Job tenure and work injuries: a multivariate analysis of the relation with previous experience and differences by age

Background One of the consequences of the increasing flexibility in contemporary labour markets is that individuals change jobs more frequently than in the past. Indeed, in many cases, through collecting a lot of contracts, individuals work in the same economic sector or even in the same company, doing the same job in the same way as existing colleagues. A very long literature has established that newly hired workers – whatever the contract type – are more likely to be injured than those with longer job tenures. The objectives of this paper are: 1) to study the relationship between job tenure and injury risk taking into account past experience as a possible confounder; and 2) to evaluate how the effects of past experience and job tenure are modified by age. Methods Using a longitudinal national database, we considered only job contracts starting in 1998–2003 held by men working as blue collars or apprentices in the non-agricultural private sector. We calculated injury rates stratified by job tenure and age. Multivariate analyses were adjusted for background variables and previous experience accrued in the same economic sector of the current job. Results In the study period 58,271 workers who had experienced 10,260 injuries were observed. These people worked on 115,277 contracts in the six years observed (1.98 contracts per worker). Injury rates decrease with job tenure; the trend is the same in each age group; young workers have both the highest injury rate (9.20; CI 95%: 8.95-9.45) and the highest decrease with job tenure. Previous experience is associated with a decreasing injury rate in all age groups and for all job tenures. Multivariate analyses show that, even after checking for previous experience, workers with job tenure of less than 6 months show always higher relative risks compared with job tenure > 2 years: relative risk is 41% higher among under-thirty workers; it is 22% higher among people over forty. Previous experience is protective against injury risk in workers over thirty: after checking for all other variables, relative risk is lower in workers who have accumulated more than 5 years of experience. Conclusions In a context in which career fragmentation is increasing, workers find themselves more and more in the "high risk" period and only individuals who are able to build their career with similar jobs may mitigate the higher risks thanks to their past experience. If institutions don’t adopt appropriate prevention policies, injury risk is likely to increase, especially among young people.


Background
Temporary employment is increasing throughout the world. In Italy, the share of temporary contracts among workers between 15 and 64 years of age rose from 9.8% in 1999 to 13.4% in 2011.
In 2011, half of workers between 15 and 24 years of age had a temporary contract. The percentages are comparable with other European countries [1].
Most published research documents the adverse effects of temporary work on health, although null findings have also been reported [2]. A recent review suggests that temporary employment might be associated with lower sickness absence rates than permanent employment [3], even if some studies did not find this association after controlling for individual situations and job characteristics [4,5]. In the case of work injuries, the review concluded that only seven out of thirteen studies have shown higher risk for fixed-term workers compared to permanent ones.
Among the factors that have been identified as explaining the association between temporary employment and work injuries are the assignment of temporary workers to occupations with more hazardous working conditions and a lack of safety training in workplaces [6][7][8]. Another common explanation states that the higher injury incidence is a result of less work experience. This second explanation suggests a more general link between the increasing flexibility of contemporary labour markets, where individuals change jobs more frequently than in the past, and workplace safety. Indeed, a very long literature has established that newly hired workerswhatever the contract typeare more likely to get injured than those with longer job tenures [9,10]. The increased flexibility however is a contextual factor that modified the way in which this relation operates: 1. On the one hand, the increased flexibility often results in the fragmentation of the entire career. Short-term job tenure is no longer mainly associated with young people entering the labour force [11]: individuals now frequently collect many contracts during their entire working life and this may put them continuously in the "high risk" period of starting a job. 2. On the other hand, a worker starting a new contract after having already had many jobs should not be considered tout court (wholly) without experience.
Most studies have focussed mainly on investigating whether the short job tenures of temporary employment are associated with higher injury risks, without considering the potential protective effect of past experience. Yet, in many cases, through collecting a lot of contracts, individuals work in the same economic sector or even in the same company, doing the same job in the same way as colleagues with permanent contracts. This may confound the relation between job tenure and injuries, particularly at different ages and may lead to an overestimation of the public health concerns about temporary employment if career experience proves to be a substitute for within-the-job experience.
The aim of this paper is to study whether the experience accumulated prior to the beginning of a contract may have a protective role against work injuries. In particular the objectives are: 1) to study the relationship between time on the job (that is job tenure) and injury risk, taking into account past experience as a possible confounder, and 2) to evaluate how the effects of past experience and job tenure are modified by age.

The WHIP-INAIL data base
The study is based on the WHIP-INAIL archive, described in detail elsewhere [12].
In short, the Work History Italian Panel (WHIP) is the result of data processing of a 1% sample of individuals taken from the archives of the Italian National Social Security Institute (INPS). INPS insures approximately 15 million people: employees in private companies, semidependent workers and self-employed people (artisans and traders) with the exclusion of professionals like lawyers or architects. This group effectively represents all activities in manufacturing, construction and services in Italy; while public employment, predominant in education and health sectors, is not covered. Moreover there are other sectors, such as agriculture, which are not entirely comprised in INPS. For each person sampled the employment history was reconstructed, including all employment spells, retirement, and any period in which the individual received social security benefits such as unemployment subsidy. The period covered is from 1985 to 2004. The data regarding dependent employment (used in this paper) include basic demographic information (age, sex, country of birth), information about the job (such as weeks worked, skill level, type of contract, temporary leaves, dates of job start and end), plus basic data about the employer (date of constitution, economic sector, yearly average number of employees).
The Italian Workers Compensation Authority (INAIL) collects information on occupational injuries with an absence from work of more than three days certified by a physician (always required). The injuries occurred between 1994 and 2003 were extracted from the national database with the same sampling frame used for the WHIP archive. Injuries were deterministically linked with the WHIP sample using an encrypted version of the Italian tax code, which is issued by the Italian tax office to unambiguously identify all individuals residing in Italy. All activities, regardless of their complexity or depth, were conducted in accordance with Italian regulations on privacy and with the approval of the national institutes involved. The accuracy of the WHIP-INAIL estimate of the injury risk was demonstrated by a comparison with Eurostat statistics and by the consistency of preliminary results with published data [12].

Statistical analysis
In the analysis we considered only job contracts in private companies starting in the period 1998-2003, held by men working as blue collars or apprentices in the nonagricultural private sector. We chose to focus the analysis in the last period available to avoid any bias related to changes occurring in the labor market regulation over the years (in 1997 there was an important reform in the regulation of the labor marketlaw num. 196/1997). However, for the workers selected in the study, we use all the information in the database, considering their career also prior to 1998.
In total we selected 115,277 contracts (representing 74% of contracts in the data set), both permanent and temporary. The latter included fixed-term contracts, seasonal work and on-the-job training contracts, while jobs obtained through temporary work agencies were excluded, since for these it is impossible to identify important confounders such as economic sector and firm size at the time of the injury as they are recorded in the INPS archives under the name of the agencies rather than the employer.
We distinguished two types of experience: the first is the one accrued within the job contract currently held by a worker, which we call the job tenure; the second is the one accrued before the current job contract, which we call the previous experience.
The job tenure is a time dependent variable calculated as the time elapsed from the beginning of the contract. Its length is maximum 72 months. It was categorized into 4 classes (< 6 months; 6-12 months; 13-24 months; >24 months To classify workers according to their previous experience, workers never registered in the data base prior to their current job were defined as "workers with no previous experience". For all other workers the previous experience is calculated as the number of years spent as an employee or self-employed in the same economic sector of the current job. It was categorized into 3 classes (<1 year; 1-5 years; >5 years).
Time at risk was calculated on the basis of months actually worked, which were obtained subtracting from paid months all periods of absence from work due to illness or injuries, temporary-lay-off and maternity. All analyses were stratified by age class (< 30 years; 31-40 years; > 40 years).
In our analysis we considered only injuries occurred in 1998-2003. Injury rates per 100 person years were calculated (with the relative confidence intervals at 95%) for job tenure and previous experience stratified by age.
Injury risks for job tenure, stratified by age, were calculated using a Poisson distribution for panel data that takes into account time-dependent variables. The 95% confidence intervals were calculated applying the correction for repeated events [13]. We tested three models: Finally, we calculated injury risks for previous experience stratified by age and adjusted for background variables and job tenure.
Analyses were performed using SAS 9.2 and Stata 10.

Results
In the study period 58,271 workers who had experienced 10,260 injuries were observed. These people worked on 115,277 contracts in the six years observed (1.98 contracts per worker). Table 1 shows the main characteristics of population included in the study. As expected, among workers who began a new contract (Table 2), there were mostly people under 30 years of age (47% of person years); however, the phenomenon was relevant also for workers over forty (23% of person years). Young people changed contract more frequently: among workers with a job tenure shorter than 6 months, 52% of person years were under thirty; those with a duration > 24 months accounted for 40%. Young people had less previous experience: among the workers with no previous experience, 75% of person years was under 30. On the contrary 48% of person years with previous specific experience longer than 5 years were over 40.   Table 4 shows the results of the multivariate analysis stratified by age: compared with job tenure > 2 years, workers with job tenure of less than 6 months show always higher relative risks (first model): relative risk is 46% (CI 95%: 1.33-1.59) higher among workers under 30 years of age; it is 21% (CI 95% 1.07-1.73) higher among people over forty. The relative risks do not change much after controlling for background variables (second model). The most important confounding factors are firm size and economic sector (data not shown). The relative risks do not change even after controlling for previous experience (third model).
Previous experience is protective against injury risk in workers over thirty (Table 5): after controlling for all other variables, RR is lower in workers with more than 5 years experience. Workers with no previous experience have a lower RR in all age groups, but only among workers over forty this becomes statistically significant (RR: 0.68; CI 95%: 0.53-0.88).

Discussion
Our findings show an inverse relationship between job tenure and injury risk: results estimated on the Italian cohort are consistent with studies conducted in other countries [7,9,10]. This relationship concerns all ages, but the risk differences are more evident among people under 30 years of age. The relationship persists despite the effect of confounding due to economic sector, firm size, country of birth and firm geographic area (model 2). This work is the first of this kind in Italy: the studies available up to the time of our research were limited to workers hired from temporary work agencies and did not take into account confounding factors [14][15][16].
The present study makes some important contributions to understanding the relationship between work injuries and careers characterized by many work contracts of short duration. The two factors that shape this relation are job tenure and past experience: to our knowledge, there have been no published papers assessing the role of the latter factor in determining injuries. Usually studies assessing the influence of experience on safety take into consideration only on-the-job experience (that is job tenure); Saloniemi emphasizes the importance of reconstructing the work histories of fixed-term workers to evaluate experience as a whole [17].
Firstly, previous experience has a protective role against injuries for workers over 30 years old (see Table 5). Compared to the colleagues with more than 5 years experience, workers with less than 5 years of experience have a 20% higher injury risk. Below 5 years of experience, there appear to be no differences among workers with less than 1 year or between 2 to 5 years of experience. This result seems to exclude a gradient-like shape of the effect, instead it points to a bipartition between workers with or without a long (five years and over) specific experience. This is coherent with the absence of an effect on workers under 30 years of age, whodue to their agewill rarely have a specific experience significantly higher than five years.
Our results show that increased mobility across jobs as well as amongst workers under 30 years of age may have different effects, depending on whether workers are able to re-employ themselves in jobs requiring their specific work experience.
Previous experience, however, is not a modifier of the relation between injury risk and job tenure (see Table 4): in all age classes, newly hired workers show a relative risk of injury which is the same with or without controlling for previous work experience (model 3). This result shows that previous experience is only a partial substitute of the experience accrued within the job. Although it is protective factor, it is unable to support workers in facing all the demands of a new organization: it would take at least one year of work in a new company to acquire the familiarity needed to work in a comparable way to colleagues employed the longest time. The increased injuries risk would be related to work organization and to interaction with colleagues [18]. Newly hired workers tend to perform unfamiliar work or tasks that were not part of their previous duties. In a previous study conducted in the USA, "using a different method to do a task" and "doing an unusual task" were found to significantly increase the relative risk of an occupational acute hand injury [19]. The newly hired are less familiar with the formal and informal rules governing safety on site and their presence may affect inter-worker communications. In other words, it becomes more difficult to coordinate decisions, to anticipate dangers and to recognize the consequences of particular individuals or groups controlling materials and processes.
Another explanation for high injury risk in job tenures of less than 1 year may be due to the assignment of newly hired workers to high-risk activities that existing workers refuse to do [18]. To contrast such risks, the management of a company should outline an initial phase of adaptation during which new employees are properly introduced to the company's organization with the support of colleagues who already have experience in the company and during which they are assigned to simpler and less dangerous tasks. A study based on a survey conducted on 300 Italian workers employed by temporary work agencies found similar results [14]. Workers under 30 years of age show a stronger job tenure/injury association than older workers. Our multivariate analysis shows that neither background characteristics nor previous experience change this distinction. There are no obvious mechanisms that would help explain this fact. Further research is needed to investigate whether in Italy there are differences by age in safety training at the beginning of a new job as reported in other countries [6].
Another important issue concerns the existence of age differences in risk perception that could put young workers in situations of greater danger compared to older colleagues [20]. The two most relevant reviews of age differences [21,22] indicate that a substantial part of the elevated injury risk experienced by young workers appears to be due to differences in the types of jobs held by young people and adults (and the associated hazard exposures). Although temporary work is a phenomenon that concerns the entire workforce, it is more common among young people and our results suggest it is important to include job tenure among the control variables when studying injury risk in young workers [23]. Further research is needed to focus on this age class, in particular on the fact that certain  age groups may be more likely to have unfavourable working conditions. All analyses show that injury rates are always lower among workers with no previous experience. We think that this is the result of two different selection processes.
Among workers under 30 years of age, those with no previous experience were mostly represented by young people entering the labour market for the first time: they had an average age of 21, a contract of apprenticeship in 40% of cases and an average wage of 873 euro per month. For all those entering the market, the best worker-firm matches on average will survive longer and individuals in these (presumably safer) jobs will remain classified as "workers with no previous experience". On the other hand, workers with worse jobs and/or worse health are more likely to leave their first job and have another or many other contracts. This sort of "healthy worker survivor effect" will imply that on average the young who are no longer in their first job will probably be working in worse and more dangerous jobs.
Among people over thirty, workers with no previous experience are mostly immigrants (< 30 years: 28%; 31-40 years: 78%; > 40 years: 64%). In this case there is probably a selection mechanism of healthy workers with experience gained in foreign firms. In fact, some studies showed evidence of a "healthy immigrant effect" whereby immigrants, and especially recent immigrants, are less likely than indigenous population to have poorer health [24]. In regards to workers over 30 years of age born in Italy who enter in the INPS data-base for the first time (very few indeed in our sample, see the confidence intervals), they mostly belong to particular categories previously recorded in other public social security funds. So, for the large part these are fake entries: the typical situation is that of a worker employed in a large public company, which is then privatized. He /she, on average, will be working in a safer work environment.
Interpretation of our findings must be made in light of study limitations.
Our research was restricted to male workers in the private sector, employed as blue collars or apprentices, mainly in manufacturing, construction and services: this way we reduced the likelihood that other job characteristics and selection processes might bias our estimates [3]. The results for this workers' category are quite reliable but, in order to test whether our conclusions can be generalized to the entire Italian population, we are conducting further analyses on women, the self-employed and white-collars. Farm workers and public employees are insured by other social security funds. Since the Italian government has recently assigned the management of these funds to INPS, we hope in the future to be able to extend our analysis to these workers too. This paper does not consider the type of injuries. The risks due to age, experience and job tenure may be different, for example, for fall and handling machines/ objects. Further analyses devoted to specific injury types would be advisable.
Previous experience was calculated without distinguishing the length and number of contracts that had generated it. For example, a worker with a previous two-year contract has been equated to a person with four previous contracts of six months each. Furthermore, we didn't consider whether the experience was gained in the same company or in different companies and we didn't take into account periods out of work, where the possibility of a depreciation of human capital should be considered. In order to assess in more detail if these differences in worker careers have a specific effect on injury risks, we are reconstructing the careers of every single worker in the WHIP-INAIL archive.
Jobs obtained through temporary work agencies are excluded because they are recorded in INPS archives under the agencies, so it's impossible to identify the economic sector and firm size at the time of the injury. We believe that injury risks aren't underestimated since they were not very numerous in the study period (in 2008 they represented 4.5% of workers with temporary contracts [25]; 1.6% of time at risk in our analyses). Although we could not include workers that obtained jobs through temporary work agencies in the study, our results raise great concern about these subjects: they work in the same firm only for very short periods, hence incurring the high risks that are linked to the short tenure. Moreover, since they are mostly of young age and may often change type of job, it is unlikely that they would have been able to acquired the long specific experience needed to countervail the effects of the short tenure.

Conclusions
Our analysis shows that job tenure is inversely associated with injury risks even taking into account background variables and previous experience. People under 30 years of age show higher risk rates but the phenomenon does not spare other age groups; previous experience is protective against injury risk among workers over thirty but not among young people. These results, in a context in which labour market flexibility is increasing and job switches are more common, mean that workers find themselves more and more in the "high risk" period and only individuals who are able to build their career with similar jobs may mitigate the higher risks thanks to their past experience. If institutions do not adopt appropriate prevention policies, injury risk is likely to increase, especially among young people. It is also evident that it is important to adopt appropriate strategies at company level: management should outline safety systems in which new employees are properly introduced in the company's organization with the support of colleagues who already work in the company and they should also be assigned to simpler and less dangerous tasks. The current economic crisis could modify these results. Research has shown an inverse relationship between unemployment rate and morbidity among temporary workers: in this situation, employers are more likely to find and recruit healthy workers (into temporary jobs) from the reserve of unemployed people [3]. Surveillance activities on injury risk will need to take into account the presence of contextual effects like this, which may hide medium term structural trends.
The causal effect of career fragmentation on injury and the relative contribution of several selection processes should be examined with prospective study designs, tracking the change from one employment status to another. The WHIP-INAIL data-base allows the tracking of careers and, in the near future, will provide data allowing the extension of the samples and the inclusion of new morbidity data. This will allow the study of other effects of labour market flexibility on health in Italy.