Association between non-communicable diseases and work productivity and health-related quality of life among Japanese workers: a cross-sectional study using a claims database and survey data

Abstract

Objectives

To assess the impact of non-communicable diseases (NCDs) by number and pattern on productivity loss costs (PLCs) and health-related quality of life (HRQoL) in a large working-age population in Japan.

Design

Cross-sectional analysis of linked claims—survey data with sex-stratified multivariable regression.

Setting

Employment-Based Health Insurance Associations in Japan. Administrative claims using International Classification of Diseases, Tenth Revision codes were linked to an online questionnaire (November 2022).

Participants

Insured workers and families aged 18–65 years who completed the survey (n=18 236; mean age 48.7 years; 66.2% men).

Primary and secondary outcome measures

Primary: PLC, estimated from the Work Productivity and Activity Impairment instrument using the Human Capital Method (annual US$).

Secondary

HRQoL measured using the EuroQoL 5-Dimension 5-Level index. Exposures were NCD categories, multimorbidity counts and high-impact disease combinations. Models adjusted for sex, age and the number of other conditions.

Results

Multimorbidity (≥2 conditions) was present in 58.4% (56.3% men; 62.5% women); 29.3% had ≥5 conditions. Women had higher NCD prevalence (76.9% vs 70.0%), higher median PLC (US$3683 vs US$117) and lower median HRQoL (0.89 vs 1.00) than men. In adjusted analyses, neuropsychiatric conditions were associated with higher PLC in both sexes (men: US$2393 (95% CI 1966 to 2820); women: US$2097 (US$1442 to US$2751)). Musculoskeletal disorders were statistically significant in men but not in women (US$809 (US$381 to US$1236)), whereas genitourinary disorders were statistically significant in women but not in men (US$701 (US$121 to US$1282)). Neuropsychiatric disorders were associated with lower HRQoL in both sexes (men: −0.040 (−0.045 to −0.034); women: −0.044 (−0.053 to −0.035)); musculoskeletal disorders showed similar decrements (men: −0.031 (−0.037 to −0.025); women: −0.040 (−0.049 to −0.031)). Combinations involving sleep and neuropsychiatric disorders produced the highest PLC (eg, depressive episode and sleep disorder: men: US$5219.9 (US$3646.4 to US$6793.4); women: US$5858.5 (US$2634.4 to US$9082.5)). Each additional condition was independently associated with higher PLC and lower HRQoL.

Conclusions

In this cross-sectional study of Japanese workers, specific combinations of chronic conditions, particularly neuropsychiatric conditions co-occurring with sleep disorders and musculoskeletal disorders, were associated with greater productivity loss and lower HRQoL than predicted by the number of co-occurring diseases alone. These findings suggest that workplace health strategies addressing high-impact disease combinations may warrant further investigation in longitudinal studies.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The first Japanese study to link nationwide employment-based health insurance claims (International Classification of Diseases, Tenth Revision coded) with patient-reported outcomes (the Work Productivity and Activity Impairment-General Health questionnaire and the EuroQoL 5-Dimension 5-Level questionnaire) in a large working-age cohort, enabling simultaneous analysis of co-occurring non-communicable disease patterns, productivity loss costs and health-related quality of life.

• The cross-sectional design with a 3-month claims window limits causal inference and the low survey participation rate (3.0% of eligible individuals) among kencom application users may introduce selection bias towards individuals with higher health literacy, which may limit the generalisability of our findings to the broader Japanese working-age population.

Introduction

Non-communicable diseases (NCDs) are chronic, disabling conditions caused by a combination of genetic, physiological, environmental and behavioural factors. According to the WHO, NCDs are the leading cause of death globally, accounting for 75% of non-pandemic-related deaths worldwide (43 million deaths annually).^{1 2} Among these, cardiovascular diseases account for 44% of NCD-related deaths, followed by cancer (23%), chronic respiratory diseases (10%) and diabetes mellitus (4%). In the working-age population (below age 70), NCDs are responsible for 17 million premature deaths.² Additionally, NCDs place a heavy burden on individuals through healthcare costs, reduced work capacity and financial insecurity—all of which reduce quality of life (QoL).³

The coexistence of multiple chronic conditions, often referred to as multimorbidity, is increasing in prevalence.⁴ This phenomenon imposes a growing burden, characterised by reduced health-related QoL (HRQoL),⁵ increased healthcare needs and costs and elevated mortality.^{6 7} The accumulation of these overlapping conditions typically progresses from midlife onward, though exposure to cardiovascular risk factors in early adulthood can significantly accelerate this process.⁸ This clinical progression is further driven by a combination of sociodemographic variables (older age, female sex, lower education) and modifiable lifestyle factors (obesity, excessive alcohol consumption, physical inactivity).^{9 10} Yet despite this extensive characterisation of the aetiology and growing public health impact of multimorbidity, no clear or consistent patterns for its management have been identified to date.¹¹

The necessity of establishing these management strategies is particularly acute in nations facing severe demographic shifts. In Japan, for instance, the population is ageing rapidly; in 2024, 29.1% of its citizens were aged 65 or older, and this proportion is projected to reach 38.5% by 2070.¹² Meanwhile, the working-age population (15–64 years) is decreasing owing to low birth rates and is projected to decline to 52.75 million by 2050 (a 29.2% decrease from 2021 levels).^{13 14} It is therefore crucial for Japan, and for any country in a similar situation, to maximise the health and productivity of their working-age populations. A recent study estimated that approximately 30% of working-age adults in Japan have at least one chronic disease,¹⁵ underscoring the importance of preventing and controlling chronic diseases in the workforce. However, most existing studies assess multimorbidity using simple disease counts, and evidence on how specific patterns of co-occurring NCDs affect work productivity and HRQoL remains limited, particularly in Asian working-age populations. Furthermore, few studies have linked administrative claims data with patient-reported outcomes (PROs) to simultaneously examine clinical diagnoses and their functional consequences in employed adults. Therefore, this study aimed to assess the impact of the quantity and specific patterns of NCDs on work productivity loss and HRQoL among working adults in Japan, using linked claims and survey data.

Methods

Study design and data sources

We conducted a cross-sectional, secondary analysis using the DeSC database (established in 2021), which contains health insurance claims, health examination data, and survey data (DeSC Healthcare, Tokyo, Japan). The DeSC database includes data from the three main insurance schemes of Japan: National Health Insurance, Social Health Insurance and the Medical Care System for the Elderly. The database covers approximately 15.4 million patients,^{16 17} providing a broadly representative sample of the Japanese population. For this study, we included only employer-sponsored health insurance associations, covering employees of large companies and their dependents. As of June 2023, the DeSC database contained claims data for 1 142 905 individuals from these associations, spanning April 2014 to November 2022. Of these, data on approximately 950 000 employees aged 19–74 years and their dependent family members were available for secondary use.

The claims data within the DeSC database include the following information: (1) unique identifiers; (2) age and sex; (3) diagnoses based on the International Classification of Diseases, Tenth Edition (ICD-10) codes (online supplemental table 1); (4) medical procedures; (5) medications dispensed in accordance with the Anatomical Therapeutic Chemical Classification System and (6) coverage period, indicating the duration of coverage. The DeSC database links data from online surveys (conducted by each insurer for health promotion purposes) to insured individuals, a process facilitated by DeSC Healthcare.

The survey data used in the present study came from an online questionnaire conducted by DeSC in 1–30 November 2022. This survey was part of a health-promotion initiative and was offered to insured individuals who had voluntarily registered with ‘kencom’,¹⁸ a health support application managed by DeSC.

Approximately 230 000 insured individuals were registered users of kencom. The survey included the following validated instruments: the Work Productivity and Activity Impairment—General Health questionnaire, V.2.2 (WPAI-GH V.2.2)¹⁹ and the EuroQoL 5-Dimension 5-Level questionnaire (EQ-5D-5L).²⁰ In addition, the survey included questions on lifestyle (smoking, alcohol consumption, physical activity, sleep habits) and symptoms of depression, anxiety and constipation. For each survey respondent, their claims data for the survey month and the two preceding months (1 September to 30 November 2022) were extracted. A 3-month window was chosen because it is a standard practice in Japan to visit a physician for prescription renewal every 90 days for typical chronic conditions.²¹ These claims provided individual-level information on demographics and any healthcare services used (inpatient, outpatient or pharmacy). The complete list of survey items used in the kencom questionnaire is provided in the online supplemental appendix. Of the 603 620 eligible insured workers in the underlying employment-based health insurance associations, 21 613 (3.6%, survey respondents) completed the survey, corresponding to approximately 7.9% of the kencom registered users.

Study population

The study population included insured participants aged 18–65 years as of 1 November 2022, registered in the DeSC database and who responded to the questionnaire during the study period. Insurance enrolment/withdrawal, age and sex were determined from the insurer enrolment information database. Individuals outside the age range, those who did not complete the questionnaire themselves or gave contradictory answers and those with missing WPAI data were excluded. The detailed study design is presented in online supplemental figure 1.

Variables

Diseases and their categories

Diagnoses were identified by their ICD-10 codes in the claims data, excluding any entries labelled as ‘suspected’. All ICD-10 coded diagnoses recorded during the observation period were included rather than limiting diagnoses to a pre-specified list. In the analysis, diseases were classified at three hierarchical levels: (a) the ICD-10 three-digit category level, (b) the two-digit chapter level and (c) the specific full ICD-10 code. NCD categories were defined using ICD-10 codes according to WHO definitions^{22 23} (online supplemental table 1).

Multimorbidity is generally defined as the coexistence of two or more chronic diseases, but there is still no worldwide agreement on which chronic diseases should be counted, nor is there a consensus on the minimum number of concurrent diseases required for classification.²⁴ Therefore, each study selected diseases according to the purpose of the study and characteristics of the data to be used.^{1 25–27} A systematic review of studies on the prevalence of multimorbidity reported that estimates varied from <5% to 95% depending on the definition.²⁸ Hence, we defined multimorbidity as the presence of two or more disease categories at the level of the top three digits of the ICD-10 code.

Outcomes

The primary outcomes were work productivity loss and HRQoL, both obtained from the online survey. Work productivity was assessed using the WPAI-GH V.2.2,¹⁹ which quantifies how health problems affect work and daily activities over the prior 7 days. The WPAI yields four domain scores: absenteeism (percentage of work time missed owing to health problems), presenteeism (percentage reduction in on-the-job effectiveness), overall work impairment (combined absenteeism and presenteeism) and activity impairment (percentage reduction in daily activities other than work). Higher WPAI scores indicate greater impairment. Productivity loss costs (PLCs) were defined as the estimated annual monetary value of work productivity lost owing to health-related absenteeism and presenteeism, calculated from the WPAI-GH instrument using the Human Capital Method. The 7-day WPAI scores were annualised by multiplying by 52 weeks. Annual PLC was calculated as: ((hours missed owing to absenteeism)+(hours worked×presenteeism percentage))×52×average hourly wage (Japanese yen) following the Human Capital Method.^{29 30} These costs were converted from Japanese yen to US$ at the 1 November 2024 exchange rate of 153.02 JPY/US$. HRQoL was measured using the EQ-5D-5L,²⁰ which assesses the current health state of the respondent. EQ-5D index scores were calculated using the validated Japanese value set.³¹ Scores range from −0.025 to 1, where 1 indicates full health and 0 represents a state equivalent to death.

Covariates

Covariates included age and sex (at the time of survey) and the presence of any diagnosed diseases during the observation period. Age, sex and insurance enrolment status were obtained from insurer records.

Sample size

All eligible participants were invited to participate in the survey; therefore, no a priori sample size calculation was performed.

Data analysis

Data on age were summarised using mean and SD, whereas WPAI scores, PLC and HRQoL had skewed distributions and were summarised with median and IQR. Participants with missing WPAI or HRQoL data were excluded from analyses. The overall proportions of those who fell into each status were calculated, as were their WPAI scores and HRQoL. Participants were classified into the following categories: ‘No diseases’, ‘one or more diseases without an NCD category’, ‘one or more of the NCD category’ and ‘NCD category’, ‘multimorbidity of more than two diseases, counted by ICD-10 code 3 digits’ and ‘multimorbidity of two, three, four and five, or more diseases’. Median and IQR were calculated as a PLC.

Multiple linear regressions were performed to determine the association between disease status and outcomes (PLC and HRQoL). Analyses were stratified by sex a priori, given the well-documented sex differences in NCD prevalence, multimorbidity patterns and work productivity.^32–34 Explanatory variables (all coded as binary presence/absence) included the following: no chronic disease (reference), any non-NCD chronic condition, any NCD, each specific NCD category and indicators for multimorbidity (having ≥2, ≥3, ≥4 or ≥5 conditions). Models were adjusted for age and the total number of conditions (and for sex in combined analyses). Additionally, a detailed analysis of specific disease combinations was conducted. For the combination analysis, we selected the seven NCD categories that showed significant associations with PLC in the initial regression analysis (in either sex): diabetes/endocrine, neuropsychiatric, sense-organ, cardiovascular, genitourinary, musculoskeletal and oral disorders. The pairwise combinations of conditions within these categories were examined at the 3-digit ICD-10 level. These combination models were adjusted for age and included a covariate representing the count of other co-occurring diseases (defined as the number of additional 3-digit ICD-10 conditions aside from the combination of interest). A formal statistical comparison of sex differences (eg, interaction tests) was not performed; any observed differences between sexes should be interpreted descriptively.

Data processing was performed using Amazon Athena (V.3+), and statistical analyses were conducted in R (V.4.3.2+).

Results

Characteristics of the study population

Out of 603 620 eligible insured persons, 18 236 (3.0%) met the inclusion criteria (figure 1). The mean (SD) age was 48.7 (9.8) years, and 66.2% of participants were men. Multimorbidity (≥2 chronic conditions) was common, affecting 58.4% of the sample; 29.3% of all participants had five or more conditions, whereas 25.3% had none. The most common disease categories overall were oral, endocrine (including diabetes), digestive, cardiovascular and sense-organ disorders (table 1). When stratified by sex, women had higher NCD prevalence than men (76.9% vs 70.0%) and higher multimorbidity prevalence (62.5% vs 56.3%). Women also had greater median annual PLC (approximately US$3683 vs US$117 in men) and slightly lower median HRQoL (0.89 vs 1.00).

Figure 1. Flow diagram.

Table 1. Characteristics of study population.

	Overall		Male		Female
	18 236	100%	12 080	100%	6156	100%
Sex, male			12 080	66.2%	6156	33.8%
Age, mean (SD)	48.7	(9.8)	49.9	(9.6)	46.4	(9.9)
WPAI scores*, median (IQR)
Presenteeism	10.0	(30.0)	0.0	(20.0)	10.0	(30.0)
Absenteeism	0.0	(0.0)	0.0	(0.0)	0.0	(0.0)
Productivity loss costs†, median (IQR)	2985.6	(8 720.4)	117.2	(7 715.2)	3682.7	(11 048.2)
HRQoL‡, median (IQR)	0.89	(0.17)	1.00	(0.13)	0.89	(0.18)
No diseases	4612	25.3%	3335	27.6%	1277	20.7%
Those who do not fall under any NCD categories	438	2.4%	293	2.4%	145	2.4%
Those who fall under one or more of the NCD categories	13 186	72.3%	8452	70.0%	4734	76.9%
Malignant neoplasms	429	2.4%	224	1.9%	205	3.3%
Other neoplasms	995	5.5%	343	2.8%	652	10.6%
Diabetes mellitus and endocrine disorders	4313	23.7%	3080	25.5%	1233	20.0%
Neuropsychiatric conditions	2756	15.1%	1813	15.0%	943	15.3%
Sense organ diseases	3281	18.0%	1944	16.1%	1337	21.7%
Cardiovascular diseases	3243	17.8%	2517	20.8%	726	11.8%
Respiratory diseases	2902	15.9%	1862	15.4%	1040	16.9%
Digestive diseases	4270	23.4%	2893	23.9%	1377	22.4%
Genitourinary diseases	2031	11.1%	846	7.0%	1185	19.2%
Skin diseases	2748	15.1%	1572	13.0%	1176	19.1%
Musculoskeletal diseases	3010	16.5%	1908	15.8%	1102	17.9%
Congenital anomalies	146	0.8%	80	0.7%	66	1.1%
Oral conditions	6555	35.9%	4120	34.1%	2435	39.6%
Multimorbidity§ ≥2	10 655	58.4%	6805	56.3%	3850	62.5%
2	2298	12.6%	1491	12.3%	807	13.1%
3	1717	9.4%	1133	9.4%	584	9.5%
4	1296	7.1%	822	6.8%	474	7.7%
≥5	5344	29.3%	3359	27.8%	1985	32.2%

The values represent n (%) unless otherwise stated.

^*The Work Productivity and Activity Impairment—General Health (WPAI-GH) Scores.

^†Calculated in Japanese yen and converted to 153.02 yen per US$ as of 1 November 2024.

^‡Health-related quality of life (HRQoL): measured using the EuroQoL Group 5 Dimension (EQ-5D-5L) questionnaire.

^§Counted by ICD-10 code 3 digits.

ICD-10, International Classification of Diseases, Tenth Revision; NCD, non-communicable disease.

Burden of WPAI/HRQoL

Participants with neuropsychiatric disorders exhibited the highest PLC in both sexes (table 2 and online supplemental table 2). For all diseases, the PLC was higher for women than for men, with the figures being particularly high for genitourinary diseases. The median HRQoL (EQ-5D index) was approximately 0.89 overall. Only women with neuropsychiatric or musculoskeletal disorders had a slightly lower median HRQoL (~0.87).

Table 2. Proportion of participants in each status, productivity loss costs and HRQoL, stratified by sex.

	Male						Female
	Proportion in overall		Productivity loss costs*		HRQoL†		Proportion in overall		Productivity loss costs*		HRQoL†
Overall	12 080	(100.0%)	117.2	(7715.2)	1.00	(0.13)	6156	(100.0%)	3682.7	(11 048.2)	0.89	(0.18)
No diseases	3335	(27.6%)	0.0	(7682.0)	1.00	(0.13)	1277	(20.7%)	3532.9	(9972.2)	0.89	(0.17)
All those who fall under one or more without the NCDs category	293	(2.4%)	2906.8	(7715.2)	1.00	(0.13)	145	(2.4%)	3682.7	(11 048.2)	0.89	(0.18)
All those who fall under one or more of the NCDs category	8452	(70.0%)	2354.5	(7715.2)	0.89	(0.17)	4734	(76.9%)	3841.0	(11 048.2)	0.89	(0.18)
Malignant neoplasms	224	(1.9%)	2772.2	(8720.4)	0.89	(0.18)	205	(3.3%)	3841.0	(11 523.0)	0.89	(0.18)
Other neoplasms	343	(2.8%)	2906.8	(8369.2)	0.89	(0.17)	652	(10.6%)	3841.0	(11 523.0)	0.89	(0.18)
Diabetes mellitus and endocrine disorders	3080	(25.5%)	0.0	(7715.2)	0.89	(0.17)	1233	(20.0%)	3841.0	(11 523.0)	0.89	(0.18)
Neuropsychiatric conditions	1813	(15.0%)	3841.0	(11 572.8)	0.89	(0.18)	943	(15.3%)	7365.4	(14 730.9)	0.87	(0.23)
Sense organ diseases	1944	(16.1%)	2883.1	(7715.2)	0.89	(0.17)	1337	(21.7%)	3841.0	(11 048.2)	0.89	(0.18)
Cardiovascular diseases	2517	(20.8%)	0.0	(7715.2)	0.89	(0.17)	726	(11.8%)	3682.7	(10 598.6)	0.89	(0.18)
Respiratory diseases	1862	(15.4%)	2985.6	(9972.2)	0.89	(0.18)	1040	(16.9%)	3857.6	(13 187.5)	0.89	(0.18)
Digestive diseases	2893	(23.9%)	2906.8	(8720.4)	0.89	(0.18)	1377	(22.4%)	3857.6	(11 572.8)	0.89	(0.18)
Genitourinary diseases	846	(7.0%)	2906.8	(8470.0)	0.89	(0.17)	1185	(19.2%)	5813.6	(12 541.7)	0.89	(0.18)
Skin diseases	1572	(13.0%)	2985.6	(8720.4)	0.89	(0.18)	1176	(19.1%)	3857.6	(11 523.0)	0.89	(0.18)
Musculoskeletal diseases	1908	(15.8%)	3532.9	(11 048.2)	0.89	(0.18)	1102	(17.9%)	3857.6	(11 572.8)	0.87	(0.22)
Congenital anomalies	80	(0.7%)	3428.5	(7715.2)	0.89	(0.18)	66	(1.1%)	3682.7	(11 441.6)	0.89	(0.21)
Oral conditions	4120	(34.1%)	23.3	(7715.2)	1.00	(0.13)	2435	(39.6%)	3841.0	(11 048.2)	0.89	(0.18)

Values represent n (%) for proportions and median (IQR) for productivity loss costs and HRQoL.

Mean values (male/female): 751.4/934.3, 803.6/997.9, 801.2/1089.5 and 798.0/1054.7 for ‘No diseases’, ‘Diabetes mellitus and endocrine disorders’, ‘Cardiovascular diseases’ and ‘Oral conditions’, respectively.

^*Calculated in Japanese yen and converted to 153.02 yen per US$ as of 1 November 2024.

^†Health-related quality of life (HRQoL): measured using the EuroQoL Group 5 Dimension (EQ-5D-5L) questionnaire.

^‡Counted by ICD-10 code 3 digits.

ICD-10, International Classification of Diseases, Tenth Revision; NCD, non-communicable disease.

Association with PLC and HRQoL

Neuropsychiatric conditions were associated with higher PLC in both sexes (β for men=US$2393 (95% CI US$1966 to US$2820); β for women=US$2097 (95% CI US$1442 to US$2751)). Musculoskeletal disorders were significantly associated with increased PLC in men (β=US$809 (95% CI US$381 to US$1236)) but not in women, whereas genitourinary disorders were significantly associated with increased PLC in women (β=US$701 (95% CI US$121 to US$1282)) but not in men. Each additional chronic condition was associated with a significant increase in PLC. Neuropsychiatric disorders were associated with lower HRQoL in both sexes (β=−0.040 in men; β=−0.044 in women), as were musculoskeletal disorders (men: β=−0.031; women: β=−0.040). Respiratory disorders were associated with a small HRQoL decrease in women (β=−0.010) but showed no significant association in men. Diabetes/endocrine, cardiovascular, sense-organ, skin and oral disorders were associated with comparatively lower PLC or higher HRQoL. As expected, having more chronic conditions was consistently associated with worse outcomes (higher work impairment and lower HRQoL) (online supplemental table 3).

High-impact disease combinations

We further investigated the specific combinations of conditions within seven broad NCD categories (table 3): diabetes/endocrine, neuropsychiatric, sense-organ, cardiovascular, genitourinary, musculoskeletal and oral disorders. Conditions in these categories were analysed at the three-digit ICD-10 level to identify high-impact pairs (table 4). In male participants, seven specific disease combinations were associated with a significantly higher PLC. The greatest cost impacts were observed in combinations involving a sleep disorder with another neuropsychiatric condition: depressive episode combined with sleep disorder (β=US$5220 (95% CI US$3646 to US$6793)) and anxiety disorder combined with sleep disorder (β=US$5611 (95% CI US$2586 to US$8636)). Other notable high-PLC combinations in men included migraine (β=US$3518 (95% CI US$1113 to US$5924)), other arrhythmia (β=US$3183 (95% CI US$777 to US$5589)), chronic back pain and having a sleep disorder alone. Among female participants, the highest PLC was again found for depressive episode combined with sleep disorder (β=US$5859 (95% CI US$2634 to US$9083)). Additionally, significant cost impacts were observed for otitis externa (β=US$3756 (95% CI US$1028 to US$6485)) and other anxiety disorder (β=US$3475 (95% CI US$745 to US$6206)). Certain endocrine and genitourinary conditions were costly: volume depletion (hypovolemia) in a patient with diabetes (β=US$4075 (95% CI US$465 to US$7684)) and menopausal disorders (β=US$1507 (95% CI US$76 to US$2939)).

Table 3. Association of those who fall into each status and their productivity loss costs and HRQoL, stratified by sex.

	Adjusted β¶ (95% CI)		Adjusted β¶ (95% CI)
	Male		Female
	Productivity loss costs*	HRQoL†	Productivity loss costs	HRQoL
No diseases	−189.4 (−538.3 to 159.6)	0.003 (−0.002 to 0.008)	−194.7 (−757.6 to 368.2)	0.005 (−0.003 to 0.013)
Those who do not fall under any NCD categories	138.4 (−745.2 to 1022.1)	−0.008 (−0.021 to 0.004)	1026.4 (−338.4 to 2391.3)	−0.022 (−0.041 to −0.002)
Those who fall under one or more NCD categories	166.9 (−181.1 to 515.0)	0.002 (−0.007 to 0.003)	19.3 (−532.0 to 570.6)	−0.001 (−0.009 to 0.007)
Malignant neoplasms	−154.5 (−1,177.2 to 868.2)	0.010 (−0.004 to 0.025)	95.0 (−1,078.9 to 1269.0)	0.012 (−0.005 to 0.029)
Other neoplasms	−9.6 (−843.0 to 823.9)	0.006 (−0.006 to 0.018)	−98.0 (−807.4 to 611.4)	0.007 (−0.003 to 0.017)
Diabetes mellitus and endocrine disorders	−722.9 (−1,106.8 to −339.0)	0.011 (0.006 to 0.016)	−322.1 (−915.4 to 271.2)	0.005 (−0.004 to 0.013)
Neuropsychiatric conditions	2393.0 (1965.7 to 2820.3)	−0.040 (−0.045 to −0.034)	2096.8 (1442.2 to 2751.4)	−0.044 (−0.053 to −0.035)
Sense organ diseases	−457.0 (−866.4 to −47.7)	0.015 (0.009 to 0.020)	−399.4 (−961.1 to 162.4)	0.007 (−0.001 to 0.015)
Cardiovascular diseases	−571.5 (−977.4 to −165.7)	0.005 (−0.001 to 0.010)	−1,077.1 (−1,797.9 to −356.3)	0.001 (−0.009 to 0.012)
Respiratory diseases	268.9 (−149.0 to 686.9)	0.003 (−0.003 to 0.008)	520.4 (−92.9 to 1133.7)	−0.010 (-0.019 to -0.001)
Digestive diseases	224.1 (−172.0 to 620.3)	−0.001 (−0.006 to 0.005)	274.9 (−319.6 to 869.5)	−0.007 (-0.016 to 0.001)
Genitourinary diseases	−336.7 (−905.6 to 232.2)	0.016 (0.008 to 0.024)	701.4 (120.7 to 1282.1)	−0.003 (-0.011 to 0.006)
Skin diseases	−239.3 (−672.8 to 194.1)	0.006 (−0.000 to 0.012)	−283.5 (−868.3 to 301.3)	0.015 (0.007 to 0.024)
Musculoskeletal diseases	808.6 (381.1 to 1236.0)	−0.031 (−0.037 to −0.025)	498.8 (−129.4 to 1127.0)	−0.040 (-0.049 to -0.031)
Congenital anomalies	−607.8 (−2,279.6 to 1064.1)	0.001 (−0.023 to 0.024)	−301.0 (−2,314.9 to 1712.8)	−0.007 (-0.036 to 0.022)
Oral conditions	−562.2 (−859.0 to −265.5)	0.012 (0.008 to 0.016)	−216.3 (−652.5 to 219.9)	0.013 (0.007 to 0.019)
Adjustment variables§
Age	−114.5 (−129.1 to −99.9)	0.000 (0.000 to 0.001)	−99.9 (−121.0 to −78.7)	0.001 (0.000 to 0.001)
Number of diseases‡	197.3 (163.0 to 231.6)	−0.004 (−0.005 to −0.004)	193.1 (146.9 to 239.2)	−0.004 (-0.005 to -0.004)

^*Calculated in Japanese yen and converted to 153.02 yen per US$ as of 1 November 2024.

^†Health-related quality of life (HRQoL): measured using the EuroQoL Group 5 Dimension (EQ-5D-5L) questionnaire.

^‡Disease combinations are counted by ICD-10 code three digits.

^§Age and number of diseases are continuous values; the others are binary.

^¶Adjusted by age and number of diseases.

ICD-10, International Classification of Diseases, Tenth Revision; NCD, non-communicable disease.

Table 4. Association of relevant disease combinations for productivity loss costs and HRQoL, stratified by sex.

Male
	NCD category		Disease combination patterns	Proportion
				12 080	(100.0%)	Adjusted β	95% CI
Productivity loss costs	Neuropsychiatric conditions	F41:G47	Other anxiety disorders: Sleep disorder	24	(1.32%)	5611.0	(2586.0	to	8636.1)
	Neuropsychiatric conditions	F32:G47	Depressive episode: Sleep disorder	89	(4.91%)	5219.9	(3646.4	to	6793.4)
	Neuropsychiatric conditions	G43	Migraine	38	(2.10%)	3518.2	(1112.7	to	5923.7)
	Neuropsychiatric conditions	F32	Depressive episode	43	(2.37%)	3200.1	(938.5	to	5461.8)
	Cardiovascular diseases	I49	Other arrhythmias	38	(1.51%)	3183.1	(777.4	to	5588.8)
	Musculoskeletal diseases	M54	Back pain	294	(15.41%)	1541.2	(658.6	to	2423.8)
	Neuropsychiatric conditions	G47	Sleep disorders	429	(23.66%)	841.9	(106.5	to	1577.4)
HRQoL	Neuropsychiatric conditions	F41:G47	Other anxiety disorders: Sleep disorder	24	(1.32%)	−0.094	(−0.136	to	−0.052)
	Musculoskeletal diseases	M17	Knee osteoarthritis (knee joint disease)	43	(2.25%)	−0.070	(−0.102	to	−0.039)
	Neuropsychiatric conditions	F32:G47	Depressive episode: Sleep disorder	89	(4.91%)	−0.064	(−0.086	to	−0.042)
	Neuropsychiatric conditions	G64	Other disorders of the peripheral nervous system	56	(3.09%)	−0.047	(−0.074	to	−0.019)
	Musculoskeletal diseases	M51	Other intervertebral disc disorders	47	(2.46%)	−0.044	(−0.074	to	−0.014)
	Neuropsychiatric conditions	F32	Depressive episode	43	(2.37%)	−0.037	(−0.069	to	−0.006)
	Oral conditions	K05:K12	Gingivitis and periodontal disease: Stomatitis and related disorders Type 2 (non-insulin-dependent) diabetes mellitus (NIDDM): Lipoprotein (protein) metabolism disorders and other dyslipidaemias	45	(1.09%)	−0.034	(−0.065	to	−0.003)
	Musculoskeletal diseases	M79	Other soft tissue disorders, unspecified	60	(3.14%)	−0.029	(−0.056	to	−0.002)
	Neuropsychiatric conditions	G98	Other disorders of the nervous system, not elsewhere classified	60	(3.31%)	−0.028	(−0.054	to	−0.001)
	Musculoskeletal diseases	M47	Spondylosis	93	(4.87%)	−0.025	(−0.047	to	−0.004)
	Musculoskeletal diseases	M54	Back pain	294	(15.41%)	−0.022	(−0.034	to	−0.010)
	Diabetes mellitus and endocrine disorders	E11:E78	Type 2 (non-insulin-dependent) diabetes mellitus (NIDDM): Lipoprotein (protein) metabolism disorders and other dyslipidaemias	180	(5.84%)	−0.018	(−0.033	to	−0.002)
	Oral conditions	K04:K05	Diseases of the dental pulp and periodontal tissues: Gingivitis and periodontal disease	215	(5.22%)	−0.018	(−0.032	to	−0.004)

Female
				Proportion
	NCD category		Disease combination patterns	6156	(100.0%)	Adjusted β	95% CI
Productivity loss costs	Neuropsychiatric conditions	F32:G47	Depressive episode: Sleep disorder	25	(2.65%)	5858.5	(2634.4	to	9082.5)
	Diabetes mellitus and endocrine disorders	E86	Decreased body fluid volume (syndrome)	20	(1.62%)	4074.7	(465.2	to	7684.3)
	Sense organ diseases	H60	Otitis externa	35	(2.62%)	3756.1	(1027.7	to	6484.5)
	Neuropsychiatric conditions	F41	Other anxiety disorders	35	(3.71%)	3475.4	(744.5	to	6206.4)
	Genitourinary diseases	N95	Menopause and other perimenopausal disorders	132	(11.14%)	1507.4	(76.0	to	2938.9)
HRQoL	Musculoskeletal diseases	M17	Knee osteoarthritis	29	(2.63%)	−0.121	(−0.164	to	−0.079)
	Sense organ diseases	H60	Otitis externa	35	(2.62%)	−0.057	(−0.096	to	−0.018)
	Musculoskeletal diseases	M75	Shoulder injury	43	(3.90%)	−0.055	(−0.091	to	−0.020)
	Neuropsychiatric conditions	G98	Other disorders of the nervous system, not elsewhere classified	23	(2.44%)	−0.055	(−0.103	to	−0.007)
	Neuropsychiatric conditions	G43	Migraine	63	(6.68%)	−0.049	(−0.078	to	−0.020)
	Diabetes mellitus and endocrine disorders	E14	Diabetes, unspecified	29	(2.35%)	−0.047	(−0.090	to	−0.004)
	Neuropsychiatric conditions	F32:G47	Depressive episode: Sleep disorder	25	(2.65%)	−0.047	(−0.093	to	−0.001)
	Oral conditions	K05:K07	Gingivitis and periodontal disease: Craniofacial (congenital) anomalies (including malocclusion)	37	(1.52%)	−0.039	(−0.077	to	−0.001)
	Genitourinary diseases	N80:N94	Endometriosis: Pain and other conditions related to the female reproductive organs and menstrual cycle	50	(4.22%)	−0.034	(−0.067	to	−0.001)
	Sense organ diseases	H10	Conjunctivitis	132	(9.87%)	−0.027	(−0.047	to	−0.006)
	Musculoskeletal diseases	M54	Back pain	168	(15.25%)	−0.022	(−0.040	to	−0.003)

Calculated in Japanese yen and converted to 153.02 yen per US$ as of 1 November 2024.

Health-related quality of life (HRQoL): measured using the EuroQoL Group 5 Dimension (EQ-5D-5L) questionnaire.

Disease combinations are counted by ICD-10 code three digits.

Age and number of diseases are continuous values; the others are binary.

Adjusted by age and number of diseases.

ICD-10, International Classification of Diseases, Tenth Revision; NCD, non-communicable disease.

In male participants, 13 specific disease combinations were associated with a significantly lower HRQoL. The largest HRQoL decreases were observed in the same sleep-related neuropsychiatric pairs: anxiety disorder combined with sleep disorder (β=–0.094 (95% CI –0.136 to –0.052)) and depressive episode combined with sleep disorder (β=–0.064 (95% CI –0.086 to –0.042)). Musculoskeletal conditions were prominent: for example, knee osteoarthritis (β=–0.070 (95% CI –0.102 to –0.039)) and intervertebral disc disorders (β=–0.044 (95% CI –0.074 to –0.014)) were associated with substantial HRQoL deficits. Certain endocrine (diabetes-related) and oral conditions likewise showed significant HRQoL impairment in men. In female participants, 11 combinations were associated with a lower HRQoL. The greatest reduction was observed for knee osteoarthritis (β=–0.121 (95% CI –0.164 to –0.079)). Other notable HRQoL decrements were linked to otitis externa (β=–0.057 (95% CI –0.096 to –0.018)), other nervous system disorders (β=–0.055 (95% CI –0.103 to –0.007)), shoulder disorders (β=–0.055 (95% CI –0.091 to –0.020)), migraine (β=–0.049 (95% CI –0.078 to –0.020)) and the combination of depressive episode with sleep disorder (β=–0.047 (95% CI –0.093 to –0.001)). Several other conditions (eg, unspecified diabetes, endometriosis, lower back pain, other neurological disorders, craniofacial/oral anomalies, conjunctivitis) were associated with smaller HRQoL declines.

The complete regression results for all of the evaluated disease combinations, including those associated with better outcomes, are available in online supplemental tables 4.1 and 4.2.

Discussion

In this cross-sectional study of 18 236 employed adults enrolled in Japanese employment-based health insurance associations, we found that specific patterns of co-occurring NCDs, particularly those involving neuropsychiatric and musculoskeletal conditions, were associated with greater work productivity loss and lower HRQoL. These associations were independent of the total number of co-occurring conditions, suggesting that the specific combination of diseases, rather than the simple count, may be relevant to understanding the functional burden of multimorbidity in working populations.

Over half of the working population in this study had multimorbidity, including nearly 30% with five or more conditions. The relatively high prevalence of multimorbidity compared with other studies may partly reflect our method of counting diseases at the ICD-10 three-digit level. Simply having more conditions was associated with greater productivity loss and lower HRQoL, consistent with prior reports.^{5 35–37} However, our findings suggest that the specific interplay of certain diseases, rather than the count alone, is associated with poor outcomes. These findings may inform the development of integrated care models³⁸ or trajectory-based interventions^{6 34} that identify disease combinations warranting further investigation in longitudinal studies, although longitudinal evidence is needed to confirm causal pathways, particularly for combinations involving neuropsychiatric or musculoskeletal conditions.

Neuropsychiatric and musculoskeletal conditions showed the strongest associations with work productivity loss and reduced HRQoL in our study, despite their relatively lower prevalence.³⁹ This is consistent with prior reports that mental health and musculoskeletal disorders are associated with diminished QoL.^40–43 In particular, neuropsychiatric disorders, particularly sleep disorders, were associated with substantial productivity loss. When sleep disorders co-occurred with depression or anxiety, the associated burden was even greater. This finding aligns with longitudinal research showing that depressive symptoms in young adults predict faster accumulation of chronic conditions,⁸ and that short sleep duration is linked to higher risks of chronic disease and later multimorbidity.^44–46 Our cross-sectional findings suggest that the co-occurrence of sleep disorders with neuropsychiatric conditions is associated with substantially greater productivity loss, highlighting a potential target for future interventional research. Musculoskeletal conditions (eg, chronic back pain and knee osteoarthritis) were individually associated with reduced HRQoL, consistent with evidence that these conditions impair work performance in midlife.^47–49 Interestingly, knee osteoarthritis did not show a significant association with PLC in our working sample, possibly reflecting a ‘healthy worker’ survivor effect whereby individuals with severe osteoarthritis may have already exited the workforce.⁵⁰

The results revealed other notable disease associations. We observed that oral health conditions were the most common diagnoses in both sexes, which is consistent with other studies.⁵¹ Notably, nearly 10% of Japanese workers report missing work or being less productive owing to oral health problems,⁵² and poor periodontal health has been linked to a reduction of approximately 8% in work performance.⁵³ We found that other arrhythmias were associated with increased PLC in men, but not in women. This may reflect sex differences in return to work after cardiovascular events⁵⁴; women who experience serious cardiovascular conditions may be less likely to remain in the workforce^{55 56} and thus would not appear in our employed sample.

Additionally, female-specific conditions like menopausal disorders can negatively affect work performance.⁵⁷ We noted that some ailments are more common in women than in men; for example, otologic disorders such as otitis externa, as well as common conditions such as conjunctivitis, affect women to a greater extent and can impair productivity and HRQoL.^58–61 One unexpected finding was the association of volume depletion (hypovolemia) with high PLC in women. A previous study noted that the nutritional status of young Japanese females has been declining, possibly owing to the prevalent desire for thinness⁶²; this context might partly explain our observation, although further investigation is needed.

We observed differences between sexes in NCD prevalence and associated outcomes. Female participants had higher rates of NCDs and multimorbidity than male participants, consistent with prior research reporting that women are more susceptible to multiple chronic conditions.^32–34 However, as formal interaction tests were not performed, these sex-specific patterns should be interpreted descriptively rather than as formal evidence of effect modification.

Our findings have several implications. For future research, longitudinal studies are needed to confirm whether the high-impact disease combinations identified here (particularly neuropsychiatric–sleep disorder pairs) have causal effects on work productivity and to examine whether early identification and treatment of sleep disorders can mitigate the development of multimorbidity. For workplace practice, these findings highlight the potential value of integrated occupational health screening that considers specific disease combinations rather than relying solely on single-disease management approaches. For public health policy, the substantial associations between neuropsychiatric–musculoskeletal multimorbidity patterns and productivity loss suggest that investment in mental health and musculoskeletal programmes targeting working-age populations may have economic as well as health benefits. However, interventional evidence is required before specific recommendations can be made.

Our study has several limitations. First, our models adjusted for age, sex and number of co-occurring conditions; however, socioeconomic factors (eg, income, education, occupation type) and occupational variables (eg, job demands, shift work) are not routinely captured in administrative claims data and were therefore unavailable in our dataset. These unmeasured variables are strongly associated with health outcomes and work productivity, and their omission introduces residual confounding that may bias our estimates. The direction of this bias is uncertain; for example, lower socioeconomic status is associated with higher NCD prevalence and lower productivity, which could lead to overestimation of disease–productivity associations. Future studies incorporating socioeconomic and occupational variables are warranted. Second, only 3.0% of the eligible population participated in the survey, and the respondents were self-selected kencom users, raising concerns about selection bias. The respondents may have higher health literacy and therefore may not be representative of the broader working population. This potential selection bias risks underestimating NCD prevalence and associated productivity loss. Nevertheless, the primary aim of this study was to examine associations between disease patterns and outcomes rather than to estimate population prevalence, and internal validity may be less affected. Additionally, the worker population tends to be healthier than the general population (the healthy worker effect).⁶³ Some conditions in our study (eg, diabetes, cardiovascular disease) showed associations with better-than-average productivity, possibly because individuals with these conditions who remain employed have them well-controlled or because those with severe disease have left the workforce.

Third, conditions that were present but untreated or undiagnosed would not appear in the claims data, potentially leading to underestimation of the true prevalence of chronic conditions. We could not account for disease severity or treatment stage; for example, some individuals with a history of cancer or cardiovascular disease may have been in remission and fully productive at the time of the survey. Fourth, as this is a cross-sectional analysis, we cannot infer causality between having certain diseases and the observed productivity or QoL outcomes; the observed associations may be confounded by factors not measured in our study. Fifth, we observed a clustering of responses at the ‘no impairment’ end for PLC and HRQoL. The EQ-5D-5L instrument is known to have a ceiling effect,⁶⁴ which likely contributed to the high average HRQoL in our sample, potentially masking some differences between groups.

Lastly, our results should not reinforce the stigma regarding low productivity among patients with neurological disorders. Likewise, our results do not imply that serious conditions, such as cancer or cardiovascular disease, are unimportant; acute events in these diseases have profound impacts and continued efforts in their prevention and management are still warranted.

Conclusions

This cross-sectional study found that specific patterns of co-occurring NCDs were associated with reduced work productivity and HRQoL among Japanese workers, and that these associations varied by disease combination. Particularly, combinations involving a sleep disorder paired with another neuropsychiatric condition, as well as neuropsychiatric-musculoskeletal multimorbidity patterns, were associated with the greatest productivity loss and HRQoL deficits, beyond what would be expected from the number of co-occurring diseases alone. These findings suggest that workplace health management by companies and occupational physicians may benefit from considering specific high-impact combinations of chronic conditions, especially those involving mental health, sleep and musculoskeletal conditions, rather than focusing solely on disease count. However, given the cross-sectional design, these associations require confirmation in longitudinal studies before informing specific clinical or policy recommendations. To our knowledge, this is among the first Japanese reports using linked claims and PRO data to demonstrate that specific multimorbidity patterns, beyond simple disease counts, are associated with PLCs and HRQoL in working-age adults.

Data availability statement

Data may be obtained from a third party and are not publicly available. No data are available.