Objective To evaluate the impact of persistent precarious employment (lasting 12+ months) on the health of working age adults, compared with more stable employment. Persistent precarity reflects a shift towards less secure forms of employment and may be particularly important for health.
Methods Nine databases were systematically searched to identify quantitative studies that assessed the relationship between persistent precarious employment and health outcomes. Risk of bias (RoB) was assessed using an adaptation of the Effective Public Health Practice Project tool. Narrative synthesis and random effects meta-analysis were conducted. Certainty of evidence was assessed using the Grades of Recommendations, Assessment, Development and Evaluation (GRADE) approach.
Results Of 12 940 records screened, 50 studies met the inclusion criteria and 29 were included in meta-analyses. RoB was generally high (n=18). The most reported outcome domain was mental health; with evidence also reported relating to general health, physical health,and health behaviours. Of GRADE assessed outcomes, persistent precarious employment was associated with increased risk of poor self-rated health (OR 1.53, 95% CI 1.09 to 2.14, I2=80%) and mental health symptoms (OR 1.44, 95% CI 1.23 to 1.70, I2=65%). The association with all-cause mortality was imprecisely estimated (OR 1.10, 5% CI 0.91 to 1.33, I2=73%). There was very low GRADE certainty across all outcomes.
Conclusions Persistent precarious employment is associated with poorer health, particularly for outcomes with short time lags, though associations are small and causality is highly uncertain. Further research using more robust methods is needed but given potential health harms of persistent precarious employment, exploration of precautionary labour regulations and employment policies is warranted.
- PUBLIC HEALTH
- SYSTEMATIC REVIEW
Data availability statement
Data are available upon reasonable request. Extracted data are available on request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
WHAT IS ALREADY KNOWN ON THIS TOPIC
Persistent precarious employment reflects a shift towards less secure forms of employment and may be particularly important for health.
WHAT THIS STUDY ADDS
Despite concern about labour markets increasingly featuring less secure forms of employment, no systematic reviews have investigated the relationship between persistent precarious employment and health. We found persistent precarious employment to be associated with poorer health for a number of outcomes.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Further research is needed but, given our review findings, exploration of precautionary labour regulations and employment policies is warranted.
Work is arguably ‘the most important determinant of population health and health inequalities in advanced market democracies’, being associated with improved physical and mental health, and reduced mortality.1 Paid work is an important protective factor for health and well-being. However, there is a growing recognition of the importance of psychosocial work environments and the concept of ‘good work’ characterised by job security, sufficient income, a safe physical work environment, clear responsibilities and procedures, a positive mental health environment, and employment rights.2 A systematic review posed the question ‘is no work better than “bad work”?’ and found comparable health risks associated with the two.3 Other evidence suggests job characteristics such as income4 and occupation5 lead to differences in health.
Precarious employment is typically viewed as ‘bad work’. There is no single definition of precarious employment but it commonly includes the following dimensions: employment instability, low material rewards, erosion of employee rights and protection, non-standard working arrangements, limited opportunities for training and employability, erosion of employee representation, and imbalanced power relations.6 7 Proposed mechanisms through which precarious employment impact health include hazardous working environments (including psychosocial stressors) and material deprivation.6 A political economy approach is helpful for understanding the driving forces behind the trend towards more precarious forms of employment.1 8 Dominant political and economic ideologies from the late 20th century onwards have shaped the way in which labour is organised and structured in high-income countries, and the scope for the mitigation of adverse effects provided via the welfare state.
Precarious employment is often viewed as a modern phenomenon. However, secure employment only became commonplace (though by no means universal) in high-income countries in the aftermath of the economic crises of the 1930s, following the post-war full-employment consensus, the formation of welfare states in most European countries, and the emergence of collective bargaining and action via trade unions.9–12 This post-war consensus held until the economic recession and subsequent stagnation of the labour market during the 1970s, when the emergent neoliberal doctrine sought to both limit the reach and power of the state and also increase economic growth by relying on free market forces.
The organisation of work thus began to change in ‘novel and unorthodox’ ways during this period.13 There have been structural changes to workforce organisation that have increased the precarious nature of employment (including, eg, the employers’ reduction of their core workforce and increased outsourcing of specific services to external providers). These structural changes impacted the experience of individual workers. A central criticism of labour market flexibility is that it has resulted in risk and insecurity being transferred from employers to workers.14
Following the 2008 Great Recession, there has been concern that labour markets have seen less secure forms of employment become more prevalent. For example, in the UK, precarious employment has been framed a ‘low pay-no pay cycle’ of low paid, insecure work interspersed with periods of unemployment,15–17 suggesting a conceptual shift from precarious employment as a short-term exposure towards one where exposure is persistent. While unemployment spells and low-pay are associated with ‘scarring’ effects where they predict future low-pay and unemployment,18–20 there is also evidence that persistently precarious employment might be time-limited and act as a stepping stone towards more stable employment21 and higher pay.22
Despite the concern about less secure forms of employment, no systematic reviews have investigated the relationship between persistent precarious employment and health. Measures of precarious employment have tended to focus on cross-sectional exposures such as current employment status or sometimes a transition between employment states. Several reviews have found evidence relating precarious employment to adverse physical and mental health outcomes.3 23–37 However, conclusions have differed.38 Nine of these reviews included meta-analysis and four used the Grades of Recommendations, Assessment, Development and Evaluation (GRADE) framework to assess certainty of evidence. None looked specifically at persistent precarious employment.
The aim of this systematic review is to investigate the association between persistent precarious employment, compared with more stable employment histories, on the health and health behaviours of working age adults.
We followed a prespecified review protocol registered in the PROSPERO database.39 Review criteria were developed using the framework Population, Exposure, Comparisons, Outcomes and Study Design (PECOS).40 Our population of interest was working age adults (16–64 years) in high-income countries. Our exposure of interest was persistent precarious employment, based on perceived job insecurity, temporary or non-standard employment contract, underemployment or exposure to job insecurity such as downsizing; and covering at least 12 months. Although precarious employment is best thought of as a multidimensional determinant of health, relatively few studies have investigated multiple aspects of precarity as a joint measure. We therefore operationalised precarious employment within this review by identifying a range of indicators that act as proxies for precarious employment, and hypothesise that although the specific exposures may vary they influence health through similar mechanisms (primarily via stress responses to an individual’s employment being in some way insecure). Comparison was made with a reference group representing more secure employment. Included outcomes related to physical health, mental health, general health (related to both mental and physical health such as overall self-rated health, all-cause mortality) or health behaviours. Study designs included were longitudinal, or cross-sectional quantitative studies where participants provided detailed employment history information that met the exposure criteria above.
A structured search strategy was developed with an information scientist, including terms related to precarious employment and additional health-related terms that were included in searches of non-health databases (online supplemental additional file 1). We searched the following databases: MEDLINE, Embase, Applied Social Sciences Index & Abstracts (ASSIA), Scopus, International Bibliography of the Social Sciences, PsycINFO, EconLit, SocINDEX and Sociological Abstracts, up to March 2019. Relevant grey literature identified as part of a wider scoping review was also considered for inclusion.
Study selection and data extraction
Two reviewers screened titles and abstracts and compared decisions for the first 200 records to check consistency of approach (17 conflicting decisions). The title and abstract of all remaining records were screened for relevance by one reviewer, with a low threshold for inclusion into the next stage.
Full-text screening of potentially relevant records was performed by two reviewers, with any disagreements resolved by discussion with a further review team member. Reasons for exclusion for all full-text articles are provided in online supplemental additional file 2.
A data extraction template was produced in Excel for the full-text screening stage. The lead reviewer undertook extraction, which was checked for accuracy by a second reviewer. Risk of bias was assessed independently by two reviewers using an amended version of the Effective Public Health Practice Project (EPHPP) Quality Assessment tool for Quantitative Studies (online supplemental additional file 3).41 The EPHPP assesses: selection bias, study design, confounders, blinding, data collection and methods, withdrawals and drop-outs, intervention integrity, and analyses. Each domain is rated as strong, moderate or weak, and combined to derive a global rating. Conflicting assessments were discussed with a further review team member.
Data analysis and presentation
We tabulated extracted data by exposure and outcome. Due to the wide range of outcome measures included in this review, a narrative synthesis was undertaken for all exposure/outcome groupings using vote-counting based on the direction of effect following Cochrane guidance.42 Narrative synthesis methods and findings are presented in online supplemental additional file 4.
Forest plots were created and meta-analyses undertaken where extracted data included two or more data points for an exposure–outcome comparison, to provide a pooled estimate and confidence interval (CI) for the association between persistent precarious employment and health outcomes. Random effects models were chosen a priori due to anticipated high levels of heterogeneity in study designs, measures and settings. Binary and continuous exposures and outcomes were meta-analysed separately. We used three-level meta-analysis grouped by study and comparator group in cases where more than one data point used the same reference group within a study. We pooled studies across exposure topic to provide an overall estimate on the assumption that the different exposure topics all represent the underlying phenomenon of persistent precarious employment. We also include nested subgroup meta-analyses by exposure topic. This includes where only one study is included in the sub-group, although these single study nested subgroups are not reported as findings. Nested subgroup analysis was not possible in cases where one exposure topic subgroup included only multiple data points from the same reference group within a single study. For these cases, we present three-level meta-analysis without grouping by exposure topic, and include meta-analyses stratified by subgroup in online supplemental additional file 5. We conducted subgroup analyses by sex (online supplemental additional file 6) and sensitivity analyses excluding high risk of bias studies (online supplemental additional file 7). We present the meta-analyses for self-rated health, all-cause mortality and common mental health disorders in the main paper and other outcomes are presented in online supplemental additional file 8.
Certainty of evidence was assessed using the GRADE framework which assesses how strong evidence is for associations reflecting causal relationships.43 GRADE domains include risk of bias; consistency of relative effect estimates; directness of comparison of exposures, outcomes and populations to the review question of interest; precision of effect estimates; and risk of publication bias. An overall rating was scored to indicate the certainty in estimates of effect (high, moderate, low or very low) for each outcome. We prioritised the following outcomes for GRADE assessment based on their burden of disease44 and the plausibility of being impacted by persistent precarious employment: all-cause mortality, self-rated health and mental health symptoms are presented in this paper; diastolic blood pressure, alcohol consumption and current smoking status are presented in online supplemental additional file 9. If a study reported a binary and continuous outcome we chose the former, as these were more commonly reported. Absolute risk estimates were calculated by applying the pooled estimates from meta-analysis to UK prevalence estimates.
After deduplication, 12 940 records were subject to title and abstract screening, with 555 full-texts assessed and 50 studies (reflecting 46 separate datasets and 236 independent data points) included (Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart presented in online supplemental additional file 1). Twenty-nine studies were included in ten meta-analyses.
A descriptive summary of the included studies is presented in online supplemental additional file 1. Most studies reported findings from cohort studies (n=44), and a small number from cross-sectional studies (n=4) or case–control studies (n=2). Based on the amended EPHPP assessment, 16 studies were classified as low, 17 medium and 18 high risk of bias. Most studies were published in the 21st century, with only four studies published earlier. Studies were included from the following countries: USA (n=11), Finland (n=6), Sweden (n=6), UK (n=6), Italy (n=5), South Korea (n=4), Canada (n=3), Germany (n=3), Japan (n=3), Denmark (n=1), Norway (n=1), Spain (n=1) and Switzerland (n=1). The duration of exposure measurement ranged from 1 year to 51 years. The most investigated exposure was having an insecure employment contract (n=19), followed by perceived job security (n=14), number of employment spells (n=9), multiple precarity indicators (n=5), underemployment (n=2) and exposure to job insecurity such as downsizing (n=1). The most reported outcome domain was mental health,45–71 with evidence also reported relating to general health,45 56 57 61 64 66 72–85 physical health52 57–59 72 73 84 86–90 and health behaviours.57–59 72 84 91–94
General health outcomes
For studies where poor self-rated health had been treated as a binary outcome (n=5), persistent precarious employment was associated with higher odds of poor self-rated health (OR 1.53, 95% CI 1.09 to 2.14, I2=80%) (figure 1A). When stratified by exposure topic, only perceived job security was clearly associated with higher odds of poor self-rated health (OR 2.09, CI 1.61 to 2.71, I2=0%) (online supplemental additional file 5). For studies where self-rated health was measured as a scale variable (n=5), persistent precarious employment was again associated with poorer self-rated health relating to an average reduction of 0.19 on a five-point scale (regression coefficient −0.19, 95% CI −0.30 to −0.08, I2=37%) (figure 1B). When stratified by exposure topic, only perceived job security had more than one estimate that could be pooled (regression coefficient −0.25, 95% CI −0.45 to −0.004, I2=46%) (figure 1B). However, similar estimates were found across the three exposure topics included in this meta-analysis.
No clear association between persistent precarious employment and all-cause mortality was found (OR 1.10, 95% CI 0.91 to 1.33, I2=73%). Studies in this meta-analysis covered three exposure topics, though each was represented by only one study (figure 1C).72 81 82
Mental health outcomes
Persistent precarious employment was associated with symptoms of poor mental health both as a binary outcome pooled from 11 studies (OR 1.44, 95% CI 1.23 to 1.70, I2=65%) (figure 2A) and as a continuous outcome using the Center for Epidemiological Studies Depression Scale (CES-D) pooled from two studies (regression coefficient 0.26, 95% CI 0.13 to 0.38, I2=0%) (figure 2B). The CES-D scale is a 20-item scale designed to identify depressive symptoms in the general population, with higher scores indicating greater symptomology.95 Meta-analyses of studies that used a binary outcome for poor mental health, stratified by exposure topic, showed an association for having a non-permanent employment contract (OR 1.69, 95% CI 1.25 to 2.29, I2=76%), a greater number of employment spells (OR 1.18, 95% CI 0.99 to 1.40, I2=31%) and perceived job security (OR 1.80, 95% CI 1.45 to 2.22, I2=0%); but not measures of persistent precarious employment that incorporated multiple exposures (OR 1.01, 95% CI 0.58 to 1.76, I2=37%) (figure 2A).
Summary of findings
All three outcomes that we assessed using the GRADE framework were downgraded to very low overall certainty (table 1). No publication bias was detected based on visual assessment of funnel plots for outcomes with ten or more data points (see online supplemental additional file 10).
There is consistent evidence that persistent precarious employment is associated with a range of adverse health outcomes, but there remains considerable uncertainty about whether this reflects causal relationships. Our meta-analyses found the clearest evidence of negative associations with persistent precarious employment for the following outcomes: 26% increased odds of poor self-rated health, equating to an illustrative increase from 25% to 30% prevalence among the English adult population; 42% increased odds of symptoms of poor mental health, equating to an illustrative increase from 19% to 24% prevalence among the English adult population. Pooled effect estimates suggested little association with all-cause mortality. The findings of our review were not qualitatively changed by stratifying our meta-analyses by sex or by excluding high risk of bias studies.
The findings of our review, therefore, suggest fairly consistent evidence of health harms relating to persistent precarious employment. However, based on the general low certainty of the studies included in the systematic review, the certainty of effect size estimates is also low. Time lag between exposure and outcome may be a factor in the associations reported in our review. Proximal outcomes such as symptoms of poor mental health were found to be associated with persistent precarious employment, while no association was found with all-cause mortality—the most distal outcome included in the review. This reflects the findings of previous reviews that have investigated the relationship between precarious employment and health, which have reported adverse outcomes for mental health,23 25 26 28–30 32 33 physical health,24 28 31 32 38 and health behaviours.30 Our review provides clearer evidence around the lagged effects of precarious employment persisting for 12 months or longer. Only two previous review presented a meta-analysed estimate that could be compared with our meta-analyses. Kim and von dem Kneseback estimated an OR of 1.29 (95% CI 1.06 to 1.57) for the risk of perceived job insecurity on depressive symptoms.26 This is slightly smaller than our review’s overall pooled estimate for persistent precarity (OR 1.42, 95% CI 1.25 to 1.61) and for persistent perceived job insecurity (OR 1.80, 95% CI 1.45 to 2.22), potentially reflecting the persistent nature of our exposures of interest. Similarly, Rönnblad et al reported slightly smaller pooled estimate for the risk of perceived job insecurity on all mental health outcomes (OR 1.52, 95% CI 1.35 to 1.70).37 Unlike our review, Rönnblad et al found an increased risk of multidimensional measures of precarious employment on all mental health outcomes (OR 2.01, 95% CI 1.60 to 3.64). Additionally, our review adds meta-analysed evidence on two common measures of general health (self-rated health and all-cause mortality), which were absent from previous reviews.
This review builds on previous research highlighting the importance of ‘good work’ to public health,1–5 and emphasises emerging notions of precarious employment as a persistent exposure driven by social, political and economic forces dominant from the latter part of the 20th century.15–17 It is noteworthy that the extensive search undertaken for our systematic review found no studies that explicitly evaluated the effectiveness of policies or interventions to reduce workers’ exposure to persistent precarious employment. A protocol has since been published for a systematic review that will investigate interventions addressing precarious employment and its impact on workers’ health and well-being.96 There are, however, general principles about what constitutes ‘good work’ that may be applicable to preventing or mitigating health harms associated with precarious employment. The persistent nature of precarious employment suggests that policy responses are required not just at the level of employer/contractor practice but also interventions focused on labour market regulations and social support structures. This could include strengthened employee protection, representation and collective bargaining, and social support.97 98 In addition, our review highlights the need for further research using causal study designs to investigate the relationship between specific exposures indicative of persistent precarious employment and health outcomes.
Strengths and limitations
Our review builds on previous research into the association between precarious employment and health by framing precarious employment as a persistent exposure. Although the causal relationship between persistent precarious employment and health is unclear, the temporal ordering of exposure and outcome in the studies included in our review provides stronger evidence than simple cross-sectional associations. We were able to meta-analyse pooled effect sizes for a number of outcomes of importance to public health and specific exposures indicative of persistent precarious employment. In addition to the overall pooled estimates which draw on a range of different proxy measures indicative of precarious employment, we also explored specific dimensions of precarity within our meta-analyses.
There are a number of limitations to our systematic review that should be considered when interpreting the findings. First, only one reviewer did the initial title and abstract screening. Any potential bias introduced at the stage was mitigated by adopting an inclusive screening policy where any paper that looked like it could meet the criteria was included in the full-text screening, which was done independently by two reviewers with a good level of inter-reviewer agreement. Second, precarious employment is best considered as a multidimensional concept.6 7 14 While it is a strength that we were able to include different exposure categories of precarious employment and single exposures are likely to be more amenable to policy intervention; it should be noted that most studies included in the review only focused on a single dimension of precarious employment and our focus on persistence has necessitated the synthesis of a heterogeneous evidence base. On their own, these may be insufficient to adequately measure precarity and may individually underestimate the overall effect of multiple dimensions of precarious employment on health. However, the direction and magnitude of effect was quite consistent across outcomes. Third, we have defined persistent precarious employment in a way that allowed us to operationalise persistence within the scope of our review but acknowledge that other definitions of persistence could be used. We were limited in the extent to which we could investigate dose-response effects by length of exposure. The length of exposure time was not always clearly reported and may relate to the study period rather than time spent in precarious employment. Fourth, some exposure–outcome combinations had no or few estimates, which resulted in meta-analysis not being possible or relying on a small number of data points. This resulted in uncertainty about pooled point estimates and high levels of heterogeneity.
Despite identifying evidence assessed as being very low certainty across all outcomes of interest, our findings suggest that persistent exposure to precarious employment is associated with health harms; in particular those with relatively short time lag such as symptoms of poor mental health and self-rated general health. Further research would be beneficial to provide more certain evidence on the extent to which such associations are causal. The trend towards less secure forms of work in many countries suggests that persistent precarious employment is an increasingly prevalent risk factor for poor health and while it may pose a small relative risk to health, it may have a large adverse impact at a population level. It also supports the increasing need for policy responses that go beyond individual employer/contractor practices to broader policy areas addressing social, political and economic trends that could include, as general principles, strengthened employee protection, representation and collective bargaining, and social support. There would be value in further natural experiment studies that can investigate the impact of such policy changes.
Data availability statement
Data are available upon reasonable request. Extracted data are available on request.
Patient consent for publication
The authors wish to thank Valerie Wells (Research Associate - Information Scientist, MRC/CSO Social and Public Health Sciences Unit, University of Glasgow) for her advice on the development of our search strategy; and Ms Mhairi Campbell (Systematic Reviewer, MRC/CSO Social and Public Health Sciences Unit, University of Glasgow), Dr Michele Hilton-Boon (Research Fellow, Glasgow Caledonian University) and Dr Hilary Thomson (Senior Research Fellow, MRC/CSO Social and Public Health Sciences Unit, University of Glasgow) for their advice on review methodology.
Twitter @AndrewPulford, @AlastairLeyland, @vkatikireddi
Contributors This review was conceived by AP with advice from FP, AL and SVK. The review protocol was written by AP with comments from RMT, FP, AL and SVK. Screening, data extraction and quality assurance was undertaken by AP, AT, RMT, AG and SVK. Review synthesis was undertaken by AP with advice from RMT, MJG, AL and SVK. The manuscript was drafted by AP with comments from all co-authors. AP acts as guarantor for this review.
Funding AP, RMT, AG, MJG, AL and SVK acknowledge funding from the Medical Research Council (MC_UU_00022/2) and the Scottish Government Chief Scientist Office (SPHSU17). SVK acknowledges funding from a NRS Senior Clinical Fellowship (SCAF/15/02).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Author note The primary affiliation of the first author, Andrew Pulford, is affiliation 2 (MRC/CSO Social & Public Health Sciences Unit, University of Glasgow, Glasgow, UK).
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.