submitted to the insurer have been charged monthly according to the insured salary, generally about 70%–90% of the employee's regular salary. ²⁷ , ²⁸ To increase efficiency, the NHI Administration divides monthly salary into about 50 strata for premium submission, and every individual is charged by the stratum to which her salary belongs. For example, in 2017, the bottom stratum was US$ 861, while the highest was US$ 6881. Because we constructed the function of the productivities by month, ²² we consider the dynamic fluctuations in salary would reflect the productivity or contribution to the labor market for women living with BC. As we have obtained both the relative loss of LED and productivity in BC patients after comparison with the corresponding referents, we could compare these two losses to look for the potential presence of presenteeism if the relative loss of LP is larger than that of LED.

2.5. Reductions by mammography screening

We interlinked the cohort with the Mammography Screening database (2010–2018), which covers all women who ever received screening. We determined the proportion of stage 0 to stage IV, including stage unknown for each group. We multiplied these by the stage‐specific LE, LED, and LP estimates to calculate the losses for a woman with BC detected through mammography or non‐screening. The estimated reductions of LE loss, LED loss, and LP loss from mammography were derived from the differences in stage‐proportion‐weighted losses. ³ , ²⁹ This approach might partially account for the health literacy differences. ³ Healthcare professionals performing the screening program must inform and persuade all women with positive mammography results to receive further diagnostic work‐ups within 2 months for confirmation. We allowed a 6‐month interval for those screened positive to obtain a confirmatory diagnosis because of the universal coverage of Taiwan's NHI. Namely, if the date of diagnosis of a woman in the cohort was within 6 months after a positive screening result, she was classified as detected by mammography; diagnoses beyond this interval were considered non‐screening detections. ³ We also used a 1‐year interval for sensitivity analysis. The flow diagram of estimating LE loss, LED loss, LP loss, and the estimated related reductions by mammography is shown in Supplementary Figure 1.

3. RESULTS

3.1. Cohort

In total, we have recruited 113,169 women with BC diagnosis during 2002–2017. As summarized in Table 1, about 70%–80% of women aged 18–54 with BC earlier than stage III were employed and in the private sector (60.8%–68.1%). The proportion of unemployment seemed to increase from age 50–54 (about 24.4%–30.4%) up to around 57.9%–60.5% in age 60–64. Generally, women with stage IV and unknown showed higher unemployed proportions regardless of different age groups. The medians of monthly insured salary, including the unemployed women, were US$ 658–715 for women with stages II or more advanced, and they were US$ 715–903 for women with stages 0 and I. The differences appeared to diminish if we considered only employed women, with the average salaries of different stages ranging from US$ 1152–2192 to US$ 1253–2038, respectively (Table 1).

3.2. LE, LE loss, LED, LED loss, LP, and LP loss of BC women by stages

As the BC stages progressed from stages II to IV, the LE, LED, and corresponding forgone earnings became higher and higher (Tables 2 and 3). The above impacts were more evident among younger women during their productive years in the labor market. For women aged younger than 50–54, the LED losses were about twofold increases between stages II (0.7–4.1 years) and III (2.2–8.5 years) and threefold between stages III and IV (5.1–16.7 years) with the relative losses being 9.7%–20.5%, 30.6%–42.9%, and 72.3%–82.8%, respectively. Their corresponding productivity losses were US$ 7531–46,762, US$ 28,811–58,569, and US$ 64,988–232,611 for stages II, III, and IV, respectively (Table 3). The relative losses of LP between stages II and III increased to about threefold for those younger than 40–44 and elevated to 4–6 fold for 45–54. For women with stage IV younger than 50–54, the relative losses of LP were above 70%. Women with an unknown stage exhibited LED and LP distributions between stages III and IV. With most 95% CIs including zero, women with BC stage 0 or I showed no significant LED loss compared to the corresponding matched referents. With the same trend, there were no lifetime earning losses or negative values compared with matched referents (Table 3). Presenteeism seemed to appear only in women with stage III and above and diagnosed younger than 50–64 (Supplementary Table 1).

TABLE 2.

Life expectancy (LE), lifetime employment duration (LED), and the relative loss of LED of women with breast cancer (BC) and their corresponding age‐, sex‐, and calendar year‐mated referents (reference).

Stages, ages	LE, year (95% CI)			LED, year (95% CI)			Relative loss, % (95% CI)
(Censored rate, %)	BC	Reference	LE loss	BC	Reference	LED loss	LED loss/Reference's LED
Stage 0
18–39	48.7	49.8	1.1	20.7	20.4	−0.3	−1.5
(97.9)	(47.5, 55.9)	(49.5, 49.9)	(−6.2, 2.1)	(19.7, 21.6)	(20.2, 20.5)	(−1.3, 0.6)	(−6.4, 3.1)
40–44	42.2	42.3	0.2	15.6	14.4	−1.2	−8.4
(98.2)	(22.9, 49.7)	(42.2, 42.4)	(−7.5, 18.9)	(14.7, 16.2)	(14.3, 14.4)	(−1.9, −0.4)	(−13.5, −2.8)
45–49	37.4	37.8	0.5	11.0	10.8	−0.2	−2.3
(98.1)	(23.4, 44.4)	(37.8, 37.9)	(−6.6, 13.5)	(10.5, 11.6)	(10.8, 10.8)	(−0.8, 0.2)	(−7.9, 1.8)
50–54	36.3	33.3	−3.1	7.3	7.1	−0.2	−2.3
(97.4)	(27.2, 39.7)	(33.2, 33.3)	(−6.5, 5.7)	(7.0, 7.6)	(7.1, 7.1)	(−0.5, 0.1)	(−7.4, 1.5)
55–59	30.8	28.6	−2.2	3.7	3.6	0.0	−1.0
(96.7)	(22.5, 34.8)	(28.6, 28.7)	(−6.3, 6.1)	(3.5, 3.9)	(3.6, 3.7)	(−0.2, 0.1)	(−6.3, 2.5)
60–64	28.1	24.2	−3.9	1.0	1.1	0.0	4.3
(96.5)	(20.5, 30.8)	(24.1, 24.3)	(−6.7, 3.5)	(1.0, 1.1)	(1.1, 1.1)	(0.0, 0.1)	(−3.1, 9.5)
Stage I
18–39	44.4	49.4	5.0	18.8	20.0	1.3	6.2
(94.3)	(38.2, 46.5)	(49.0, 49.3)	(2.5, 10.8)	(18.1, 19.7)	(20.0, 20.1)	(0.2, 1.9)	(1.1, 9.4)
40–44	37.7	42.4	4.7	14.3	14.4	0.2	1.3
(95.6)	(29.6, 40.5)	(42.3, 42.4)	(1.8, 12.1)	(13.8, 14.6)	(14.4, 14.5)	(−0.2, 0.6)	(−1.7, 4.2)
45–49	34.3	37.8	3.6	10.5	10.8	0.3	2.8
(95.5)	(27.7, 36.7)	(37.8, 37.9)	(1.1, 9.9)	(10.3, 10.8)	(10.8, 10.8)	(0.0, 0.5)	(0.0, 4.8)
50–54	31.6	33.2	1.6	7.2	7.1	0.0	−0.6
(94.8)	(28.2, 32.4)	(33.2, 33.3)	(0.8, 4.4)	(7.1, 7.4)	(7.1, 7.1)	(−0.3, 0.0)	(−4.6, 0.2)
55–59	27.3	28.6	1.3	3.6	3.6	0.1	1.8
(94.6)	(23.1, 28.8)	(28.5, 28.7)	(−1.7, 5.0)	(3.5, 3.7)	(3.6, 3.7)	(−0.1, 0.1)	(−2.0, 3.8)
60–64	25.8	24.2	−1.7	1.0	1.1	0.1	7.4
(93.9)	(18.3, 28.4)	(24.1, 24.2)	(−4.2, 5.3)	(1.0, 1.1)	(1.1, 1.1)	(0, 0.1)	(3.4, 11.2)
Stage II
18–39	40.2	49.4	9.2	16.0	20.1	4.1	20.5
(86.5)	(31.4, 41.6)	(49.2, 49.4)	(7.5, 16.4)	(14.7, 16.8)	(20.0, 20.2)	(3.2, 5.3)	(16.1, 26.1)
40–44	35.8	42.3	6.5	12.5	14.5	1.9	13.4
(88.5)	(31, 37.9)	(42.3, 42.4)	(4.4, 11.2)	(12.1, 13.1)	(14.5, 14.5)	(1.3, 2.3)	(9.2, 16.1)
45–49	32.5	37.9	5.4	9.8	10.8	1.0	9.7
(89.0)	(29.1, 34.3)	(37.7, 37.9)	(3.5, 8.5)	(9.5, 10.2)	(10.8, 10.9)	(0.6, 1.3)	(5.7, 11.9)
50–54	27.8	33.3	5.4	6.4	7.1	0.7	9.8
(86.6)	(22.4, 29.6)	(33.2, 33.3)	(3.5, 10.6)	(6.3, 6.6)	(7.1, 7.2)	(0.5, 0.8)	(6.5, 11.4)
55–59	23.4	28.7	5.3	3.4	3.7	0.3	9.0
(85.8)	(20.9, 25.5)	(28.6, 28.7)	(3.1, 7.7)	(3.3, 3.5)	(3.7, 3.7)	(0.2, 0.4)	(6.0, 11.5)
60–64	21.0	24.1	3.1	1.0	1.1	0.1	9.3
(85.4)	(19.4, 22.0)	(24.1, 24.2)	(2.2, 4.6)	(1.0, 1.1)	(1.1, 1.2)	(0.1, 0.1)	(5.4, 12.4)
Stage III
18–39	24.2	49.2	25.0	11.4	19.9	8.5	42.9
(65.9)	(17.6, 29.4)	(48.7, 49.1)	(19.4, 30.9)	(10.3, 13.1)	(19.8, 20.0)	(6.6, 9.6)	(33.3, 47.9)
40–44	21.9	42.3	20.4	9.7	14.5	4.8	33.0
(70.4)	(14.7, 29.6)	(42.2, 42.4)	(12.4, 27.5)	(9.0, 10.6)	(14.4, 14.5)	(3.8, 5.4)	(26.5, 37.5)
45–49	22.6	37.7	15.1	7.4	10.8	3.4	31.6
(70.0)	(17.7, 26.2)	(37.7, 37.8)	(11.5, 19.5)	(6.9, 7.8)	(10.7, 10.8)	(2.9, 3.7)	(26.7, 34.6)
50–54	21.2	33.1	11.9	4.9	7.1	2.2	30.6
(65.4)	(18.4, 22.5)	(33.1, 33.2)	(10.5, 14.2)	(4.6, 5.1)	(7.0, 7.1)	(1.9, 2.4)	(27.4, 33.4)
55–59	18.7	28.7	10.0	2.9	3.7	0.8	22.7
(66.1)	(17.3, 20.0)	(28.6, 28.7)	(8.6, 11.2)	(2.7, 3.0)	(3.7, 3.8)	(0.7, 1.0)	(18.6, 26.2)
60–64	14.6	24.1	9.5	0.9	1.1	0.2	20.0
(65.1)	(12.1, 16.6)	(24.0, 24.2)	(7.4, 12.0)	(0.8, 1.0)	(1.1, 1.2)	(0.2, 0.3)	(15.1, 25.3)
Stage IV
18–39	5.4	49.6	44.1	3.5	20.2	16.7	82.8
(29.4)	(4.8, 9.1)	(48.7, 49.5)	(39.9, 44.4)	(2.9, 4.0)	(19.9, 20.4)	(16.2, 17.3)	(81.1, 84.7)
40–44	7.1	42.3	35.3	3.3	14.5	11.2	77.2
(27.7)	(4.9, 9.7)	(42.2, 42.5)	(32.5, 37.2)	(2.7, 4.0)	(14.4, 14.5)	(10.4, 11.6)	(72.5, 79.8)
45–49	5.6	37.8	32.1	2.7	10.7	8.0	74.6
(26.6)	(4.7, 7.8)	(37.6, 37.8)	(29.7, 32.9)	(2.4, 3.1)	(10.7, 10.8)	(7.6, 8.3)	(71.4, 76.7)
50–54	5.2	33.2	28.0	1.9	7.0	5.1	72.3
(24.1)	(4.1, 7.0)	(33.0, 33.3)	(25.8, 29.0)	(1.7, 2.2)	(7.0, 7.1)	(4.8, 5.3)	(69, 74.2)
55–59	6.0	28.7	22.7	1.4	3.7	2.3	62.3
(25.4)	(5.2, 6.8)	(28.6, 28.8)	(21.7, 23.5)	(1.3, 1.5)	(3.6, 3.7)	(2.1, 2.4)	(59.1, 64.4)
60–64	4.8	24.2	19.4	0.5	1.1	0.6	50.6
(26.5)	(4.0, 5.9)	(24.0, 24.3)	(18.2, 20.1)	(0.5, 0.6)	(1.1, 1.1)	(0.5, 0.6)	(45.6, 55.5)
Stage unknown
18–39	26.4	49.5	23.2	10.2	20.1	9.9	49.4
(61.9)	(17.0, 30.5)	(48.4, 49.7)	(18.3, 30.2)	(8.6, 11.9)	(19.8, 20.4)	(8.1, 11.3)	(41.1, 55.5)
40–44	24.6	42.5	17.9	8.8	14.6	5.7	39.3
(63.3)	(18.3, 28.1)	(42.3, 42.6)	(14.0, 23.6)	(7.7, 10.1)	(14.5, 14.7)	(4.3, 6.8)	(29.6, 46.5)
45–49	23.6	37.7	14.1	7.0	10.8	3.8	35.1
(66.6)	(17.7, 26.4)	(37.6, 37.8)	(11.2, 19.4)	(6.2, 7.7)	(10.7, 10.9)	(3.1, 4.5)	(28.7, 41.5)
50–54	21.9	33.2	11.2	4.8	7.2	2.4	32.8
(66.7)	(16.4, 24.5)	(33.1, 33.4)	(8.5, 16.5)	(4.3, 5.5)	(7.1, 7.3)	(1.6, 2.8)	(22.8, 37.9)
55–59	17.4	28.7	11.4	2.4	3.8	1.4	37.8
(56.6)	(15.0, 19.1)	(28.6, 28.8)	(9.6, 13.5)	(2.1, 2.6)	(3.7, 3.9)	(1.2, 1.7)	(32.5, 44.9)
60–64	15.0	24.2	9.2	0.9	1.2	0.3	25.0
(59.1)	(12.7, 16.7)	(24, 24.3)	(7.4, 11.3)	(0.7, 1.1)	(1.1, 1.3)	(0.1, 0.4)	(10.6, 35.4)

Note: Mean of the estimates with 2.5%, 97.5% confidence interval (CI) estimated from bootstrap method with 100 times. Numbers in red indicate the negative estimates or negative 95% CI, which means the women with BC have a longer life expectancy and lifetime employment duration after diagnoses than the age‐, sex‐, and calendar year‐matched referents. The numbers in blue indicate that 95% CI contains zero, implying no differences in the LED between women with BC and the matched referents.

TABLE 3.

The lifetime productivity (estimated by earnings) of women with breast cancer (BC) and their sex‐, age‐, and calendar year‐matched referents, along with the relative loss.

	Lifetime productivity, US$ (95% CI)		Forgone earnings, US$ (95% CI)	Relative loss, % (95% CI)
Stages, ages	BC	Reference	Reference‐BC	Forgone earnings/reference
Stage 0
18–39	304,105 (285,258, 316,093)	274,377 (272,191, 276,291)	−29,728 (−42,905, −11,647)	−10.8 (−15.8, −4.2)
40–44	248,353 (235,247, 262,170)	191,992 (191,150, 192,598)	−56,361 (−72,110, −43,270)	−29.4 (−37.7, −22.5)
45–49	166,109 (157,488, 175,890)	142,187 (141,686, 142,672)	−23,922 (−34,064, −16,568)	−16.8 (−24.0, −11.6)
50–54	103,091 (97,180, 108,291)	91,123 (90,600, 91,524)	−11,968 (−17,734, −6310)	−13.1 (−19.6, −6.9)
55–59	49,390 (46,993, 52,742)	45,095 (44,666, 45,459)	−4296 (−7820, −2059)	−9.5 (−17.5, −4.5)
60–64	12,849 (11,615, 13,924)	12,487 (12,126, 12,805)	−362 (−1632, 623)	−2.9 (−13.5, 4.9)
Stage I
18–39	264,415 (254,522, 285,180)	269,698 (268,782, 270,692)	5283 (−17,263, 13,915)	2.0 (−6.4, 5.1)
40–44	203,362 (195,821, 210,685)	192,406 (191,983, 192,929)	−10,957 (−18,855, −4006)	−5.7 (−9.8, −2.1)
45–49	150,455 (145,038, 153,669)	141,414 (140,899, 141,744)	−9041 (−12,487, −4100)	−6.4 (−8.9, −2.9)
50–54	100,346 (97,904, 105,866)	90,561 (90,250, 90,931)	−9785 (−15,297, −8271)	−10.8 (−16.9, −9.1)
55–59	47,736 (45,971, 49,709)	44,639 (44,378, 44,927)	−3098 (−5246, −1284)	−6.9 (−11.8, −2.9)
60–64	12,751 (12,025, 13,309)	12,583 (12,400, 12,746)	−169 (−850, 427)	−1.3 (−6.9, 3.3)
Stage II
18–39	223,340 (204,023, 240,573)	270,102 (269,314, 271,183)	46,762 (27,300, 65,181)	17.3 (10.1, 24)
40–44	171,419 (163,396, 179,013)	191,917 (191,594, 192,405)	20,499 (12,142, 27,566)	10.7 (6.3, 14.3)
45–49	132,422 (127,290, 138,254)	140,609 (140,117, 140,910)	8187 (1472, 13,075)	5.8 (1.1, 9.3)
50–54	82,254 (80,326, 85,114)	89,784 (89,331, 90,113)	7531 (4648, 9092)	8.4 (5.2, 10.1)
55–59	41,303 (39,901, 42,889)	45,146 (44,848, 45,409)	3843 (2340, 5124)	8.5 (5.2, 11.3)
60–64	11,725 (11,091, 12,357)	12,781 (12,566, 13,008)	1057 (427, 1565)	8.3 (3.4, 12.0)
Stage III
18–39	144,189 (126,417, 168,860)	267,356 (265,671, 268,886)	123,167 (98,940, 138,917)	46.1 (37.2, 51.7)
40–44	132,712 (121,793, 150,141)	191,280 (190,536, 191,954)	58,569 (36,442, 68,079)	30.6 (19.1, 35.5)
45–49	94,901 (88,837, 100,985)	139,413 (138,880, 139,942)	44,511 (37,144, 50,188)	31.9 (26.7, 35.9)
50–54	60,121 (56,819, 63,744)	88,932 (88,409, 89,451)	28,811 (25,169, 31,978)	32.4 (28.5, 35.7)
55–59	35,085 (32,628, 36,481)	45,271 (44,818, 45,681)	10,186 (8416, 12,218)	22.5 (18.8, 26.7)
60–64	10,326 (9424, 11,088)	12,702 (12,293, 13,027)	2377 (1600, 3196)	18.7 (13.0, 24.5)
Stage IV
18–39	39,318 (33,146, 45,176)	271,929 (268,923, 275,064)	232,611 (225,036, 239,094)	85.5 (83.7, 86.9)
40–44	44,558 (33,928, 57,642)	192,203 (190,824, 193,177)	147,646 (132,459, 157,092)	76.8 (69.4, 81.3)
45–49	33,755 (29,323, 40,142)	139,830 (138,963, 140,795)	106,075 (97,511, 110,544)	75.9 (70.2, 78.5)
50–54	24,599 (21,699, 29,139)	89,588 (88,751, 90,316)	64,988 (59,453, 67,753)	72.5 (67.0, 75.0)
55–59	16,934 (15,014, 18,905)	44,907 (44,285, 45,571)	27,972 (25,965, 29,625)	62.3 (58.6, 65.0)
60–64	6547 (5564, 7496)	12,455 (12,015, 12,880)	5908 (4883, 6815)	47.4 (40.6, 52.9)
Stage unknown
18–39	128,757 (104,295, 152,735)	268,342 (264,344, 272,394)	139,586 (103,073, 162,603)	52.0 (39.0, 59.7)
40–44	119,300 (97,813, 134,765)	191,768 (190,439, 193,701)	72,468 (54,955, 93,493)	37.8 (28.9, 48.3)
45–49	84,283 (74,278, 93,730)	138,300 (137,026, 139,704)	54,016 (44,857, 63,393)	39.1 (32.7, 45.4)
50–54	60,268 (52,395, 71,086)	89,119 (87,653, 90,332)	28,851 (17,512, 35,212)	32.4 (20.0, 39.0)
55–59	29,588 (25,256, 33,660)	45,383 (44,479, 46,342)	15,795 (11,611, 19,840)	34.8 (26.1, 42.8)
60–64	10,420 (8311, 12,237)	12,936 (12,225, 13,712)	2516 (623, 4306)	19.5 (5.1, 31.4)

Note: Mean of the estimates with 2.5%, 97.5% confidence interval (CI) estimated from bootstrap method with 100 times. US$1 = NT$31.9, the average exchange rate of the Central Bank of the Republic of China (Taiwan) from 2002 to 2017. Numbers in red indicate the negative estimates or negative 95% CI, which means the women with BC have higher lifetime productivity after diagnoses than the age‐, sex‐, and calendar year‐matched referents. The numbers in blue indicate that the 95% CI contains zero, implying no differences in the LP between women with BC and the matched referents.

Figure 1 illustrates LED and LP and their losses for women with stages II and III. Both curves showed marked fluctuation for women with BC younger than 49, especially those with stage III rather than stage II in the same age stratum. The fluctuation trend appeared larger in LP curves than in LED curves within the same stage and age stratum. The results of other stages are illustrated in Supplementary Figure 2.

FIGURE 1.

Curves illustrating trends after breast cancer (BC) diagnosis for survival, conditional EMRATIO (employment‐population ratio), conditional insured salary, and their losses compared to the matched referents until age 65 stratified by stages II and III and ages. Abbreviations: LED, lifetime employed duration; LP, lifetime productivity. The dark gray area is the LED or LP estimate of women with BC, and the light gray area is that of the age‐and‐calendar‐year‐matched referent. Therefore, losses of LED and LP are the light gray area, also meaning the differences between those of BC and the matched referents, respectively. There were substantial differences between the LED and LP losses of women with BC stage II and stage III.

3.3. Reductions by mammography screening

After weighting by stage shifts, we found that mammography screening would reduce a woman with BC 2.9 (95% CI, 1.7–4.2) years in LE, 0.6 (95% CI, 0.4–0.7) years in LED, and US$ 7169 (95% CI, 1374–12,963) in LP, which corresponded with lifetime relative losses of 8.9%, 7.1%, and 7.1%, respectively (Table 4). The estimates remained similar in the sensitivity analysis, expanding the interval of detecting BC by screening to 1 year (Supplementary Table 2). These estimates were calculated under the assumption of no statistically significant loss for BC women with a stage of 0 or I, and they could be considered a lower bound.

TABLE 4.

Estimated losses of life expectancy (LE), lifetime employment duration (LED), lifetime productivity (LP), and their relative losses (RL) reduced by mammography screening—proportionally weighted by the stages.

Stage	LE loss (95% CI) RL% (95% CI)	LED loss (95% CI) RL% (95% CI)	LP loss (95%CI) RL% (95% CI)	Detected by mammography				Detected not by mammography
				n (%) of BC	LE loss, years (RL%)	LED loss, years (RL%)	LP loss, US$ (RL%)	n (%) of BC	LE loss, years (RL%)	LED loss, years (RL%)	LP loss, US$ (RL%)
0	−1.6 (−10.2, 7.1) −5.6% (−8.0%, −3.2%)	−0.3 (−0.7, 0.2) −2.0% (−2.5%, −1.5%)	−18,827 (−26,703, −10,951) −14.4% (−15.1%, −13.7%)	8730 (23.2%)	0.0 (0.0%) a	0.0 (0.0%) a	0.0 (0.0%) a	8946 (10.9%)	0.0 (0.0%) a	0.0 (0.0%) a	0.0 (0.0%) a
I	2.1 (−1.9, 6.1) 5.2% (4.0%, 6.5%)	0.1 (−0.1, 0.4) 2.3% (2.0%, 2.6%)	−7018 (−11,200, −2837) −6.6% (−7.0%, −6.2%)	13,890 (36.9%)	0.8 (1.9%)	0.0 (0.8%)	0.0 (0.0%) a	22,928 (28.0%)	0.6 (1.5%)	0.0 (0.6%)	0.0 (0.0%) a
II	5.3 (2.5, 8.0) 15.6% (14.8%, 16.4%)	0.8 (0.6, 1.1) 10.2% (9.9%, 10.5%)	8304 (3783, 12,826) 8.2% (7.8%, 8.5%)	10,420 (27.6%)	1.5 (4.3%)	0.2 (2.8%)	2295 (2.3%)	28,379 (34.7%)	1.8 (5.4%)	0.3 (3.5%)	2881 (2.8%)
III	13.3 (9.4, 17.1) 39.1% (38.1%, 40.2%)	2.3 (1.9, 2.7) 28.0% (27.6%, 28.4%)	29,387 (22,039, 36,735) 27.9% (27.4%, 28.4%)	3511 (9.3%)	1.2 (3.6%)	0.2 (2.6%)	2737 (2.6%)	12,721 (15.6%)	2.1 (6.1%)	0.4 (4.4%)	4570 (4.3%)
IV	27.1 (25.6, 28.6) 82.4% (82.0%, 82.8%)	5.1 (4.8, 5.4) 67.4% (67.1%, 67.8%)	66,159 (60,270, 72,048) 67.2% (66.7%, 67.6%)	694 (1.8%)	0.5 (1.5%)	0.1 (1.2%)	1218 (1.2%)	6830 (8.4%)	2.3 (6.9%)	0.4 (5.6%)	5524 (5.6%)
Unknown	12.8 (9.2, 16.4) 37.9% (36.9%, 38.9%)	2.8 (2.1, 3.5) 34.3% (33.5%, 35.1%)	35,884 (25,563, 46,206) 33.5% (32.6%, 34.4%)	453 (1.2%)	0.2 (0.5%)	0.0 (0.4%)	431 (0.4%)	1994 (2.4%)	0.3 (0.9%)	0.1 (0.8%)	875 (0.8%)
Sum				37,698 (100%)	4.1 (11.8%)	0.6 (7.9%)	6681 (6.5%)	81,798 (100%)	7.0 (20.8%)	1.2 (15.0%)	13,850 (13.6%)
Reductions by mammography		LE loss: 7.0–4.1 = 2.9 (95% CI: 1.7, 4.2) years, converted into lifetime RL savings of 20.8–11.8 = 8.9% (95%CI: 7.6%, 10.2%) LED loss: 1.2–0.6 = 0.6 (95% CI: 0.4, 0.7) years, converted into lifetime RL savings of 15.0–7.9 = 7.1% (95%CI: 6.6%, 7.6%) LP loss: 13,850–6681 = 7169 (95% CI: 1374, 12,963) US dollars, converted into lifetime RL savings of 13.6–6.5 = 7.1% (95%CI: 6.7%, 7.5%)

Note: The estimates of each stage were pooled based on the age proportions between 40 and 64 in Tables 2 and 3. The stage distributions were calculated from the women who had received the mammography screening within 6 months prior to their diagnosis or beyond during 2010–2018. Reductions by mammography were calculated by subtracting the sums of stage‐proportion weighted estimates detected by mammography from those not detected.

Abbreviations: LE, life expectancy in years; LED, lifetime employment duration in years; LP, lifetime productivity in US$; RL, relative loss, which refers to the loss divided by the referents estimates and presents in %. US$1 = NT$31.9, which is the average exchange rate of the Central Bank of the Republic of China (Taiwan) during 2002–2017.

^a When calculating the stage‐proportion weighted estimates, we assumed that the LE and LED losses of women with BC stage 0 were zero. This also applied to the productivity loss of women with BC stage 0 and I.

4. DISCUSSION

Based on analyzing real‐world data, we have corroborated with previous studies showing that mammography reduced BC mortality and LE loss ³ , ³⁰ , ³¹ , ³² and further demonstrated that it also saved the losses of LED and LP. Before further discussion, we provide the following arguments to support the accuracy of these estimations. First, as Taiwan's NHI waives all co‐payments related to BC, all cases must provide valid documents, including pathology, for the diagnosis and are assured by two specialists. The coverage of NHI has been over 98% since 2004, and our datasets are nationwide, comprehensive, and have negligible selection bias. Second, because the premiums of NHI are collected based on the insured salary, the dynamic status of individual employment is reflected in monthly real‐world data. We also constructed a model to extrapolate the employment status based on its association with matched referents and hazards. Third, since the insured salary was about 70%–90% proportional to the actual working salaries ²⁷ , ²⁸ and we did not include other incomes from property rentals and firm profits, our estimation of the LP loss would be underestimated or considered a lower bound of human capital loss. To improve the accuracy, we estimated the relative loss (in percentages) of insured salary of BC women by comparing them with the overall average of the same age and calendar month. Namely, we estimated the saving of lifetime average losses of percentages for BC women stratified by stage and 5‐year strata for the overall weighted sum. Thus, we tentatively conclude that under a conservative estimation, the mammography program would reduce about 2.9 years of LE loss, 0.6 years of LED loss, and 7.1% loss of forgone earnings (or >US$ 7169) per woman with BC. By multiplying these estimates shown in Table 4 with the 5250 women with BC detected by the mammography screening program in 2024, ³³ we obtained the potential savings of 15,225 years of LE, 3150 years of LED, and about 38 million US dollars in productivity. Adding the contribution of unpaid work by women, the above real‐world evidence could serve as part of incentives not only for eligible women themselves but also for all stakeholders in the national policy decisions concerned with the sustainability of the universal coverage system and societal welfare.

The weighted average losses of LED of women aged 50–64 at diagnosis were 0, 0.4, 1.2, and 2.9 years for stages I, II, III, and IV, respectively. As our BC cohort and the age‐ and calendar‐month matched referents were collected from the whole population in Taiwan, the above estimates were about half to one‐third (stage IV) of those reported from Sweden (0.5, 0.9, 2.5, 8.1, respectively) for BC versus BC‐free working women during 1997 and 2012. ¹² Alternatively, if we compared the relative losses between the two countries, they appeared to be similar, namely, 2.3%, 9.4%, 25.3%, and 63% for stages I to IV, respectively, in Taiwan compared to 3.8%, 6.8%, 18.8%, and 60.9% in Sweden, ¹² indicating a high corroboration with each other. The LEDs were the summations of the dynamic trajectories of EMRATIO utilizing beneficiaries of NHI datasets considering survival probabilities. Among all available national datasets provided in public or private sectors, we tentatively think it may be the relatively appropriate surrogate to capture all the employment status of individuals, even with irregular work patterns, such as contract workers or those transitioning between jobs, because NHI was mandatory for every resident in Taiwan. The aggregated additional labor market data may be helpful for further studies to serve as the proportional weights within different employment sectors provided in the NHI. Additionally, compared to the women with end‐stage kidney disease in Taiwan, women with BC seemed to have less LED loss, LP loss, and presenteeism. ²² This slightly reflects the higher socio‐economic status of women with BC. ¹⁵

The conventional HCM to the indirect loss of disease usually incorporates cross‐sectional survey data of salaries with life‐year loss because of premature mortality before retirement, which does not consider the losses because of morbidity. More studies started to estimate BC's temporary losses at employment and home because of functional disability, but resulted in the average total annual impact ³⁴ , ³⁵ , ³⁶ or significant differences between the two commonly used methods; the FCM estimate per woman was only 4.2% of that of HCM. ³⁷ In contrast, we proposed a novel method to estimate the lifetime BC‐associated productivity loss using an existing longitudinal claim dataset, which directly depicted the dynamic real‐life impacts on salaries after BC diagnosis. Our results also found some indications of potential presenteeism in women with advanced stages at younger ages. However, the EMRATIO cannot be used to assess labor force flows. We need more information to determine whether a decrease in the EMRATIO represents more people exiting employment or fewer new entrants. Moreover, we need more data to explore whether economic recession and growth impact labor demand or demographic trends drive changes in the EMRATIO. Nonetheless, by comparing the age, sex, and calendar year of the matched reference cohort to obtain the losses of LED or LP, our estimates would partially reduce the underestimation. As we were abstracting real‐world data, the estimates allowed us to explore the presenteeism of the index cohort by comparing the relative loss percentages of LED and LP.

Since our data only included insured salaries and used the whole population as the denominator regardless of willingness to work, our estimations would be an underestimation. We were unable to measure the unpaid productivity loss in this study, which is estimated to be about 56% of the total productivity losses from women with BC in Europe ³⁸ and 62% of the share outside the labor force due to caring responsibilities in South Korea. ³⁹ We recommend that future studies consider this portion to tackle the tremendous societal impacts in more detail.

Additionally, we found no statistically significant losses of LED and LP for BC women with a stage 0 or I due to the wide 95% CIs likely from the high censored rates (Table 2). Instead, BC women below age 60 diagnosed at stage 0 and some middle‐aged at stage I showed increased LP compared with age‐ and calendar year‐matched referents. In the final estimation, we conservatively assumed that there would be no significant productivity change for women with BC at stages 0 and I, even though our models found some increase in different age strata. Within the same stage between II and IV, however, the younger the ages, the larger the losses; the gaps were even greater in advanced stages. Moreover, the differences in losses of LED and LP between women with BC stage II and stage III were quite substantial, or the former were about 20% lower than the latter. Although further studies with a longer period of follow‐up and low censored rates would improve the accuracy, our estimates would be a lower bound of LE, LED, and LP reductions. With continual advancements in BC treatment and the expansion of mammography screening, we anticipated these reductions would increase.

One of the primary purposes of this study is to quantify the overall effect that can be readily estimated from real‐world data; we compared every case with the average EMRATIO and insured salary of the general population of the same age, sex, and calendar year of diagnosis to control for these factors. As the calculation has adjusted for lifetime survival function, our estimates of the losses of LED and LP would be more accurate than simply summing up the human capital loss due to mortality. To minimize the potential confounding due to different health literacy between those who received screen versus those who did not screen, we adopted the overall estimates of different stage distributions between the two compared groups to estimate the possible benefits of the mammography screening program conservatively. Although our modeling method did not include any covariates, we alternatively used the approaches above to manage them. Nonetheless, future studies stratified by different comorbidities and/or treatment modalities are still warranted to explore detailed conditions.

At least the following limitations must be acknowledged: First, most estimated losses of LED and LP of stages 0 and I were negative. They also showed wider 95% CIs, including zero, which can be interpreted as no differences between women with BC and the matched referents. To avoid overestimation, we assumed all of them to be zero when calculating the reductions by mammography screening. Second, although most previous reports usually collected employment‐related data directly from those employed, we calculated the employment ratios, including employed and non‐employed in the denominator, to examine the concerned population. In this way, constructing models to extrapolate throughout life would not require the variable of a person's willingness to work, which has not been collected regularly in claim data. As women with BC may drop out and re‐enter the job market dynamically, the longitudinal real‐world data would be the most accurate to reflect reality. Since the calculations are equally applied to those with BC and matched referents, the relative losses would be valid and practically feasible. The results would still be underestimated as we have not included the unpaid productivity losses. ⁴⁰ , ⁴¹ Future research needs to integrate unpaid work into macroeconomic policies to clearly view the dynamics of the expanded macroeconomy. Moreover, further studies are warranted to explore how BC diagnosis and treatment affect different occupational categories to formulate effective interventions for specific employment sectors (Table 1) and job types. Third, although Taiwan NHI has mandated the premium paid by each individual based on the actual salary since 2013, ⁴² which usually reflects about 70%—90% of the personal income, this leads to a general underestimation of the LP after diagnosis. Although we recruited matched referents to reduce biases and weighted by different stages and age strata, future studies are warranted to establish the association between insured salary and real income in the real world for a more accurate estimation. Fourth, we did not discount the productivity loss to the present value. Quantifying the relative losses may generate an overestimation, but the magnitude may be eliminated and compensated by using matched referents to quantify the relative losses. Fifth, Taiwan has a unique universal health coverage (UHC) system with strong support from each public health agency down to the township level, which may not be generalizable to other countries. However, the quantification of LP loss potentially reduced by mammography and/or other preventive healthcare services deserves more attention and efforts for making different health and insurance policies for countries with or without UHC system, such as the United States, of which the US Preventive Services Task Force (USPSTF) recommends mammography ⁶ but current regulation allows versatile options for women of working age.

5. CONCLUSION

Since women's health is vital to a nation, our study is only an initial step to quantify the benefits of mammography from a societal perspective, which deserves more attention when confronting a low fertility rate crisis. We thus conclude that mammography screening reduces the losses of lifetime duration of employment and overall productivity. Moreover, these reductions benefit women, their families, and the broader economy. Future studies are warranted to include these estimations in the cost‐effectiveness evaluation.

AUTHOR CONTRIBUTIONS

Chia‐Ni Lin: Conceptualization; methodology; validation; formal analysis; visualization; investigation; writing – original draft; data curation. Jung‐Der Wang: Conceptualization; methodology; validation; funding acquisition; writing – review and editing; resources; supervision; project administration. Wen‐Yen Huang: Methodology; conceptualization; writing – review and editing. Jing‐Shiang Hwang: Software; methodology; conceptualization; writing – review and editing. Li‐Jung Elizabeth Ku: Writing – review and editing; supervision. Fuhmei Wang: Supervision; writing – review and editing; conceptualization; methodology; validation.

FUNDING INFORMATION

This study was funded by the National Science and Technology Council, NSTC, Taiwan government grant “NSTC 112‐2627‐M‐006‐002,” “NSTC 113‐2627‐M‐006‐008.” The NSTC has no role in study design, data analysis, result interpretation s, or manuscript writing. The publication of study results was not contingent on the sponsor's approval or censorship of the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest relevant to this article.

ETHICS STATEMENT

The current study was approved by the Institutional Review Board, National Cheng Kung University Hospital (NCKUH; number A‐ER‐105‐191) before commencement.

Supporting information

Data S1

Lin C‐N, Wang J‐D, Huang W‐Y, Hwang J‐S, Ku L‐JE, Wang F. Mammography screening reduced lifetime loss of employment duration and productivity for women with breast cancer: Real‐world evidence of societal impacts. Int J Cancer. 2025;157(11):2385‐2398. doi: 10.1002/ijc.70051

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the Health and Welfare Data Science Center (HWDC) of Taiwan in a confidentially secured area following an application review process and approval. Further information is available from the corresponding author upon request.