Direct cost of occupational asthma and associated factors in the private sector in central Tunisia

Abstract

Background

Occupational asthma (OA) is a significant public health concern due to its high prevalence and socioeconomic burden. In Tunisia, direct costs of OA have not been previously evaluated, highlighting an important knowledge gap for health policy planning. This study aimed to assess the direct cost of OA in the private sector of central Tunisia and identify factors influencing this cost.

Methods

A retrospective claims-based cohort study was conducted using data from the National Health Insurance Fund (CNAM) on OA cases recognised between 2015 and 2017 in the governorates of Sousse, Monastir, Mahdia and Kairouan, central Tunisia. Direct costs were estimated from the CNAM payer perspective. The costs were tracked from the date of recognition until 31 December 2020 and expressed in 2020 Tunisian dinars (TND). Statistical analyses included univariate tests and generalized linear models (GLM) with gamma family and log link for right-skewed cost data, reporting adjusted incidence rate ratios (aIRR) with Huber-White (HC1) robust standard errors.

Results

A total of 157 cases of OA were analyzed, predominantly female (75.8%), with a mean age of 43.41 ± 7.29 years. The textile sector represented 72% of cases. High-molecular weight allergens, particularly vegetable textile dust (70.7%), were the agents most frequently implicated. The median total direct cost was 4,593.52 TND (€1,467.24) per case [IQR: 3,408.00–6,871.72]. The median annualized cost was 1,114.03 TND (€355.94) per person-year. Cash benefits (96.8% of patients) dominated over in-kind benefits (22.9%). In multivariate analysis, age ≥ 40 years was the only significant independent predictor of medical costs (adjusted incidence rate ratio [aIRR] = 2.84, 95% CI: 1.59–5.07; p < 0.001), whereas no clinical or demographic factor predicted indemnity costs. Univariate analysis additionally identified male sex (p = 0.04), higher PPD rate (p < 0.001), and absence of prior occupational disease history (p = 0.001) as factors associated with higher total direct costs.

Conclusions

This study highlights the considerable direct costs associated with occupational asthma in central Tunisia’s private sector, and the resulting burden placed on the healthcare system and the affected workforce. These findings underscore the need to strengthen primary prevention strategies, particularly in the textile industry, to reduce both the health and economic impact of this disease.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12913-026-14941-0.

Introduction

Occupational asthma (OA) is a major public health problem due to its high prevalence and significant social and economic burden [1, 2]. In central Tunisia’s private sector, the incidence of OA is estimated at 24.42 per million workers covered by CNAM per year (2000–2008), based on recognized cases among actively insured private sector workers in central Tunisia. OA accounted for 6% of all recognized occupational diseases and 46.9% of all recognized respiratory occupational diseases during the study period [3]. The textile sector, which employs a large proportion of women in central Tunisia, is the most affected industry, consistent with international data showing a high incidence of OA among textile workers exposed to organic dusts [4, 5].

In Tunisia, both work-induced asthma (new onset due to occupational sensitization) and work-aggravated asthma (pre-existing asthma worsened by occupational exposures) are recognized as occupational asthma under Tunisian law (Law No. 94 − 28 of 21 February 1994) [6]. The distinction between these two phenotypes is clinically important: work-induced asthma typically requires complete avoidance of the causative agent, whereas work-aggravated asthma can be managed with reduced exposure combined with pharmacotherapy. However, the CNAM database does not distinguish between these two phenotypes; both are classified under the same recognition tables.

Clinically, OA typically manifests itself as work-related dyspnea, wheezing, chest tightness, and cough that improve when away from exposure. It is often associated with allergic rhinitis and is characterized by variable airflow obstruction and bronchial hyperresponsiveness [7–9]. Spirometric abnormalities commonly include a reduced forced expiratory volume in one second (FEV1), a decreased FEV1/FVC ratio, and a positive methacholine challenge test with a PD20 below 800 µg [10]. The diagnosis of OA is based on a combination of occupational history, the temporal relationship between exposure and symptoms, and objective confirmation through spirometry or bronchial provocation tests [8, 11].

The diagnosis of OA has major consequences for both employees and employers, including possible job changes, compensation claims, and long-term disability [12]. Early recognition and removal from exposure are crucial, as continued contact with the causative agent may lead to persistent asthma even after cessation of exposure [13].

However, OA is often underreported and underdiagnosed, particularly in developing countries where occupational health surveillance systems are limited [14, 15]. In Tunisia, inadequate awareness among healthcare providers and workers, combined with a complex administrative declaration process, likely contributes to underreporting. The true incidence of OA is estimated to be two to three times higher than the officially recognized rate [15]. Effective prevention strategies including assessment of occupational exposure, elimination or substitution of sensitizing agents, respiratory protection, and worker education are essential to reduce the burden of OA [16, 17].

In addition to its physical health implications, OA also has a substantial socioeconomic impact. Studies show that 25–38% of patients with OA experience prolonged work interruptions with 42–78% loss of income [18]. The economic burden encompasses direct medical costs (diagnostic workup, pharmacotherapy, hospitalization), direct non-medical costs (transport, home modifications), and indirect costs (productivity loss, disability payments, job retraining) [19]. In the United States, the economic burden of asthma was estimated at $81.9 billion in 2013 [20], while earlier national estimates reported direct costs of $5.1 billion in 1997 [21]. The annual cost of asthma in western countries ranges from €500 to €3,000 per patient, with estimates reaching €1,255 per patient in France (2004) [22] and €2,543 per patient for respiratory allergic diseases in Italy (2015) [23], with occupational cases generally incurring higher costs than nonoccupational cases due to younger age at onset and longer disease duration. Furthermore, asthma costs are strongly correlated with disease severity [24], underscoring the economic importance of early detection and effective management.

Socioeconomic evaluations usually distinguish direct costs (medical expenses, compensation paid to affected workers, etc.) and indirect costs (job loss, reduced productivity, premature retirement) [19]. Effective management of OA requires costly prevention strategies including early diagnosis, exposure controls, employee education, and medical surveillance [16, 17]. Although various studies have examined the costs associated with asthma in high-income countries, there is a paucity of data in low- and middle-income settings, particularly in North Africa. Quantifying these costs in the Tunisian context, where preventive infrastructure remains underdeveloped, is essential for evidence-based policy. The total cost of OA is influenced by multiple factors, including disease incidence and patient management practices [19]. To our knowledge, no Tunisian study has estimated the cost of OA to date. Therefore, this study aimed to assess the direct cost of OA in the private sector of central Tunisia and identify factors influencing this cost.

Methods

Study type

This retrospective claim-based cohort study examined asthmatic patients whose respiratory conditions were recognized as occupational diseases by the National Health Insurance Fund (CNAM) between 1 January 2015 and 31 December 2017, in the governorates of Sousse, Monastir, Mahdia and Kairouan, central Tunisia. Cost data were tracked from the date of OA recognition to 31 December 2020. The study was reported in accordance with the CHEERS 2022 statement for economic evaluations and the STROBE checklist.

Costing methods

This analysis adopted a payer perspective (CNAM), capturing all direct costs reimbursed by the National Health Insurance Fund. All costs were expressed in nominal 2020 Tunisian dinars (TND; €1 = 3.13 TND). Purchasing power parities and inflation adjustments were not applied; costs were reported in nominal 2020 TND as the price year. No discounting was performed given the relatively short observation period.

The index date was defined as the date of recognition of OA by the CNAM Medical Commission. Since individual recognition dates were not available in the database, the duration of follow-up was estimated as (2020 − year of recognition + 0.5). The estimated median follow-up was 4.5 years (IQR: 3.5–5.5 years; range: 3.5–5.5 years). To account for variable follow-up, annualized costs per person-year (PPY) were calculated as (cumulative cost / estimated follow-up years) for each patient.

Study population

Inclusion criteria

• OA cases recognized by the two occupational accidents and occupational diseases (OA/OD) medical commissions in Sousse and Monastir, which assess the rates of permanent partial disability (PPD) according to the Tunisian compensation schedule.

• Employment in the private sector at the time of recognition of OA.

• Residence or employment in one of the four target governorates (Sousse, Monastir, Mahdia, or Kairouan).

• Diagnostic confirmation required: (i) documented occupational exposure to a known asthmagen; (ii) compatible respiratory symptoms related to work (dyspnea, wheezing, chest tightness or cough improving away from the workplace); (iii) spirometric abnormalities (FEV1 < 80% predicted and/or positive bronchial hyperresponsiveness test); and (iv) recognition by the CNAM Medical Commission as an occupational disease.

• Availability of complete CNAM cost records from the date of OA recognition through 31 December 2020.

• Both compensated and non-compensated OA cases were included.

Exclusion criteria

• Workers employed in the public sector, who are not covered by the CNAM private sector scheme and, therefore, are not captured in this database.

• Asthma cases not officially recognized as occupational diseases.

• OA cases that arise outside the four target governorates.

• Incomplete CNAM cost records.

Sample size calculation

The minimum sample size was calculated using the following formula:

where:

Zα/2: is the confidence level (1.96 for a 95% confidence rate),

p: estimated proportion of OA among occupational diseases (7.17% on the basis of national data [3]),

i: is the desired precision (set at 5%).

The calculation yielded a minimum sample size of 102 cases.

Sampling

An exhaustive sampling approach was used. Data were extracted from CNAM’s Oracle database in 2021 and included 184 privately insured OA cases officially recognised between 1 January 2015 and 31 December 2017 in four governorates of central Tunisia (Sousse, Monastir, Mahdia, and Kairouan). After excluding cases that arise outside the target governorates, non‑recognised OA, and records with incomplete cost data, the final analytical sample comprised 157 patients (Supplementary Figure S1).

Data collection

Data were collected via a structured synoptic sheet divided into two main sections:

The first section captured information about the study population, including: sociodemographic details (e.g., gender and age), professional characteristics (e.g., sector of activity, job position, professional seniority, and exposure to occupational hazards), and medical and legal information (e.g., personal history, symptoms, physical examination findings, ongoing treatments, respiratory function test results, dates of initial medical certificate and consolidation, PPD rates).

The second section focused on calculating the direct costs associated with OA, which included the total payments made by CNAM for medical care coverage and both temporary and permanent work indemnities.

Data were extracted from multiple sources, including:

• Medical certificates of OA declarations submitted to regional CNAM offices,

• Technical investigation reports conducted by CNAM engineers for each case,

• Medical and administrative records from the OA/OD medical commission secretariat.

Data validation was performed through multiple steps: (i) range checks on all cost variables (no negative values; maxima verified against CNAM payment schedules); (ii) denominator verification for all reported percentages (n/157 for overall, n/153 for cost analyses); (iii) cross-referencing of total direct cost with the sum of in-kind and cash benefit components; and (iv) double data entry for a random 10% subsample with > 99% concordance.

Operational definition of variables

The total direct cost of OA was calculated from the date of the disease declaration until the end of 2020, considering the sum of the following:

• In-kind benefits: included : (i) pharmaceutical expenses, defined as the costs of asthma medications reimbursed by CNAM (maintenance inhaled therapies and rescue drugs); (ii) medical expenses, defined as reimbursement for occupational medicine consultations, spirometry, and other respiratory investigations related to OA; (iii) hospitalization fees; and (iv) transport allowances.

• Cash benefits: comprising amounts paid by the CNAM as temporary work indemnities (daily allowances for TWD) and permanent work indemnities (pension for PPD when the rate is ≥ 15% or a lump sum for rates between 6% and 14%).

As medians are not additive, the median total direct cost does not equal the sum of the component medians. Coverage proportions (any in-kind, any cash benefit) are reported to characterize the distribution of cost components across the cohort.

Ethical considerations

The study was approved by the Ethics Committee of Sahloul University Hospital (IRB approval number: [HS 21-2021]), as well as by the General Director of Medical Control and the Director of the Central Medical Control District. As the study was based on the retrospective analysis of anonymized data from the CNAM database, the requirement of informed consent was waived. The confidentiality and anonymity of all patient data were strictly maintained according to Tunisian regulations and the Declaration of Helsinki.

Statistical analysis

The Data were entered and analyzed via SPSS version 21.0 for descriptive statistics and bivariate analyses. All statistical tests were two-sided. Qualitative variables were presented as frequencies and percentages. Quantitative variables were tested for normality using the Shapiro-Wilk test. Normally distributed variables (age, FEV1%) were expressed as mean ± standard deviation and compared using the Student’s t-test. Non-normally distributed variables (costs, seniority) were expressed as median [interquartile range] and compared using the Mann-Whitney U test or the Kruskal-Wallis test, as appropriate. Correlations were assessed using Pearson’s correlation coefficient for normally distributed variables and Spearman’s rank correlation for non-normal variables.

Given the markedly right-skewed distribution of cost data (skewness = 3.99, kurtosis = 17.92), confirmed by the Shapiro-Wilk test (W = 0.651, p < 0.001), generalized linear models (GLM) with a gamma family and log link were used for multivariate analysis, reporting adjusted incidence rate ratios (aIRR = exp(β)). Huber-White (HC1) robust standard errors were applied. The Park test yielded imprecise slope estimates (α = 0.95–1.13) due to small sample size; however, the direct variance-mean relationship (log[Var]/log[Mean] ≈ 2.0), AIC comparison across candidate distributions, and a non-significant Pregibon link test (p > 0.18 for both models) confirmed the gamma-log specification. Variables with p < 0.20 in univariate analysis or established clinical relevance were included as candidates. Multicollinearity was assessed using the variance inflation factor (VIF), with VIF > 5 indicating potential concern and VIF > 10 indicating serious multicollinearity requiring variable removal.

PPD was included in Model A as a clinical severity proxy (the PPD rate is independently evaluated by the CNAM medical advisor using standardized functional criteria, not derived from medical costs), but was explicitly excluded from Model B to avoid tautological prediction of indemnity costs from the very variable used to calculate them. All GLM analyses were performed using Python (version 3.11) with statsmodels (version 0.14.1) for generalized linear modeling, SciPy (version 1.11.4) for statistical distributions and goodness of fit testing, and pandas (version 2.1.4) for data manipulation.

Missing data affected less than 5% of the variables in general. Complete case analysis was used as the proportion of missing data did not warrant multiple imputation.

Results

Descriptive analysis

A total of 157 cases of occupational asthma (OA) recognized by the CNAM between 2015 and 2017 were included in this study, with a median estimated follow-up of 4.5 years (IQR: 3.5–5.5). OA accounted for 6% of all recognized occupational diseases and 46.9% of all recognized respiratory occupational diseases.

Sociodemographic and professional characteristics

The study population was predominantly female (75.8%), with a sex ratio of 0.32. The mean age was 43.4 ± 7.3 years (range 29–61 years). The textile industry was the most represented sector (72%), followed by the food industry (7%), furniture manufacturing (5.7%), automotive/electric (3.2%), and electronics (3.2%). The mean professional seniority was 15.8 ± 8.9 years. The Monastir governorate contributed the highest number of cases (60.5%).

High-molecular-weight agents were involved in 80.9% of the cases. The most common allergen was vegetable textile dust (70.7%), followed by flour and cereals (5.7%). Isocyanates and heated rosin accounted for 4.5% each. Cost estimates for etiologic categories with n ≤ 5 (methacrylates, formaldehyde, sulfites/bisulfites, enzymes, paper industry, metallurgy) should be interpreted with caution, as they are based on very small samples and can be driven by individual outliers. For example, the high median cost for methacrylates (37,017.99 TND) was driven by a single case with extensive PPD compensation.

Clinical and paraclinical characteristics

A personal history of allergies was observed in 12.7% of the patients. Most of the patients had typical episodes of wheezing and dyspnea (87.3%). Chest tightness was reported by 44.6% of patients, and cough by 33.1%. Associated rhinitis was present in 62.4% of cases. Background maintenance asthma therapy was received by 55.4% of patients, mainly based on inhaled corticosteroids combined with long-acting beta-2 agonists. A nonspecific bronchial hyperresponsiveness test was performed in 47 patients (29.9%). Positive results were obtained in all cases, with a mean value of PD20 (Provocative Dose causing a 20% decrease in FEV1) of 0.78 ± 0.44 mg/ml.

Declaration and recognition

70% of OA cases were declared under Tunisian occupational disease Table No. 53 (“Vegetable Textile Dusts”), while 10.2% were declared under Tunisian occupational disease Table no. 58 (“Other Agents Responsible for Respiratory Allergic Mechanism Diseases”). A PPD assessment was performed for 152 patients (96.8%); of 151 patients with known PPD rates, 141 (93.4%) had a rate ≥ 15% (pension eligibility threshold). The median PPD rate was 20% [IQR: 18–25%].

Direct cost of occupational asthma

• In-kind benefits

Thirty-six patients (22.9%) received any in-kind benefit. None of the patients received hospitalization or transport expenses coverage during the study period, consistent with the CNAM reimbursement schedule for OA under Tunisian occupational disease Tables 53 and 58. The median in-kind benefit was 282.34 TND (Tunisian Dinars) [103.25–1,622.23] (Table 1).

Table 1.

Summary of in-kind and cash benefits provided to our patients throughout the study duration

		Amount (TND)	n (%)	Median	[Q1-Q3]	Minimum	Maximum
In-kind benefits		Pharmaceutical expenses	22 (14%)	597.11	[210.90–1,715.69]	40.80	2,896.11
		Medical expenses	21 (13.4%)	120.00	[49.25–289.79]	10.00	2,294.46
		Total In-Kind benefits	36 (22.9%)	282.34	[103.25–1,622.23]	10.00	2,896.11
Cash benefits	TWD (Temporary work disability)	Total TWD amount	33 (21%)	359.37	[221.83–948.64]	99.40	11,613.03
	PPD (Permanent partial disability)	Total pension amount	141 (89.8%)	4,505.08	[3,350.45–6,239.68]	830.32	52,012.76
		Lump sum	11 (7%)	2,910.52	[2,450.82–7,770.92]	1,994.05	8,031.77
		Total PPD amount	152 (96.8%)	4,500.88	[3,232.44–6,268.59]	830.32	52,012.76
		Total Cash Benefits	152 (96.8%)	4,510.70	[3,388.42–6,353.35]	583.13	52,012.76
Total direct cost			153 (97.4%)	4,593.52	[3,408.00–6,871.72]	101.86	52,109.85
Annualized cost PPY			153 (97.4%)	1,114.03	[831.58–1,523.17]	22.63	11,579.96

n: Number of patients who received coverage for expenses

TND: Tunisian Dinars

In-kind benefits include pharmaceutical and medical expenses; cash benefits include temporary and permanent work disability indemnities (pension or lump sum)

• Cash benefits

• Temporary work disability (TWD)

Twenty-one percent of the patients received temporary work indemnities. The median duration of the work stoppage was 29 days [14–61.5] (range: 0 to 715 days), with a median payment of 359.37 TND [221.83–948.64].

– Permanent work disability (PWD)

The median payment for permanent work indemnities was 4,500.88 TND [3,232.44 – 6,268.59].

• Total cash benefits

The median total cash benefits (combining TWD and PWD) was 4,510.70 TND [3,388.42 – 6,353.35] (Table 1). Figure 1 illustrates the distribution of cost components using box plots on a log10 scale, highlighting the variability between in-kind and cash benefits.

Fig. 1.

Distribution of cost components for in-kind and cash benefits among patients with occupational asthma in central Tunisia

• Total and annualized direct cost of OA

The median total direct cost was 4,593.52 TND [3,408.00–6,871.72]. The median annualized cost per person-year (PPY) was 1,114.03 TND [831.58–1,523.17]. Of 153 patients with any cost, 36 (23.5%) received any in-kind benefit, 33 (21.6%) received TWD, and 152 (99.3%) received PPD indemnity. The overlap between in-kind and cash benefit recipients was 34 patients. Four patients (2.5%) had no recorded CNAM expenditure (PPD < 15%, lump sum not yet paid by 31 December 2020). As the medians are not additive, the median total direct cost (4,593.52 TND) is not equal to the sum of the median components in-kind (282.34 TND) and cash (4,510.70 TND) (Table 1).

• Sensitivity analysis

Two sensitivity analyses were performed to assess the robustness of the GLM results. First, excluding the top 5% of annualized costs (> 4,808 TND PPY, n = 8 excluded), the medical model retained age > = 40 years as the strongest predictor (aIRR = 2.94, p < 0.001), consistent with the main findings. Second, converting all costs to 2020 TND using the official Tunisian CPI (average annual inflation: 4.7% during 2015–2020) did not materially change the direction or significance of any predictor, with the largest change in aIRR being < 8% for all covariates. These results confirm the robustness of the main parsimonious medical cost model.

Analytical section

Univariate study

Direct cost of OA in relation to sociodemographic, occupational, and medical characteristics

The analysis revealed that the direct cost of OA was significantly higher in men than in women (p = 0.04). However, no significant correlation was found between cost and age (p = 0.09). Table 2 presents detailed results according to sociodemographic and occupational characteristics. A significant association was observed between the direct cost of OA and the history of occupational diseases (p = 0.001): patients without a history of this had higher costs (Table 3).

Table 2.

Direct cost of occupational asthma based on sociodemographic and occupational characteristics of the patients

Characteristics	n	Median (TND)	[Q1 – Q3]	p
Sociodemographic characteristics
Gender
Male	37	5,944.24	[4,094.56–7,603.52]	0.04
Female	116	4,477.81	[3,361.52–6,313.46]
Age
≤ 43 years	80	4,436.54	[3,062.65–6,894.33]	0.09
> 43 years	73	5,105.20	[3,869.21–6,916.46]
Governorate
Monastir	93	4,587.29	[3,828.05–6,742.80]	0.053
Sousse	48	4,701.08	[2,976.85–7,095.91]
Mahdia	10	5,556.46	[3,099.72–21,199.91]
Kairouan	2	3,588.67	[2,129.39–5,047.96]
Professional characteristics
Activity Sector
Textile	111	4,554.09	[3,577.48–6,724.03]	0.11
Food industry	11	5,665.91	[2,162.66–7,217.95]
Furniture Industry	9	4,290.02	[3,061.50–7,520.38]
Automotive/Electrical	5	5,105.20	[3,644.99–8,334.75]
Electronic Industry	5	4,000.85	[2,783.11–14,899.25]
Plastics Industry	3	7,073.50	[2,737.39–29,591.67]
Metallurgy	2	7,553.66	[7,118.24–7,989.08]
Paper Industry	2	3,912.46	[2,493.82–5,331.10]
Professional Seniority
≤ 15 years	78	4,828.78	[3,803.66–6,724.57]	0.76
> 15 years	75	6,724.03	[3,029.02–7,039.23]
Etiological Agents
Plant Textile Dusts	109	4,554.09	[3,583.66–6,725.10]	0.07
Flour and Cereals	9	5,944.24	[4,444.21–7,800.55]
Isocyanates	7	5,105.20	[3,854.38–8,031.77]
Heated Rosin	7	4,853.75	[3,353.45–7,073.50]
Wood Dust	6	3,726.78	[2,868.54–5,168.54]
Aerosol Proteins	3	2,295.74	[899.99–2,474.77]
Enzymes	2	5,416.97	[3,745.18–7,088.76]
Methacrylates	2	37,017.99	[21,926.14–52,109.84]
Sulfites and Bisulfites	2	4,806.03	[2,493.82–7,118.24]
Formaldehyde	2	3,892.67	[2,737.39–5,047.96]
Molecular Weight
High-Molecular-Weight	125	4,587.29	[3,451.63–6,725.10]	0.99
Low-Molecular-Weight	28	6,725.10	[3,211.02–7,683.82]

* FEV1: The mean forced expiratory volume in one second

TND: Tunisian Dinar

Categories with n ≤ 5 should be interpreted with caution due to limited sample size (methacrylates n = 2, formaldehyde n = 2, sulfites/bisulfites n = 2, enzymes n = 2, paper industry n = 2, metallurgy n = 2). Sector counts reflect the cost-analysis subsample (n = 153); total sample counts differ slightly due to 4 patients with zero expenditure (2 from the textile sector)

Table 3.

Direct cost of occupational asthma based on clinical and paraclinical patient characteristics

Characteristics	n	Median (TND)	[Q1 – Q3]	p
History of occupational diseases
No	122	5,039.15	[3,766.40–7,047.80]	0.001
Yes	31	3,884.04	[2,910.52–5,362.98]
Dyspnea with wheezing
Yes	133	4,697.42	[3,408.00–6,871.72]	0.79
No	17	4,404.10	[3,258.79–6,650.27]
Chest Tightness
Yes	68	4,457.14	[3,305.92–7,075.65]	0.86
No	82	4,985.12	[3,536.78–6,452.26]
Cough
Yes	49	4,587.29	[3,495.58–7,052.65]	0.29
No	101	4,603.69	[3,400.21–6,725.10]
Nasal Symptoms
Yes	94	4,853.75	[3,451.63–7,074.94]	0.15
No	60	4,548.42	[3,398.90–6,126.22]
Tachypnea
Yes	3	7,675.88	[7,118.24–8,233.52]	0.73
No	108	4,650.56	[3,473.90–6,487.31]
Wheezing on auscultation
Yes	38	4,914.61	[3,927.17–8,016.82]	0.11
No	71	4,603.69	[3,163.54–6,268.59]
Bronchial Rales
Yes	7	4,593.52	[2,829.57–7,118.24]	0.96
No	102	4,775.59	[3,592.31–6,586.75]
FEV1 (%)
≥ 70	54	4,347.51	[2,917.57–6,167.63]	0.07
60–69	20	3,891.56	[2,591.00–6,045.96]
50–59	12	5,358.39	[4,254.57–7,878.31]
35–49	13	6,916.94	[3,994.30–13,735.10]
< 35	2	6,250.43	[4,469.10–8,031.77]
Background Treatment
Yes	87	4,512.11	[3,488.57–6,826.51]	0.42
No	14	5,814.01	[3,678.93–6,824.19]
Associated Rhinitis
Yes	98	4,598.61	[3,407.46–7,047.80]	0.23
No	55	4,587.29	[3,401.31–6,133.97]
Associated Conjunctivitis
Yes	2	12,442.80	[6,826.51–21,926.14]	0.72
No	151	4,593.52	[3,414.70–6,826.51]

* FEV1: The mean forced expiratory volume in one second

TND: Tunisian Dinars

Direct cost of OA and spirometry data

The direct cost of OA was significantly associated with the percentage of FEV1 relative to the predicted value (p = 0.02, rho = − 0.22). However, there was no statistically significant correlation between cost and the FEV1/FVC ratio (p = 0.19) (Table 3).

Direct cost of OA and the PPD rate

A significant positive association was observed between the direct cost of OA and the PPD rate (p < 0.001, rho = 0.36), indicating that the direct cost increases with higher PPD rates across governorates (Fig. 2). This association is partially tautological, as PPD directly determines the amounts of the indemnity payment. To address this endogeneity, PPD was included only in Model A (medical costs) and explicitly excluded from Model B (indemnity costs), as discussed in Section II-2.

Fig. 2.

Direct costs of occupational asthma by the PPD rate in various governorates of central Tunisia

Multivariate analytical study

Given the markedly right-skewed distribution of cost data (skewness = 3.99, kurtosis = 17.92), confirmed by the Shapiro-Wilk test (W = 0.651, p < 0.001), generalized linear models with gamma family and log link were used. Two separate-component models were fitted: Model A for annualized medical costs (n = 35) and Model B for annualized indemnity costs (n = 152). Results are presented in Table 4.

Table 4.

Generalized Linear Models (GLM, gamma family, log link) for annualized medical and indemnity costs per person-year

Model A: Annualized medical costs (parsimonious model) (n = 35 patients with any medical expenditure)
Variable	aIRR (95% CI)	p
Age ≥ 40 years (vs. < 40)	2.84 (1.59–5.07)	0.000***
PPD per 10%	1.43 (0.93–2.19)	0.099
AIC: 431.11	Max VIF: 2.09

Model B: Annualized indemnity costs (PPD excluded to avoid tautology) (n = 152 patients with any indemnity payment)
Variable	aIRR (95% CI)	p
Age ≥ 40 years (vs. < 40)	0.77 (0.56–1.06)	0.108
Male sex (vs. female)	1.36 (0.96–1.93)	0.087
Textile sector (vs. non-textile)	1.11 (0.78–1.58)	0.562
Seniority ≥ 15 years (vs. < 15)	0.85 (0.64–1.13)	0.266
Symptoms ≥ 3 (vs. < 3)	1.35 (0.88–2.07)	0.168
AIC: 2497.39	Max VIF: 3.14

***p < 0.001, *p < 0.05. Models: gamma family, log link, robust Huber-White (HC1) standard errors. aIRR: adjusted incidence rate ratio (exp(β)). CI: confidence interval. VIF: variance inflation factor

Separate component GLM analysis (Table 4) yielded the following results. In the parsimonious medical cost model (Model A, n = 35), age ≥ 40 years was the only significant predictor of annualized medical costs (aIRR = 2.84, 95% CI: 1.59–5.07, p < 0.001). PPD, included as a clinical severity proxy, showed a positive but non-significant association (aIRR = 1.43 per 10% points, 95% CI: 0.93–2.19, p = 0.099), possibly reflecting insufficient statistical power given the small sample. The model showed acceptable multicollinearity (max VIF = 2.09), representing a substantial improvement over the original full specification (max VIF = 18.29) which was discarded due to overfitting.

In the indemnity cost model (Model B, n = 152), PPD was explicitly excluded as a predictor to avoid tautological inference, since indemnity payments are directly derived from the PPD rate. None of the remaining clinical or demographic variables reached statistical significance: age ≥ 40 years (aIRR = 0.77, p = 0.108), male sex (aIRR = 1.36, p = 0.087), textile sector (aIRR = 1.11, p = 0.562), seniority ≥ 15 years (aIRR = 0.85, p = 0.266), and symptoms ≥ 3 (aIRR = 1.35, p = 0.168). The model showed acceptable multicollinearity (max VIF = 3.14). This absence of significant predictors indicates that indemnity costs are primarily determined by the administratively assigned PPD rate rather than by the clinical severity or demographic profile of the patients.

Discussion

This study achieved its objectives by providing the first comprehensive estimate of direct costs of occupational asthma in Tunisia’s private sector from the perspective of the CNAM payer. We quantified substantial economic burden, with a median total direct cost of 4,593.52 TND (€1,467.24) per case. In multivariate analysis, age > = 40 years was the only significant predictor of medical costs, while indemnity costs were found to be determined primarily administratively by the rate of PPD rather than by clinical or demographic characteristics of the patient.

In GLM with gamma distribution and log link, the aIRR represents the multiplicative change in expected mean cost associated with a one-unit increase in the predictor variable. For example, an aIRR of 2.84 for age > = 40 years indicates that older patients have 2.84 times higher expected medical costs compared to younger patients, after adjusting for PPD.

Estimation of the direct cost of OA

Occupational exposures are estimated to account for 15–20% of asthma cases [25–32]. A systematic analysis reported that approximately 16.3% of all adult-onset asthma cases are attributable to occupational exposures, with higher proportions in high-risk industries [28, 32]. In our study, the textile sector, the dominant employer in central Tunisia, represented 72% of OA cases, highlighting the need for targeted prevention in this industry.

Once diagnosed, OA has significant socioeconomic implications for employees, employers, and the healthcare system [10]. To improve international comparability, our main cost estimates are presented both in TND and euros (2020 exchange rate: €1 = 3.13 TND). However, direct cross-country comparisons of OA costs should be interpreted with caution due to differences in healthcare financing systems, price years, costing methodologies (e.g., payer vs. societal perspective) and currency fluctuations. The figures cited above span studies conducted between 1996 and 2018, with costs expressed in various currencies and price years. Our estimates (2020 TND) are provided alongside euro equivalents solely to give readers an approximate order of magnitude, not to suggest direct comparability.

For example, in 1996, the cost of OA in the United States was estimated at $1.6 billion, with 74% attributed to direct costs [33]. According to a study conducted by Birnbaum HG et al., the total expenditure per employed adult with asthma in the United States was $ 5,385 in 2003 USD [34]. In Canada, Malo et al. reported a median total compensation cost of OA of CAD 61,300 per case between 1988 and 2002 [35]. In Spain, García Gómez et al. estimated the mean direct healthcare cost of OA at € 1,738 per patient in 2008 [36]. These comparisons, while informative, are limited by substantial differences in healthcare systems, compensation schemes, and price years between countries and study periods.

In Tunisia, there is a lack of statistical data on the cost of occupational diseases. CNAM statistics indicate that technical expenses for occupational accidents and occupational diseases doubled between 2008 and 2018, from 57 million TND to 114 million TND [37]. Our finding of a median annualized cost of 1,114.03 TND PPY is substantially lower than Western estimates, reflecting differences in both healthcare cost structures and compensation levels.

Factors associated with the direct cost of OA

Our univariate analysis revealed significant associations between direct cost and factors such as sex, history of occupational diseases, percentage of PPD, and percentage of FEV1. However, separate component GLM modeling revealed a more nuanced picture. In the parsimonious medical cost model (Model A), only age > = 40 years was independently associated with higher medical costs (aIRR = 2.84, p < 0.001), while PPD showed a borderline non-significant trend (aIRR = 1.43 per 10%, p = 0.099). Critically, in the indemnity cost model (Model B), which excluded the tautological predictor of PPD, no clinical or demographic variable reached statistical significance. This finding fundamentally changes the interpretation: the strong univariate association between PPD and total cost (rho = 0.36, p < 0.001) was largely driven by the mechanical relationship between PPD and indemnity payments, not by a clinical dose-response relationship.

Furthermore, our results did not reveal a significant relationship between professional seniority and the direct cost of OA. Similarly, Miedinger D et al. reported that the number of years with an employer was not a significant predictor of OA costs [38]. Deschamps et al. also found no significant association between disease duration and cost in their 1997 French study [39].

Within our study population, we observed that the median direct cost of OA was higher for low-molecular-weight (LMW) allergens (6,725.10 TND / €2,149.07 [3,211.02–7,683.82]) compared to high-molecular-weight (HMW) allergens (4,587.29 TND / €1,465.78 [3,451.63–6,725.10]), although this difference did not reach statistical significance in the multivariate model. This is consistent with the international literature suggesting that LMW-induced OA, often caused by chemicals such as isocyanates, can result in more severe and persistent disease [40].

Although our univariate analysis showed a significant association between FEV1 and direct cost (p = 0.02, rho = -0.22), FEV1 was not included in the final parsimonious medical cost model (Model A) due to the substantial proportion of missing data (33.1%) and the limited sample size. Malo JL et al. reported that the direct costs of OA were greater among patients with lower FEV1, reflecting the increased use of healthcare facilities associated with more severe airflow obstruction [35]. The univariate association observed in our study is consistent with this literature and supports the importance of early diagnosis and prompt removal of exposure to preserve lung function and limit long-term costs. However, the missing FEV1 data precluded its inclusion in the multivariate model, representing an important limitation.

Separate component modeling was essential to address the structural endogeneity between PPD and indemnity costs. The finding that no clinical or demographic predictor was significantly associated with indemnity costs after excluding PPD (Model B) is a key result of this study. It demonstrates that indemnity costs in the Tunisian OA compensation system are primarily determined by the administratively assigned PPD rate, not by the clinical severity or demographic profile of patients. This has important policy implications [41]: interventions targeting clinical risk factors (age, lung function, symptom burden) may reduce medical costs but are unlikely to affect indemnity expenditures, which are governed by the legal compensation framework. The only modifiable predictor identified was age for medical costs (aIRR = 2.84), suggesting that early diagnosis and intervention in younger workers before 40 could reduce medical expenditure.

Targeted primary prevention in Tunisia’s textile sector (72% of OA cases) could generate substantial savings. With 29 expected annual cases of OA among 1.2 million textile workers (incidence 24.42/10⁶ workers/year), and a median annualized cost of 1,114.03 TND PPY, reducing incidence by 50% through exposure controls could save approximately 16,153 TND (€5,161) per year in direct costs for this sector alone.

Prevention and recommendations

Prevention is central to occupational medicine, encompassing both medical and technical dimensions that must be addressed in an integrated manner. Primary prevention through elimination or substitution of sensitizing agents is the most effective strategy to reduce the incidence of OA [42, 43].

Occupational physicians play a central role in the direction of collective technical prevention by recommending the elimination or substitution of allergens, as well as advising on the best ventilation and respiratory protection systems. Regular workplace exposure monitoring and health surveillance programs are essential components of an effective prevention strategy [16].

Individual technical prevention for OA involves selecting appropriate respiratory protection [2, 40]. During various medical examinations, occupational physicians assess the fitness of workers and identify those at risk of developing OA. The periodic occupational health examination is an opportunity for early detection and secondary prevention.

Furthermore, OA costs can be reduced by improving therapeutic practices and patient education programs [44]. Given that OA primarily affects young workers [45], ensuring a swift return to work without re-exposure is critical for both health outcomes and economic productivity. Legislative enforcement, worker education, and employer engagement are necessary to achieve significant reductions in the burden of OA.

Strengths and limitations

The cost of occupational asthma (OA) has received limited attention in the scientific literature. This study makes a significant contribution by addressing this gap, as it is the first study to estimate the direct costs of OA in the Tunisian private sector from the payer perspective. This study was limited to the private sector in four governorates in central Tunisia. The findings may not be generalizable to the public sector, other Tunisian regions, or countries with different occupational disease compensation systems.

Only CNAM-recognized OA cases were included, potentially excluding legitimate cases denied due to administrative issues. Underreporting may have occurred due to misdiagnosis, lack of awareness, or voluntary reporting systems [14]. Some workers may choose not to report their illness out of fear of job loss or because they believe that the compensation offered is inadequate.

Furthermore, our sample showed a marked female predominance (76%) and concentration in the textile sector. Women had lower median direct costs than men (4,477.81 vs. 5,944.24 TND), potentially biasing our overall median cost estimate downward and limiting generalizability to male-dominated industries.

Variable follow-up duration (estimated median 4.5 years, IQR 3.5–5.5) may have introduced bias, as patients recognized earlier had longer cost accrual periods. Annualized costs PPY were used to mitigate this, but residual confounders by follow-up duration cannot be excluded.

Hospitalization or transportation were not reimbursed for OA under the current CNAM schedule, which may underestimate the true burden of medical costs of OA.

In addition to direct cost limitations, indirect costs were intentionally excluded from this CNAM payer analysis. Productivity losses and job transitions were not captured, potentially underestimating the total societal burden of OA. Future studies should incorporate indirect costs, including absenteeism, presenteeism, and career change-related income losses.

All costs were expressed in nominal 2020 TND without inflation adjustment. While the median follow-up was relatively short (4.5 years), the Tunisian CPI increased by approximately 4.7% annually during 2015–2020, which may have modestly affected the comparability of costs incurred earlier versus later years. A sensitivity analysis using CPI-adjusted costs confirmed that inflation did not materially alter the direction or significance of any predictor.

Conclusion

This study provides the first estimate of direct costs of OA in the private sector of Tunisia, revealing a median cost of 4,593.52 TND (IQR: 3,408.00–6,871.72) per case, with CNAM reimbursing a total of 1,029,200.02 TND during the study period. Cash benefits, particularly PPD indemnity, constituted the dominant cost component. In multivariate analysis, age > = 40 years was the only significant predictor of medical costs (aIRR = 2.84), while no clinical or demographic factor predicted indemnity costs, confirming that these are administratively determined by the PPD rate. These findings underscore the need for targeted primary prevention in the textile sector to reduce disease incidence and medical expenditure, as well as a review of the current compensation framework to better align indemnity levels with clinical severity.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

OA

Occupational Asthma

CNAM

Caisse Nationale d’Assurance Maladie (National Health Insurance Fund)

OA/OD

Occupational accidents and occupational diseases

PPD

Permanent Partial Disability

TWD

Temporary Work Disability

TND

Tunisian Dinar

FEV1

Forced Expiratory Volume in one second

VC

Vital Capacity

PD20

Provocative Dose causing a 20% fall in FEV1

PWD

Permanent work disability

SPSS

Statistical Package for the Social Sciences

GLM

Generalized Linear Models

LMW

Low-Molecular-Weight

HMW

High-Molecular-Weight

Author contributions

CS, AF, and IK contributed to the conceptualisation, methodology, investigation, data curation, formal analysis, and writing – original draft. FC, NB, MB, and BBB contributed to methodology, validation, and writing – review & editing. SA, SC, AJM, AA, NM, and MM contributed to supervision, project administration, and writing – review & editing. All authors have read and approved the final version of the manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The data sets used or analysed during this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of Sahloul University Hospital, as well as by the General Director of Medical Control and the Director of the Central Medical Control District. As the study was based on the retrospective analysis of anonymized administrative data from the CNAM database, the requirement for informed consent was not required. Confidentiality and anonymity of all patient data were strictly maintained in accordance with Tunisian regulations and the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.