The Brazilian Portuguese version of the Work Productivity and Activity Impairment – General Health (WPAI-GH) Questionnaire

ABSTRACT CONTEXT AND OBJECTIVE: It is still difficult to measure work productivity losses caused by health problems. Despite the importance given to this issue over the last few years, most instruments for performing this task are available only in the English language. This study translated the Work Productivity and Activity Impairment - General Health (WPAI-GH) Questionnaire into Brazilian Portuguese, adapted it cross-culturally and evaluated its reliability and validity. DESIGN AND SETTING: Cross-sectional survey to test scale reliability and validity, at São Paulo Hospital and the clinic of the Rheumatology division of Universidade Federal de São Paulo — Escola Paulista de Medicina (Unifesp-EPM). METHODS: Data were obtained from a survey that incorporated the WPAI-GH, short form-36 (SF-36) and some demographic questions. The questionnaires were administered by interview to 100 subjects. RESULTS: Descriptive statistics was used to characterize the subjects. The intraclass correlation coefficient and Cronbach's alpha were used to assess the reliability and internal consistency of the instrument. Intraclass correlation coefficients from 0.79 to 0.90 indicated good reliability. Cronbach's alpha of 0.74 indicated good internal consistency. Pearson's correlation coefficient was used to assess validity. There were significant positive relationships between the WPAI-GH and SF-36. CONCLUSION: The Brazilian Portuguese version of the WPAI-GH is a reliable and valid measurement tool and may be useful for those who seek to measure the impact on productivity of health problems among populations of Brazilian employees.


INTRODUCTION
2][3] Problems with employees' health generally imply decreases in their activities and reduced productivity at work.
5][6] Most of the time, these instruments relate to individuals who are working on a formal basis or receiving regular pay and they do not include vo luntary work or housework, although most authors highlight the importance of such measurements. 7oreover, from researchers working on this issue, there is a generalized need for validated measurement tools for work productivity and techniques to translate these measurements into costs. 8he WPAI-GH (Work Productivity and Activity Impairment -General Health) questionnaire measures the effects of health in general and specifi c symptoms on work productivity and outside of work. 9It is the instrument that has been most assessed and most frequently used. 5t has already had its validity and relia bility evaluated in relation to several diseases, such as allergies, 10 dermatitis, 11 gastroesophageal refl ux, 12 nocturia, 13 asthma 14 and irritable bowel syndrome. 15This questionnaire can detect absenteeism (days or hours of work missed due to a health problem) and presenteeism (reduction of effectiveness of a person while working, due to a health problem). 16The productivity loss would thus be the total work impairment; the sum of absenteeism and presenteeism.

OBJECTIVE
The aim of our study was to translate, culturally adapt and evaluate the performance of the psychometric properties of WPAI-GH among a Brazilian population.

WPAI-GH productivity WPAI-GH productivity assessment instrument assessment instrument
The WPAI-GH is composed of six questions that ask the subject if he is currently employed, the number of nonworking hours due to health problems, the number of nonworking hours due to other reasons (e.g.vacation), the number of hours really worked, how much the health problems affected his productivity while he was working and how much the health problems affected his daily activities over the last seven days.The two last questions are evaluated on a scale of 10 points, ranging from 0 (no effect on work) to 10 (health problems prevent the person from working). 17he questions are computed according to specifi c calculation rules and have four scores: (1) percentage work time missed due to health (absenteeism); (2) percentage impairment at work due to health (presenteeism); (3) percentage overall work productivity loss due to health (absenteeism and presenteeism); and (4) percentage daily activity impairment outside of work due to health.High scores indicate prolonged sick leave or impairment and decreased productivity. 9The score from this questionnaire can be transformed into monetary value.The general percentage work limitation score is multiplied by the employee's hourly wage to determine the value of the productivity loss.

Translation and Translation and cross-cultural adaptation cross-cultural adaptation
The translation and cross-cultural adaptation of the instrument were undertaken on the basis of the directions in da Mota Falcão et al., 18 consisting of translation, back-translation and cross-cultural adaptation.The Brazilian Portuguese version of the WPAI-GH questionnaire is presented in the Annex below.

Initial translation Initial translation
The items of the WPAI-GH version were initially translated by an independent Brazilian health professional with great knowledge of the English language who was aware of the research objectives.The importance of the conceptual translation was emphasized to the detriment of a strictly literal translation.We thus obtained the fi rst Portuguese version.

Back-translation and Back-translation and evaluation of the initial evaluation of the initial translation translation
The initial translation (first version in Portuguese) was back-translated into English by an English teacher who had not participated in the previous stage.The original instrument was then compared with the new English version.A multi-disciplinary group composed of four health professionals (one rheumatologist, two dentists and a nurse) documented and analyzed the discrepancies that were found.Some verb tenses and sentences in Portuguese were rewritten until a consensus was reached.Thus, the second Portuguese version was obtained.

Evaluation of the Evaluation of the cross-cultural adaptation cross-cultural adaptation
The questionnaire was applied to a group of 10 patients randomly selected at the rheumatology outpatient clinic of Universidade Federal de São Paulo -Escola Paulista de Medicina (Unifesp-EPM).To evaluate the degree of comprehension of the questions, we selected the ones that did not appear to be easy to comprehend.The same group as before evaluated these items with the aim of replacing them with others expressing the same notion but easier to comprehend, even using suggestions from the patients themselves, but trying not to change the structure and evaluative properties of these questions.This new version was then applied to a group of 10 patients randomly selected at the same place.Its cultural equivalence was tested again until none of the items presented comprehension problems.Thus, the third Portuguese version was obtained.

Evaluation of the Evaluation of the psychometric properties psychometric properties
After fi nishing the process of translation and cross-cultural adaptation of the instrument, we assessed its reliability and validity. 19ne hundred subjects who were currently employed and complaining of a health problem that in their opinion affected their performance at work and/or outside of work were included as participants in the survey.
The study was approved by the Research Ethics Committee of Universidade Federal de São Paulo and the information was collected after obtaining written consent from the subjects.

Reliability: Reliability: test-retest evaluation test-retest evaluation
The reliability of the Portuguese version of the WPAI-GH questionnaire was assessed by means of two interviews.
A group of 100 subjects formed by employees of São Paulo Hospital and patients at the outpatient clinic of the Rheumatology division of Unifesp-EPM was selected and assessed by two interviewers.Two assessments were made by one or other of these two observers on the same day.The second assessment by the same observer was made at most one hour after the fi rst assessment.

Validity Validity
The same group of 100 subjects was interviewed in order to assess the construct validity.The WPAI-GH questionnaire was assessed by correlating it with other sociodemographic and clinical parameters related to their activities, such as length of work, absences from work and degree of damage, among others.These parameters were collected during the fi rst interview with the patient.
In addition to the abovementioned parameters, the instrument was compared with the Short Form-36 (SF-36) generic quality-of-life evaluation questionnaire (Medical Outcomes Study 36-Item Short-Form Health Survey), which evaluates the domains of: general health perceptions, physical functioning, role limitations due to physical health, role limitations due to emotional health, bodily pain, vitality, social functioning and general mental health.The score for each domain ranges from 0 to 100, such that 0 is the worst status and 100 is the best.The SF-36 instrument had already been translated into Brazilian Portuguese and its psychometric properties had already been tested. 20

Questionnaire administration Questionnaire administration
The questionnaires were individually administered to employees of São Paulo Hospital and to patients at the outpatient clinic of the Rheumatology division of Unifesp-EPM.
The employees were approached at their workplace and patients while being attended, by two postgraduate students who explained the objectives and content of the questionnaire and asked for their help and consent to participate in the survey.First of all, the researchers asked if the subject was presently employed and whether he/she had felt any symptom or health problem during the last seven days.If the person gave responses compatible with the inclusion criteria of the study and had time to answer the questionnaires immediately, the researchers then asked the subject to complete a form with sociodemographic data.Following this, they applied the WPAI-GH for the fi rst time in the form of an interview.Next, the SF-36 was administered in the form of an interview as well.The researchers then asked the interviewees to wait for about 25-30 minutes for the last questionnaire to be administered.A randomization table was then used to determine whether the second administration of the WPAI-GH questionnaire would be self-administered or interviewer-administered.When it was self-administered, the researchers waited while the participants answered the questions and the questionnaires were immediately collected after completion.No form of incentive or bonus payment was offered to the participants in the study.

Statistical analysis Statistical analysis
The data were analyzed by means of the Statistical Package for the Social Science (SPSS) program, version 11.0.Descriptive statistics was compiled for sociodemographically and clinically characterizing the study population.
The intraclass correlation coefficient and Cronbach's alpha were used to assess the reliability and internal consistency of the instrument.Pearson's correlation coeffi cient was used to assess its validity.
The analyses on the subjects of the sample were made fi rstly as a single group and then as two separate groups according to the method used for administering the second questionnaire (interviewer-administered or self-administered).

Cross-cultural adaptation Cross-cultural adaptation
No question was considered non-applicable.The patients showed good comprehension of the questions with some suggestions and it was not necessary to make any substitution or alteration of any question already assessed in the second Portuguese version.

Characteristics of the Characteristics of the population studied population studied
Table 1 shows the sociodemographic characteristics of the 100 patients included in the reliability and validity assessment of the Portuguese version of the WPAI-GH questionnaire.Eighty-two per cent of the subjects were women, with mean age of 39.22 years (standard deviation, SD: 11.17).With regard to their schooling level, 27% had completed intermediate school; 34%, high school and 39%, college degree.
Table 1 shows some clinical characteristics of the sample and their impact on the productivity of these patients at work.Even though only 16% of the subjects had missed work during the last week, and had had on average 1.44 days of absence, 77% of them said that their health problems interfered with their work activities.It is important to say that, in this population, the intensity of health interference at work was mild; of the 100 patients assessed, 61% classifi ed this interference as not much.Most patients showed musculoskeletal diseases (49%) and the medication in use had been prescribed in 40% of the cases.
Table 2 shows the values obtained for each domain of the WPAI-GH version, in the Portuguese language.Most mean values for questionnaire domains were between 40 and 50, except for the fi rst domain, which measures absenteeism.The highest mean values were the ones obtained for the domain of activity impairment due to health (49.70).This represents almost 50% impairment of the individual's capacity to perform regular activities outside of work.
Table 3 shows the values obtained for each domain of SF-36.Most mean values for the domains of this questionnaire were between 50 and 70, except for the physical functio ning and bodily pain domains, which showed lower scores.

Reliability Reliability
The overall internal consistency, as assessed by Cronbach's alpha, was 0.74.Table 4 shows the reliability results obtained by means of test-retest evaluation for each domain of the Portuguese version of the WPAI-GH questionnaire, using the intraclass correlation coeffi cient and Cronbach's alpha.
The reliability of the four domains of the WPAI-GH was considered highly satisfactory.All domains showed an excellent correlation between the fi rst and second administrations of the questionnaire, independent of the administration method used on the second occasion, except for the domain of work time missed due to health, which indicated a good intraclass correlation.
Table 5 shows the results from the reliability assessment when the second method of applying the questionnaire was modifi ed.These reliability results were also obtained by means of test-retest evaluation for each domain of the Portuguese version of the WPAI-GH questionnaire, using the intraclass correlation coeffi cient and Cronbach's alpha.
For the administration sequence of interview followed by self-administration, the reliability was shown to be very satisfactory.The domains of impairment at work due to health and overall work productivity loss due to health showed excellent correlation.The domains of work time missed due to health and of activity impairment due to health indicated good correlation.
For the administration sequence of interview followed by interview, the reliability was shown to be highly satisfactory.All domains showed an excellent correlation.As expected, the administration method of interview followed by interview produced better reliability, because the administration method was not modifi ed.

Validity Validity
Table 6 shows Pearson's correlation coeffi cients for the sociodemographic data and the four domains of the Portuguese version of the WPAI-GH questionnaire.
Statistically signifi cant correlation coeffi cients were found between the following variables: sex, schooling, years of study and weekly schedule, and the following domains: impairment at work, overall work productivity loss and activity impairment.There was also a statistically signifi cant correlation between the domain of work time missed due to health and the variable of monthly salary.
Table 6 also shows Pearson's correlation coeffi cients for the clinical data and the four domains of the Portuguese version of the WPAI-GH questionnaire.
Statistically signifi cant correlation coeffi cients were found between the following variables: absence from work, days of absence, health interference at work and intensity of the interference, and all four domains of WPAI-GH.There was also a statistically signifi cant correlation between the domain of work time missed due to health and the variable of absence from work, and between the domains of impairment at work and overall work productivity loss and the variable of intensity of the interference.
Table 7 shows the construct validity analysis, the correlations between the four domains of WPAI-GH and the eight domains of SF-36.
Statistically signifi cant correlation coeffi cients were found between the following  domains: impairment at work, overall work productivity loss and activity impairment, and all the domains of SF-36.The most notable correlation was between the domain of impairment at work and the domain of pain in SF-36.
The WPAI-GH scores range from lower to higher values according to the impairment or worsening of productivity.On the contrary, the SF-36 scores range from lower to higher value according to the improvement in the patient's general health status.This is the reason why the comparisons between the questionnaires show negative correlation coeffi cients.

DISCUSSION
It is still diffi cult to measure work productivity losses caused by health problems.Despite the importance given to this issue over the last few years, and despite the existence of several generic and specifi c instruments for performing this task, most instruments are available only in the English language.Translation, cross-cultural adaptation and validation of these instruments is a hard task because of its pioneering nature and because there are no similar instruments in Brazil with which to make the necessary comparisons for this process.
Since the WPAI-GH questionnaire has simple and objective questions, the interviewees did not have any diffi culty in understanding the questions.Another important factor is that in some questions, such as numbers 5 and 6, the instrument itself gives examples of some situations, which facilitated comprehension among the subjects who fi lled in the questionnaire.
In the reliability assessment, we observed highly satisfactory agreement scores with regard to the administration methods and internal consistency of the instrument, for the whole sample and also when it was subdivided according to the administration method on the second occasion.This shows that both administration methods (interview or self-administration) are reliable.In another study involving reliability assessments on instruments among a population with low schooling levels, it was observed that it was only better to apply the instrument in the self-administered form when the subjects had completed intermediate school, because the level of confi dence was higher in this situation. 21n the present study, we cannot prove whether the subjects with lower schooling levels would have been able to answer the questionnaire in the self-administered form, since we had a higher number of subjects with schooling levels up to eighth grade selected for interview.Although they showed that there was an excellent correlation between the administration methods, we cannot conclude that they would have had the same performance if the administration had been via the self-administered method.
In order to evaluate the construct validity of the WPAI-GH questionnaire we needed to compare it with a generic instrument for quality of life.We believed that the SF-36 was the best option for this.
We observed that the domains most related to the physical component, such as physical functioning, role limitations due to physical health, bodily pain and general health status, showed the best correlations with the WPAI-GH domains; except for the domain of work time missed due to health, which did not correlate with any domain.We attribute this fi nding to the fact that the number of nonworking days was very small in the studied sample (1.44 on average), thus giving a lower score to the domain of work time missed due to health, which showed a mean score of 8.40 on a scale from 0 to 100; very different from the other three WPAI-GH domains, which showed scores of more than 40.Many patients said that, even though they were having some health problems, they did not stop going to work and tried to accomplish their tasks in the best possible way.They thought that the highest damage occurred during tasks outside of the workplace.This shows the need for instruments that measure presenteeism, because simply measuring the number of days absent from work would not be enough to check the impact of a health problem on the subject's productivity.Absenteeism represents only a small proportion of the total indirect costs. 3valuation of productivity losses makes it possible for an assessment of disease costs to incorporate a better approach towards the indirect costs, which in some situations exceed the direct cost.It would be of great value to carry out local studies that use specifi c instruments to evaluate the impact of some diseases on productivity losses among workers, because other instruments measuring productivity losses would thus be made available for comparison purposes.

CONCLUSIONS
The Brazilian Portuguese version of the WPAI-GH questionnaire is a reliable and valid measurement tool and may be useful for those who seek to measure the impact on productivity of health problems among populations of Brazilian employees.

Annex. Brazilian Portuguese translation of the Work Productivity and Activity Impairment -General Health (WPAI-GH) questionnaire
WPAI (Produtividade e Capacidade Diminuída no trabalho)

Table 2 .
Work Productivity and Activity Impairment -General Health (WPAI-GH) questionnaire scaling test results

Table 4 .
Intraclass and Cronbach's alpha correlation coeffi cients between fi rst and second administrations of the Work Productivity and Activity Impairment -General Health (WPAI-GH) questionnaire