Artigos

Conteúdo (article):

Validade e confiabilidade da versão brasileira do Índice de Capacidade para o Trabalho - ICT: uma reavaliação

Validity and reliability of the Brazilian version of the Work Ability Index – WAI: a re-evaluation

Maria Carmen Martinez (https://orcid.org/0000-0002-4427-5976 ) a
Maria do Rosário Dias de Oliveira Latorre (http://orcid.org/0000-0002-5189-3457 ) b
Frida Marina Fischer (https://orcid.org/0000-0001-9403-6300 ) c


a WAF Informática e Saúde, São Paulo – SP – Brasil - mcmarti@uol.com.br
b Faculdade de Saúde Pública da Universidade de São Paulo, Departamento de Epidemiologia, São Paulo – SP – Brasil - mdrddola@usp.br
c Faculdade de Saúde Pública da Universidade de São Paulo, Departamento de Saúde Ambiental São Paulo – SP – Brasil - fischer.frida@gmail.com

Resumo
Objetivo: reavaliar as propriedades de medida da versão em português do Índice de Capacidade para o Trabalho (ICT). Métodos: estudo transversal junto a 3.051 trabalhadores de enfermagem do estado de São Paulo, que preencheram um formulário com questões sobre características individuais, das condições do trabalho, e o ICT. A validade foi avaliada por meio de análise discriminativa de grupos conhecidos, e da análise fatorial exploratória (AFE) com análise paralela (AP). A confiabilidade foi avaliada usando os coeficientes alfa de Cronbach e ômega de McDonald. Resultados: na análise de grupos conhecidos, todas as variáveis individuais e do trabalho apresentaram associação estatisticamente significativa com o ICT (p<0,005). A AFE evidenciou estrutura unidimensional, confirmada pela AP. O modelo foi adequado (Raiz do Erro Quadrático Médio de Aproximação=0,125; Índice de Tucker-Lewis =0,900), embora o Índice de Ajuste Comparativo =0,882 tenha sido ligeiramente inferior ao esperado. A confiabilidade foi satisfatória (coeficiente alpha de Cronbach = 0,787 e ômega de McDonald=0,819). Conclusões: o ICT continua sendo uma medida valida e confiável para avaliação da capacidade para o trabalho no atual contexto brasileiro do trabalho.

Descritores: avaliação da capacidade de trabalho; condições de trabalho; questionários; reprodutibilidade dos testes; saúde do trabalhador.

Abstract

Objective: To reevaluate the measurement properties of the Portuguese version of the Work Ability Index (WAI). Methods: A cross-sectional study was conducted with 3,051 nurses in the state of São Paulo. Participants answered a questionnaire on individual characteristics, working conditions, and WAI. Validity was assessed through discriminative analysis of known groups and exploratory factor analysis (EFA) with parallel analysis (PA). Reliability was assessed using Cronbach\'s alpha and McDonald\'s omega coefficients. Results: In the analysis of known groups, all investigated individual and work variables showed statistically significant associations with WAI (p<0.005). EFA revealed a unidimensional structure, which was confirmed by PA. The model proved to be appropriate (Root Mean Square Error of Approximation=0.125; Tucker-Lewis Index=0.900), although the Comparative Fit Index=0.882 was slightly lower than expected. Reliability was satisfactory (Cronbach\'s alpha coefficient=0.787 and McDonald\'s omega=0.819). Conclusions: The WAI continues to be a valid and reliable measure to evaluate work ability in the current context of Brazilian work.

Key words: Work capacity evaluation; Working conditions; Questionnaires; Reproducibility of results; Occupational health.

Introduction
Work ability – WA is understood as a condition resulting from the combination of human resources in relation to physical, mental, and social work demands, organizational culture and work environment1. This concept is expressed as “how well a worker is or will be in the present or near future, and how capable this worker is of performing work given the demands of the job and the health status and mental and physical capabilities of the worker”1.
WA is characterized as a relevant indicator of worker health, incorporating aspects related to physical health, psychosocial wellbeing, individual competence, and working conditions1. It has a predictive value for several outcomes, including one’s quality of life and wellbeing, absenteeism and presenteeism, performance, illness, job loss, retirement, leaving profession and mortality2-8.
The demographic and epidemiological transitions that Brazil has been going through in recent few decades intensify critical issues in terms of WA and the employability of both aging and younger workers7,9,10. The relevance of WA also increases in the context of decent work and changes in the world of work, with the fragmentation of labor relations, precarious working conditions, increased use of information and communication technologies, environmental changes, pandemics, migratory movements, and gender issues11.
The Work Ability Index (WAI) is an instrument that measures WA based on the worker\'s own perceptions, enabling early assessment and detection of changes and supporting the implementation of preventive measures at both individual and collective levels1,12,13. The WAI was structured by the Finnish Institute of Occupational Health (FIOH) in the 1980s, based on studies on aging of the working population and the sustainability of the social security system1,13. Since then, the WAI has been widely used in research worldwide and has been translated into more than 30 languages13. In recent decades, several studies have been conducted to evaluate the measurement properties of the instrument14-21. In Brazil, studies focusing on the WA began after the translation and adaptation of the WAI into Brazilian Portuguese in 19961. The first assessment of the instrument\'s validity and reliability was carried out in 200912. It was conducted with a sample of workers in the electric power sector in the state of São Paulo, who were predominantly male (90.9%) and young (median age of 37.0 years). WAI results showed satisfactory measurement properties in terms of construct validity, criterion validity, and reliability12.
One point to consider is that studies to assess WAI validity adopted different methods, demonstrating satisfactory measurement properties, but often with different results, especially regarding the factorial structure14-21. Another relevant point is that, since the creation of WAI and its adaptation for use in Brazil, many changes occurred in the country\'s working conditions and relationships, as well as in the demographic structure of the population, and health population 7,9,10. Furthermore, the psychometric performance of an instrument may change over time and be influenced by several factors, including changes in contexts and social situations22. Considering these aspects, this study aims to reassess the measurement properties of the Portuguese version of the WAI as an indicator of worker health, among a population of nurses in the state of São Paulo.

Methods
Study design, study population, and sample
A cross-sectional study was conducted with 3,051 active nurses from the state of São Paulo. There were 495,602 professionals (25.0% of the Brazilian contingent) actively registered with the Regional Nursing Council of São Paulo (Conselho Regional de Enfermagem de São Paulo – Coren-SP), distributed across 14 regional subsections, at the time of data collection. The exclusion criterion was not having a valid email address registered with Coren-SP (approximately 17.0% of the professionals). Among the 411,162 eligible individuals, 1.0% joined the study (3,993 individuals), of which 3,051 (76.4%) were active in the profession, totaling a participation rate of 0.74%. This study population was chosen because nursing is a professional category subject to significant physical and mental workload, which can compromise WA2,7. In addition, operationally, it was possible to access the entire eligible population.
The analysis of losses, using the chi-square test, showed that there were differences between participants and non-participants regarding the age group under 40 years (50.1% vs 48.3%, p=0.046), residents in the capital city of the state of São Paulo (36.8% vs 35.0%, p=0.049), and the professional category of nurses (61.6% vs 26.8%, p<0.001). Another analysis with this same sample showed a high frequency of exposure to risk factors for the occurrence of stress, illness, work incapacity, and intention to leave the profession, details of which can be found in another publication23.

Data collection
Data collection took place between October 2018 and March 2019. Coren-SP sent a message to the entire eligible population via email, containing information about the survey and the link to access a Google Form via the Internet, in addition to two additional posts encouraging participation.
The form consisted of modules. The first module contained questions about sociodemographic characteristics (sex, age, monthly family income); lifestyle (smoking, risky alcohol consumption assessed by the CAGE questionnaire, physical activity, body mass index and sleep quality); and occupational history (professional category, night work, weekly work hours, recent history of work-related illness/accident in the last 12 months, and intention to leave the nursing profession).
The second module was the Brazilian version of the WAI, consisting of 10 questions, which are summarized in 7 dimensions: (1) “current work ability as compared with the lifetime best”, obtained from 1 question with a score of 0 to 10 points; (2) “work ability in relation to the demands of the job”, obtained from two questions that consider the nature of the work (physical, mental, or mixed) and which, when weighted, provide a score of 2 to 10 points; (3) “number of current diseases diagnosed by a physician”, with the result obtained from a list of 51 diseases, defining a score of 1 to 7 points; (4) “estimated work impairment duo to diseases”, obtained from 1 question with a score ranging from 1 to 6 points; (5) “sick leave during the past year”, obtained from a question about the number of absences, categorized into 5 groups, with scores ranging from 1 to 5 points; (6) “own prognosis on work ability two years from now”, obtained from a question with a score of 1, 4, or 7 points; and (7) “mental resources”, from a score of 1 to 4 points obtained by weighting the answers to 3 questions1. The results of the 7 dimensions provide a measure of work ability that ranges from 7 to 49 points. The instructions for calculating the score are found in Tuomi et al. (2005)1. The score was categorized as excellent, good, moderate, or low. For workers up to 35 years of age, the cut-off points were: excellent – 45 to 49, good – 41 to 44.9, moderate – 37 to 40.9, and low – 7 to 36.9. For workers aged 35 or over, the cut-off points were: excellent – 44 to 49, good – 37 to 43.9, moderate – 28 to 36.9, and low – 7 to 27.924.
The next module consisted of instruments to assess psychosocial and physical work conditions. Psychosocial stress at work was measured using the Job Stress Scale (JSS), an adaptation for use in Brazil of the Job Content Questionnaire, based on the Demand-Control Model, with 17 questions covering the dimensions of demands, control and social support25. The dimensions were dichotomized by the midpoint of each score. Demand and control were combined into 4 categories: work with high strain, active job, low strain, and passive job23,25. Stressors in the psychosocial work environment were assessed using the Brazilian version of the Effort-Reward Imbalance questionnaire, structured on the theoretical model Effort-Reward Imbalance (ERI), consisting of 23 questions covering 3 dimensions: efforts, rewards, and overcommitment26. To classify the imbalance, the ratio between effort and reward is calculated and multiplied by 6/11, providing a coefficient ranging from 0.17 to 5.00 points, where values greater than 1.0 indicate an imbalance between efforts and rewards26.
The Work-Related Activities That May Contribute to Pain and/or Injury (WRAPI) questionnaire, a validated version for use in Brazil, was used to assess characteristics of the physical environment that could contribute to the development of musculoskeletal disorders, providing a score from 0 to 150 points27. The scores were categorized into tertiles23.
The last module, to assess the intention to leave the profession (ILP), adopted the translation of the NEXT-Study question "How often during the course of the past year have you thought about giving up nursing?"2. The variable was dichotomized into intend to leave the profession (considered the possibility several times a month or more often) and have no intention of leaving the profession (considered the possibility only a sometimes a year or never)2.

Statistical analyses
Initially, the Shapiro-Francia test was performed, which demonstrated that the WAI score and the other continuous quantitative variables did not adhere to a normal distribution, determining the use of nonparametric statistical tests. A descriptive analysis was then performed using means, standard deviations, minimum and maximum values, and quartiles of the scores of the quantitative variables, and proportions for the qualitative variables.
To assess the measurement properties of the WAI, three types of analysis were chosen: (1) construct validity based on hypothesis testing, through discriminative analysis of known groups, (2) construct validity through assessment of internal structure, and (3) reliability through assessment of internal consistency22,28,29. The content validity analysis was performed and is presented in a previous publication1.
For construct validity based on discriminative analysis of known groups, mean comparison tests were adopted using the Mann-Whitney test for qualitative variables with 2 categories, the Kruskal-Wallis test for variables with 3 or more categories, and the Tukey post hoc test. The analyzed variables were sex, age group, monthly family income, smoking, risky alcohol consumption, physical activity, body mass index (BMI), sleep quality, insomnia, professional category, night work, weekly work hours, recent history of work-related accidents or illnesses in the last 12 months, intention to leave the nursing profession, and working conditions (JSS, ERI, and WRAPI). These variables were selected, as their approach, association, or predictive value in relation to WA are well-known in the literature1-9,23.
For construct validity through the assessment of the internal structure of the WAI, Exploratory Factor Analysis (EFA) was performed, based on the procedures proposed by Damásio (2012)30 and Lorenzo-Seva & Ferrando (2013)31. This analysis aimed to verify whether the three factors identified in the previous study carried out in Brazil would be confirmed in this population: (1) mental resources, (2) presence of diseases and restrictions resulting from health status, and (3) worker perception of their WA12. The following steps were followed: (a) polychoric correlation matrix, robust factor analysis technique, and extraction method by robust diagonally weighted least squares (RDWLS); (b) definition of the number of factors to be retained through Parallel Analysis (PA), with random permutation of the observed data and Robust Promin rotation; (c) assessment of the suitability of the use of EFA for the data set using Bartlett\'s sphericity test and the Kaiser-Meyer-Olkin (KMO) measure; (d) the suitability of the obtained model was assessed using the Root Mean Square Error of Approximation (RMSEA), Comparative Fit Index (CFI), and Tucker-Lewis Index (TLI) indices; (e) the reliability of the factor structure and the stability of the factors were assessed using the composite reliability (CR) index, the Overall Reliability of fully-Informative prior Oblique N-EAP scores (ORION) index, the estimate of the generalized H-Index (H-Observed), and the Factor Determinacy Index (FDI); (f) the unidimensionality of the questionnaire was assessed using the Unidimensional Congruence (UniCo), Explained Common Variance (ECV), and Mean of Item Residual Absolute Loadings (MIREAL) indicators. A new reliability assessment was performed by analyzing the internal consistency of the questionnaire items, using Cronbach\'s alpha and McDonald\'s omega coefficients.
The choice of these analysis strategies is in line with that set forth in the literature22,28-31. In all analyses, a significance level of 5.0% was considered.

Software used
Data collection used the Google Forms application. The data were then exported to Excel and later to STATA software, version 14, to verify consistency (completeness and quality of data), recoding, descriptive analyses, and association and correlation tests. The Factor Program, version 11.05.01, was used to perform the EFA. The Composite Reliability Calculator31 was used to calculate the CR index. JAMOVI was used to calculate the alpha and omega coefficients.

Ethical aspects
This study was approved by the Board of Directors of Coren-SP and by the Research Ethics Committee from the Universidade de São Paulo (no. 2.614.513). It was conducted in accordance with the principles of the Declaration of Helsinki and the World Medical Association. The participation of the workers was voluntary, and the confidentiality of individual data was guaranteed. Access to the completion of the rest of the form was only granted after registering the option to accept the Free and Informed Consent Form on the electronic form. In compliance with the information security premises of Coren-SP, the researchers did not have access to the registration databases, electronic addresses, or nominal identification of the professionals, which made it impossible to verify double participation.

Results
The study population was predominantly female (85.6%). The mean age was 40.8 years (SD=10.0), ranging from 18.0 to 83.0 years, with a 1st quartile of 33.0, a median of 39.0, and a 3rd quartile of 48.0 years. Regarding the professional category, 61.6% were nurses, 32.5% were nursing technicians, and 5.9% were nursing assistants or attendants. The presence of work stressors showed 86.8% of high demands, 17.3% of low control, 16.2% of low social support, 24.3% of high effort, 17.8% of low rewards, 45.1% of overcommitment, and 36.2% with high exposure to situations that favor the occurrence of pain or injury. Workloads were 64.4% mixed, 31.6% predominantly mental, and 4.0% predominantly physical. Further details are available in another publication23.
The WAI had a mean of 36.5 points (SD=6.0), ranging from 14.0 to 49.0 points, with a 1st quartile of 33.0, a median of 37.0, and a 3rd quartile of 41.0 points. When the score was categorized, the distribution was excellent (11.0%), good (33.6%), moderate (36.4%), and poor (19.0%) WA, totaling 55.4% of people with compromised WA.
Regarding the dimensions that make up the WAI, the following results were presented: in the “current work ability as compared with the lifetime best”, 72.8% of the professionals had the 3 highest scores. In the “work ability in relation to the demands of the job”, 58.3% had the 3 highest scores. In the “number of current diseases diagnosed by a physician” 6.1% of the workers did not report the presence of illnesses (score=7), and 60.1% reported 5 or more illnesses (score=1). In the “estimated work impairment duo to diseases”, 51.8% of the workers reported that they do not have any impediment/illness to perform their work (score=6). In the “sick leave during the past year”, 54.6% of the workers denied having been away from work due to illnesses in the last 12 months (score=5). In the “own prognosis on work ability two years from now”, 70.3% of the workers considered it quite likely that they will be able to do their current job in 2 years (score=7). In the “mental resources”, 34.3% of the workers scored the highest in terms of mental resources (ability to enjoy daily activities, perception of being active and alert, feeling of hope for the future). Details are available in Supplementary Material 1.
The groups referring to the most frequent health problems cited with medical diagnosis cited in the WAI were musculoskeletal diseases in the back, arms, legs, or other part of the body (62.9%); mental disorders (60.1%); and injuries or accidents resulting from an accident (58.9%). Details are available in Supplementary Material 2.
In the analysis of known groups, all investigated variables showed a statistically significant association with the WAI, which was able to discriminate workers with worse mean among women (p<0.001), age 45 years or older (p=0.032), monthly family income below 10.0 minimum wages (p<0.001), smokers who quit smoking (p=0.026), risky alcohol consumption (p<0.001), sedentary lifestyle (p<0.001), obesity (p<0.001), poor sleep quality (p<0.001) and insomnia (p<0.001), non-nursing professionals (p<0.001), working night shifts (p<0.001), higher workloads (p<0.001), having a history of work-related accidents or illnesses (p<0.001), having the intention to leave the nursing profession (p<0.001), working conditions with high strain (p<0.001), imbalance between efforts and rewards (p<0.001), and greater exposure to activities that favor the occurrence of pain or injury (p<0.001) (Table 1).
The EFA was initially run with three factors, similar to a previous study conducted in Brazil12. The results of the current study showed three factors. The first factor consisted of questions 8, 9, and 10, which represent the “mental resources” dimension. The second factor consisted of questions 2, 4, 5, 6, and 7, which include the worker’s perception of his/her ability to meet the physical demands of the job, future ability and aspects related to the presence of diseases, and restrictions resulting from the health status. The third factor consisted of questions 1, 2, and 3, which represent the worker’s perception of his/her WA. Question 2 included two factors (Table 2). In this factorial solution, the RMSEA (0.018) showed model inadequacy, the CR index was below expectations for factors 2 (0.779) and 3 (0.609), and the assessment of the instrument\'s unidimensionality did not confirm the 3-factor structure (UniCO=0.961; MIREAL=0.289) (Table 3).
A new EFA model with 1 factor was then run, in which the 10 questions were retained in the unidimensional structure with factor loadings above 0.400, and no pattern of cross-loadings was found (Table 2). Bartlett\'s sphericity test (p=0.019) indicated that the correlation matrix was favorable, and the KMO (0.879) suggested a very good interpretability of the item correlation matrix. The assessment of the adequacy of the one-dimensional model showed a factorial structure with adequate adjustment indices: RMSEA=0.125 and TLI=0.900, but with CFI=0.922 slightly below expectations. The reliability of the one-dimensional structure showed favorable results: CR index=0.872, ORION=0.882, H-observed=0.875, and FDI=0.939 (Table 3). The parallel analysis confirmed the one-dimensional structure of the WAI, with the single factor explaining 54.4% of the variance of the observed real data (Figure 1).
Considering that the expected value should be from 0.70034, the analysis of the internal consistency of the one-dimensional structure presented satisfactory results, with Cronbach\'s alpha coefficient of 0.787 and McDonald\'s omega of 0.819; the exclusion of any of the questions did not change the result. The structure with 3 factors presented unsatisfactory results in factors 2 and 3 (Table 4).

Discussion
This study aimed to reassess the measurement properties of the Brazilian Portuguese version of the WAI. To this end, we chose to assess construct validity and reliability. The results of the study indicated that, more than four decades after its creation and in the face of constant changes in terms of work organization, the WAI continues to present satisfactory measurement properties, with good performance in the ability to discriminate known groups, in the validity of its one-dimensional structure, and in its reliability, showing that it is a valid and reliable instrument for measuring WA in Brazil.
Despite the differences in the study populations and statistical strategies adopted, the current results reaffirm the findings of the first Brazilian study, from 2009, which at that time, had already shown satisfactory performance in WAI measurement capability12. The 2009 study counted on a sample of workers from the electric power sector in the state of São Paulo, who were predominantly male (90.9%); young (median of 37.0 years); physically healthy (36.6% reported not having any medically diagnosed disease); with a high level of education, income, and benefits in relation to the general population of the state; with formal employment in a sound company; and a median WAI of 43.0 points. In the current study, the population was predominantly female (85.6%), slightly older (median of 39.0 years), with a relevant occurrence of health problems (6.1% reported not having any medically diagnosed disease), especially musculoskeletal, resulting from injuries/accidents, and related to mental illness. They also had a high educational level, but a lower income level, different types of employment relationships, and a median WAI of 37.0 points. Although the nature of the activities is different, both study populations had predominantly mixed workloads (physical and mental). It is important to note that the current study population has the perception of intense and constant exposure to work stressors. In addition, the predominance of female workers also includes a gender issue. Women often face different (worse) treatment in the workplace, as compared to men, due to gender discrimination and stereotypes. In addition, occupational exposures to different risks, such as physical and psychological risks, differ for men and women, influenced by differences in their social roles, expectations, responsibilities, biological characteristics, and employment patterns11, which can increase the occurrence of illness and WA impairment among women.
The assessment of construct validity through hypothesis testing showed adequate results in the discriminative analysis of known groups. Given that it is necessary to define prior hypotheses when assessing construct validity through discriminative analysis28, the variables were chosen because of their association and/or predictive value, or relation to WA, as demonstrated in the literature1-7,23. The results obtained in this study were compatible with the prior hypotheses, with the presence of the expected associations for all analyzed variables: sociodemographic (sex, age group, monthly family income); lifestyle (smoking, risky alcohol consumption, physical activity); occupational (professional category, night work, weekly work hours); health aspects (BMI, sleep quality, insomnia, history of work-related accidents or illnesses); working conditions (psychosocial environment and situations that may generate pain or musculoskeletal injury); and intention to leave the nursing profession. In the initial study conducted in 2009, the discriminative analysis considered only the variable of absenteeism, and the WAI showed a higher mean (42.3 points) among workers with low absenteeism than among the others (32.7 points)12. In the present study, the discriminative analysis was expanded, including a broader spectrum of variables.
The assessment of construct validity through EFA did not confirm the previous hypothesis of 3 factors, defined in the 2009 study12, and showed that the questions that make up the WAI define a unidimensional internal structure, with satisfactory model adjustment indices, while the PA confirmed the 01-factor structure.
This result differs from other studies that identified structures of 216,18 or 3 factors12,15,17,19,21. However, it is compatible with a study that identified a single factor in 2 of 10 countries in a sample of nursing professionals14 and with a study conducted with German workers35. These differences can be attributed to the different methods used and populations investigated. Some studies evaluated the structure of 10 questions12,15,17,21,36, while others focused on the seven dimensions of the WAI16,18. Some studies adopted Pearson\'s correlation matrix12,15-17, while others used polychoric or biserial correlations18,36. Likewise, some studies used orthogonal rotation methods12,15,17,36, while others adopted an oblique rotation16,19,20,36. There were also differences in the size and type of populations analyzed, which vary in terms of professional category, number, and location of the study: one study conducted with a specific population of 236 nursing professionals, aged 40 years and over, in Iran15; the 2009 study conducted with 475 workers in the electric power sector in Brazil12; another focusing on 103 workers from 103 different professions also in Brazil21; or even a study with 38,000 nurses from 10 European countries14. Thus, differences arise both from statistical stability in larger samples, as well as from the influences of different work, demographic, socioeconomic, cultural contexts, and from the factorial techniques used in the study.
Regarding the methodological aspects, the present study preferred the 10-question structure for EFA, as these questions are what make up the instrument and are then aggregated into seven dimensions and the final score. In the 2009 study, confirmatory factor analysis was used through principal component analysis, adopting the varimax process for matrix rotation12. In the present study, the polychoric correlation matrix and the RDWLS extraction method were used. The polychoric matrix is recommended for ordinal items, especially when frequency distributions are asymmetric and the assumption of normality is violated, which is also true when adopting RDWLS as a method for extracting factors31,33,37. PA was also performed, which is a robust procedure used to select the number of factors and which is considered accurate because it is based on samples, with the hypothetical matrix being factored repeatedly30,31,33. This was accompanied by the robust diagonal oblique promin rotation, since oblique rotations do not stipulate the mandatory total independence between factors, which is preferred in the areas of human sciences. Moreover, promin offers an interpretable and stable rotation solution, which is recommended when the extraction method is DWLS30,37. The evaluation indexes of the factor solution showed that the chosen technique presented adequate results.
The unidimensional structure identified by EFA is empirically aligned with the WA construct, which is understood as a condition resulting from the combination of human resources in relation to the physical, mental, and social demands of work, organizational culture, and work environment1. In this sense, the WA as a whole is an outcome resulting from a complex multifactorial process involving elements that are internal and external to the worker and that interact with each other1,12,13. Furthermore, the WAI was originally structured without subdimensions; rather, it integrated, in a single score, items that assess different relevant aspects of WA (perception of WA, illness and limitations due to illness, prognosis, and mental resources)1. Another aspect is that the 3 initially predicted factors would have 3 or 4 items each, which may provide minimal statistical stability.
The results of this study showed that the WAI continues to present satisfactory construct validity in its version for use in Brazil, both in discriminative analysis and in the evaluation of its factorial structure. This type of construct validity is considered the main form of validation of an instrument, verifying the adequate representation of the construct28,38.
The WAI structure with 3 factors presented unsatisfactory internal consistency in 2 factors, while the unidimensional structure showed adequate results. The poor performance of the 3-factor structure can be explained by the limited number of questions in each factor. A scale should include a sufficient number of questions to measure the underlying factor; and the alpha coefficient is influenced by the absence of normal distribution in the correlations and by the number of questions, and may increase according to the increase in items, with omega coefficient being less influenced by these aspects34.
Previous studies conducted in different work groups have demonstrated favorable results in the evaluation of the reliability of the WAI when using the 10 questions, with Cronbach\'s alpha and/or McDonald\'s omega above 0.70. Examples include European nurses14, Brazilian nursing professionals36, the 2009 study of Brazilian electricians12, Iranian petrochemical workers17, Swedish workers on disability leave39, or Egyptian civilian workers40, among others.
Thus, both from a conceptual point of view and in terms of the composition of the scales and subscales, the unidimensional version is more favorable, a result that is also corroborated by the UniCO and MIREAL indexes of the EFA.
Despite criticism regarding the theoretical basis of the WA and the dimensionality of the WAI41, the concept of WA was based on the stress-strain theory12,36 and on the theoretical model of the Work Ability House, which presents evidence of validity42,43. This is particularly relevant when considering the recognized predictive value of the WAI for worker health outcomes2-8.
The implications of confirming the unidimensional structure of the WAI are its application as a “proxy” indicator of worker health, as a quick, simple, and low-cost instrument12. This is because the results of the WAI can be used at an individual level to identify workers with impaired WA, as well as direct actions to restore or improve this condition. The results at the collective level allow us to identify situations of risk and establish priorities to guide preventive measures1,12,13.
The WAI, as a proxy for worker health, can be adopted as a key performance indicator, integrating worker health into the business vision and organizational strategy. In addition, collective results can aid in decision-making in public policies for specific labor groups, as well as in the review of social security and health systems and in promoting employability1,2,19.
As a limitation, this study was developed in a specific population (nursing professionals), which has its own occupational insertion and work characteristics; therefore, the generalization of these results should be done with caution. Another limitation is that, in the present study, it was not possible to perform the analysis of the criterion validity, due to the lack of a representative variable of a comparison criterion for WA, unlike the 2009 study. In this study, a generic health assessment, using the SF-36 questionnaire, showed a positive and significant correlation of WAI with all of the evaluated health dimensions12.
It is important to note that, even after more than 40 years since the creation of the WAI and its widespread use, there are still no national Brazilian studies to validate the score\'s cutoff points. The available cutoff points were defined based on results from Finnish workers in the aging phase, close to 50 years of age1. There is also a proposal for younger Finnish workers, near 35 years of age, to avoid overestimating WAI in this age group23. Since Brazilian workers differ from Finnish workers in terms of living and working conditions, it is recommended that preference be given to using the overall WAI score or its categorization adjusted by the proposal of Kujala et al. (2005)24 for younger workers12.
It is suggested that new studies be conducted with other work groups, seeking to strengthen the external validity of the WAI, as well as studies with representative samples of the Brazilian working population, in order to verify the cutoff points of the WAI score in different age groups.

Conclusions
The ICT continues to present satisfactory measurement properties in the current Brazilian work context, with good performance in the ability to discriminate known groups, in the validity of its one-dimensional structure, and in its reliability, thus corroborating that it is an adequate option for assessing WA both in both individual approaches and population surveys.

References
1. Tuomi K, Ilmarinen J, Jahkola A, Katajarinne L, Tulkki A. Índice de capacidade para o trabalho. São Carlos: EDUFSCAR; 2005.
2. Hasselhorn H-M, Müller BH, Tackenberg P, editors. Nurses Early Exit Study – NEXT. NEXT Scientific Report—July 2005. EU-Project No QLK6-CT-2001-00475. Wuppertal: University of Wuppertal [report on the Internet] 2005 [cited 2024 Apr 26]. Available from: https://www.researchgate.net/publication/260592268_Next_Scientific_Report_July_2005
3. Reeuwijk KG, Robroek SJW, Niessen MAJ, Kraaijenhagen RA, Vergouwe Y, Burdorf A. The Prognostic Value of the Work Ability Index for Sickness Absence among Office Workers. PLoS ONE. 2015;10(5):e0126969.
4. Wagenaar AF, Kompier MAJ, Houtman ILD, van den Bossche SNJ, Taris TW. Who gets fired, who gets re-hired: The role of workers’ contract, age, health, work ability, performance, work satisfaction and employee investments. Int Arch Occup Environ Health 2015;88(3):321–334.
5. Lundin A, Leijon O, Vaez M, Hallgren M, Torgén M. Predictive validity of the Work Ability Index and its individual items in the general population. Scand J Work Environ Health 2017;45(4):350-356.
6. Boissonneault M, de Beer J. Work ability trajectories and retirement pathways: a longitudinal analysis of older American workers. J Occup Environ Med 2018;60(7):e343–e348.
7. Martinez MC, Fischer FM. Work ability and job survival: four-year follow-up. Int J Environ Res Public Health 2019;16(17):3143.
8. Elovainio M, Laaksonen M, Sakari K, Aalto A-M, Jääskeläinen T, Rissanen H, Koskinen S. Association of short poor work ability measure with increased mortality risk: a prospective multicohort study. BMJ Open 2022;12:e065672.
9. Felix J. O idoso e o mercado de trabalho. In: Alcântara AO, Camarano AM, Giacomon KC, editores. Política nacional do idoso: velhas e novas questões. Rio de Janeiro: Instituto de Pesquisa Econômica Aplicada – Ipea [livro na Internet] 2016. p. 241-263 [cited 2024 Apr 26]. Disponível em: https://repositorio.ipea.gov.br/handle/11058/7253
10. Santos GT, Salcedo GM, Guimarães LHV, Baêta PVL, Braga JO. A integralidade no mercado de trabalho: desafios e potencialidades. Revista Ibero-Americana De Humanidades, Ciências E Educação 2022;8(12):505-528.
11. Schulte PA, Iavicoli I, Fontana L, Leka S, Dollard MF, Salmen-Navarro A, Salles FJ, Olympio KPK, Lucchini R, Fingerht M, Violante FS, Seneviratne M, Oakman J, Lo O,Alfredo CH, Bandini M, Silva-Junior JS, Martinez MC, Cotrim T, Omokhodion F, Ficher FM. Occupational safety and health staging framework for decent work. Int J Environ Res Public Health 2022;19(17):10842.
12. Martinez MC, Latorre MRDO, Fischer FM. Validade e confiabilidade da versão brasileira do Índice de Capacidade para o Trabalho. Rev Saúde Pública 2009;43(3):525-532.
13. Ilmarinen J. From work ability research to implementation. Int J Environ Res Public Health 2019;16(16):2882.
14. Radkiewicz P, Widerszal-Bazyl M. Psychometric properties of Work Ability Index in the light of comparative survey study. Int Congr Ser 2005;1280:304–309.
15. Abdolalizadeh M, Arastoo AA, Ghsemzadeh R, Montazeri A, Ahmadi K, Azizi A. The psychometric properties of an Iranian translation of the Work Ability Index (WAI) questionnaire. J Occup Rehabil 2012;22(3):401–408.
16. Alexopoulos E, Merekoulias G, Gnardellis C, Jelastopulu E. Work Ability Index: validation of the Greek version and descriptive data in heavy industry employees. Br J Med Med Res 2013;3(3):608–621.
17. Adel M, Akbar R, Ehsan G. Validity and reliability of Work Ability Index (WAI) questionnaire among Iranian workers; a study in petrochemical and car manufacturing industries. J Occup Health 2019;61(2):165-174.
18. Freyer M, Formazin M, Rose U. Factorial Validity of the Work Ability Index among employees in Germany. J Occup Rehabil 2019;29(2):433-442.
19. Bascour‑Sandoval C, Soto‑Rodriguez F, Munoz‑Poblete C, Nasri Marzuca‑Nassr GN. Psychometric properties of the Spanish version of the Work Ability Index in working Individuals. J Occup Rehabil 2020;30(2):288-297.
20. Rodríguez IM, Knox ECL, Hernández CO, Codina AD, The esTAR Group. Psychometric properties of the Work Ability Index in health centre workers in Spain. Int J Environ Res Public Health 2021;18(24):12988.
21. Pucci RLA, Silva AM, Padula RS. Factorial analysis of the Brazilian-Portuguese version of the Work Ability Index, reproducibility and validity of the single item and the short version for online application. Braz J Phys Ther 2024;28(2):101060.
22. American Educational Research Association; American Psychological Association; National Council on Measurement in Education. Standards for educational and psychological testing. Washington, DC: American Educational Research Association; 2014. 230 p.
23. Martinez MC, Latorre MRDO, Fischer FMF. Capacidade para o trabalho e intenção de saída da profissão na enfermagem de São Paulo. Rev Enferm UERJ 2021;29:e57941.
24. Kujala V, Remes J, Ek E, Tammelin T, Laitinen J. Classification on Work Ability Index among young employees. Occup Med 2005;55:399-401.
25. Alves MGM, Chor D, Faerstein E, Lopes CS, Wenerck GL. Short version of the “job stress scale”: a Portuguese-language adaptation. Rev Saúde Pública 2004;38(2):164-171.
26. Chor D, Werneck GL, Faerstein E, Alves MGM, Rotenberg L. The Brazilian version of the effort reward imbalance questionnaire to assess job stress. Cad Saúde Pública 2008;24(1):219-224.
27. Coluci MZO, Alexandre NMC. Adaptação cultural de instrumento que avalia atividades do trabalho e sua relação com sintomas osteomusculares. Acta Paul Enferm 2009;22(2):149-154.
28. Polit DF. Assessing measurement in health: Beyond reliability and validity. Int J Nurs Stud 2015; 52(11):1746-1753.
29. Mokkink LB, Prinsen CAC, Patrick DL, Alonso J, Bouter LM, de Vet HCW, Terwee CB. COSMIN Study Design checklist for patient-reported outcome measurement instruments – version July 2019. Netherlands: COnsensus-based Standards for the selection of health Measurement INstruments – COSMIN [book on the Internet] 2019 [cited 2024 Apr 26]. Available from: https://www.cosmin.nl/wp-content/uploads/COSMIN-study-designing-checklist_final.pdf
30. Damásio BF. Uso da análise fatorial exploratória em psicologia. Aval Psicol 2012;1192):213-228.
31. Lorenzo-Seva U, Ferrando PJ. Terragona: Manual of the program FACTOR v. 9.20. Tarragona: Universitat Rovira i Virgili [manual on the Internet] 2013 [cited 2024 Apr 26]. Available from: https://psico.fcep.urv.cat/utilitats/factor/documentation/Manual-of-the-Factor-Program-v92.pdf
32. Colwell SR. The composite reliability calculator [calculator on the Internet] 2016. [cited 2023 Jun 08]. Available from: http://www.thestatisticalmind.com/composite-reliability/
33. Ellis JE. Conducting and reporting factor analysis. In: Factor analysis and item analysis. Nijmegen: Applying statistics in behavioural research [book on the Internet]; 2017 [cited 2024 Apr 26]. Available from: https://www.applyingstatisticsinbehaviouralresearch.com/documenten/factor_analysis_and_item_analysis_version_11_.pdf
34. Stensen K, Lydersen S. Internal consistency: from alpha to omega? Tidsskr Nor Laegeforen 2022;142(12).
35. Bethge M, Radoschewski FM, Gutenbrunner C. The Work Ability Index as a screening tool to identify the need for rehabilitation: longitudinal findings from the Second German Sociomedical Panel of Employees. J Rehabil Med 2012;44(11):980–987.
36. Silva Jr SHE, Vasconcelos AGG, Griep RH, Rotenberg L. Validade e confiabilidade do índice de capacidade para o trabalho (ICT) em trabalhadores de enfermagem. Cad Saúde Pública 2011;27(6):1077-1087.
37. Lorenzo-Seva U, Ferrando PJ. Robust Promin: a method for diagonally weighted factor rotation. Liberabit 2019;25(1):99-106.
38. Kerlinger FN. Foundations of behavioral research. New York: CBS College Publishing; 1986. 667 p.
39. Rashid M, Heiden M, Nilsson A, Kristofferzon ML. Do work ability and life satisfaction matter for return to work? Predictive ability of the work ability index and life satisfaction questionnaire among women with long-term musculoskeletal pain. BMC public health 2021;21(1):584.
40. Abdelrehim MG, Eshak ES, Kamal NN. The mediating role of work-family conflicts in the association between work ability and depression among Egyptian civil workers. J Public Health (Oxf) 2022;45(2):e175-e183.
41. McCarthy GS, Truxillo DM, O’Shea DE, Brady GM, Cadiz DM. The development and validation of a multidimensional perceived Work Ability Scale. J Occup Health Psychol 2024;29(2):90-112.
42. Martinez MC, Latorre MRDO, Fischer FM. Testando o Modelo da Casa da Capacidade para o Trabalho entre profissionais do setor hospitalar. Rev Bras Epidemiol 2016;19(02):403-418.
43. Ilmarinen J, Gould R, Jäevikoski A, Järvisalo J. Diversity of work ability. In: Gould R, Ilmarinen J, Järvisalo J, Koskinen S, editors. Dimensions of work ability: Results of the Health 2000. Vaasa: Waasa Graphics Oy [report on the Internet] 2008 p. 13-24 [cited 2024 Apr 26]. Available from: https://www.julkari.fi/handle/10024/78055