Application of mixed methods for exploration of the association of job stress and hypertension among software professionals in Bengaluru, India.

CONTEXT: Quantitative and Qualitative studies have been widely used in isolation to estimate several associations in developing countries, but little is known about combining both methods in a given study and ascertains validity. AIMS: The objective of the following study was to accurately measure the constructs and to check for the internal consistency whereas measuring occupational stress among software professionals. We compared contextual stressors developed from the qualitative study with occupational stress index (OSI) among Information Technology/Information Technology Enabled Service (IT/ITES) professionals in India. SETTINGS AND
DESIGN: We employed mixed methods sampling strategy for selecting the IT/ITES professionals for the study. The first stage involved a qualitative study followed by a cross-sectional study among 1071 workers in the IT and ITES sector in Bengaluru.
MATERIALS AND METHODS: There were two types of stress domains used in the questionnaire. First, contextual stress domains, which were constructed based on the results of the qualitative study. Second, we used OSI for computer workers. STATISTICAL ANALYSIS USED: Data from the cross sectional survey was analyzed using SAS 9.1.(SAS Institute Inc., Cary, NC, USA). We used Cronbach's coefficient alpha for analyzing latent constructs of OSI and contextual stress domains.
RESULTS: The results indicate that OSI doesn't correlate well with contextual stressors.
CONCLUSIONS: OSI is a generic questionnaire designed for a computer worker and using the contextual stressors based on the results from the qualitative study might capture the occupational stressors more reliably.

INTRODUCTION

Cardiovascular diseases are considered to be among the major causes of morbidity and mortality in the Indian subcontinent, causing more than 25% of deaths.[1] It has been predicted that occurrence of these diseases will increase rapidly in India and this country will be the locale of more than half the cases of cardiovascular disease in the world within the next 15 years. It is established that any modest reductions in the risk factors such as smoking, obesity and hypertension are associated with dramatic reductions in development of non-communicable diseases (NCD). However, we are yet to establish the socio-economic and environmental factors at the macro level, which promote these risk factors.[234]

The increasing burden of morbidity in the developing countries is attributable to the increasing incidence of NCD, perhaps related to urbanization. Further, some NCD manifest at a relatively early age affecting a large proportion of the population especially young adults and the middle-aged in these countries. Workers in the Information Technology (IT) industry are prone to almost all the recognized risk factors for NCDs. The health profile of IT and Information Technology Enabled Service (ITES) professionals is significantly changing as a result of urbanization and other factors. IT/ITES professionals start working at age 25-30. This age group has increased prevalence of coronary risk factors as established by studies done in India and elsewhere.[5678] IT/ITES professionals may have to sit for long hours (sometimes including day and night completing the project), cannot get regular sleep as they have to work in odd hours, do not have sufficient time and resources to do exercises. Most of the professionals do time bound work and cannot be involved actively in family activities and thus may not get adequate social support and respect.[9101112]

It can thus be expected that there will be a higher prevalence of job stress among IT/ITES professionals indicating a need to study such workers and suggest interventions. The review of the literature suggests that there is an association between job related factors such as stress and strain at work with poor general health, musculoskeletal diseases and cardiovascular dysfunction.[21314] There are several measures of job stress employed in quantitative and qualitative method based studies. The qualitative study methodologies studies are carried out in isolation and have not been used to prepare questionnaires for the quantitative studies in India. Furthermore, very little is known about combining both methods in a given study and ascertains validity while measuring job stress. The objective was to accurately measure the constructs of job stress and to check for the internal consistency while measuring among software professionals. This is important to explore as there is limited evidence on verifiable measures of job stress available from studies done in developing countries such as India. Hence, we compared contextual stressors developed from qualitative study with occupational stress index (OSI) among IT/ITES professionals in India. The current paper focuses on methods involved in implementation of the study and their validity.

MATERIALS AND METHODS

We employed mixed methods sampling strategy for selecting the IT/ITES professionals for the study. Specifically, the first phase of the study comprised of a qualitative study followed by a second phase of cross-sectional study among workers in the IT and ITES sector in Bengaluru. The “study population” was defined as all persons aged 50 years or under who began to work as IT/ITES professionals as of 1st January 2010, at 21 sites involved in the IT and ITES sector in Bengaluru.[1516] The objective of the qualitative study was to identify specific contextual risk factors (exposure to job stress related factors) for ill-health including hypertension and to understand their influence among IT/ITES professionals. This was followed by a cross-sectional study to determine the prevalence of the contextual occupational stressors (found from the qualitative study) and its influence on among IT professionals [Figure 1].

Figure 1

Description of the study methodology- Bengaluru Information Technology workforce study, 2011-12

Details on methods followed in the qualitative study are described elsewhere.[1718] In brief, the qualitative study was conducted between August 2010 and March 2012 in the city of Bengaluru. We conducted 32 in-depth interviews with IT/ITES workers, recruited with the assistance of supervisors and Human resources (HR) Managers in IT and ITES organizations. Participants were recruited from workers holding different job titles, team leaders and administrative staff of informal groups. Recruitment of volunteers was done through personal communication as well as with the help of HR managers. The qualitative study was anonymous.

For quantitative study, we chose stratified purposive sampling (quota sampling).[19] First step involved a stratified sampling procedure in which the IT/ITES were selected randomly from strata of groups of the complete list of IT companies (characteristic of probability sampling) and the next step involves selecting a small number of units within each strata (characteristic of purposive sampling).[192021] The advantage of this method of sampling is that it allows the researcher to identify and have a detailed description of the participant characteristics that are similar across the strata or subgroups.[1920] Our study was done such that we covered each of these main zones. We took at least one IT/ITES sector from each of the major zones in the first stage of sampling and this involved stratification of the locations of IT/ITES companies. The second stage involved selecting companies within the strata of zones having IT/ITES companies. The sampling was done such that three companies each were selected from Electronic City (all IT) and Whitefield (2 IT, 1 ITES), areas that had the maximum number of IT/ITES companies. The remaining 15 sites were selected such that there was at least one company representing each geographical area having IT/ITES companies in Bengaluru.[22] We included 1071 subjects in the total sample, 509 subjects in the IT sector and 472 subjects in the ITES sector.[23] The details on sample size calculations are provided elsewhere.[18]

After obtaining the informed consent, the questionnaire was self-administered, to collect following information:

Demographic and social characteristics such as age, marital status, profession, education and socio-economic status (SES)

Components from the OSI questionnaire to assess job stress

Components from the contextual stressors based on the results from the qualitative study.

The objective of the analysis was to derive valid measures of job stress as the principal outcome of interest. Job stress was calculated by combining different combinations of the job stressors used in the study. “Job stressors” were defined as “working conditions that may lead to acute reactions, or strains in the worker.” We considered three validated questionnaires for the measurement of job stress in the proposed study. We chose the OSI with permission from Belkic and Savic.[24] The questionnaire collected the information to cover the job stressors, buffer factors, non-work activities and acute physiologic responses as suggested by the theoretical model described in the development of OSI.[242526272829303132333435] OSI is an additive burden model that focuses on work stressors relevant to the cardiovascular system.[36] In the OSI, each stressor has a set of coordinates, localizing it to the type of stress and the level at which it affects the worker. The OSI contains 58 items, including more specific questions than standard job stressor questionnaires. Each of the seven subscales of the OSI have shown acceptable internal consistency reliability, with the exception of “strictness” and “extrinsic time pressure.”[1011122426282935]

We derived contextual stress domains based on the results from qualitative study [Table 1]. The details of information on calculation of contextual stress domains are discussed in our earlier paper and are summarized in Table 1.[17]

Table 1

Components of contextual stress domains

The data from the cross sectional survey was analyzed using SAS 9.1.3104.[37] We used Cronbach's coefficient alpha for analyzing latent constructs of OSI and contextual stress domains to accurately measure the constructs and to check for the internal consistency. Interrelated items can be summed to obtain an overall score for each participant. Cronbach's coefficient alpha estimates the reliability of this type of scale by determining the internal consistency of the test or the average correlation of items within the test.[38]

The effort to detect and exclude variables with multi-collinearity is reported here. First, we used dummy variables to exclude categories and variables that had the same meaning in analysis. Due caution was applied while excluding a variable that was computed from other variables in the equation (e.g., income was part of SES, hence it was not entered as a separate variable for the adjustment). We ensured that same or almost same variables were not entered twice (For e.g., height in cm and body mass index having height2 was not entered simultaneously). After ensuring the above, we ran regression diagnostics to detect internal consistency and collinearity.

The study was anonymous and no names or other personal identifying information were collected from the participants. Each participant was assigned a unique number that linked individual information of the participants for the purpose of the analyses. Quiet private rooms provided by IT/ITES companies were used for collection of information for both qualitative and quantitative phases. The study was reviewed and approved by the University of California, Los Angeles Institutional Review Board (IRB, #G09-12-002-01, IRB#10-001348) and the Ethics Committee of The Public Health Foundation of India (TRC-IEC 40/10).

RESULTS

We invited 1369 IT/ITES professionals to participate in the study; the refusal rate was 4.7% (n = 64). Among the 1305 professionals who accepted the questionnaire for answering, 171 (13%) didn’t return them. Among the people who returned the questionnaire (1134), we found 51 (4.6%) to be ineligible based on the inclusion criteria (duration < 1 year). There were 12 subjects with missing data on inclusion criteria who were included for analysis. By treating non-responders as refusals, the refusal rate is 22%. If we exclude non-responders, the refusal rate was 18%. Among the eligible subjects (1071), we conducted the analysis regarding job stress and hypertension that included 599 IT professionals and 472 ITES professionals [Figure 2].

Figure 2

Diagram describing the flow of participants during the quantitative phase

Table 2 provides description of continuous variables used in the study.

Table 2

Diagnoses of Hypertension, 7th report of JNC on classification[86]

We had two different types of stress domains. The first domain comprised of contextual stress factors. The results of internal consistency and correlation among the contextual stressors are provided in Table 3. The results of internal consistency and correlation among the contextual stressors are provided in Tables 4 and 5.

Table 3

Descriptive statistics: Continuous variables in the study

Table 4

Internal consistency of contextual stress domains and OSI stress domains

Table 5

Internal consistency of OSI stress domains

Similar analysis has been performed on the sub domains of OSI and their results of collinearity. However, they are not relevant here as they are already subsumed within OSI. OSI is an additive model incorporating all the subdomains based on the results from theoretical models and results from cognitive psychology experiments. Hence, it would not be meaningful to split the sub domains and see their correlation amongst each other and further use this information to include in our models. Hence, whereas using OSI, it is “All or None” principle; either all the subdomains incorporated into one single OSI score is included in the model or OSI is not included at all. However, for the results are provided in Table 2 to explore the nature of correlation amongst each domain and with overall OSI.

The results indicate that stress domains are poorly correlated with each other [Tables 4 and 7]. OSI doesn’t correlate well with contextual stressors [Tables 4 and 6]. This is because OSI is a generic questionnaire designed for computer worker, while the contextual stressors are designed based on the results from qualitative study.

Table 7

Collinearity among the sub-domains of OSI

Table 6

Collinearity among the contextual stressors included in the study

DISCUSSION

India is on the verge of tacking a major epidemic of cardiovascular diseases among young population.[1356723394041424344] Hence, it is very important to ascertain the possible risk factors affecting the young people in developing hypertension. In the absence of clear evidence from India, it is very important that locally contextual stressors are included in the study. We used both OSI and contextual stressor domains as part of the validation study to include the questionnaires for the Bengaluru IT workforce study.

OSI is a questionnaire designed for computer worker mostly in different parts of the world, while the contextual stressors are designed based on the results from earlier qualitative study. The contextual stressors thus, reflect the predominant local stress conditions, which are important according to the local contexts of work environment. Hence, it is important to explore the association of each of the contextual stress domain with outcome separately. Based on these results, it is also justified to include all the contextual stress domains if one intends to do multivariate analysis, as the results would give most important factors among them. However exploring individual stress domains is important for the reasons of designing interventions to reduce, the need be, based on the results of associations.

In the pursuit of objective questionnaires of job stress, there has been considerable and inconclusive debate over the use of job-specific against generic measurements.[45] It is also difficult to develop questionnaires based on one theoretical model, as researchers would be running the risk of leaving out best features of other models.[45] The difficulty to select questionnaires enhances complexity in the absence of evidence on details on job components, work environment, type of supervision and cultural contexts. Our methodology allowed for using OSI whilst we also added another short questionnaire that sought information on contextual stressor domains as well. This was necessitated as results from our study indicated that OSI doesn’t correlate well with contextual stressors.

India is undergoing epidemiological transition. It is important to identify the environmental risk factors for developing hypertension in young professionals. There are several studies in India estimating the high prevalence of hypertension in different parts of the country. However, there is hardly any study done to estimate the association of job stress of hypertension. Hence, it is important to perform validation study to show how best the information on job stressors can be captured and maximize the utility of available indexes for assessing job stress.

Our study had some limitations. First, it had to be noted that not all companies were located in the identified IT parks. We conducted the study in confined areas of the IT parks, wherein several companies of IT sector were located. It is possible that there might be other companies that are outside the park and hence could not be included in the sampling frame of our study. The core-functions of IT companies were very specifically defined and it is very hard to imagine that companies would differ greatly within Bengaluru merely because they were outside the IT park from those who were inside. Most of the big IT Companies had their office in the vicinity of IT Park. We argue that stress levels in small IT companies might be more compared with the larger companies.[46] Hence, by limiting the chances of smaller IT companies to get included because they were located outside, if any bias got introduced at all that is expected to result in under-estimation of job stress and hypertension.

Second, OSI doesn’t discriminate between missing values and no values. There was a possibility that information was misclassified. We studied prevalent cases of hypertension and hence might under-estimate the effects of job stress if it truly caused hypertension but also caused subjects to leave the job. The same bias could occur if instead of terminating the employment, those who were affected by job stress had got transferred into occupation that did not have much stress. Hence, workers moving out of IT/ITES workforce could explain the negative or no association.

CONCLUSION

Several studies have included generic and occupation specific questionnaires reasonably standardized and validated in developed countries to estimate job stress. However, there seems to be little consensus on which one of these indices can be referred to as the “gold standard,” which can be used universally. Further, researchers in low and middle-income countries (LMICs) will also have to consider limitation of extending the constructs and questions derived from developed countries. In such scenarios, use of mixed methods can be very useful. Use of qualitative studies can identify contextually meaningful local stressors, which can strengthen the existing questionnaires.

Performing validation of the questions generated by qualitative studies will guide the formulation of more efficient questionnaires. In LMICs, where no existing questionnaires for assessing job stress are available, we consider that mixed methods will be a valid option. Notwithstanding some of the important findings, like any other cross-sectional study, caution will have to be applied before conclusive inferences can be drawn before applying the stressors domain reported in this paper or finding their association with high blood pressure. We recommend supplementing occupation specific job stressor questionnaires with questions specific and local to the target groups being studied. Effort to obtain specific and contextually relevant data is especially useful to guide intervention research and for advocacy to bring in desired changes in local participants.