Young Healthcare Workers’ Employment Status and Mental Distress over SARS-CoV-2 in Bolivia

Study Design

A prospective cohort study was conducted beginning in 2018. Nursing and medical students at the Universidad San Francisco Xavier de Chuquisaca (USFX) Sucre, Bolivia were recruited. Participants were in their last year of academic training before the internship (third year in the career of nursing and fifth year in the career of medicine), over 18 years of age, and Spanish-speaking (21). A total of 663 out of 809 invited students (82%) participated in the baseline study. Participants were followed up in 2022.

For the baseline study, study coordinators visited students’ classes and explained the objectives and procedures of the study. Participants were given a tablet to complete the digital questionnaire on site. For the follow-up study, emails, and WhatsApp messages with a link to the questionnaire were distributed to participants who had agreed to be contacted again at baseline (n = 526, 79%). A reminder was sent after one, three, and four weeks after the initial mailing to ensure a high response rate. As an incentive, participants who completed the questionnaire received five Farmacorp (Farmacy) vouchers worth 200 Bolivianos (approx. 29 USD) each, which were distributed through a lottery. To further promote the follow-up study, it was announced on the USFX University television and radio as well as on Bolivia’s public radio station ’La Bruja’. In the end, 116 (22%) participants completed the follow-up questionnaire (Figure A1).

The survey was pseudo-anonymous, and participants could withdraw from the study at any time without giving a reason. All participants gave written informed consent to participate in the baseline and follow-up study. The study was approved by the Ethics Committee of the Faculty of Medicine at the Universidad Mayor de San Simon (no project number assigned; November 8, 2021) and by the Ethics Committee of the Faculty of Medicine at the Ludwig-Maximilians Universität in Munich (project number 22-0451; June 11, 2022).

Questionnaire Instruments and Variable Definition

The baseline questionnaire included 44 questions to assess the participants’ socio-demographic characteristics, mental distress, lifestyle, employment status, working conditions, academic training, coping strategies, and social support. The questions were provided in Spanish and were administered as an online questionnaire using SurveyMonkey (Momentive Europe UC, Dublin, Ireland) at baseline and follow-up.

Employment status and socio-demographic characteristics were assessed using parts of the Quality of Life and Employment, Labor, and Health Conditions Questionnaire of the First National Survey of Workers in Chile (ENETS) (22) and parts of the Job Prospects and Employability in Nursing and Psychology Questionnaire (23) developed in Mexico.

Employment status was assessed as exposure. It had to be defined differently at baseline and follow-up, as participants had not yet completed the HCW training at baseline. At baseline, employment was defined as a positive response to the question ’Are you currently working?’. At follow-up, participants were considered employed as HCW if they answered ’yes’ to the two questions ’Are you currently working?’ and ’Is your work related to your profession?’. If the answer to wither question was ’no’, the participant was considered ’unemployed or employed insecurely’. In addition, those who were formally employed at follow-up were asked if they worked with COVID-19 patients. Those who were unemployed or insecurely employed at follow-up were asked why they were not formally employed (options: ’I stopped looking for a job because I couldn’t find one’, ’I am currently studying or in training’ or ’other reasons’).

Socio-demographic characteristics included sex, age, marital status, economic situation, and whether participants had children. Because sex and econimic situation are likely to be associated with both exposure (employment status) and outcome (mental distress), they were considered as potential confounders (4, 7). Age was not included as a potential confounder because of its limited range (97% 19−30 years).

Mental distress was assessed using the validated Spanish version of the 12-item General Health Questionnaire (GHQ-12) (24). The GHQ-12 was originally developed in English by Goldberg and colleagues. It is a well-established self-report measure of mental distress designed to screen for psychological morbidity and psychiatric disorders in community and non-psychiatric settings. On a 4-point scale from ’much more than usual’, ’rather more than usual’, ’no more than usual’ to ’not at all’, participants indicated how often they had experienced various psychological symptoms in the past four weeks. The outcome was a binary threshold score on the GHQ-12. The threshold measure was scored using the GHQ scoring method with 0-0-1-1 for positive items and 1-1-0-0 for negative items, and summed (range 0–12) (25). As suggested by a Chilean validation study (26), participants with a GHQ-12 score > 4 (threshold score 4/5) were defined as having a clinically relevant level of mental distress (hereafter referred to as ’mental distress’).

Statistical Analysis

Data analysis was performed using SPSS version 28.0 (IBM, Armonk, NY, USA). To assess selective losses in the follow-up study, we compared subjects who participated in both studies with those who dropped out. Differences between these two groups were examined using chi-square tests for binary variables and paired t-tests for continuous variables. To examine systematic differences between being employed as an HCW and being unemployed or insecurely employed as an HCW, we compared these groups at follow-up. Nominal and ordinal variables were described as absolute and relative frequencies. Metric variables were described as mean and standard deviation (SD). We used chi-square tests to assess the association between socio-demographic characteristics and the employment status with mental distress in bivariate analysis.

Missing data on outcome, exposure, and potential risk factors were multiply imputed (n = 20) for participants who answered at least one question in the baseline and follow-up questionnaires. Assuming that these data were missing at random, the Fully Conditional Specification (FSC) method was used with the Predictive Mean Matching (PMM) scaling model, as implemented in SPSS software (27).

To analyze associations between exposure and outcome, Generalized Estimating Equations (GEE) were used to estimate odds ratios (OR) and corresponding 95% confidence intervals (95% CI) using the logit link function. Participants were used as the clustering factor. Time (T0 = baseline, T1 = follow-up) was considered as an inner subject effect. First, unadjusted results were estimated for time, exposure (employment status), and all potential risk factors. In the next step, the model was adjusted for the potential confounders sex and economic situation. Associations of change in mental distress with employment over time were analyzed by including an interaction term between time and employment status. An interaction was suspected because the employment status changed between T0 and T1 due to the participants’ graduation in the meantime and the influence of the COVID-19 pandemic (7, 8, 28).

Several sensitivity analyses were performed. First, we used a threshold score of 5/6 for the binary threshold score of the GHQ-12. Second, the continuous mean GHQ-12 score was used as the outcome. For this, the GHQ-12 items were scored according to the Likert scoring method (0-1-2-3 for positive items and inverted for negative items), resulting in a score of 0–36 (29). In the GEE, the identity link function was used to analyze the continuous outcome. Third, we changed the definition of the exposure variable (employment status) at baseline. Employment was defined the same as at follow-up as an affirmative response to the question, ’Is your work related to your profession?’. Fourth, those who were still in academic training at follow-up were excluded from the analysis. Finally, fully imputed data were analyzed, taking into account all those who participated at baseline.

Consistency with other Studies

Comparing our findings with other studies of mental distress in Bolivia, we found that young HCW had a higher prevalence of mental distress than Bolivian teachers (43%) and a lower prevalence than Bolivian miners (81%) (30, 31). Given the extreme working conditions, such as very high and very low temperatures, noise, and shift work, the higher prevalence of mental distress among minders is not surprising (31). The comparatively lower prevalence of mental distress among teachers may be due to the use of a higher threshold score of 5/6 for the GHQ-12, which results in a lower estimate of mental distress (30). However, using a threshold score of 5/6 in our sensitivity analyses did not change the association between the employment status and mental distress (Table A1). The difference in prevalence could also be attributable to the impact of the COVID-19 pandemic on HCW. As pointed out by Ferreira, L.C. et al. (32), graduates faced unprecedented challenges related to the pandemic, such as changes in internships, loss of space for practice scenarios, as well as cancellation of graduation rites and ceremonies. Participants entered a highly demanding workplace and, simultaneously, a work environment with immense employment insecurity (32). These conditions can lead to a high prevalence of mental distress (7, 8, 28). Comparing our results with the COVID-19 Healthcare Workers Study (HEROES), a prospective cohort study evaluating the impact of the COVID-19 pandemic on the mental health of HCW in 26 LMIC and HIC, including Bolivia (33), the mean GHQ-12 scores were higher in our study (34). This heterogeneity in results may stem from the older age of participants in the HEROES study compared with our study population, as young HCW are more likely to develop mental distress (2, 5, 6).

Prior to the pandemic, students were shown to be at increased risk for mental distress (35). Consequently, under normal circumstances, mental distress decreases again with the onset of the professional career (12). However, as a result of the COVID-19 pandemic, our results showed that mental distress among young HCW did not decrease with entry into formal employment but has remained at a consistently high level. Our findings are supported by several studies that found an increase in mental distress among HCW over the course of the COVID-19 pandemic (2, 36, 37). Other studies contradict these findings (38, 39). Possible reasons for these inconsistent findings include differences in the operationalization of mental distress, cultural factors, and differences in the study population with respect to age, sex and occupation.

Among unemployed and insecurely employed HCW the prevalence of mental distress increased after graduation during the COVID-19 pandemic. This finding is consistent with other studies (6, 7, 40): In a secondary analysis of a longitudinal population-based study it was found that unemployment and insecure employment negatively affected mental health in the general population of the United States during the SARS-CoV-2 pandemic (7). Medina Fernández, I.A. et al. conducted a cross-sectional study of the mental health of Mexican graduate students in the context of the COVID-19 pandemic. Consistent with our results, they found that those who work as HCW were less likely to experience mental distress than those who did not work as HCW during the COVID-19 pandemic (6). Likewise, another cross-sectional study investigating the mental health of recentuniversity graduates in Brazil found that graduates were less affected by mental distress when they were formally employed (40).

However, a study frm the United Kingdom found different results: Pierce, M. et al. examined the changes in adult mental health in the general population before and one month into the COVID-19 pandemic. They found an increase in the mean GHQ-12 score in the employed population and no change in the mean GHQ-12 score in the unemployed population (4). However, it must be said that the results may not be comparable to ours for the following reasons: First, the general population was examined, not recently graduated HCW. Second, insecure employment has fundamentally different consequences in the UK (and other HICs) than in Bolivia (and other LMICs). Without a social welfare system to support the unemployed and insecurely employed, having a secure and well-paying job is essential for survival in Bolivia (10, 13). These circumstances influence the impact of the employment status on mental distress (9–11). In addition, the initial increase in mental distress one month into the pandemic found by Pierce, M. et al. may represent a spike in emotional response that stabilizes or declines as people adjust (4). The present study reported the long-term effects of the pandemic on participants’ mental health status three years into the pandemic. At this advanced stage of the pandemic, HCW may have gained resilience (38, 39).

Limitations and Strengths

Although we had a large group of participants at baseline with a high response rate (82%), the response rate at follow-up was comparatively low (22%). Therefore, our results cannot be considered representative of the study population. Therefore, selection bias may be introduced. However, the non-responder analysis showed that participants who were lost to follow-up were not statistically significantly different from responders who remained in the study at both time points. Nevertheless, low statistical power at follow-up may have resulted in failure to detect changes in important factors between baseline and follow-up.

We recognize that our exposure definition differs between baseline and follow-up because employment status varies depending on whether participants were still in training (T0) or were trained HCW (T1). This needs to be taken into account when using ’not employed at baseline’ as reference group. There may also be other aspects that changed between T0 and T1, such as participants moving out of Sucre, Bolivia after graduation. However, sensitivity analysis showed that the association between employment status and mental distress did not change when employment at baseline was defined identically to employment at follow-up as employment as HCW (Table A2). Excluding participants at follow-up who stated to still being in training also did not change the association (Table A3).

Misclassification of exposure and outcome may have occurred because they were assessed only by questionnaire. The GHQ-12 is a validated screening tool that correlates strongly with the presence of mental illness and with future clinical diagnosis of a psychiatric disorder. However, the GHQ-12 is not equivalent to a clinical diagnosis (41). This study relied on self-report, which is valid primarily because the effect of employment status on mental distress depends largely on personal assessment of the situation. This leads us to the conclusion that the possibility of measurement error is low (42). Bias in the negatively worded items of the GHQ-12 may have occurred due to inattentive reading by responders, i.e., the respondent did not notice that the response format had changed (positive and negative items) (43).

This study cannot conclusively identify the reasons why economically disadvantaged participants are more likely to be formally employed than participants in an economically advantaged situation. In addition to policy reasons, explanations could be that participants with greater economic needs applied at an early stage for formal employment and did not, e.g., complete further training. Early entry into employment allowed them to develop soft- and work-related skills, which made them more attractive to the formal labor market.

The present study did not distinguish between the different forms of underemployment and unemployment. Underemployment can be income-, skill-or hours-based (14). Whereas unemployment can be frictional, structural, or cyclical (44). Likewise, we combined unemployed and insecurely employed HCW into one group because there is evidence that insecurely employed people mirror unemployed people in terms of their mental health (7). We also did not distinguish between frontline HCW, which are HCW providing direct care for COVID-19 patients, and normal HCW, although there is indication that the prevalence may be higher among frontline HCW (45). Consequently, a dose-response relationship could not be assessed.

Due to the prospective design of the study the risk of recall bias was limited because the temporal sequence of exposure and outcome could be evaluated. Nevertheless, the possibility of reverse causation cannot be excluded. For example, childhood mental health problems could not be included as a potential confounder because such information was not available. Potential confounding by sex and economic situation was taken into account, although we did not consider aspects such as the impact of measures such as home-schooling, other occupations, such as caring for relatives, whether participants had been diagnosed with COVID-19 or other serious and/or chronic diseases, and whether participants had been vacinated against the SARS-CoV-2 virus (7).