We conducted a multicenter cross-sectional study between July and August 2020 in three Social Security (EsSalud) hospitals located in the northern regions of Piura and Lambayeque, Peru—areas severely affected during the first wave of the COVID-19 pandemic.

The present analysis uses data from the baseline assessment of a longitudinal study. Although the parent study has a longitudinal design, the current manuscript analyzes only the first measurement, corresponding to a cross-sectional dataset collected between August and December 2020.

The study population consisted of HWHW (doctors, nurses, obstetrician, nursing technicians, medical technologist, biologist, psychologist, among others) who were working in the Social Security Hospitals (EsSalud) of the Lambayeque region (Almanzor Aguinaga National Hospital and Hospital Luis Heysen Incháustegui) and Piura region (Hospital Regional Cayetano Heredia), 2020 during the first wave of the COVID-19 pandemic between the months of August-December 2020.

Eligible participants were health workers who had worked for at least one month during the COVID-19 emergency in 2020, either on the front line (COVID-19 wards, ICUs, emergency) or non-front line (general services). Inclusion required active employment at the time of data collection and voluntary participation with signed informed consent.

Exclusion criteria included: health workers on temporary medical leave or remote work due to high-risk conditions; those with incomplete or invalid responses in the PHQ-9 or GAD-7 questionnaires; and refusals to participate after initial contact. Participants were selected using systematic random sampling from the official list of health workers scheduled for mandatory COVID-19 screening.

The sample size was estimated to detect a minimum difference of 35% in the prevalence of depression and anxiety symptoms between exposed and unexposed groups, with 80% power and a 95% confidence level, using the reference prevalence observed in Lai et al. (2020) (27). Based on this, 62 participants per hospital were required. Considering a 10% non-response and 10% rejection rate, the total required sample was 152. During recruitment, 202 individuals were approached: 9 declined participation and data from another 11 were excluded due to incomplete or inconsistent responses (missing key variables). The final analytical sample comprised 182 participants.

A systematic random sampling was carried out, selecting the first participant randomly and the next with a fixed jump (every 10 participants). The list of health personnel in each hospital center was requested, differentiated by exposure groups (working and not working on the front line of COVID-19 health care). The choice of hospital centers was based on convenience, in the EsSalud Hospitals of Lambayeque and Piura. This research was based on primary data collected specifically for this study through direct field application of validated instruments, and does not correspond to a secondary data analysis.

All variables were collected through structured face-to-face interviews conducted by trained personnel during the first wave of the COVID-19 pandemic (May to August 2020). The survey incorporated standardized psychometric instruments and a structured questionnaire designed to capture sociodemographic, occupational, physical health, and psychosocial characteristics.

A structured and systematic recruitment process was employed to minimize selection bias. Healthcare workers (HW) selected through systematic random sampling from institutional screening lists were invited to participate. Those who agreed provided electronic informed consent after receiving detailed information about the study. Although participation was voluntary, the recruitment followed a closed and sequential process based on predefined sampling frames.

Data were collected through in-person interviews using structured questionnaires administered by trained field personnel. Interviewers—health professionals trained by the research team—applied the instruments directly at each healthcare worker’s workplace (HWHW) within EsSalud facilities. The questionnaire included sociodemographic, psychosocial, occupational, and physical health variables.

Mental health outcomes were evaluated using a specific temporal framing: “Have you ever experienced any of the following symptoms (…) due to your care work since the beginning of the COVID-19 health emergency?”

REDCap^® software was used to design the questionnaires, collect data digitally, and implement real-time quality control. Interviews were conducted in designated care modules that ensured biosafety standards, including adequate ventilation, physical distancing, and the use of personal protective equipment. Each interview lasted approximately 20 minutes.

All statistical analyses were performed using STATA v18^®. The primary objective was to examine whether depressive and anxiety symptoms were associated with sociodemographic, occupational, psychosocial, and health-related factors among HWs.

Missing data was handled through a complete-case approach, as the dataset was nearly complete for all primary variables. No imputation procedures were performed because the outcome variables were 100% complete. Only minimal missingness was observed in secondary covariates. Given the low proportion and random distribution of missing data, listwise deletion was applied without compromising statistical validity.

Univariate statistics were used to describe the sample. Categorical variables were summarized using absolute and relative frequencies (%). For continuous variables, means and standard deviations (SD) were reported for normally distributed data, while medians and interquartile ranges (IQR) were presented for non-normally distributed variables.

To explore crude associations between each outcome (depression or anxiety) and independent variables: Chi-square or Fisher’s exact tests were applied for categorical comparisons. Mann–Whitney U tests were used for continuous variables due to non-normal distribution. This exploratory step allowed the identification of candidate variables for multivariate modeling, based on a significance threshold of p < 0.05.

To estimate adjusted associations, we fitted Generalized Linear Models (GLM) of the Poisson family with a log link and robust standard errors. Hospital site was included as a clustering variable to adjust for intragroup correlation.

Variables were selected based on both conceptual and statistical criteria. Theoretical and empirical evidence from previous studies guided the inclusion of factors with established or plausible associations with depression and anxiety among healthcare workers during the COVID-19 pandemic. From the statistical standpoint, variables showing a p-value < 0.05 in simple Poisson regression models were entered into the multivariable model.

A third model was constructed to assess the robustness of associations by adjusting for theoretically relevant confounders.

Model 3 included all variables statistically significant in Model 2, plus key covariates identified in the literature as potential confounders—regardless of their bivariate significance—including: age, sex, resilience, insomnia, physical activity, and tobacco use. Model 3 was estimated with the same GLM-Poisson approach and hospital-level clustering.

To visually summarize the associations between the covariates and the outcomes (depressive and anxiety symptoms), forest plots were generated from the multivariable Poisson regression model 1. All forest plots were produced using the R statistical.

This study adds novel insights to the growing literature on HW mental health during COVID-19, particularly within Latin America. Unlike prior studies that often relied on convenience samples or descriptive analyses (71), our research employed institution-based probabilistic sampling, robust multivariable models, and sensitivity analyses using different clinical thresholds. It is, to our knowledge, the first multicenter study conducted in northern Peru during the early pandemic wave, and one of the few to integrate underexplored psychosocial dimensions such as perceived neighborhood stigma, mistreatment in care, and fear of being asymptomatically infectious. These contextual variables strengthen the explanatory power of our models and highlight structural vulnerabilities in overstretched public healthcare systems.

In our study, 42.9% of HW exhibited depressive symptoms—31.9% with mild and 9.3% with moderate levels. These figures are consistent with studies in Peru [Arequipa: 30.2% (2), Lima: 53.3–46.2% (3)] and Colombia [40.8% (23)], but lower than findings from multicountry Latin American studies [62.2% (8)] and Chile [65% (7)]. By contrast, lower rates were reported in China [34.3% (72)] and the U.S. [38% (11)], aligning with meta-analyses estimating pooled prevalence around 23 (73)–37% (74).

Regarding anxiety, 50.6% of participants showed symptoms (37.4% mild, 6.0% moderate, 7.1% severe), comparable to multicountry Latin American data [40.1% (8)], but higher than pooled estimates from global meta-analyses [22.8 (5)–29.6% (73)] and studies in China [18.1% (72)] or Arequipa, Peru [32.1% (2)]. Chile reported the highest prevalence [74% (7)]. Finally, to a meta-analysis that included health personnel who cared for patients with COVID-19 in whom I identified that the cumulative prevalence of anxiety was 25.8% (74).

The high frequency of symptoms of anxiety and depression could be explained by the sudden change from traditional work to the confinement of working in silence and without contact with peers, the constant fear of infection, poor social support, residing in regions with high rates. of exposure and infection by COVID-19 as was the Latin American continent, experiencing financial insecurities (75, 76). However, this difference with the continents of Europe and Asia could be attributed to the fact that LATAM presents inadequacies in health systems, absences in prevention or anticipation programs for the promotion of good mental health, and the cultural stigma associated (77) and socioeconomic inequalities, generating negative repercussions on people’s quality of life (78).

We adopted a PHQ-9 and GAD-7 cut-off of ≥5 based on both clinical relevance and managerial implications. Clinically, this threshold captures early or subthreshold symptoms, which may not yet meet diagnostic criteria but still impair emotional well-being and work performance (79, 80). From a health services perspective, this lower threshold enables broader mental health surveillance, guiding timely support and preventing deterioration. It also generates actionable data for human resources planning, including the design of workplace interventions and psychological support programs. Importantly, we do not interpret this cut-off as a diagnostic label but as a pragmatic tool for early detection during public health emergencies.

Each additional year of age was associated with a slightly lower prevalence of depressive symptoms (approximately 1%). This finding aligns with studies reporting lower depression rates in older adults, possibly due to greater emotional regulation or resilience developed with age (81–83). However, other studies have reported either a direct age-depression relationship (84) or higher risk of depressive symptoms among adolescents and young adults (85). Although the biological mechanism remains unclear, resilience and accumulated experience in managing stress may explain this inverse association in healthcare settings (86–88).

Having diabetes as a comorbidity was associated with a 41% higher prevalence of depressive symptoms. This is consistent with studies linking chronic comorbidities with increased emotional distress in healthcare personnel (89, 90). Likewise, Lucas-Hernandez et al., in Mexico during the pandemic identified a greater frequency of depressive symptoms in people with diabetes (9). The psychological burden of managing a chronic illness, combined with pandemic-related stress, may contribute to this association (91–93).

Having a previous personal history of mental illness was associated with a 32% higher prevalence of depressive symptoms. This is similar to what was reported by Valencia A. et al., who among Colombian health personnel during the pandemic identified that those with previous depression were more likely to present depression again (OR = 2.5) (23). This association could be due to the fact that health personnel, such as those who cared for patients with COVID-19, are more emotionally vulnerable and are more likely to seek mental support for their problems; it may also be explained by a lack of access to mental health services (94).

Having experienced abuse during health care was associated with a 35% higher prevalence of depressive symptoms. This finding is consistent with what was described by Want et al. who in their study carried out on health workers from 31 provinces in China found that violence reported in the work environment was more likely to have mental health symptoms during the COVID-19 pandemic (b=8.248) (95). Likewise, it agrees with what was reported in systematic reviews which suggest a significant increase in violence in health workers during the pandemic (96–98). Mistreatment may exacerbate stress, lower self-esteem, and contribute to feelings of hopelessness, thereby heightening the risk of depression (97, 99).

Moderate to extreme concern about working in a COVID-19 environment was associated with a 23% higher prevalence of depressive symptoms, in line with findings from southern Peru (100). Moderate or extreme concern about the work environment can expose healthcare personnel to a greater burden of stress and anxiety, as they fear facing challenging or risky situations in their work environment (101). Uncertainty about job security (102), Fear of exposure to infectious diseases, scarcity of resources, and emotional exhaustion are factors that could contribute to the appearance of depressive symptoms. Furthermore, constant worry can have a negative impact on the psychological and physical well-being of healthcare workers, which in turn increases the likelihood of experiencing depressive symptoms (103–105). These stressors highlight the need for psychosocial support interventions.

Extreme concern about neighborhood stigma related to COVID-19 exposure was associated with a two-fold higher prevalence of depressive symptoms. This is supported by studies from Japan (106) and Colombia (23, 107) showing that perceived discrimination strongly correlates with depression among healthcare workers. This association could be explained by the fact that exposure to a pandemic such as COVID-19 can generate stigmatization and discrimination towards those health workers who are in the first line of defense against the disease (108). Concern about being marginalized or rejected by their neighborhood environment due to their exposure can generate a constant state of alert and anxiety in health personnel (109). Additionally, uncertainty about the future and the possibility of being stigmatized can lead to excessive worry and increased anxious reactivity (110). The chronic stress associated with this worry can also have a negative impact on the psychological and physical well-being of health personnel, contributing to the development of anxious symptoms (110).

Having insomnia was associated with a 62% higher prevalence of depressive symptoms. This is consistent with that of Zhang et al, who in research conducted in China found that depression and insomnia were significantly correlated, after controlling for the effects of sex, age, education, smoking and alcoholism in medical personnel. front line of care against COVID-19 (111). Likewise, Valencia A. et al., in Colombian health personnel identified that those with previous insomnia were more likely to present depression (OR = 2.0) (23). This bidirectional relationship may be exacerbated in high-stress occupational settings, such as during the COVID-19 pandemic, where increased workloads, shift work, and concerns about safety may disrupt sleep patterns. Insomnia not only compromises emotional regulation but also reduces resilience to stress, potentially increasing vulnerability to mental health disorders. These findings highlight the need for institutional support strategies to address sleep hygiene and mental well-being in healthcare settings.

Health personnel with burnout syndrome were associated with a 42% higher prevalence of depressive symptoms. This is contrary to what was reported by García-Torres et al., who in their study of Mexican health personnel identified a correlation between SB and depressive symptoms (r= 0.217) (112). Similarly, Babamiri et al., in their study conducted on Iranian health personnel, identified that emotional exhaustion (EB) increased the probability of having negative mental health symptoms (OR = 6.92) (113). Likewise, Karadag et al., in their study conducted in Turkey during the pandemic in health personnel in the intensive care unit, identified a moderate and significant correlation between SB and depression symptoms (r = 0.519) (114). It is highlighted that there is little research on these variables in health personnel, therefore, with the present result it adds to the current literature. The high prevalence could be due to the fact that being a health personnel means exposing oneself to the main factors that contribute to the development of BS, such as the stress generated by hospital conditions with excessive workload, exposure to suffering and death of patients (115). Emotional and physical demands, lack of equipment and materials, long hours and days, among others (116). However, the biological mechanism between both variables has not been identified.

Health personnel who reported having little fear of contracting COVID-19 were associated with a lower prevalence of anxiety symptoms in the simple model, however: this was diluted in the multiple model. Similar patterns have been reported in Iran (117), Jordan (118), and Turkey (119), where fear of infection correlated positively with anxiety symptoms among healthcare staff. The lower prevalence of anxious symptoms among those with little fear of becoming infected can be explained by several psychological and emotional factors (120). Excessive or chronic fear can generate high levels of anxiety and worry in healthcare workers, negatively affecting their emotional and physical well-being (121). Conversely, those with less fear may have a more realistic and balanced perception of the risks associated with exposure to the virus, which could help reduce anxiety (122). Additionally, healthcare personnel who feel less fear may have a greater sense of control and security in their work environment, which may contribute to a lower likelihood of developing anxious symptoms. It is also possible that those with less fear are better equipped to handle the stress and challenging situations associated with the pandemic (123).

Concern about inadequate access to PPE was associated with higher anxiety prevalence, consistent with studies from Colombia (23, 124) and Peru (22), where lack of PPE increased risk of mental health symptoms among health personnel. These results are related to what we know about prevention measures to reduce the risk of becoming infected by SARS-CoV-2, which would make health personnel calmer and more confident if they work with PPE. It has been shown that Peru has had weaknesses in its response to the pandemic due to the fact that in recent years it has experienced a political crisis and a fragmented health system (22, 125). Situation that had greater risk in state establishments, with which health personnel could significantly experience the development of these negative psychiatric symptoms (126).

Perceived marginalization by neighbors due to COVID-19 exposure was associated with a higher prevalence of anxiety symptoms, echoing findings from Jordan (127) and Iran (128). However, it differs from what was reported in Nepal, in which no association was found between anxiety and stigma in health personnel (129). Health personnel who have been marginalized by people in their neighborhood due to their exposure to COVID-19 infection may experience feelings of isolation, rejection, and stigmatization (130). This perception of being excluded or judged for their work in addressing the pandemic can generate additional emotional burden and increase anxiety levels in these workers (131). Additionally, a lack of support and understanding from the community can lead to fears about their own safety and that of their loved ones, contributing to anxiety (132). Stigmatization and marginalization can have a negative impact on the emotional and psychological well-being of health personnel, affecting their self-esteem and self-confidence (133). These factors may predispose them to developing anxious symptoms and affect their ability to cope with stress related to their work in COVID-19 care (133).

Having moderate concern about working in a COVID-19 environment was associated with a 92% higher prevalence of anxiety symptoms, while having great/extreme concern was associated with a 125% higher prevalence. This is similar to what was reported in Peruvian studies, which have identified that fear of COVID-19 is positively associated with the development of anxious symptoms in the general population (134). and health workers (36). This association could be explained because the work environment in COVID-19 care presents unique and stressful challenges for health personnel (135). Continued exposure to the virus, uncertainty about your own health and that of your loved ones, the need to follow strict safety protocols, and the emotional burden of treating patients with a life-threatening illness can lead to high levels of anxiety and worry in patients. healthcare workers (120). Additionally, the constant concern about the possibility of contagion and the responsibility of providing quality care in the midst of the pandemic can lead to a state of permanent alert and tension. Furthermore, the fear of becoming infected and the perception of being at constant risk can affect the emotional and physical well-being of health personnel, predisposing them to develop anxious symptoms (120).

Moderate tobacco consumption was associated with a 54% lower prevalence of anxiety symptoms. This differs from those reported by Lucas-Hernández et al., who in their study carried out on Mexican health personnel during the pandemic identified a greater frequency of depressive symptoms in those who used tobacco (70% vs. 30%) (9). Contrary to what was reported by Stanton et al., who in Australian adults during the pandemic identified that the negative change in tobacco consumption was associated with the greater development of anxiety symptoms (OR = 1.12) (136). Finally, Bassi et al., in their study in the general population of India during the pandemic, identified that tobacco users presented more anxiety symptoms by 20.7% more (137).

Having moderate worry about being an asymptomatic COVID-19 patient was associated with a 49% higher prevalence of anxiety symptoms, and having very/extreme worry was associated with an 80% higher prevalence. However, in the multiple model the association was diluted. Although no previous studies have been identified that have exactly analyzed this association, there is solid evidence that has identified an association between fear of COVID-19 and anxiety (117, 119, 120). This association could be due to the fact that the concern about being asymptomatic implies a constant fear of carrying and transmitting the virus without presenting visible symptoms, which generates a feeling of uncertainty and lack of control over one’s own health and the possibility of infecting others (138–140). Persistent worry about this possibility can lead to continued alertness and anxiety (138–140). Furthermore, health personnel are on the front line of combating the disease, constantly facing the virus in their work, which can intensify their worry and stress. The fear of becoming infected and the responsibility of staying healthy to continue providing care to patients can put additional pressure on these workers (138–140).

The comparison between both models reinforces the validity of core risk factors and highlights how the choice of psychometric thresholds can shape epidemiological patterns. While a PHQ-9 cut-off of ≥5 is useful for identifying early or mild symptoms in occupational screening, using ≥10 captures more severe cases, often requiring clinical attention. The observed loss of significance in some demographic and clinical variables under the ≥10 cut-off may reflect their influence on subclinical or transitional symptom states. This dual analytical approach provides a more comprehensive understanding of the depressive symptom spectrum among healthcare workers and strengthens the public health and occupational implications of our findings. This sensitivity analysis reinforces the central role of certain risk factors in clinically significant anxiety and illustrates the influence of the chosen threshold on the interpretation of epidemiological data. While a GAD-7 cut-off of ≥5 allows for early detection of mild or subthreshold anxiety symptoms—important for occupational health surveillance—the ≥10 threshold highlights more severe clinical manifestations. The shift in significance and emergence of new associations under the higher threshold suggests that some psychosocial and clinical variables are more predictive of moderate-to-severe anxiety. The use of both cut-offs provides a broader epidemiological perspective, integrating early detection with clinically meaningful outcomes, and strengthening the practical implications for mental health support among healthcare workers.

Additionally, the inclusion of a third model adjusting for theoretically plausible confounders (e.g., age, sex, resilience, insomnia, physical activity, and tobacco use) further enhanced the robustness of our findings. This model confirmed the stability of key associations while accounting for residual confounding, offering a more rigorous assessment of the factors linked to mental health outcomes in healthcare workers.

Our findings provide timely evidence to inform occupational health policies within the Peruvian public health system, one of the most severely impacted in Latin America during the COVID-19 pandemic (141). The high prevalence of depressive and anxiety symptoms observed among public-sector healthcare workers especially those experiencing workplace mistreatment, high perceived marginalization, or burnout highlights the urgent need for institutional responses tailored to this population’s needs.

Based on our results, we recommend implementing structured mental health screening programs within hospitals, prioritizing frontline staff and those with known occupational vulnerabilities (e.g., limited job security, high workload, or comorbidities). Interventions should include regular psychological evaluations, on-site mental health services, and referral pathways for specialized care. In addition, promoting protective factors such as physical activity, organizational justice, and healthy workplace environments should be integrated into occupational health protocols.

These measures must be aligned with existing national mental health frameworks (e.g., the Plan Nacional de Fortalecimiento de Salud Mental), but operationalized at the institutional level to ensure coverage, continuity, and cultural relevance. Preparing contingency mental health support systems for future health emergencies is also essential, especially in under-resourced public facilities. Ultimately, ensuring the mental well-being of healthcare workers is not only an ethical imperative, but also a critical component of health system resilience in Peru and similar middle-income countries.

Our findings support WHO’s updated recommendations on protecting the mental health of healthcare workers during and beyond the COVID-19 pandemic, particularly in vulnerable systems such as Peru’s public health sector (142).

This study presents several limitations that should be considered when interpreting the results. First, due to its cross-sectional design, causal inferences cannot be drawn. The identified associations are exploratory and should be interpreted as correlational, even though we adjusted for potential confounders using multivariable regression. The study serves to generate hypotheses for future longitudinal research.

Second, self-administered questionnaires may have introduced measurement bias. Additionally, unmeasured confounding variables—such as postgraduate education, housing-related financial stress stress (137), or specific work units (e.g., ICU, emergency) (143) could have influenced the outcomes, particularly in the context of depression and anxiety.

Third, although systematic random sampling was used from COVID-19 screening rosters, only individuals present and consenting at the time of data collection were included. This may have excluded those absent due to illness, psychological distress, or other factors, introducing potential selection bias. As a result, prevalence estimates might be over- or underestimated.

Fourth, although our models incorporated hospital-level clustering using robust standard errors, we did not account for clustering at the unit or professional level. This may have led to underestimated standard errors and overestimated precision in some associations. Future studies should consider hierarchical or multilevel modeling to better reflect the nested structure of institutional settings.

Fifth, the absence of pre-pandemic mental health data prevents direct attribution of findings to the COVID-19 emergency. However, we compared our estimates to pre-pandemic literature in both healthcare and general populations in Peru, supporting the observed symptom increase during the early pandemic.

Sixth, the multivariable models included numerous covariates, which may have reduced statistical power to detect associations for predictors with low prevalence. Although the sample size was calculated based on a detectable effect size of 35%, the degrees of freedom required for modeling may have diminished sensitivity. Thus, non-significant associations should be interpreted cautiously.

Finally, some estimates showed wide confidence intervals due to low cell counts in certain exposure categories, which may affect the stability of these specific PR estimates. These results should therefore be interpreted with caution.

Despite these limitations, this preliminary study provides novel information using validated instruments to obtain a variety of characteristics that may trigger mental problems. Our findings can serve to lay the foundations for the development of new, better-designed research, which will ultimately drive the design of suitable preventive strategies. Another strength of the study is that it has captured a large and diverse sample from the two largest health systems in Peru. Finally, as mentioned in the Instruments, the Anxiety and Depression questionnaires only provide information about symptoms in patients, not a psychiatric diagnosis. Therefore, the results provided should be taken with caution.