Trajectories of sickness absence in healthcare workers: a cohort study in a Spanish hospital between 2018 and 2023 Authors: Carlos Lopez-Muley, Fernando G Benavides, José María Ramada, Consol Serra, Laura Serra, Mireia Utzet Published: 2026 ABSTRACT BackgroundSickness absence (SA) is a key indicator of health and job performance. The COVID-19 pandemic increased workload and stress, potentially affecting healthcare workers’ (HCWs) health and retention. This study aimed at identifying vulnerable HCW groups by analysing SA trajectories before and after the pandemic.MethodsA retrospective longitudinal study was conducted using administrative data from HCWs at Hospital del Mar, Barcelona, Spain, 2018–2023. SA was measured as cumulative days per semester. Latent Class Growth Analysis with zero-inflated Poisson models identified SA trajectories. The associations between socio-demographic and employment variables (occupation, shifts, care setting) and SA trajectories was analysed using bivariate analysis and multinomial logistic regression, yielding adjusted odds ratios (aOR) and 95% confidence intervals (95% CI).ResultsAn overall increase in accumulated SA days per semester was observed among HCWs between 2018 and 2023. Five SA trajectories were identified: zero (7.8%), low (53.7%), degrowing (16.4%), growing (15%), and high (7%). Women were more likely to belong to growing (aOR: 1.54; 95% CI: 1.16–2.05) and high (aOR: 1.77; 95% CI: 1.14–2.74) trajectories. Aides were strongly associated with the high trajectory (aOR: 30.7; 95% CI: 9.55–98.7). Working in non-direct care areas was associated with the zero trajectory (aOR: 2.86; 95% CI: 1.79–4.58) and negatively to the high trajectory (aOR: 0.54; 95% CI: 0.31–0.93). Night shifts were associated with the high trajectory (aOR: 1.78; 95% CI: 1.21–2.60).ConclusionsSA among HCWs increased 2018–2023, following distinct trajectories related to gender, care level, and occupation. Persistent pandemic effects and key sociodemographic and work-related factors shape these patterns. Targeted occupational health interventions could mitigate SA in high-risk groups and enhance healthcare system resilience.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12913-026-14821-7. Background Healthcare workers (HCWs) are pivotal assets of health systems but also a vulnerable occupational group exposed to specific hazards [1]. HCWs are daily exposed to heavy workloads, unpredictably long working hours, high patient involvement, and low job control [2, 3] leading to elevated rates of musculoskeletal injuries [4], mental health problems [5], and sickness absence (SA) [2, 3]. Moreover, failure to ensure the health and safety of HCWs has broader system-level impacts, including reduced health services satisfaction, patient safety breaches and increased financial costs [1, 6] thereby hindering staff retention and exacerbating staff shortages [7]. The COVID-19 pandemic highlighted the critical role of HCWs as essential workers. In turn, the pandemic entailed a major psychosocial and occupational disruption that strained the occupational health of HCWs [5]. Not only was there an elevated risk of COVID-19 infection, but also increased workloads, longer working hours, and larger volumes of new information and practices [8, 9] became major occupational hazards that HCWs had to cope with. Work-related distress among HCWs in that time has been linked to recent surges in rates of burnout, and other mental health conditions [10, 11], as well as plans to quit [12]. This is in line with effects described in previous similar outbreaks such as severe acute respiratory syndrome (SARS) (2003), Middle East respiratory syndrome (MERS) (2013–2016) or Ebola (2014 − 2016) [13, 14]. In this context, there has been a growing research interest in changes in SA among HCWs in Europe, with studies showing changes in SA rates and patterns during and after the COVID19 pandemic [15–17]. Besides reflecting temporary work disability [18], SA is also a key social benefit of welfare states aimed at ensuring social protection for workers [19]. There is also a substantial and expanding literature on SA among healthcare workers using longitudinal designs. However, most studies have focused on specific occupational exposures such as shift work [20] or on overall levels of absence [21], rather than on the temporal evolution of SA patterns as a dynamic process across occupational groups within hospital settings. In a previous study using the same cohort [22], we reported a significant rise in SA incidence among HCWs during and after the pandemic and identified some vulnerable groups. However, that analysis focused on incidence and did not describe longer-term patterns of SA over time across occupational groups. A more detailed understanding of SA dynamics may therefore contribute to a better characterization of how sickness absence evolves in hospital-based populations. Thus, a dynamic, person-centered longitudinal approach is needed to fully capture how SA behaviours evolve among HCWs [23]. Such an approach may identify groups of HCWs with similar SA trajectories, providing insights for emergency preparedness, workforce planning, and targeted interventions to prevent excessive SA, particularly among younger or more vulnerable workers [24]. Within this broader context, Spain’s healthcare system faces additional structural challenges, including population aging, increasing chronic disease burden, and limited investment in healthcare workforce resources. These pressures have been exacerbated by the pandemic, intensifying workload and staff shortages. Younger HCWs may represent a particularly vulnerable group, as they often face high job demands, job instability, and evolving expectations regarding work and well-being. Understanding SA trajectories in this context may provide insights that are not only relevant for Spain but also for other countries facing similar workforce pressures, emerging mental health challenges, and the need to develop targeted strategies to support retention, well-being, and sustainable staffing. Thus, the aim of this study is to identify and describe the trajectories of SA and to assess the role of sociodemographic and working conditions as determinants of SA trajectories among healthcare workers at a tertiary hospital in Barcelona, Spain, during the period 2018–2023. Methods Study population and design The study population was based on a sample of HCWs hired by Hospital del Mar from January 1st, 2018, to December 31st, 2023, involved in both direct and indirect patient care. Hospital del Mar is a tertiary healthcare institution located in Barcelona, Spain. It comprises around 500 beds, 33,000 annual discharges, and eight health centres for acute and long-term care. This sample is part of the HCWs cohort of Hospital del Mar (COSAMar), a retrospective dynamic cohort including every HCW of the institution since 2018 who had registered for at least one working week. Information on HCW was available from the Human Resources Department databases. Currently, the COSAMar cohort includes a total of 8,368 HCWs over the study period. For each HCW, we retrieved sociodemographic and occupational variables, as well as all SA episodes information. A participant identification number for the study was created to link all the information while ensuring confidentiality. Privacy and data safety were guaranteed, and the study was approved by the HMar Ethical Committee (2020/9379/I). Inclusion and exclusion criteria Participants included HCWs, aged 18 to 70 years, who were engaged in either direct or indirect patient care and had been employed for at least 60 months during the 65-month study period. This employment criterion ensured the inclusion of a stable workforce, minimizing variability from employment transitions instability that could affect attitudes and behaviours towards SA. Staff working in the hospital through an employment contract with an external company (mainly kitchen, security, maintenance and cleaning staff) were not included as administrative data was not available. After applying exclusion criteria, the final analytical sample comprised 2,852 HCWs. Variables The main outcome was SA episodes due to any health problems, excluding absences for other reasons such as pregnancy and other work leaves, holidays, unexpected absence, etc. SA was assessed by the number of SA days registered by the Human Resources department. Mean number of accumulated days of SA per semester were computed from the first semester of 2018 (S1) to the last semester of 2023 (S12). As covariates, sociodemographic variables such as sex (man and woman) and age (both continuous and categorical in tertiles) were included. In addition, the following occupational variables were also considered: shift schedules which were categorized into two groups as day/afternoon (encompassing full-day, morning, and afternoon shifts) and night/other shifts (including night shifts, non-specified shifts, and back-to-back shifts); weekly mean working hours, computed and categorized into two groups as < 36 h and 36–40 h; occupational categories, divided into four groups including physicians, nurses, technicians and administrative personnel, and aides; the employee’s patient care level, classified into two categories as direct care (encompassing hospitalized and critical care, surgery, emergencies and ambulatory care) and non-direct patient care (i.e. administrative services, technicians); and workplace, included acute hospitals centres (Hospital del Mar and Hospital de l’Esperança) and others (mid and long term care of geriatrics, acute, mid and long term care of psychiatric care, primary care, research and health professional schools centres). To address potential changes in these employment conditions over time, when necessary, we assigned individuals to the category in which they spent most of their time during the follow-up period [25]. Statistical analysis Mean number of accumulated days of SA per semester were computed from 2018 to 2023 to estimate SA trajectories using Latent Class Growth Analysis (LCGA) with zero-inflated Poisson model by semester [26]. This methodology assigns individuals who follow similar outcome patterns over time to different subgroups. LCGA provides an estimation of class membership probabilities for each individual being the optimal number of trajectories defined by different criteria. Firstly, the Bayesian Information Criterion (BIC) and the Akaike information criterion (AIC) which provide numerical indicators to assess the model’s goodness of fit (the lower the better) [27]. Besides, average posterior class probability and entropy (> 0.8 indicates good fit) was analysed. Finally, for each trajectory group, a minimum size of 5% was also preferred as trajectories with small size could give inaccurate estimates [28]. Eventually, researchers’ criteria were also applied to determine the number of trajectories attending to theoretical consistency, considering the interpretability of trajectory shapes, the size of each trajectory group, and theoretical consistency with prior knowledge in sickness absence research. Then trajectories were analyzed based on sociodemographic and occupational variables. First, a bivariate analysis by using Pearson’s Chi Square test in categorical variables and ANOVA or Kruskal-Walli’s test in continuous variables was conducted. Then, successive multinomial models to assess the association between SA trajectories and sociodemographic and working conditions were fitted. Low trajectory was used always as the reference group for trajectories. First model included sex and age. Then, weekly work hours, shift, care level and occupation were successively added. Reference categories for each variable are detailed both in Fig. 2 and Table S2. Finally, a sensitivity analysis excluding physicians were conducted to explore trajectories and associations in more homogeneous occupations according to sociodemographic and working conditions. All statistical analyses were conducted using R V.3.2.5. Fig. 1Mean accumulated days of sickness absence by semester for each trajectory class. Continuous lines mean values of predicted data, while dashed lines mean values of observed data. Healthcare workers at Hospital del Mar of Barcelona (Spain) 2018–2023 Mean accumulated days of sickness absence by semester for each trajectory class. Continuous lines mean values of predicted data, while dashed lines mean values of observed data. Healthcare workers at Hospital del Mar of Barcelona (Spain) 2018–2023 Fig. 2Adjusted odds ratio (OR) and 95% Confidence Intervals (95% CI) for the association of socioeconomic and working conditions with sickness absence trajectories from the final model Healthcare workers at Hospital del Mar of Barcelona (Spain) 2018–2023 Adjusted odds ratio (OR) and 95% Confidence Intervals (95% CI) for the association of socioeconomic and working conditions with sickness absence trajectories from the final model Healthcare workers at Hospital del Mar of Barcelona (Spain) 2018–2023 Results The cohort was predominantly female (74.9%) with a median age of 41 years (Table 1). Most participants worked day or afternoon shifts (83.5%) and were hired for 36–40 h per week (72.7%). The majority were employed in direct patient care roles (82.9%), primarily in acute hospital settings (75.5%), and the most common occupation was nursing (30.5%). The mean number of accumulated SA days per HCW increased over time, from 6.5 days in the first semester (S1) to 17.2 days in the final semester (S12) (Figure S1, Supplementary Material). Table 1Descriptive statistics (n, %) for variables of interest by trajectory class and total. Healthcare workers at Hospital del Mar of Barcelona (Spain), 2018–2023Zero(N = 217)Low(N = 1487)Degrowing(N = 507)Growing(N = 417)High(N = 224)P-valueOverall(N = 2852) Sex Man89 (41.0%)412 (27.7%)103 (20.3%)78 (18.7%)33 (14.7%)< 0.001715 (25.1%) Woman128 (59.0%)1075 (72.3%)404 (79.7%)339 (81.3%)191 (85.3%)2137 (74.9%) Age in years (continuous) Mean (SD)45.6 (9.92)39.3 (10.8)42.5 (10.2)40.0 (10.9)46.7 (8.99)< 0.00141.0 (10.8) Median [Min, Max]46.0 [23.0, 63.0]39.0 [18.0, 64.0]42.0 [19.0, 61.0]40.0 [18.0, 59.0]48.0 [20.0, 60.0]41.0 [18.0, 64.0] Age in years (categorical) <=3643 (19.8%)609 (41.0%)147 (29.0%)165 (39.6%)31 (13.8%)< 0.001995 (34.9%) 37–4778 (35.9%)481 (32.3%)180 (35.5%)126 (30.2%)73 (32.6%)938 (32.9%) >4796 (44.2%)397 (26.7%)180 (35.5%)126 (30.2%)120 (53.6%)919 (32.2%) Shift Day/Afternoon203 (93.5%)1289 (86.7%)406 (80.1%)327 (78.4%)157 (70.1%)< 0.0012382 (83.5%) Night/Other14 (6.5%)198 (13.3%)101 (19.9%)90 (21.6%)67 (29.9%)470 (16.5%) Weekly work hours <36 h29 (13.4%)379 (25.5%)159 (31.4%)129 (30.9%)78 (34.8%)< 0.001774 (27.1%) 36–40 h188 (86.6%)1105 (74.3%)348 (68.6%)288 (69.1%)145 (64.7%)2074 (72.7%) Occupation Physician110 (50.7%)374 (25.2%)64 (12.6%)27 (6.5%)5 (2.2%)< 0.001580 (20.3%) Nurse24 (11.1%)469 (31.5%)195 (38.5%)117 (28.1%)66 (29.5%)871 (30.5%) Admin & Technician77 (35.5%)384 (25.8%)115 (22.7%)111 (26.6%)50 (22.3%)737 (25.8%) Aide6 (2.8%)260 (17.5%)133 (26.2%)162 (38.8%)103 (46.0%)664 (23.3%) Area Non-direct patient care96 (44.2%)257 (17.3%)58 (11.4%)55 (13.2%)22 (9.8%)< 0.001488 (17.1%) Direct patient care121 (55.8%)1230 (82.7%)449 (88.6%)362 (86.8%)202 (90.2%)2364 (82.9%) Centre Acute167 (77.0%)1163 (78.2%)371 (73.2%)298 (71.5%)153 (68.3%)0.0032152 (75.5%) Others50 (23.0%)322 (21.7%)136 (26.8%)119 (28.5%)71 (31.7%)698 (24.5%) Employed months Mean (SD)70.9 (2.86)70.9 (2.65)71.4 (2.23)70.9 (2.79)70.6 (2.59)0.00771.0 (2.62) Median [Min, Max]72.0 [60.0, 72.0]72.0 [60.0, 72.0]72.0 [60.0, 72.0]72.0 [60.0, 72.0]72.0 [60.0, 72.0]72.0 [60.0, 72.0]*p-value significance set at < 0.05 Table 1Descriptive statistics (n, %) for variables of interest by trajectory class and total. Healthcare workers at Hospital del Mar of Barcelona (Spain), 2018–2023Zero(N = 217)Low(N = 1487)Degrowing(N = 507)Growing(N = 417)High(N = 224)P-valueOverall(N = 2852) Sex Man89 (41.0%)412 (27.7%)103 (20.3%)78 (18.7%)33 (14.7%)< 0.001715 (25.1%) Woman128 (59.0%)1075 (72.3%)404 (79.7%)339 (81.3%)191 (85.3%)2137 (74.9%) Age in years (continuous) Mean (SD)45.6 (9.92)39.3 (10.8)42.5 (10.2)40.0 (10.9)46.7 (8.99)< 0.00141.0 (10.8) Median [Min, Max]46.0 [23.0, 63.0]39.0 [18.0, 64.0]42.0 [19.0, 61.0]40.0 [18.0, 59.0]48.0 [20.0, 60.0]41.0 [18.0, 64.0] Age in years (categorical) <=3643 (19.8%)609 (41.0%)147 (29.0%)165 (39.6%)31 (13.8%)< 0.001995 (34.9%) 37–4778 (35.9%)481 (32.3%)180 (35.5%)126 (30.2%)73 (32.6%)938 (32.9%) >4796 (44.2%)397 (26.7%)180 (35.5%)126 (30.2%)120 (53.6%)919 (32.2%) Shift Day/Afternoon203 (93.5%)1289 (86.7%)406 (80.1%)327 (78.4%)157 (70.1%)< 0.0012382 (83.5%) Night/Other14 (6.5%)198 (13.3%)101 (19.9%)90 (21.6%)67 (29.9%)470 (16.5%) Weekly work hours <36 h29 (13.4%)379 (25.5%)159 (31.4%)129 (30.9%)78 (34.8%)< 0.001774 (27.1%) 36–40 h188 (86.6%)1105 (74.3%)348 (68.6%)288 (69.1%)145 (64.7%)2074 (72.7%) Occupation Physician110 (50.7%)374 (25.2%)64 (12.6%)27 (6.5%)5 (2.2%)< 0.001580 (20.3%) Nurse24 (11.1%)469 (31.5%)195 (38.5%)117 (28.1%)66 (29.5%)871 (30.5%) Admin & Technician77 (35.5%)384 (25.8%)115 (22.7%)111 (26.6%)50 (22.3%)737 (25.8%) Aide6 (2.8%)260 (17.5%)133 (26.2%)162 (38.8%)103 (46.0%)664 (23.3%) Area Non-direct patient care96 (44.2%)257 (17.3%)58 (11.4%)55 (13.2%)22 (9.8%)< 0.001488 (17.1%) Direct patient care121 (55.8%)1230 (82.7%)449 (88.6%)362 (86.8%)202 (90.2%)2364 (82.9%) Centre Acute167 (77.0%)1163 (78.2%)371 (73.2%)298 (71.5%)153 (68.3%)0.0032152 (75.5%) Others50 (23.0%)322 (21.7%)136 (26.8%)119 (28.5%)71 (31.7%)698 (24.5%) Employed months Mean (SD)70.9 (2.86)70.9 (2.65)71.4 (2.23)70.9 (2.79)70.6 (2.59)0.00771.0 (2.62) Median [Min, Max]72.0 [60.0, 72.0]72.0 [60.0, 72.0]72.0 [60.0, 72.0]72.0 [60.0, 72.0]72.0 [60.0, 72.0]72.0 [60.0, 72.0]*p-value significance set at < 0.05 Descriptive statistics (n, %) for variables of interest by trajectory class and total. Healthcare workers at Hospital del Mar of Barcelona (Spain), 2018–2023 *p-value significance set at < 0.05 LCGA identified five trajectories of accumulated SA over time, labelled based on trend patterns: zero (7.6% of participants) including 217 HCW with no SA along the period; low (52.1%) with a moderate increase in SA days, from 0.84 in S1 to 2.88 in S12; degrowing (17.8%) from S1 to S5 to similar SA days as the low trajectory in S12; growing (14.6%) with a sharp increase from S5 to the highest level of all trajectories at S12; and high (7.9%) from high levels which continued to increase until reaching its highest numbers in S6, and then began to decrease slightly until S12 (Fig. 1). All trajectories had an average posterior probability above 0.9, indicating high classification accuracy (Table S1). Model fit indices (BIC and AIC) were optimized with five classes, and entropy exceeded 0.9. These five trajectories were also replicated in sensitivity analyses (Figure S2). The analysis of trajectories by sociodemographic and occupational characteristics showed that the zero trajectory had the highest proportion of male HCWs (41%). The mean age was lowest in the low and growing groups. Night/other shifts were most common in the high trajectory (29.9%) and less represented (6.5%) in the zero trajectory, similar to the proportion of those who worked less than 36 h per week (34.8% vs. 13.4%). Only 2.2% of HCW in the high trajectory were physicians, compared to 50.7% in the zero trajectory. Direct patient care accounted for 55.8% in the zero trajectory, and for more than 82% in the other trajectories (Table 1). Adjusted multinomial regression analyses revealed significant differences between SA trajectories with the low trajectory as the reference group. The zero trajectory was mainly associated to higher age [aOR 1.86 (95%CI 1.24 to 2.78 for 37–47 years; 2.54 (1.71 to 3.79) for > 47 years], indirect care and physicians. The degrowing trajectory was related to higher age [1.66 (1.28 to 2.13) for 37–47 years; 2.08 (1.6 to 2.69) for > 47 years], direct care level [1.62 (1.15 to 2.3)] and non-physician occupations, particularly aides [2.56 (1.79 to 3.67)] and nurses [2.26 (1.63 to 3.15)]. Growing trajectory was associated to women [1.4 (1.05 to 1.87)], and non-physician occupation, particularly aides [7.23 (4.6 to 11.37)] and technicians and administrative [4.45 (2.79 to 7.08)]. Lastly, the high trajectory was associated to higher age groups: for 37–47 years [3.28 (2.09 to 5.15), and for more than 47 years [6.65 (4.31 to 10.24)], fewer weekly work hours [0.7 (0.5 to 0.96)], night/other shifts [1.90 (1.33 to 2.72)], direct care [1.86 (1.1 to 3.15)] and non-physician occupations, particularly aides [20.62 (8.12 to 52.36)] (Fig. 2 and Table S2). When physicians were excluded in the sensitivity analyses, the associations for aides in the growing [2.29 (1.74 to 3.03)] and high [2.41 (1.63 to 3.56)] trajectories remained significant, albeit with reduced effect sizes (Table S3). Discussion Our main findings show an increase of SA accumulated days by semester in HCWs between 2018 and 2023. Five different trajectories were identified, each associated with different sociodemographic characteristics and working conditions, most notably, the emergence of growing and high trajectories mainly related to older HCWs, night/other shift, direct care and non-physician occupations. This dramatic increase might be rooted to different causes. First, logically, our results should be placed in the context of COVID-19 pandemic. However, SA duration shows a constant increasing general trend in the study period. When analysing specific trajectories, the observed increase coincides with the onset of the COVID-19 pandemic in the low, growing and high trajectories with no return to previous levels. This is in line with previous research suggesting a huge impact of the pandemic on the health of HCWs well beyond COVID-19 infection [5]. Bjorn-Atle et al. (2023) found a substantial impact of COVID-19 pandemic on SA among HCWs in Norway, particularly on areas with high COVID-19-related workloads [17]. Similarly, a recent report in Spain showed a significant increase in SA from 2019 to 2023 from 4.0% to 5.7% in the whole Spanish working population, especially in healthcare [29]. Nevertheless, beyond the impact of the COVID-19 pandemic, other causes should be considered. As it has been suggested in the literature, aging of the working population or the effect of the economic cycle and changes in workers’ preferences between work demands and leisure time could have contributed to this longer increasing trend observed in SA rates [29, 30] Specifically, in the context of Hospital del Mar, human resources company policy changes in SA coverage may also have played a role in shaping these trends. Collective agreement between trade unions and hospital management provided whole salary and social benefits during all SA episodes. However, in this context of generalized increase in SA rates among Spanish working population, its effect would be expected to be limited. To our knowledge, this is the first study to characterize SA trajectories in HCW in Spain. Although trajectories with similar shapes have previously been described for general salaried workers, the proportion of workers observed in the zero or low categories was higher [31–33]. This is in line with studies observing higher rates of SA in HCWs compared to the general population [3, 34], especially in Spain, being the 3rd sector (non-profit entities dedicated to social action, human rights advocacy, and the promotion of general well-being) with the highest SA prevalence in 2023 [29]. Compared to the low trajectory, the zero and degrowing trajectories were linked to older HCWs, differing by care roles and occupations. The growing group included mostly women and non-physicians, with no age link. The high trajectory stood out, associated with older HCWs, part-time, night/other shifts, direct care, and especially aides. Gender, age, care level, and occupation emerged as key factors in trajectory membership. Gender was mainly associated to the growing trajectory, with the proportion of women being significantly higher. Gender is a known determinant of SA [35], with consistently observed higher rates in women than in men [36], also among HCWs [37]. Furthermore, our results show that the growing trajectory includes mainly women in reproductive age. Previous studies have suggested the differential gender impact of the presence of children at home as a main reason for SA in women, pointing towards the adverse effect of women’s role as caregivers in society, the so-called double-burden hypothesis [38, 39]. In addition, studies have also observed that the more senior in the organizational structure, the lower the level of SA in women, possibly due to woman’s ability to afford domestic support because of a higher salary [40] which would be in line with our results in age and occupation. However, causes of this gender-gap in SA remain poorly understood, warranting further research from an intersectional perspective. Age exhibited a paradoxical pattern, being older workers associated with both the zero and high trajectories. This may be partially explained by differences in occupational group and care level distribution. The zero trajectory was linked to older age, physician, and indirect care—likely representing health managers. In contrast, the high trajectory was more strongly linked to non-physician occupations, particularly aides, and to direct care activities. Two main explanations may underlie these differences. First, managers have better employment conditions, including the possibility of teleworking, and a higher socioeconomic position than other healthcare workers. There is a wide literature relating improved socioeconomic conditions to better health status, and consequently to lesser SA [41]. Besides, higher autonomy and non-direct care level have also been described as protective factors on SA onset [42]. Secondly, it has been observed that physicians and management teams have higher rates of presenteeism than other groups [43], which might have had an impact on the differences in the SA pattern shown by our results. Additionally, the observed paradoxical association may be influenced by underlying health status and the presence of chronic diseases among older workers, which could not be accounted for in our analyses due to lack of individual health data. The study population also consists of actively employed healthcare workers, which may introduce a healthy worker effect, potentially underestimating the true association between age and sickness absence trajectories. Conversely, the age distribution within the growing trajectory did not differ significantly from that of the low trajectory, with a higher proportion of younger HCWs. This result aligns with previous research on SA that reported higher short-term SA spells among younger age groups [38] a trend that may have been exacerbated by the broader societal effects of the COVID-19 pandemic. Furthermore, this result may reflect a generational turning point in relation to SA behaviour, perhaps associated to what prior literature has called The Great Resignation [44]. This phenomenon, basically observed in the United States, refers to a wave of voluntary resignations after the COVID-19 pandemic [44], due to wage stagnation, job dissatisfaction or worsened work environments among others [45]. This trend was highly reported among the healthcare sector [45]. However, since our analysis did not account for age cohorts, further research is needed to clarify whether those differences are the result of social generational changes or a cohort bias. Nonetheless, further qualitative research might shed light on the expectations and motivations regarding SA patterns by generation. Regarding care level and occupation, direct care level and non-physician occupations, particularly aides, had the highest magnitude effects in growing and high trajectories. The level of direct care has previously been associated with higher SA rates [42], particularly during the pandemic, when frontline workers faced a higher risk of both COVID-19 infection and psychosocial risk exposures [5]. Moreover, aides constitute a known risk group for SA, given their exposure to high-demand and low-control working environments [22]. Particularly, aides are also a group with higher rates of temporary work and replacement in Hospital del Mar (data not shown). Since we are analysing stable HWCs, the effects of teams with high-turnover rates could be leading stable HWCs to SA as a coping strategy. It is also worth mentioning that although associations in the growing group in non-physician occupations were significant, they had a smaller magnitude than in high groups. This may reflect a shift in SA accumulation toward physicians and, given the younger ages observed in the growing trajectory, support the hypothesis of a generational turning point described above. Beyond these main variables, other working conditions arise as determinants of SA trajectory membership. Effects of weekly working hours and shift were slight and mainly concentrated in the high trajectory. First, high trajectory was related to lesser weekly working hours. This association could reflect reverse causality, whereby workers at higher risk of SA are more likely to hold reduced-hours contracts. Alternatively, short-time contracts are often more precarious and associated with lower salaries, potentially adversely affecting workers’ health status, as mentioned earlier. High trajectory was associated with night/other shift as observed in previous research [46]. However, when care level and occupation were added to the final models, effect of working hours and shift nearly disappeared, both in main and sensitivity analyses. This finding is consistent with the literature [42] and suggests that care level and occupation may have greater explanatory potential for SA trajectories among HCWs. Limitations and strengths This study has some limitations. First, neither administrative nor SA data prior to January 2018 were available. Variables such as seniority or previous SA patterns would have been useful in multinomial analyses to fit final models and avoid confounding. Other variables of interest would be employment conditions (such as salary) or medical diagnosis of SA. In addition, regarding physicians, data on the 24-hour shift was not available. The lack of information on these variables does not influence the existence of the described trajectories and the similarity of the results of our multinomial analysis with previous literature yields robust conclusions. Second, the selection of stable workers precludes generalizability of our results to other types of workers such as temporary or junior HCWs and does not capture the high turnover that is common in the healthcare sector. However, as behaviour towards SA may be completely different in temporary or junior HCWs, we only considered a relatively stable healthcare workforce during the study period as being more homogeneous. Further research is needed to explore trajectories in SA in other groups such as temporary and junior or high-turnover HCWs, but also by medical diagnosis groups, especially musculoskeletal and mental disorders. In addition, this study may also be subject to the healthy worker effect, as it includes only actively employed healthcare workers, who tend to be healthier than the general population. Furthermore, the requirement of at least 60 months of employment over the 66-month study period may introduce a related form of selection bias, known as the healthy worker survivor effect. However, this requirement does not imply continuous, uninterrupted employment, as participants may still experience periods of sickness absence during follow-up. Overall, these mechanisms may underestimate observed association. Finally, we only analysed HCWs from one single hospital so extrapolation to other HCWs should be done cautiously. Nevertheless, our study also has relevant strengths. To our knowledge, this is the first study to characterize trajectories of SA in HCWs in a large healthcare institution before, during, and after the COVID-19 pandemic. Our results could provide new insights into the impact of major societal changes in recent years on labour markets in general and on the healthcare workforce in particular, leading to a better understanding of current patterns of SA. In addition, the use of administrative databases with health-related data derived from official medical diagnostic certificates provided not only robust validity compared with self-reported measures, as demonstrated in other studies [25, 47], but also real-time data that could be used to guide institutions’ policies. Conclusion Differences in trajectories were mainly related to gender, age, occupation and care level. Women were related to growing category, whereas age had a paradoxical effect with associations in both zero and high trajectories, but not in the growing trajectory. Working conditions such as weekly working hours or shift effect were mainly related to occupations, with aides being the occupational group mostly related to growing and high trajectories. These new insights may be useful for healthcare institutions to allocate resources more efficiently, such as adjusting night/other shift schedules or providing ergonomic and psychosocial interventions in direct patient care roles such as aides and nurses. The results concerning younger workers highlight the need for early-career support programs and workload management, among other interventions, to prevent early burnout and long-term disengagement from the workforce. Health managers should consider occupation-specific approaches to address SA in their facilities and teams. In this sense, occupational health services could play a key role as a mediator between health managers and HCW. Future research should replicate the trajectories described in a broader context such as in multicentre studies and delve into the occupational pathways that influence levels of SA, not only through quantitative methods, but also including qualitative methodologies that allow understanding subjective perspectives and attitudes towards SA. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1