Web application development using multiple regression analysis to predict silicosis risk among stone carvers in Nakhon Ratchasima, Thailand

Abstract

Purpose

To develop a web-based application that uses multiple regression analysis to predict the risk of silicosis among stone carvers in Nakhon Ratchasima, Thailand.

Methods

Data from 243 stone carvers were used to construct a multiple regression model incorporating key associated variables key associated variables: concentration of respirable silica dust, daily working hours, presence of underlying diseases, and residential proximity to the workplace. Model performance was assessed via R², adjusted R², RMSE, and the significance of associated variables. The model was integrated into a user-friendly web application and deployed for real-time risk assessment among 362 stone carvers. Silicosis risk scores were categorized into five levels to facilitate interpretation and targeted interventions. The Mann‒Whitney U test was applied to compare silicosis risk scores before and after application.

Results

The regression model explained 66.2% of the variance in silicosis risk scores (adjusted R² = 0.662), with strong predictive accuracy (RMSE = 2.59). All predictor variables were statistically significant (p < 0.05). The web application assigned silicosis risk scores ranging from 12 to 25, with 77.1% of participants classified as “very high risk.” However, no statistically significant difference was observed between the model and web application silicosis risk scores (p = 0.155); nonetheless, the observed trend suggests potential benefits in enhancing worker awareness and promoting protective behaviors.

Conclusions

The developed multiple regression model and web application provide an effective tool for real-time silicosis risk prediction and stratification in stone carving communities. This digital health tool shows promise for early risk detection and prevention of silicosis in workers.

Introduction

Silicosis is one of the most common occupational lung diseases and is caused by prolonged inhalation of respirable crystalline silica (RCS) dust, which induces progressive inflammation and fibrosis in the lungs [1, 2]. Workers exposed to high concentrations of silica dust over extended periods, typically several years, are at increased risk of developing chronic silicosis, which can severely impair respiratory function and increase susceptibility to pulmonary infections and other comorbidities. The International Agency for Research on Cancer (IARC) has classified crystalline silica as a Group 1 human carcinogen, indicating its established potential to induce lung cancer in exposed populations [3, 4]. Despite advances in occupational health regulations and protective measures, silicosis continues to pose a major public health challenge worldwide, particularly in high-risk industries such as stone carving, mining, and construction. Recent reports indicate that new cases of silicosis are still being diagnosed regularly, emphasizing the ongoing burden of disease [5]. In Thailand, approximately 1,779 cases of silicosis have been officially reported since 2013, with Nakhon Ratchasima Province accounting for the highest cumulative burden 538 cases among stone carvers, representing nearly 30% of the national total [6]. The province is well known for its stone-carving industry, which primarily involves sandstone, granite, and marble used in religious sculptures, architectural decorations, and household ornaments. Among these materials, sandstone typically contains 70% to 90% crystalline silica (SiO₂) by weight [7], making it a major source of respirable crystalline silica (RCS) exposure during carving and polishing processes. Most workers rely on electrically powered rotary tools, grinders, and chisels operated in open or semi-enclosed workshops with limited dust-control systems and inconsistent use of personal protective equipment. These working conditions, coupled with prolonged daily exposure and insufficient mechanized dust suppression, substantially increase the risk of silicosis among stone carvers in the region. Many of these workers operate outside the formal labor protection and welfare system. Although basic personal protective equipment (PPE) such as respirators and dust masks is generally available in local markets, its consistent use remains limited due to factors such as affordability, discomfort in hot working environments, lack of employer enforcement, and low awareness of silica-related health risks. Consequently, workers are regularly exposed to respirable silica dust without adequate protection or systematic health monitoring. Despite these well-documented hazards, early detection and preventive interventions remain limited. Currently, accessible, practical tools that can provide individualized assessments of silicosis risk on the basis of quantifiable occupational and environmental exposures are lacking. The absence of such tools not only hinders timely medical intervention but also reduces the effectiveness of occupational health strategies aimed at minimizing disease progression. This highlights an urgent need for innovative approaches that can bridge the gap between exposure assessment, risk prediction, and practical occupational health management.

Various studies have reported the development of associated risks model for silicosis risk, such as the rapid associated risks model developed by Chaiyadej and Ketsakorn (2024) [8], which estimates silicosis risk on the basis of factors such as the concentration of silica dust exposure, working hours per day, underlying diseases, and separation of residence from a workplace, and Myers and Thompson (2022) [9] used a logistic regression model for the likelihood of silicosis, which is based on potential associated risks: cumulative exposure to respirable dust, age, years since first exposure, years of life lost prematurely, vital status, and a history of tuberculosis diagnosis. Another example is the associated risks models and risk assessment for silicosis using a retrospective cohort study among workers exposed to silica in China by Tse et al. (2015) [10], which provides the six associated risks into the model (age at entry of the cohort, mean concentration of respirable silica, net years of dust exposure, smoking, illiteracy, and no. of jobs). These associated risks models aim to enhance early diagnosis, inform workplace interventions, and support regulatory compliance. However, many of these associated risks models have not yet been implemented as web-based tools for silicosis risk assessment, unlike other fields such as cardiovascular disease [11, 12], diabetes [13], security risk assessment [14], cancer risk level [15], earthquakes [16], and chronic kidney disease [17], including job burnout [18], where accessible online tools are widely available for public health use.

Recent advancements in data analytics and web-based technologies offer new opportunities to improve occupational health monitoring. Although “separation of residence from workplace” is not widely reported as an associated risk factor in silicosis literature, it was identified as a statistically significant variable in our previous study [8]. In the Thai stone-carving context, many workers reside within or adjacent to their carving areas, leading to continuous environmental exposure to airborne silica dust beyond working hours. Conversely, individuals who live separately from their workplaces may experience lower cumulative exposure. Thus, this variable reflects the unique occupational residential overlap characteristic of informal artisanal settings rather than a universally recognized determinant of silicosis risk. The associated risks variables used in the silicosis risk model are based on findings from the study by Chaiyadej and Ketsakorn (2024) [8]. Therefore, this study aims to develop a web-based application that uses multiple regression analysis to predict the risk of silicosis among stone carvers in Nakhon Ratchasima, Thailand. This study is novel in that it integrates a multiple regression-based silicosis associated risks model into a web-based application specifically designed for stone carvers, a population with high occupational exposure to respirable crystalline silica. Unlike previous models, which have focused primarily on offline calculations or retrospective cohort analyses, the proposed application enables real-time, individualized risk assessment, allowing users to input their specific exposure and health data to receive immediate feedback. This digital approach not only facilitates early identification of high-risk individuals but also promotes proactive occupational health management, including the adoption of protective measures and timely medical consultation. By providing an accessible, user-friendly platform, the study addresses a critical gap in occupational health monitoring and offers a scalable tool that can be adapted for other high-risk industries. The web application is intended to serve as a practical tool for health professionals, occupational safety officers, and stone carvers, enabling the early identification of high-risk individuals and supporting timely intervention measures. Ultimately, this initiative is intended to contribute to a reduction in silicosis cases while strengthening occupational health systems across the region.

Materials and methods

Study area

This study was conducted in Sikhio District, Nakhon Ratchasima Province, Thailand, as shown in Fig. 1 Nakhon Ratchasima Province is home to an abundance of sandstone mountains, making sandstone a widely available material for local artisans. This resource is commonly used for carving various items, including Buddha statues, consecrated boundary stones (Sema stones) for Buddhist temples, and decorative pieces for home use. According to the study by Chaiyadej and Ketsakorn (2024) [8], the stone carving process follows a structured lifecycle that begins with the extraction of sandstone from quarries. The sandstone is obtained in various sizes, tailored to match the proportions specified in design drawings for carved items. Once extracted, workers proceed to cut and carve the sandstone into rough forms. For decorative items intended for home use, the process includes an additional stage in which the sandstone is split into smaller pieces. The final step involves surface chipping to achieve the desired texture and detail. Throughout these processes, particularly during cutting, splitting, and chipping, significant amounts of respirable crystalline silica (RCS) are generated, posing a serious health risk to the workers involved.

Fig. 1.

Nakhon Ratchasima Province map of Thailand

Study design and participants

Participants were recruited from local stone carving communities in Nakhon Ratchasima Province. All the participants were actively engaged in stone carving work at the time of data collection, during which a web application was utilized to assess their risk of exposure to RCS. The recruitment targeted workers actively engaged in the stone carving production cycle, including tasks such as mining, cutting, knocking, and sandstone carving. These stages were selected because they generate the highest concentrations of respirable crystalline silica (RCS) due to the use of electrically powered grinders and rotary tools. To enhance transparency and reproducibility, a summary of participants’ roles within the production cycle has been included in Table 1. A total of 362 participants aged between 18 and 60 years who had worked for more than one year in Sikhio District and were Thai nationals capable of communicating in the Thai language were selected for inclusion in the study. Individuals who refused to provide informed consent were excluded. Data was collected via a web application designed to assess the risk of silicosis. Silicosis risk was defined as the probability of developing radiographically confirmed silicosis (ILO profusion category ≥ 1/0) based on individual exposure and personal health factors. In this study, silicosis risk was quantified using a multiple regression model that estimated a continuous risk score derived from key associated risks variables identified in the study by Chaiyadej and Ketsakorn (2024) [8]. The following independent variables were included in the model: concentration of respirable silica dust (mg/m³), daily working hours, presence of underlying diseases, and separation of residence from the workplace.

Table 1.

Distribution of participants according to stone carving work cycle (n = 362)

Job cycle stage	Description of job function	No. of participants	Percentage (%)
Mining (rough carving)	Initial shaping of sandstone blocks using chisels and power grinders	30	8.3
Cutting or splitting (fine carving)	Initial stage of shaping raw sandstone blocks into smaller, workable pieces or rough forms that correspond to the intended sculpture or product design	141	39.0
Knocking	Intermediate stage of sandstone carving in which workers use hammers, chisels, or pneumatic tools to remove excess material and roughly shape the stone into its intended form	137	37.8
Sandstone carving	Multistage process involves the shaping, detailing, and finishing of sandstone blocks into desired artistic or architectural forms. The process typically includes splitting, knocking, cutting, carving, and polishing stages, each performed using manual tools (such as chisels and hammers) or electrically powered rotary grinders.	54	14.9

Web application development and data collection

The development of the proposed web application began with the collection of original data from 243 stone carvers in Nakhon Ratchasima Province, as reported by Chaiyadej and Ketsakorn (2021) [8], representing the initial phase of data acquisition. Detailed demographic and occupational information were obtained from each participant, including age, work experience, and work cycle characteristics. Participants were aged 18–60 years (mean ± SD: 38.9 ± 12.5 years) and had engaged in stone carving for more than one year in Sikhio District. The average duration of work experience was 8.8 ± 9.0 years, with 54.3% having worked in the occupation for over five years. A typical work cycle consisted of seven working days per week, with daily work durations of 8–12 h depending on production demand. Most participants reported limited rest periods and infrequent use of respiratory protective equipment. The majority performed carving and polishing tasks in open or semi-enclosed workshops using electrically powered grinders and rotary tools, which generated high concentrations of respirable dust. After preprocessing, which involved cleaning, standardizing, and validating the data, multiple regression analysis was performed to examine the relationships between these independent variables and silicosis risk, which was expressed as a continuous silicosis risk score. The web-based application quantifies individual silicosis risk using the multiple regression equation derived from the study dataset of 243 stone carvers in Nakhon Ratchasima Province. The dependent variable was the silicosis risk score, while the independent associated risks included respirable silica concentration (mg/m³), daily working hours, presence of underlying diseases (no = 1, yes = 2), and separation of residence from workplace (no = 0, yes = 1). All variables used in the multivariate regression model were coded according to standardized criteria to ensure consistency and interpretability. The dependent variable, silicosis risk score, was a continuous outcome derived from the regression equation. The independent variables were defined and coded as follows: concentration of respirable crystalline silica (RCS): continuous variable measured in mg/m³, geometric mean concentrations of RCS ranging from 0.005 to 0.039 mg/m³ were derived from the authors’ previous exposure assessment among stone carvers in Sikhio District, Nakhon Ratchasima Province, Thailand [8]. Personal air sampling was performed in the breathing zone of workers across major stone carving workstations, followed by gravimetric analysis and X-ray diffraction (XRD) to quantify crystalline silica. The reported exposure range reflects task-specific variability observed across different work processes within the stone carving production cycle; daily working hours: continuous variable representing the average number of hours worked per day; presence of underlying disease: binary variable indicating whether the participant had a pre-existing chronic medical condition (e.g., respiratory disease, cardiovascular disease, diabetes). Coding: 1 = No underlying disease; 2 = Presence of underlying disease; and separation of residence from workplace: binary variable coded as 0 = Residence not separated from workplace (living onsite) and 1 = Residence separated from workplace (offsite living). In the regression model, RCS concentration was treated as a continuous variable to capture exposure variability across different work tasks, allowing incremental increases in RCS intensity to be directly reflected in the predicted silicosis risk score. For the web-based application, RCS input values were assigned using empirically measured, task-specific exposure ranges derived from field measurements and defaulted according to job characteristics, as most artisanal workers lacked individual exposure monitoring data. Although reliance on task-based exposure estimates may limit generalizability, this approach provides a pragmatic and context-specific method for estimating silicosis risk in informal stone carving settings. The contribution (weight) of each associated risk was determined by its standardized β coefficient in the regression model, reflecting the relative influence of each factor on the overall risk score (1 to 25). Continuous variables (e.g., respirable silica concentration, daily working hours) were entered directly into the model, while categorical variables (e.g., presence of underlying diseases, separation of residence from workplace) were encoded numerically to allow integration into the same associated risks equation. Each variable was weighed according to its regression coefficient (β), as shown in Table 2. Risk scores were then normalized to a 1–25 scale, categorized as insignificant (1–5), low risk (6–8), moderate risk (9–15), high risk (16–20), and very high risk (21–25) risk levels. RCS concentration was modeled as a continuous variable ranging from 0.005 to 0.039 mg/m³, based on measured exposure data from the study population. In the multivariate regression model, each 0.01 mg/m³ increase in RCS concentration (X₁) was associated with an estimated 3.17-unit increase in the predicted silicosis risk score, after adjusting for other covariates. These threshold values were determined based on quartile distribution of predicted probabilities and validated through expert consensus to ensure clinical interpretability and practical applicability in occupational health screening. The resulting regression formula was programmed into the web application to calculate individual risk scores interactively. A schematic overview of variable integration and relative weighting is presented in Fig. 2, adapted conceptually from the cumulative exposure framework proposed by Howlett et al. (2024) [19]. In particular, the model emphasizes the contribution of RCS exposure (mg/m³·years) to cumulative risk (%) estimation.

Table 2.

Summary of multiple regression analysis for predicting silicosis risk among stone carvers (n = 243)

Predictor variable	Unstandardized coefficientsa		t-value	p-value*
	β	Std. Error
Constant	9.481	1.771	5.353	< 0.001
Concentration of respirable silica dust (mg/m3): X1	317.267	25.308	12.536	< 0.001
Working hours per day (hours): X2	0.712	0.141	5.035	< 0.001
Presence of underlying diseases (1 = no, 2 = yes): X3	-2.863	8.220	-3.485	0.001
Separation of residence from workplace (0 = no, 1 = yes): X4	1.374	0.478	2.875	0.005

Model summary

R = 0.822, R² = 0.675, adjusted R² = 0.662, Std.Error = 2.442, F = 53.521, RMSE = 2.59, p-value < 0.001:

Indicates strong model fit and predictive reliability

^aUnstandardized coefficients are the default values returned by all the statistical programs. For a coefficient value of β = 317.267, for example, a unit increase in respirable crystalline silica (RCS) concentration (mg/m³) was associated with an average 317.267-unit increase in the silicosis risk score. Accordingly, each 0.01 mg/m³ increment in RCS concentration corresponded to an estimated 3.17-unit rise in the predicted silicosis risk score, after adjusting for other covariates

*p-value < 0.05

Fig. 2.

Conceptual framework illustrating integration of continuous and categorical predictors into the web-based silicosis risk model

Model performance was evaluated via several metrics, including the coefficient of determination (R²), adjusted R², root mean square error (RMSE), and the statistical significance (p-value) of each predictor variable, to ensure both the accuracy and reliability of the regression model. The model that demonstrated the highest predictive accuracy and statistical robustness was selected as the basis for risk prediction in the web application, as shown in Fig. 3. Data preprocessing was performed prior to model development to ensure the accuracy, completeness, and consistency of all variables. The initial dataset comprised records from 243 stone carvers. Missing or incomplete responses (< 5% of total entries) were handled using mean substitution for continuous variables and mode substitution for categorical variables. Outliers were identified through interquartile range (IQR) analysis and verified against original data forms before removal. All continuous variables (e.g., respirable crystalline silica concentration and daily working hours) were standardized using z-score normalization to minimize scale bias during regression analysis. Following data cleaning, the dataset was randomly divided into training (70%) and testing (30%) subsets to evaluate model performance and prevent overfitting. The regression model was trained using the development dataset and validated on the testing dataset. Model accuracy and predictive performance were assessed using the coefficient of determination (R²), adjusted R², and root mean square error (RMSE). Statistical significance of predictor variables was evaluated at p < 0.05. The development of the associated risks web application involves three main steps: front-end, back-end, and model integration. In the front-end stage, a user-friendly interface is designed using modern web technologies, including Next.js, Tailwind CSS, Chakra UI, and Chart.js, to ensure accessibility, visual clarity, and ease of use for the target users. The back-end stage focuses on implementing regression model logic via server-side programming languages such as Node.js, Next.js API Routes, and NextAuth.js. In the final step, model integration, the validated multiple regression equation is embedded into the back-end system to process the user input. This functionality allows the system to generate real-time associated silicosis risks and provide immediate, data-driven feedback to users. The application uses a backend-as-a-service database to securely store and retrieve user data in real time. The web application is deployed via Vercel, a cloud platform optimized for Next.js, enabling seamless continuous deployment, fast performance, and reliable global accessibility, as shown in Fig. 4. Usability testing was conducted with a separate group of stone carvers who were not included in the original model development dataset. This was done to evaluate the functionality of the web application and assess the clarity and usability of the user interface.

Fig. 3.

Workflow diagram of web application development

Fig. 4.

User interface of the web application

Statistical analysis

Descriptive statistics, including percentages, means, standard deviations (SDs), minimums, and maximums, were used to summarize participants’ demographic characteristics and silicosis risk levels. Chest radiographs were obtained to screen for silicosis and other respiratory abnormalities. All images were evaluated according to the International Labour Organization (ILO) International Classification of Radiographs of Pneumoconioses. Two certified radiologists, blinded to participants’ exposure information, independently reviewed each film, and any discrepancies were resolved by consensus. The diagnosis of silicosis was defined primarily by radiographic evidence consistent with ILO criteria, in conjunction with characteristic occupational exposure histories. The web-based risk score functioned as a supplementary tool to estimate disease likelihood and identify individuals requiring further clinical evaluation, thereby providing complementary support to the radiographic findings in assessing participants’ respiratory health status. The Mann–Whitney U test, a nonparametric alternative to the independent samples t test, was used to compare differences in numerical parameters between two independent groups (model development: n = 243; web application: n = 362), particularly when the data did not meet the assumption of normality. A p-value less than 0.05 was considered to indicate a statistically significant difference between the two groups, whereas a p-value greater than 0.05 indicated no statistically significant difference. All the statistical analyses were performed via the Statistical program Statistical Package for Social Sciences (SPSS version 29).

Results

The development of the web-based application began with the collection of primary data from 243 stone carvers in Nakhon Ratchasima Province. This dataset was used to construct a multiple regression model for predicting associated silicosis risks based on four key variables: concentration of respirable silica dust, daily working hours, presence of underlying diseases, and residential proximity to the workplace. The model demonstrated strong explanatory power, with a coefficient of determination (R²) of 0.675 and an adjusted R² of 0.662, indicating that approximately 66.2% of the variance in silicosis risk scores could be explained by the selected associated risks predictors. These values suggest a robust fit of the model while minimizing the potential for overfitting, as shown in Fig. 5.

Fig. 5.

The relationship between observed and predicted silicosis risk scores with adjusted R² and RMSE

Model performance was evaluated using error-based metrics, yielding a root mean square error (RMSE) of 2.59, which indicates a low deviation between observed and associated risks scores. All four associated risks variables demonstrated statistical significance (p < 0.05), confirming their independent contributions to the model’s predictive accuracy (Table 2). The selected silicosis risks predictors respirable silica dust concentration (mg/m³), daily working hours, presence of underlying diseases, and separation of residence from workplace were identified based on empirical evidence from a previous study by Chaiyadej and Ketsakorn (2024) [8], which established their relevance to silicosis risk among stone carvers. These results collectively support the reliability and validity of the model for integration into a predictive web application. Following model development, the web application was applied in real-world settings to generate silicosis risk scores for an expanded cohort of 362 stone carvers. The individual risk scores ranged from 12 to 25 (mean = 22.15, SD = 3.168), suggesting substantial heterogeneity in the predicted risk levels within the population. Risk scores were stratified into five categories: insignificant, low, moderate, high, and very high risk, providing a structured framework for interpretation and intervention (Table 3). Notably, most participants (59.4%) fell within the “very high risk” category, reflecting a substantial occupational health burden in this community and underscoring the urgency of preventive interventions.

Table 3.

Silicosis risk scores from the web application (n = 362)

Silicosis risk scores	Silicosis risk levels	Number	Percent
1–5	Insignificant	0	0
6–8	Low risk	0	0
9–15	Moderate risk	26	7.2
16–20	High risk	121	33.4
21–25	Very high risk	215	59.4

The Mann‒Whitney U test was conducted to compare silicosis risk scores before and after the implementation of the web application. Prior to the application’s introduction, the mean rank of risk scores was 314.94, whereas following its development, the mean rank declined to 294.99. As presented in Tables 4 and 5, this difference approached the threshold of statistical significance (U = 41,085, p = 0.155), suggesting a modest but noteworthy reduction in predicted silicosis risk scores among the study population. Although the effect size was relatively small, the consistent downward trend in scores highlights the potential of the web-based approach to positively influence occupational health outcomes.

Table 4.

Comparative analysis of mean ranks for silicosis risk scores during model development and subsequent web application evaluation

Silicosis risk score groups	N	Mean rank	Sum of ranks
Model development	243	314.93	76,527
Web application	362	294.99	106,788
Total	605

Table 5.

Comparison of silicosis risk scores before and after web application development via the Mann‒Whitney U test

	Silicosis risk scores
Mann‒Whitney U test	41,085
Wilcoxon W	106,788
Z	-1.423
Mean of U (µu)	43,983
Standard deviation of U (σu)	2,107
Critical U (lower bound, U*)	39,854
Critical U (upper bound, U*)	48,112
Asymptotic significance (p-value)	0.155

Note: Statistical significance was assessed using the asymptotic p-value of the Mann–Whitney U test, which represents the probability of observing the data under the null hypothesis of no difference between groups. A p-value < 0.05 was considered statistically significant. In this study, the observed U value fell within the non-rejection region (39,854 − 48,112), consistent with the asymptotic p-value (p = 0.155), indicating no statistically significant difference in silicosis risk scores between groups

Discussion

This study developed and validated a web application to predict associated risks among stone carvers in Nakhon Ratchasima Province via a multiple regression model based on key exposure and health-related variables. The model demonstrated strong predictive ability, with an adjusted R² of 0.662, indicating that approximately 66.2% of the variance in silicosis risk scores could be explained by the concentration of respirable silica dust, working hours, underlying diseases, and residential proximity to the workplace. This high explanatory power underscores the relevance of these variables in assessing silicosis risk within this occupational group [8]. The performance metrics, including an RMSE of 2.59, further support the model’s accuracy and reliability for associated risks prediction. The statistically significant coefficients of all associated risks variables confirm their meaningful contributions, aligning with literature that highlights silica dust exposure and duration of work as critical determinants of silicosis development [20–22]. Notably, the negative coefficient for underlying diseases suggests a complex relationship warranting further investigation, possibly reflecting variations in self-reports or disease management among participants [23–25]. Although individual characteristics such as tobacco use or metabolic disorders may influence respiratory health outcomes, the present analysis emphasizes that silica exposure remains the predominant and most consistent determinant of silicosis risk among stone carvers. The variability observed in other comorbid factors across studies (including tobacco use and diabetes mellitus) underscores the importance of maintaining silica exposure as the principal focus when evaluating and mitigating occupational risk in this population.

Applying the model within the web application allowed real-time generation of silicosis risk scores for 362 stone carvers, with scores ranging from 12 to 25 on the 25-point scale. Most participants (59.4%) were classified in the “very high risk” category, indicating a substantial occupational health burden in this community. This risk stratification facilitates targeted health interventions, allowing resources to be prioritized for those at greatest risk [8]. The Mann–Whitney U test comparing silicosis risk scores before and after web application deployment revealed a marginally significant reduction in mean rank scores (U = 41,085, p = 0.155). The modest reduction in mean rank silicosis risk scores observed after implementation of the web-based application should be interpreted with caution. The application was designed as a risk assessment and communication tool rather than a direct exposure-reduction intervention. Several key predictors in the regression model, including RCS concentration, presence of underlying disease, and residential proximity to the workplace, are not readily modifiable in the short term. Consequently, the observed change in rank scores does not imply a true reduction in measured RCS exposure or reversal of disease-related risk. Instead, the downward trend in predicted risk scores is more plausibly attributable to increased awareness, improved risk perception, and more accurate reporting of job characteristics following individualized feedback [26, 27]. Although the model incorporates RCS concentration as a primary determinant of risk, the web application relies on task-based exposure estimates rather than real-time personal monitoring. Heightened awareness may have prompted some workers to temporarily adjust work practices (e.g., increased use of wet methods or local ventilation), which could influence task-level exposure assumptions used by the model without necessarily reflecting sustained reductions in actual exposure intensity [28, 29]. Importantly, the web application does not assume that workers reduced working hours, relocated their residences, or modified non-reversible health conditions in response to feedback. Rather, its primary value lies in facilitating early identification of high-risk individuals, improving understanding of cumulative risk drivers, and supporting informed decision-making regarding protective behaviors and medical follow-up. As such, the observed reduction in mean rank scores should be viewed as an indicator of improved risk awareness rather than evidence of causal exposure reduction.

The developed silicosis risk score integrates multiple exposure and work-related variables into a composite index that quantifies the combined influence of these predictors on an individual’s probability of developing silicosis. Unlike conventional clinical risk models that emphasize non-modifiable biological factors, this model focuses on modifiable occupational and behavioral determinants, enabling practical risk reduction strategies. A visual representation of the classification outcomes (Fig. 6) demonstrates the predominance of high- and very high-risk categories within the study population and workflow of the web-based silicosis risk prediction system. This stratification enhances the interpretability of risk levels for both workers and occupational health practitioners and provides a data-driven foundation for targeted interventions, such as prioritizing medical surveillance, reinforcing the use of respiratory protective equipment, and implementing dust control measures for individuals at highest risk.

Fig. 6.

A visual representation of the classification outcomes of silicosis risk: (a) workflow of the web-based silicosis risk prediction system; (b) silicosis risk prediction results

Importantly, the integration of the regression model into a digital platform enabled real-time, individualized risk assessments tailored to workers’ specific exposure and health profiles. This feature represents a significant advancement over traditional risk assessment methods, which often rely on retrospective health outcomes or generalized exposure metrics. By providing immediate feedback, the application facilitates timely recognition of at-risk individuals, thereby supporting early medical evaluation and potentially mitigating the progression of silicosis.

This study has several strengths, including the use of a well-characterized dataset for model development and rigorous validation through real-world applications [30]. However, limitations include reliance on self-reported data for some variables [31], potential selection bias, and the cross-sectional design limiting causal inferences [32]. Additionally, further refinement of the model incorporating additional environmental and genetic factors may increase its predictive accuracy [22, 33–35]. In conclusion, the integration of a multiple regression-based risks model into a web application provides a valuable tool for silicosis risk assessment among stone carvers. The application offers an accessible platform for occupational health monitoring, enabling early risk detection and tailored interventions. Continued development and deployment of such digital tools have the potential to significantly reduce the burden of silicosis in high-risk populations.

The observed reduction may be explained by several interrelated mechanisms. First, providing individualized, real-time feedback on silicosis risk has likely increased workers’ awareness of their own vulnerability, a factor known to play a critical role in motivating protective behaviors in occupational health settings [36]. By visualizing their personal risk level and receiving tailored recommendations, stone carvers may have been more inclined to adopt safer work practices, such as consistent use of respiratory protective devices, improved ventilation practices, or limiting daily exposure duration. Second, the web application’s risk stratification framework may have facilitated the identification of workers in higher risk categories, enabling both self-initiated protective actions and early referral for medical consultation. This form of risk communication aligns with preventive health behavior models, where timely recognition of elevated risk acts as a catalyst for behavioral change and compliance with safety protocols. Third, the digital format of the application ensured accessibility and convenience, allowing workers to repeatedly engage with the tool and track changes in their risk scores over time. This iterative process may have reinforced positive health behaviors, providing feedback loops that sustain risk-reducing practices. Despite the reduction being modest, its significance lies in demonstrating the potential for scalable, low-cost, and user-friendly digital tools to complement traditional occupational health interventions. These findings suggest that even small improvements in awareness and protective behavior, when applied across large populations of at-risk workers, could yield substantial reductions in disease burden over time.

Limitations

While the web application demonstrated a modest reduction in mean silicosis risk scores following its use, this should not be interpreted as a causal decrease in actual exposure or disease risk. Given the cross-sectional design, the observed change likely reflects increased risk awareness and improved input accuracy rather than true modification of exposure determinants, as variables such as respirable silica concentration, exposure duration, residential proximity to the workplace, and underlying health conditions remain relatively stable in the short term. Nevertheless, the interactive nature of the tool may have enhanced users’ understanding of their personal risk profiles, leading to more accurate self-assessment and heightened awareness. Several limitations should be acknowledged. First, the reliance on self-reported data for certain variables, such as underlying diseases and working hours, may introduce recall bias or reporting inaccuracies. Second, the cross-sectional design limits the ability to establish causal relationships between associated risks variables and silicosis risk. Third, the sample was limited to stone carvers in Nakhon Ratchasima Province, which may affect the generalizability of the findings to other regions or occupational groups, its applicability to other regions or different occupational settings may be limited. Therefore, external validation of the web application in other stone-carving communities or occupational groups is recommended for future research. Fourth, silicosis diagnosis relied solely on risk scores, and the absence of ILO-classified chest radiographs and high-resolution CT scans may have underestimated early or subtle cases. Future studies should include both diagnostic methods to improve accuracy and detect early-stage silicosis. Finally, the model did not account for potential environmental or genetic factors that may influence silicosis susceptibility, indicating opportunities for future refinement. Moreover, as the intervention primarily targets personal-level behavioral change using protective equipment, the lowest tier in the hierarchy of occupational health controls, it should be regarded as a complementary strategy to enhance awareness and compliance alongside engineering and administrative measures that mitigate silica exposure at its source.

Recommendations

Based on these findings, we recommend broader implementation of the web application as an accessible tool for early silicosis risk assessment in stone carving communities. Efforts should be made to promote user education on occupational health and safe work practices informed by the application’s outputs. Future research should focus on longitudinal studies to evaluate the long-term impact of the application on silicosis incidence and progression. Additionally, integrating environmental monitoring data and exploring genetic predispositions could enhance the model’s predictive accuracy. Collaboration with local health authorities to facilitate medical follow-ups for high-risk individuals identified by the application is also encouraged.

Conclusions

This study developed and validated a multiple regression model to predict silicosis risk among stone carvers in Nakhon Ratchasima Province, integrating key factors such as respirable silica dust concentration, working hours, underlying health conditions, and residential proximity to the workplace. The model demonstrated good explanatory power (adjusted R² = 0.662) and was successfully implemented as a web-based, real-time risk assessment tool. This platform stratifies workers into five risk levels, supporting targeted risk communication and preventive interventions.

Although the statistical analyses revealed no significant difference in the silicosis risk scores between the model development and post application phases, a downward trend in predicted silicosis risk was observed, suggesting that personalized digital feedback may motivate protective behaviors. The web-based tool’s scalability and accessibility highlight its potential for integration into national occupational health programs and for adaptation to other silica-exposed industries. In summary, this digital approach represents a promising and practical innovation for silicosis prevention, enabling early identification, worker empowerment, and enhanced occupational health surveillance in high-risk settings.

Author contributions

Conceptualization: Ratchapong Chaiyadej and Arroon Ketsakorn; Methodology: Ratchapong Chaiyadej and Arroon Ketsakorn; Software: Ratchapong Chaiyadej and Arroon Ketsakorn; Validation: Arroon Ketsakorn; Formal analysis: Arroon Ketsakorn; Investigation: Ratchapong Chaiyadej and Arroon Ketsakorn; Resources: Ratchapong Chaiyadej and Arroon Ketsakorn; Data curation: Ratchapong Chaiyadej and Arroon Ketsakorn; Writing – original draft: Ratchapong Chaiyadej; Writing – review & editing: Arroon Ketsakorn; Visualization: Ratchapong Chaiyadej and Arroon Ketsakorn; Supervision: Arroon Ketsakorn; Project administration: Arroon Ketsakorn; Funding acquisition: Arroon Ketsakorn.

Funding

This research project is supported by the Thai Health Promotion Foundation (ThaiHealth): 000705/64.

Data availability

The authors are willing to provide the raw data used in this study upon request.

Declarations

Ethical approval

The Human Research Ethics Committee of Thammasat University reviewed and approved the aims and procedures of this study. Ethical approval no.061/2566, and the date of approval was 24 July 2023.

Informed consent

Informed consent was obtained from all individual participants included in the study. The authors affirm that human research participants provided informed consent for publication.

Competing interests

The authors declare no competing interests.