Open access Original research Measuring precarious employment in Sweden: translation, adaptation and psychometric properties of the Employment Precariousness Scale (EPRES) Johanna Jonsson,‍ ‍ 1 Alejandra Vives,2,3 Joan Benach,3,4 Katarina Kjellberg,1,5 Jenny Selander,1 Gun Johansson,1,5 Theo Bodin1,5 To cite: Jonsson J, Vives A, ABSTRACT Benach J, et al. Measuring Strengths and limitations of this study Objectives Precarious employment (PE) is a determinant precarious employment in of poor health and health inequality. However, the Sweden: translation, adaptation ►► First translation and adaptation of the Employment evidence of health consequences and mechanisms and psychometric properties of Precariousness Scale (EPRES)-2010 to Swedish and underlying the associations, are still limited due to a the Employment Precariousness the Swedish context. Scale (EPRES). BMJ Open lack of a comprehensive multidimensional definition and ►► First assessment of the psychometric properties of 2019;9:e029577. doi:10.1136/ measurement instrument. The Employment Precariousness the EPRES-Se. bmjopen-2019-029577 Scale (EPRES) is a Spanish, multidimensional scale, ►► Relatively small sample restricted to non-standard developed to measure degree of PE. The aim of this study ►► Prepublication history and employees. was to translate the EPRES-2010 into Swedish, adapt it additional material for this ►► Limited generalisability of results. to the Swedish context and to assess the psychometric paper are available online. To view these files, please visit properties of the Swedish EPRES. the journal online (http://​dx.​doi.​ Method EPRES was translated, adapted and implemented org/​10.​1136/​bmjopen-​2019-​ for data collection within the research project PRecarious inequality.1–3 PE is present in both low-in- 029577). EMployment in Stockholm (PREMIS). During 2016–2017, come and middle-income as well as devel- questionnaire data were collected from 483 non-standard oped countries.4 5 However, evidence of the Received 31 January 2019 employees in Stockholm, Sweden, sampled with web- health consequences of PE, and by which Revised 08 August 2019 based respondent-driven sampling. Analyses included mechanisms PE harm workers’ health, is still Accepted 16 August 2019 item descriptive statistics, scale descriptive statistics and limited due to the lack of a comprehensive exploratory factor analysis. multidimensional definition and measure- Results The final EPRES-Se (Swedish version of the ment instrument.6 This lack also brings about EPRES),consisted of six dimensions and 23 items. There challenges in terms of capturing the size of was a high response rate to all items and response options. the population in PE, conducting occupa- Global Cronbach’s alpha was 0.83. Subscales ‘vulnerability’, ‘rights’ and ‘exercise rights’ had reliability coefficients tional health and safety surveillance,6 7 as well between α=0.78–0.89 and item-subscale correlations as cross-country comparisons. The Employ- between r=0.48–0.78. ‘Temporariness’ had poor reliability ment Precariousness Scale (EPRES) is a (α=−0.08) and inter-item correlation (r=−0.04), while Spanish questionnaire developed to measure ‘disempowerment’ showed acceptable psychometric six dimensions of PE. EPRES has previously properties (α=0.5; r=0.34). Exploratory factor analysis been validated in Spanish and Chilean popu- confirmed the original EPRES factor structure. lations,8–11 and also applied to the popula- Conclusions ‘Vulnerability’, ‘wages’, ‘rights’, ‘exercise tion of Catalonia,12 but as of yet there is no rights’ and ‘disempowerment’ worked in the Swedish Swedish translation or adaptation. © Author(s) (or their context; however, ‘temporariness’ would need revising employer(s)) 2019. Re-use before implementing the EPRES-Se in further research. Precarious employment permitted under CC BY-NC. No Continued work and validation of EPRES-Se is encouraged. commercial re-use. See rights During the past decades, neoliberal economics In order to enable international comparisons and and permissions. Published by and policies together with increased global- BMJ. multinational studies, similar studies in other European isation, trade competition, technological countries are also called for. For numbered affiliations see innovation and financial crises, has had a end of article. considerable impact on the dynamics of the Introduction Correspondence to labour market.5 13 14 These impacts have had Precarious employment (PE) is considered a Johanna Jonsson; several implications, including an increase social determinant of poor health and health ​Johanna.​Jonsson@​ki.​se in privatisation, downsizing, outsourcing, a Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 1 Open access weakening of union input and collective regulation, and a not yet fully understood but pathways that have been more competitive and uncertain context for workers with suggested include: more harmful working conditions, increases in flexible work, unemployment and non-stan- limited control over one’s professional and personal lives, dard employment arrangements.5 13 14 Furthermore, feelings of insecurity and income below the subsistence there has been a decline in attachment to employers, risk level, which consequently can affect other social determi- shifting from employer to employees, growth in perceived nants of health such as housing quality, lifestyles and so and actual job insecurity and work-based stress, as well as on.19 diminished bargaining power and rights.13 14 Non-standard arrangements, in comparison with stan- Employment Precariousness Scale dard employment contracts (ie, open-ended full-time EPRES is a Spanish, multidimensional theory-based scale, contracts), includes part-time work (underemploy- developed to measure degree of PE among waged and ment), temporary work, temporary agency work, zero salaried workers.8 EPRES is comprised of 22 items and hour contracts, ‘gig’ work arrangements and self-em- six subscales corresponding to six dimensions: ‘tempo- ployment.1 4 7 14 15 Non-standard work can also include rariness’ (contract duration; two items), ‘wages’ (low or holding multiple jobs.16 A comprehensive term used to insufficient; possible economic deprivation; three items), describe forms of non-standard employment is ‘PE’.17 ‘disempowerment’ (level of negotiation of employment PE does not, however, solely refer to the type of employ- conditions; two items), ‘vulnerability’ (defenselessness ment, but also to unfavourable employment conditions, to authoritarian treatment; five items), ‘rights’ (entitle- such as vulnerability, low pay, low levels of social security ment to workplace rights and social security benefits; four and rights.3 4 14 17–21 As these elements are not exclusively items) and ‘exercise rights’ (powerlessness, in practice, found in non-standard employments per se, employees in to exercise workplace rights; six items).8 9 EPRES items a standard employment are also at risk of experiencing are scored on a 5-point or 3-point scale, depending on precariousness.14 22 23 Thereby, it is important to move item, and all items taken together gives a global score beyond a simplistic categorical grouping of employment, ranging between 0 (least precarious) and 4 (most precar- such as temporary versus permanent, and instead work ious).33 EPRES has demonstrated good acceptability, towards a comprehensive multidimensional approach good internal consistency and evidence of construct that enables a better understanding of PE.16 23 validity in Spanish and Chilean populations.8 9 The orig- Several definitions and attempts to create multidimen- inal EPRES scale was revised in 2015 (hereafter known as sional constructs capturing PE already exist. In terms of EPRES-2010), which showed good metric properties and the previous, Rodgers and Rodgers include employment improved sensitivity to worker vulnerability and employ- instability, employment insecurity, lack of protection and ment stability.10 Further details about EPRES and its economic/social vulnerability as components in their defi- development has been described elsewhere.33 nition of PE17; and the International Labour Organization include low wage, poor protection from termination of PE in Sweden employment, lack of access to social protection/benefits In Sweden, much of previous research has focused on (usually associated with full-time standard employment) health outcomes of PE and related exposures. Some and lack of/limited access to exercise rights at work.21 To studies have focused on exposures such as temporary our knowledge, two validated questionnaires have been employment,34 peripheral employment35 and temporary developed for the purpose of measuring PE: The Employ- employment and job insecurity.36 Fewer studies have ment Precarity Index, identifying employment precarity created proxies of multidimensional exposures of PE, for by 10 questions and dividing scores in four groups: instance by combining previous unemployment, tempo- secure, stable, vulnerable and precarious23; and the rary/permanent employment and perceived job insecu- EPRES with its six dimensions: ‘temporariness’, ‘wages’, rity37; or by identifying multiple indicators of PE (eg, type ‘disempowerment’, ‘vulnerability’, ‘rights’ and ‘exercise of contract, income, working times etc).26 It is challenging rights’, where precarity ranges from low to high.10 Several to put Sweden in a larger comparative context of PE as studies have also used a combination of indicators as there is no consensus on its operationalisation. Sweden proxy measures in order to identify PE.24–26 Despite these has, according to some definitions (as defined by a typo- efforts, there is of yet no universally accepted definition logical approach of 11 indicators), one of the smallest or operationalisation. proportions of precarious employees among the Scandi- By the means of a cross-national multidimensional defi- navian countries.19 However, in other measures (defined nition and measurement instrument of PE, comparative by involuntary part-time work, temporary employment and more precise estimations of health effects would be and fear of job-loss) Sweden has the largest proportion made possible. Previously, PE has been linked to an array of precariously employed individuals24 and the largest of health issues including mental and physical health2 27 proportion of fixed-term employment contracts38 in the and occupational injuries.28 Previous research on related same context. Reports show that the Swedish labour concepts such as job insecurity and temporary employ- market is growing increasingly more insecure, especially ment also show consistent associations with various health for temporary employees.39 40 The proportion of tempo- outcomes.29–32 Mechanisms linking PE and health are rary employees has been stable around 15%–17% since 2 Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 Open access the late 1990’s.40 41 There has, however, been reports on Translation and adaptation of EPRES a shift within this group where longer-term positions The translation and adaptation process of the EPRES- have been replaced by a larger proportion of on-demand 2010 consisted of five steps: (1) translation from Spanish employees and day labourers, which are more precarious to Swedish; (2) cultural adaptation to fit the Swedish by nature.39 40 42 Especially women (19%, compared with context; (3) back translation to Spanish and adjustments; 15% among men), 16–24 years old (56%, compared with (4) pilot testing; (5) final adaptations based on user feed- 21% among 25–34 years old and 9% among 35–44 years back in pilot. old) and foreign born (24%, compared with 15% among 1. The Spanish version of the revised EPRES-2010 scale, individuals born in Sweden) are likely of holding a tempo- which has been published elsewhere,10 was translated rary employment.40 These are groups that reportedly are into Swedish. The translation was done by a bilingual exposed to high employment precariousness.19 Further, member of the research team (TB) in close discussion around 10% of employees in Sweden are not covered with AV, a native Spanish speaker with previous experi- by collective bargaining agreements and around 9% ence of validation studies of EPRES. have multiple jobs. The latter has seen an increase with 2. Several adaptations of the questionnaire were imple- 1.5 percentage points since 2005.42 According to a defi- mented in order to fit Swedish labour market condi- nition by the Swedish Labour Policy Council, the group tions. The questionnaire and its translation was dis- of atypical employees is constituted by those that fulfil cussed during workshops in the project team, which one of the following: not being covered by a collective consisted of Swedish, Spanish and Chilean researchers bargaining agreement, have a temporary employment, within public-—and occupational health; as well as are employed by a temp agency or are self-employed, have within the reference group involved in PREMIS, which their own company, hold multiple jobs or are working consisted of labour union members and workers with in the informal sector. This group of atypical workers is experience of PE. estimated to be around 35%–39% of the Swedish work- 3. Several drafts of the Swedish questionnaire were trans- force,42 and likely of experiencing precariousness. lated back to Spanish during the adaptation process Thereby, despite the stabile levels of temporary employ- and discussed until the final translation was decided ment in Sweden, it could be assumed that the level of on. precariousness has increased on the Swedish labour 4. A two-stage pilot testing was performed: first face-to- market. A study from Finland, however, indicate that the face with five volunteers from the reference group, precariat (as defined by atypical employment, previous and thereafter online with six volunteers who were unemployment, fear of job-loss, poor prospects of currently working but without a permanent full-time employment and low earnings) has not seen an increase employment. The latter were also asked to participate the past couple of decades.43 Without longitudinal studies in an evaluation of the survey either via the phone or with precise measures of PE assessing changes over time, online. evidence in Sweden remains inconclusive. 5. With the input from the pilot, a few minor adaptations To our knowledge, this is the first study in Sweden were made to the EPRES. aiming at translating and adapting a validated multidi- Further, in order to offer non-Swedish speaking partic- mensional measurement of PE, EPRES, to Swedish and ipants an opportunity to participate in the PREMIS-study, the Swedish context. It is an important step in the direc- the PREMIS-survey, including the Swedish version of tion to more precisely and fully comprehend the distribu- EPRES, was translated into English by an external, profes- tion and trends of PE in the Swedish population, as well as sional, translator. After the translation, minor changes in it will allow for future studies on health outcomes of PE. terms of style and terminology was made by the research Further, such an instrument will also enable well-needed group. The English translation of the EPRES-Se was not cross-country comparisons. validated in this study, nor has it been validated in any previous studies. Aim The aim of this study was to translate the EPRES-2010 into Implementation of the EPRES-Se Swedish, adapt it to the Swedish context, and to assess the PREMIS is an ongoing, longitudinal, web-based study psychometric properties of the Swedish EPRES. conducted in Stockholm county, Sweden. PREMIS aims at methodological development in terms of sampling strategies and assessment of PE, as well as at Method studying health outcomes of PE. In 2016–2017, 483 In order to fulfil the study aims, the EPRES-2010 was non-standard employees were sampled with web-based first translated into Swedish and subsequently adapted respondent-driven sampling (webRDS). WebRDS uses to suit the Swedish context (hereafter known as the peer-to-peer recruitment to build a sample from popu- Swedish version of the Employment Precariousness Scale, lations that are hard-to-reach and therefore lacks a EPRES-Se). Thereafter, EPRES-Se was piloted and imple- sampling frame.44 Questionnaire data were collected mented as a part of the survey used in the research project through an online survey tool specifically developed for PRecarious EMployment in Stockholm (PREMIS). respondent-driven sampling (RDS).45 The PREMIS-survey Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 3 Open access included all the items of the EPRES-Se, as well as ques- (design of survey and formulation of questions not other- tions on employment type, occupational environment, wise standardised) and data collection process (deciding health outcomes and background. The survey could be on the appropriate compensation for participation, completed in Swedish or English. recruitment of participants and testing the survey soft- As one of the aims of PREMIS was to sample precar- ware) through active discussions and workshops. Results ious employees with webRDS, a process which has been will be disseminated to study participants through the described elsewhere,46 participation in the study was website of the PREMIS-study. restricted to individuals considered particularly vulner- able to PE conditions, such as individuals with tempo- rary employment, on-demand employment, involuntary part-time employment and involuntary self-employment. Results The inclusion criteria for participants were: living and/ Adaptations of EPRES-Se or working in Stockholm County, being aged 18–65 years, The following adaptations were made to the EPRES-Se in having and indicating a Swedish personal identification comparison with EPRES-2010. number, and having a current employment. Exclusion 1. The response alternatives in ‘temporariness’ were stat- criteria were: having a fixed, full-time, employment, being ed as categories, as opposed to free text values of num- voluntarily self-employed or being a student. Out of the ber of days, months or years in EPRES-2010, in order 483 participants included in the sample, 68 participants to increase usability in the web-survey context. For the were excluded due to not matching criteria of county same reason, the question on contract duration which (n=6), re-using or giving an incorrect personal number contained three dependent items in EPRES-2010, was (n=8 and n=17, respectively), being underage (n=1) or collapsed into one question in EPRES-Se. In addition, suspected cheating (ie, systematic repeated participation; the response options ‘I do not have a contract’ was n=36), giving a final sample of 415 participants. added in order to capture informal work and ‘Do not know’ was added in order to capture poor contractual Statistical analysis relationship. Item descriptive statistics (mean, SD, response frequen- 2. In ‘wages’, the question on income was presented in cies, missing responses and Pearson item-subscale correla- local currency (Swedish krona, SEK) and intervals tions) and scale descriptive statistics (mean, SD, missing were set to ~€300 as the €150–200 intervals used in items, range, floor and ceiling effects, and reliability the EPRES-2010 version were perceived as too narrow (Cronbach’s alpha coefficient)) were assessed for the in the Swedish context. In EPRES-2010, intervals were entire sample. Participants answering ‘No answer’ on the €150–€200 below a monthly income of €1200, and question on income (question three in EPRES-Se) were €300 above an income of €1201. excluded from the analyses of this question due to the 3. In EPRES-2010, the two response alternatives captur- ambiguity of the response alternative. Exploratory factor ing working conditions that were decided unilaterally analysis was conducted in order to determine the under- by the employer, were merged into one response op- lying factor structure of the data. Principal axis factoring, tion in the Swedish adaptation in order to enhance with varimax rotation, extracting eigenvalues >1 was used, usability. Further, as some workers, especially freelanc- which is the method used in previous EPRES-studies.8–10 ers, are given a fixed budget with no hours specified, Subanalyses were conducted without informal workers, the response alternative ‘Not applicable. I work proj- ie, participants answering ‘I have no contract’ to the ques- ect-based’ was added for the question on how salary tion on contract duration (question one in EPRES-Se). was decided on. Further, as the sample was recruited with RDS, weighted 4. One of the items in ‘vulnerability’, ‘afraid to demand analyses were conducted in addition to the unweighted better working conditions …’, was taken from the orig- analyses. RDSII weights47 were calculated in RDS Analyst inal EPRES8 according to the recommendations made 0.42 for Windows (Los Angeles, California, USA). In by Vives et al.10 short, weights are based on the self-reported network size 5. In ‘rights’, the question on pension in EPRES-2010, in the target population (degree) of the participant and which contained both pension due to old age and dis- applied in order to account for over-sampling of individ- ability, was split in two as these are distinct systems uals included in large social networks. Participants with unrelated to one another in the Swedish context. large social networks are given a smaller weight, and vice Retirement pension (ie, pension due to old age) was versa.47 48 The results from the weighted analyses will be kept in EPRES-Se and disability pension was removed. shown in full in the online supplementary material. All However, a new item assessing the right to sickness analyses were performed using SPSS V.23. benefit was added in the subscale instead, capturing both long term sick leave and shorter spells of sick- Patient and public involvement ness absence. In PREMIS, a reference group consisting of individuals 6. In ‘exercise of rights’ the item on taking a day off for with experience from PE and labour union representa- family reasons was clarified by adding ‘care of a sick tives, was involved in the design of the PREMIS-survey child, care of a sick relative and so on’ within brackets. 4 Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 Open access EPRES-Se Table 1 Demographic characteristics of the study The final version of EPRES-Se consisted of a total of 23 population (frequencies and percentages), n=415 items and six dimensions: ‘temporariness’ (two items), ‘wages’ (three items), ‘disempowerment’ (two items),   N % ‘vulnerability’ (five items), ‘rights’ (five items) and Age ‘exercise rights’ (six items). As in the EPRES-2010, the  18–24 122 29 response scales were 5-point ordinal scales for ‘tempo-  25–29 185 45 rariness’, 5-point ordinal and 5-point frequency scales  30–64 108 26 for ‘wages’, 5-point frequency scales for ‘vulnerability’ and ‘exercise rights’, and 5-point and 3-point categorical Sex scales for ‘disempowerment’ and ‘exercise rights’, respec-  Men 190 46 tively. See EPRES-Se in online supplementary material A.  Women 225 54 Employment Coding of EPRES-Se  Temporary employment 121 29 Similarly as EPRES-2010, subscale scores were computed as averages and transformed into a 0–4 scale. The  On demand/by the hour 243 59 employment global EPRES score is the average of the six subscales, ranging from 0 to 4, where 0 represent the lowest level  Self-employment (involuntary) 13 3 of precariousness and 4 represent the highest level of  Intern 2 1 precariousness.  Part-time employment 36 9 For questions in ‘temporariness’, response options (involuntary) were coded slightly different compared with EPRES- 2010 in order to accommodate the changes made in the Swedish version. For instance, in the question on duration 0 to 3 (ie, the same coding as the response alternative of contract, response options ‘Do not have a contract’ <1-month contract). Those with any other employment and ‘Do not know’ were coded as 4 (most precarious), type and an indefinite contract (n=36) remained coded in comparison with the Spanish version where a contract as 0. See results of subscale-average comparisons in online length of less than 6 months was coded as 4. In the supplementary material C. question on income, intervals were larger and consis- tent in size, as compared with the Spanish version. The Psychometric properties of EPRES-Se cut-offs for income were based on the Swedish median The demographic characteristics of the sample is shown net income for 2016, which was just above SEK 18 000 in table 1. The sample consisted of a larger propor- for individuals 20–64 years of age.49 Merging of items in tion 25–29 years old (45%) compared with the other ‘disempowerment’ did not affect the coding. The coding age-groups, and a larger proportion women (54%) as of the individual items along with the English translation compared with men. Further, the sample was dominated of EPRES-Se can be found in the online supplementary by workers employed on demand/by the hour (59%). material B. Table 2 shows the item-descriptive statistics. There was a small proportion of missing values (<3%). Item means Issues leading to recoding of the ‘temporariness’ dimension were similar within subscales, with the greatest mean differ- During the data analysis phase, it was revealed that 79% ence found within ‘wages’ (item mean difference=1.6). (n=139) of the respondents answering ‘indefinitely’ to the All response options within the items were used by partici- EPRES question on contract length (‘How long is your pants, although to a varying extent. Item-subscale correla- current employment contract valid?’) also answered that tions were around 0.6–0.8 in ‘vulnerability’ and ’exercise they were employed on demand/by the hour on a ques- of rights’; and around 0.4–0.6 in ‘wages’ and ‘rights’. tion assessing employment type included in the PREMIS- There was a weak correlation between item and subscale survey. We suspected that this combination could be a in ‘disempowerment’ and no correlation between items type of ‘zero hour’ contract, in which the employer is in ‘temporariness’. With exception of the latter, all items not required to offer the employee any fixed number of correlated higher with their corresponding subscale hours of work at all per day, week or month.38 Thereby, compared with other subscales. this type of employment situation could be regarded as Table 3 shows the scale descriptive statistics. The contingent with a high degree of precariousness. This was subscale mean scores ranged between 1.4 and 2.3, with confirmed as the group indicating an indefinite contract a global average of 1.9. The proportion of participants length and on-demand/by the hour employment, were with any missing values in the subscales were around 1%, most similar (in terms of the other EPRES subscales) to except in the case of ‘wages’ where it was 3.1%. The latter employees with a contract lasting < 1 month and least also included participants answering ‘No answer’ (n=9). alike employees with a fixed-term contract >2 years. Subscale scores ranged between 0 and 4, and global Consequently, we recoded the group with an indefinite scale score ranged between 0.09 and 3.07. Both floor contract and on demand/by the hour-employment from and ceiling effects were generally low (<5%), with floor Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 5 6 Table 2 Item descriptive statistics and Pearson item-subscale correlations of EPRES-Se Response value frequency (%)† Pearson item subscale correlations‡ Item Missing %* Mean SD 0 1 2 3 4 T W D V R ER Temporariness§ Open access  Length of contract 0.0 2.6 1.1 36 (8.7) 24 (5.8) 75 (18.1) 228 (54.9) 52 (12.5) −0.04** 0.04 0.05 0.06 0.08 0.04  Time working 0.0 2.1 0.9 17 (4.1) 92 (22.2) 171 (41.2) 119 (28.7) 16 (3.9) −0.04** 0.06 0.11 −0.04 0.31 0.13 employer Wages  Income monthly 2.2 2.4 1.2 43 (10.6) 45 (11.1) 87 (21.4) 171 (42.1) 60 (14.8) 0.09 0.39¶ 0.14 −0.07 0.29 0.17  Cover basic need 1.0 0.8 0.9 175 (42.6) 171 (41.6) 44 (10.7) 14 (3.4) 7 (1.7) −0.02 0.60¶ 0.07 0.21 0.13 0.26  Cover unforeseen 1.0 1.8 1.1 60 (14.6) 113 (27.5) 136 (33.1) 70 (17.0) 32 (7.8) 0.08 0.55¶ 0.08 0.39 0.14 0.26 expenses Disempowerment  Working hours settled 0.0 2.1 1.2 84 (20.2) 8 (1.9) 110 (26.5) 190 (45.8) 23 (5.5) 0.12 0.17 0.34¶ 0.10 0.26 0.23  Salary settled 0.0 2.2 1.2 77 (18.6) 8 (1.9) 101 (24.3) 210 (50.6) 19 (4.6) 0.07 0.03 0.34¶ 0.02 0.13 0.09 Vulnerability  Demand work 1.0 1.6 1.2 79 (19.2) 129 (31.4) 113 (27.5) 63 (15.3) 27 (6.6) 0.01 0.13 0.09 0.75¶ 0.04 0.28 conditions  Unfair treatment 1.0 1.3 1.1 114 (27.7) 159 (38.7) 67 (16.3) 50 (12.2) 21 (5.1) −0.03 0.12 0.08 0.78¶ −0.01 0.26  Afraid fired 1.0 1.5 1.2 106 (25.8) 130 (31.6) 81 (19.7) 60 (14.6) 34 (8.3) 0.06 0.20 0.10 0.78¶ 0.04 0.35  Treated authoritarian 1.0 1.3 1.1 107 (26.0) 160 (38.9) 87 (21.2) 36 (8.8) 21 (5.1) −0.04 0.20 −0.01 0.63¶ 0.02 0.23  Easily replaced 1.0 1.6 1.3 91 (22.1) 124 (30.2) 92 (22.4) 55 (13.4) 49 (11.9) 0.06 0.22 0.02 0.74¶ 0.04 0.27 Rights  Right parental leave 1.2 1.2 0.8 100 (24.4) 130 (31.7) 180 (43.9) 0.18 0.16 0.26 0.05 0.58¶ 0.16  Right retirement 1.2 1.2 0.8 107 (26.1) 113 (27.6) 190 (46.3) 0.26 0.22 0.23 0.07 0.60¶ 0.16  Right unemployment 1.2 1.0 0.9 156 (38.0) 100 (24.4) 154 (37.6) 0.15 0.24 0.15 −0.01 0.60¶ 0.08  Right severance pay 1.2 1.4 0.6 20 (4.9) 223 (54.4) 167 (40.7) 0.11 0.06 0.03 −0.02 0.48¶ 0.03  Right sickness 1.2 0.9 0.8 160 (39.0) 110 (26.8) 140 (34.1) 0.26 0.18 0.17 0.03 0.53¶ 0.12 benefits Exercise of rights  Take weekend off 1.0 1.6 1.1 86 (20.9) 93 (22.6) 154 (37.5) 48 (11.7) 30 (7.3) −0.01 0.23 0.18 0.18 0.10 0.62¶  Take vacation 1.2 1.9 1.1 50 (12.2) 88 (21.5) 163 (39.8) 77 (18.8) 32 (7.8) 0.14 0.22 0.20 0.30 0.15 0.68¶  Take day off 1.2 1.6 1.1 88 (21.5) 85 (20.7) 166 (40.5) 50 (12.2) 21 (5.1) 0.15 0.19 0.21 0.24 0.16 0.75¶  Take day off, pers. 1.0 1.9 1.2 63 (15.3) 78 (19.0) 159 (38.7) 69 (16.8) 42 (10.2) 0.11 0.21 0.08 0.36 0.07 0.74¶  Sick leave 1.2 1.6 1.1 98 (23.9) 74 (18.0) 163 (39.8) 50 (12.2) 25 (6.1) 0.06 0.22 0.20 0.26 0.12 0.69¶ Continued Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 Open access effects being highest for ‘disempowerment’ (9.2%) and 0.64¶ ‘vulnerability’ (11.7%), and ceiling effects being highest ER for ‘rights’ (18.0%). Cronbach’s alpha coefficients were around 0.7 or higher for ‘wages’, ‘vulnerability’, ‘rights’ 0.14 and ‘exercise of rights’. Only ‘temporariness’ exhibited a Pearson item subscale correlations‡ Cronbach’s alpha coefficient close to 0. The global alpha R †Frequency (%) of participants indicating the specific response. 0 indicates the lowest precariousness score, 4 indicates the highest precariousness score. coefficient was 0.83. 0.25 The exploratory factor analysis extracted six factors with eigenvalues >1 (eigenvalues=5.3; 3.2; 2.3; 1.6; 1.3; V 1.1). The emerging factors were the same as in EPRES- 2010, thereby confirming the original factor structure. 0.09 Together, the six factors explained 64.1% of the vari- D ance. The six factors and their rotated factor loadings are shown in table 4. All loadings were above 0.35, except in 0.26 the case of ‘length of contract’. W Subanalyses were conducted in order to investigate the potential effect of including informal workers in 0.11 the sample (n=35), which had minor influence on the T correlation between temporariness-items (r=0.002) and reliability of the subscale (−0.003). However, in the factor 14 (3.4) analysis seven factors with eigenvalues >1 emerged (5.3; 3.2; 2.3; 1.6; 1.3; 1.1; 1.0), explaining 68.5% of the vari- 4 ance. The seventh factor was caused by a split of the temporariness dimension, grouping items on length 25 (6.1) of contract (duration) and time working for employer 3 (tenure) in separate factors, with factor loadings of 0.33 D, disempowerment; ER, exercise rights; R, rights; T, temporariness; V, vulnerability; W, wages. and 0.42, respectively. 106 (25.8) Analyses for the weighted population sample resulted 0 indicates the lowest precariousness score, 4 indicates the highest precariousness score. Response value frequency (%)† in virtually the same results in regard to item-subscale 2 correlations, Cronbach’s alpha coefficients, floor and ceiling effects, as well as factor loadings from the explor- atory factor analysis of EPRES-Se, and did thereby not 126 (30.7) ‡Corrected for overlap, ie, item is removed from the corresponding subscale. affect the interpretation of the results. See the weighted results in online supplementary material 1. 1 140 (34.1) Discussion Key findings and summary 0 EPRES-Se generally performed well, with a small propor- tion of missing values across all subscales, usage of all SD 1.1 response options and good global reliability. The factor *Proportion of participants with any missing item. structure established in the Spanish EPRES-201010 Missing %* Mean was confirmed. The subscales ‘vulnerability’, ‘wages’, 1.1 ‘rights’ and ‘exercise of rights’ generally worked well in the Swedish context, with high item-subscale correla- tions, subscale reliability and factor loadings. However, ‘temporariness’ did not perform as expected and would need revision. In addition, although ‘disempowerment’ 1.0 showed acceptable psychometric properties, the subscale might benefit from additional adaptation. §Inter-item correlations. Continued ‘Temporariness’ As opposed to previous studies,8–10 temporariness yielded  Go to doctor very poor psychometric properties. The items in the ¶ p <0.001 subscale did not correlate, there was a negative Cron- **p >0.05 Table 2 bach’s alpha coefficient and a low factor loading for Item the item on contract duration. Only minor changes in terms of correlation and Cronbach’s alpha was initiated Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 7 Open access Table 3 Scale descriptive statistics: range, mean, SD, floor and ceiling effects and Cronbach’s alpha coefficient Cronbach’s Subscale Items Mean SD Missing (%)* Obs. range Floor %† Ceiling %† alpha Temporariness 2 2.3 0.7 0.0 0–4 0.5 1.0 −0.08 Wages 3 1.6 0.8 3.1 0–4 4.2 0.7 0.69 Disempowerment 2 2.1 1.0 0.0 0–4 9.2 1.4 0.50 Vulnerability 5 1.4 1.0 1.0 0–4 11.7 1.9 0.89 Rights 5 2.3 1.2 1.2 0–4 2.9 18.0 0.78 Exercise of rights 6 1.6 0.9 1.2 0–4 6.8 1.0 0.88 EPRES-Se 23 1.9 0.5 3.4 0.09–3.07 0.2 0.2 0.83 *Proportion of participants with any missing item. †Proportion of participants with lowest (floor) and highest (ceiling) EPRES-Se scores. by the removal of informal workers. This subanalysis did, to be dependent on context. Given the proposed mech- however, split the temporariness dimension in two sepa- anism of temporariness leading to ill-health mediated rate factors, both of which had higher factor loadings. On via feelings of insecurity, temporariness is most relevant the basis of these results, we believe that the temporari- in labour markets which do not offer regulatory protec- ness dimension needs further development and evalua- tion for certain groups of workers, such as permanent tion in a population of both standard and non-standard employees in most European countries, which does not employees. Based on the results from the present study, apply to the same extent in the USA.51 we offer the following thoughts on this matter. However tempting it would be suggest the inclusion First, it is important to acknowledge the sample of questions regarding ‘contract type’ in a future devel- selection. As the sample was restricted to non-standard opment of EPRES, the continuous flexibilisation of the employees (ie, permanent, full-time, employees were labour market and fast changes in employment practices excluded), the lower end of the precariousness scale in combination with contextual differences, makes it had a smaller proportion than what would be expected increasingly difficult—at least if international compar- if standard employees with longer duration and tenure ison using similar scales is sought after. would have been included in the sample. This limitation From a mechanistic standpoint and with an aspiration is likely to have contributed to the lack of correlation to develop a scale which could be used in international between the items. However, considering that the sample comparison independent of context, we believe that an was intentionally recruited in order to capture a popu- item that measures the future employment opportuni- lation of precariously employed individuals, the poor ties with the current employer as objectively as possible psychometric properties of temporariness also shows that should be developed. EPRES-Se and other translations these items are not necessarily related in a meaningful could further be adapted, for example, by combining way when measuring precariousness among non-standard the contract duration and tenure-items with an item employees. assessing the number of repeated contracts with the Second, in the current context, Swedish legisla- same employer or an item assessing how often during a tion (the Employment Protection Act SFS 1982:80) specific time interval the employment contract is up for prevents an employer to hire an employee for more renewal. Contract duration could also be complemented than 2 years during a 5-year period (consecutive or in or replaced by a question more explicitly assessing shorter repeated contracts) without having to employ the remaining duration of the contract at the time of (or dismiss) the employee on a permanent contract.50 answering the survey. Thereby, an employee with an 18-month tenure might be more precarious than an employee with a 6-month tenure ‘Wages’ as the latter has longer time left before being forced in or The income-item correlated moderately with the other out. Further, approximately 50% of temporary employees items in the subscale (items assessing how well the income in Sweden has had repeated contracts with the same covers basic needs and unforeseen expenses). Previous employer,40 which is an additional reason that could studies report similar findings for this item.8–10 One contribute to the lack of correlation between tenure and explanation for these results is the quantitative nature of duration of contract. the income-item compared with the other items. Slightly Thirdly, we found that several participants were more than half, 57%, of the participants scored a precari- employed by the hour or on demand while still indicating ousness level of 3 or 4 on income, which is not surprising an indefinite contract length. This highlights the difficul- as the sample is constituted by young non-standard ties in assessing temporariness only by contract duration employees. However, only 5% and 25% of the sample (and tenure). Temporariness is the dimension most likely scored a precariousness level of 3 or 4 on the other two 8 Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 Open access Table 4 Factor loadings from exploratory factor analysis of the EPRES-Se Factor Exercise rights Vulnerability Rights Wages Disempowerment Temporariness Temporariness  Length of contract 0.12  Time working 0.12 0.36 −0.36 employer Wages  Income monthly 0.14 −0.16 0.23 0.49 0.35  Cover basic need 0.17 0.13 0.71  Cover unforeseen 0.12 0.32 0.79 −0.10 expenses Disempowerment              Working hours settled 0.17 0.21 0.52 0.15  Salary settled 0.60 Vulnerability              Demand work 0.14 0.80 conditions  Unfair treatment 0.11 0.84  Afraid fired 0.20 0.80  Treated authoritarian 0.11 0.66 0.12 0.13  Easily replaced 0.12 0.77 0.15 Rights  Right parental leave 0.67 0.21  Right retirement 0.69 0.16  Right unemployment 0.69 0.11 0.29  Right severance pay 0.55 −0.10  Right sickness 0.60 0.11 benefits Exercise rights  Take weekend off 0.65 0.10 0.14  Take vacation 0.70 0.18 0.12 −0.13  Take day off 0.80 0.10 0.10 0.13 −0.13  Take day off, pers. 0.78 0.24 −0.10  Sick leave 0.72 0.11 0.12 0.15  Go to doctor 0.66 0.11 0.13 Table showing factor loadings >0.1. items, respectively. How well one can cover basic needs would expect a larger proportion of participants with a and unforeseen expenses could depend on more than high income. income, such as family support. Approximately 24% of Aside from the income-item, item-subscale correla- adults aged 20–27 years in Sweden still live at home. In the tions for the remaining two items and subscale reliability majority of municipalities in large city regions, this figure were acceptable and only a fraction of the sample did not is more than 50%. About half of those living at home pay provide an answer to the item on income (2.2%). There- nothing in rent.52 As 74% of the sample is between 18 and fore, as in the other EPRES scales, we believe the subscale 29 years old, it is likely that at least a part of the partici- can be used in its current form in future studies. pants still live at home and receive help from their family. ‘Disempowerment’ Including standard employees in the sample could poten- The items in ‘disempowerment’ had acceptable item-sub- tially have increased the item-subscale correlation as we scale correlations and reliability. However, some response Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 9 Open access options (‘my working hours/salary was decided within my response options (for all items) and few missing values. working team’ and ‘do not know’) were hardly used at all The psychometric properties for five out of six subscales in, indicating that these options might not be appropriate were satisfactory, considering this being the first transla- for the current population in the Swedish context. Further, tion and adaptation. However, one subscale, ‘temporar- the remaining response options were also inadequate from iness’, worked poorly and would need revision before an adaptational point of view. For example, the working implementing the scale in further research. As employ- hours/salary being in line with collective bargaining agree- ment precariousness is an emergent determinant of ments does not imply that the working hours/salary was not health it is important that PE can be properly measured. set by the employer; these options are not mutually exclu- The EPRES-Se is an important step in this direction. We sive. In a revised EPRES-Se, the disempowerment-items therefore encourage others to continue working with would benefit from revision and clarification. Combining EPRES-Se and to validate it further in populations of both response options not mutually exclusive could be consid- standard and non-standard employees. Our hope is for ered as one way of improving the subscale, such as ‘my a revised EPRES-Se with satisfactory psychometric prop- working hours are in accordance with Swedish law and the erties to be implemented in research both as an inde- collective agreement, decided by my employer’ and ‘my pendent tool and within nation-wide surveys. In order working hours are not in accordance with Swedish law and to enable international comparisons and multinational the collective agreement, decided by my employer’. studies, similar studies in other European countries are also called for. ‘Rights’ The new item assessing sickness benefit had an accept- Author affiliations able item-subscale correlation and factor loading, similar 1 Institute of Environmental Medicine, Unit of Occupational Medicine, Karolinska to the other item-subscale correlations and factor load- Institutet, Stockholm, Sweden 2 ings in the subscale, Further, the subscale reliability was Department of Public Health, Pontificia Universidad Catolica de Chile Escuela de good. Taken together, these results point towards that the Medicina, Santiago, Chile 3 Department of Political and Social Sciences, Health Inequalities Research Group, new item worked well in the subscale. Employment Conditions Knowledge Network (GREDS-EMCONET), Universitat Pompeu Fabra, Barcelona, Spain Strengths and limitations 4 Transdisciplinary Research Group on Socioecological Transitions (GinTRANS2), This study finds strength in the fact that it is the first study Universidad Autonoma de Madrid, Madrid, Spain translating and adapting EPRES-2010 to the Swedish 5 Center for Occupational and Environmental Medicine, Stockholm County Council, context, as well as the first study to assess the psycho- Stockholm, Sweden metric properties of the resulting EPRES-Se. In addition, this work provides context-specific recommendations for Acknowledgements The authors are grateful to the PREMIS-project members and reference group members who provided valuable input during the formative work future research using EPRES-Se. This study is, however, of the study. not without limitations. Contributors The study was conceived by TB and GJ. All authors, TB, GJ, JJ, AV, The main limitation is the sample. As this study lies JB, KK and JS, contributed to the design of the study. TB translated and back- within the frame of the PREMIS-project,46 the sample was translated the EPRES, and AV supported the translation process and examined restricted to employees with a non-standard employment. the back translation. JJ collected the data with support from TB and GJ. JJ and EPRES, however, is developed to measure precariousness TB conducted the analyses. JJ wrote the draft(s) of the manuscript. All authors reviewed the draft(s) and provided feedback on its content. All authors approved the independent of the type of employment.8 By only assessing contents of the final manuscript. the psychometric properties of the scale in a population Funding This work was supported by grant 2015-00933 from The Swedish of non-standard employees, the scale properties cannot be Research Council for Health, Working Life and Welfare (FORTE). The funder had no directly compared with similar studies as the heterogeneity role in study design, data collection and analysis, decision to publish, or preparation of the sample is limited. A sample representative of the of the manuscript. entire work force would have provided better insights as to Competing interests None declared. how the scale behaves among Swedish employees. A next Patient consent for publication Informed consent was attained from all step which this study has provided strong grounds for. participants by the respondent clicking ‘Yes’ to the question ‘I understand the Finally, the relatively small sample size could influence information given above and want to participate’. the reliability of the results, which should be kept in mind Ethics approval Permission from the Regional Ethics Committee of Stockholm when interpreting and generalising the results. Further, was given for the study, with dnr: 2016/1291-31/5. being a convenience sample limits the generalisability of Provenance and peer review Not commissioned; externally peer reviewed. the results. However, as the weighted results confirmed Data availability statement Deidentified participant data are from the PREMIS the psychometric properties of the scale, we could expect study whose authors may be contacted at ​johanna.​jonsson@​ki.​se. Data cannot be made publicly available for ethical and legal reasons, but could be made available similar results in a representative sample applying the to researchers who meet the criteria for access to confidential data after approval same sample restrictions. from the Regional Ethics Committee of Stockholm. Open access This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, Conclusion which permits others to distribute, remix, adapt, build upon this work This study found that EPRES-Se worked well in the current non-commercially, and license their derivative works on different terms, context, with high global reliability, endorsement of all provided the original work is properly cited, appropriate credit is given, 10 Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 Open access any changes made indicated, and the use is non-commercial. See: http://​ 26. Oke A, Braithwaite P, Antai D. Sickness absence and precarious creativecommons.​org/​licenses/​by-​nc/​4.​0/. employment: a comparative cross-national study of Denmark, Finland, Sweden, and Norway. Int J Occup Environ Med 2016;7:125–47. 27. Rönnblad T, Grönholm E, Jonsson J, et al. Precarious employment and mental health: a systematic review and meta-analysis of References longitudinal studies. Scand J Work Environ Health 2019. 1. CSDH. Closing the gap in a generation: health equity through action 28. Koranyi I, Jonsson J, Rönnblad T, et al. Precarious employment and on the social determinants of health. Final report of the Commission occupational accidents and injuries – a systematic review. Scand J on social determinants of health. Geneva: World Health Organization, Work Environ Health 2018;44:341–50. 2008. 29. Kim TJ, von dem Knesebeck O, insecurity Pjob. Perceived job 2. Benach J, Vives A, Amable M, et al. Precarious employment: insecurity, unemployment and depressive symptoms: a systematic understanding an emerging social determinant of health. Annu Rev review and meta-analysis of prospective observational studies. Int Public Health 2014;35:229–53. Arch Occup Environ Health 2016;89:561–73. 3. Benach J, Muntaner C, Solar O, et al. Employment, work, and health 30. Virtanen M, Nyberg ST, Batty GD, et al. Perceived job insecurity as a inequalities: a global perspective. Geneva: WHO, 2007. risk factor for incident coronary heart disease: systematic review and 4. Benach J, Muntaner C. Precarious employment and health: meta-analysis. BMJ 2013;347. developing a research agenda. J Epidemiol Community Health 31. Kivimäki Met al. Temporary employment and risk of overall and 2007;61:276–7. cause-specific mortality. Am J Epidemiol 2003;158:663–8. 5. Quinlan M, Mayhew C, Bohle P. The global expansion of precarious 32. De Witte H, Pienaar J, De Cuyper N. Review of 30 years of employment, work disorganization, and consequences for longitudinal studies on the association between job insecurity and occupational health: placing the debate in a comparative historical health and well-being: is there causal evidence? Aust Psychol context. Int J Health Serv 2001;31:507–36. 2016;51:18–31. 6. Benach J, Vives A, Tarafa G, et al. What should we know about 33. Vives A. A multidimensional approach to precarious employment: precarious employment and health in 2025? framing the agenda for measurement association with poor mental health and prevalence in the next decade of research. Int J Epidemiol 2016;45:232–8. the Spanish workforce. Universitat Pompeu Fabra, 2010. 7. Howard J. Nonstandard work arrangements and worker health and safety. Am J Ind Med 2017;60:1–10. 34. Waenerlund A-K, Virtanen P, Hammarström A. Is temporary 8. Vives A, Amable M, Ferrer M, et al. The employment precariousness employment related to health status? analysis of the Northern scale (EPRES): psychometric properties of a new tool for Swedish cohort. Scand J Public Health 2011;39:533–9. epidemiological studies among waged and salaried workers. Occup 35. Waenerlund A-K, Gustafsson PE, Virtanen P, et al. Is the core- Environ Med 2010;67:548–55. periphery labour market structure related to perceived health? 9. Vives-Vergara A, González-López F, Solar O, et al. Precarious findings of the Northern Swedish cohort. BMC Public Health employment in Chile: psychometric properties of the Chilean version 2011;11:956. of employment precariousness scale in private sector workers. Cad 36. Virtanen P, Janlert U, Hammarström A. Exposure to temporary Saude Publica 2017;33:e00156215. employment and job insecurity: a longitudinal study of the health 10. Vives A, González F, Moncada S, et al. Measuring precarious effects. Occup Environ Med 2011;68:570-4. employment in times of crisis: the revised employment 37. Canivet C, Bodin T, Emmelin M, et al. Precarious employment is a precariousness scale (EPRES) in Spain. Gac Sanit 2015;29:379–82. risk factor for poor mental health in young individuals in Sweden: 11. Vives A, Amable M, Ferrer M, et al. Employment precariousness and a cohort study with multiple follow-ups. BMC Public Health poor mental health: evidence from Spain on a new social determinant 2016;16:687. of health. J Environ Public Health 2013;2013:978656 38. International Labour Organization. Non-Standard employment around 12. Benach J, Julià M, Tarafa G, et al. La precariedad laboral medida the world: understanding challenges, shaping prospects. Geneva, de forma multidimensional: distribución social Y asociación Con La 2016. salud en Cataluña. Gac Sanit 2015;29:375–8. 39. Alfonsson J. Nomaderna på den svenska arbetsmarknaden. Det 13. Arnold D, Bongiovi JR. Precarious, informalizing, and flexible otrygga livets framväxt och effekter på klasstrukturen [In Swedish], work: transforming concepts and understandings. Am Behav Sci 2018. 2012;57:289–308. 40. Sweden S. Trends for persons in temporary employment [In 14. Broughton A, Green M, Rickard C, et al. Precarious employment in Swedish], 2015. Europe. Strasbourg: European Parliament, 2016. 41. Eurofund. Non-Standard forms of employment: recent trends and 15. Muntaner C, Platforms D, Economy G. Digital platforms, Gig future prospects. International Publications, 2017. economy, precarious employment, and the invisible hand of social 42. The Swedish Labour Policy Council (AER). Olika vägar till jobb [In class. Int J Health Serv 2018;48:597–600. Swedish] 2018. 16. Vosko LF, Zukewich N, Cranford C. Precarious jobs: a new typology 43. Pyöriä P, Ojala S. Precarious work and intrinsic job quality: evidence of employment. Perspect labour income 2003;15. from Finland, 1984–2013. The Economic and Labour Relations 17. Rodgers G, Rodgers J. Precarious jobs in labour market regulation: Review 2016;27:349–67. the growth of atypical employment in Western Europe. International 44. Salganik MJ, Heckathorn DD. 5. Sampling and estimation in hidden Labour Organisation, 1989. populations using Respondent-Driven sampling. Sociol Methodol 18. Standing G. Global labour flexibility: seeking distributive justice. 2004;34:193–240. London: Macmillan, 1999. 45. Bengtsson L, Lu X, Nguyen QC, et al. Implementation of web-based 19. Julià M, Vanroelen C, Bosmans K, et al. Precarious employment and respondent-driven sampling among men who have sex with men in quality of employment in relation to health and well-being in Europe. Vietnam. PLoS One 2012;7:e49417. Int J Health Serv 2017;47:389–409. 46. Jonsson J, Stein M, Johansson G, et al. A performance 20. Scott-Marshall H, Tompa E. The health consequences of precarious assessment of web-based respondent driven sampling among employment experiences. Work 2011;38:369–82. workers with precarious employment in Sweden. PLoS One 21. International Labour Organization. From precarious work to decent 2019;14:e0210183. work: outcome document to the workers’ symposium on policies and 47. Volz E, Heckathorn DD. Probability based estimation theory for regulations to combat precarious employment. Geneva: ILO, 2012. Respondent driven sampling. J Off Stat 2008;24:79–97. 22. Julià M, Vives A, Tarafa G, et al. Changing the way we understand 48. Johnston LG, Sabin K. Sampling hard-to-reach populations with precarious employment and health: precarisation affects the entire Respondent driven sampling. Methodological Innovations Online salaried population. Saf Sci 2017;100:66–73. 2010;5:38.1–48. 23. Lewchuk W. Precarious jobs: where are they, and how do they affect 49. Statistics Sweden. Inkomstrapport 2016 Hushåll och Individer [In well-being? Econ Labour Relat Rev 2017;28:402–19. Swedish], 2016. 24. Kretsos L, Livanos I. The extent and determinants of precarious 50. Svensk författningssamling. Lagen(1982:80) om anställningsskydd employment in Europe. Int J Manpow 2016;37:25–43. (SFS 2017:363), 2017. 25. Van Aerden K, Moors G, Levecque K, et al. Measuring employment 51. OECD. Detailed description of employment protection legislation, arrangements in the European labour force: a typological approach. 2012-2013. OECD countries, 2013. Soc Indic Res 2014;116:771–91. 52. Hyresgästföreningen. Unga vuxnas boende [In Swedish], 2018. Jonsson J, et al. BMJ Open 2019;9:e029577. doi:10.1136/bmjopen-2019-029577 11