Survey-based longitudinal analysis of burnout and behavioral profiles in a single Brazilian pediatric intensive care unit during and after the COVID-19 pandemic

Abstract

Burnout, characterized by emotional exhaustion, depersonalization, and reduced personal accomplishment, is common among healthcare workers, particularly in pediatric intensive care units (PICUs). Longitudinal data from low- and middle-income countries are limited. Repeated cross-sectional surveys were conducted among PICU staff at a Brazilian public hospital from 2020 to 2023 using the Maslach Burnout Inventory–Human Services Survey (MBI-HSS). Participants were classified into behavioral profiles (engaged, ineffective, overextended, disengaged, burnout). Emotional exhaustion increased from 2.47 ± 1.17 in 2020 to 3.33 ± 1.30 in 2023 (p = 0.027), while depersonalization rose from 1.13 ± 1.07 to 1.86 ± 0.98 (p = 0.014). Personal accomplishment declined from 4.62 ± 0.79 to 4.11 ± 0.84 (p = 0.048). Overextended staff increased from 22.5% to 44%, whereas engaged workers fell from 45% to 16%. Burnout worsened over time among PICU staff in this Brazilian tertiary center. Protected breaks, peer support, and AI-assisted monitoring may help preserve staff wellbeing and workforce stability in resource-limited settings.

Introduction

The concept of burnout traces back to Graham Greene’s 1960 novel A Burnt-Out Case¹, with Freudenberger later defining it clinically in 1974². Christina Maslach’s seminal work established burnout as a three-dimensional syndrome: Emotional exhaustion which is the core component of Burnout Syndrome (BS) representing the feeling of being emotionally overextended and depleted of emotional and physical resources. Depersonalization, the impersonal response towards the recipients of care, manifesting as cynicism and detachment. Reduced personal accomplishment, the diminished feelings of competence, productivity and success at work. These dimensions are operationalized through the Maslach Burnout Inventory (MBI), which remains the diagnostic gold standard³. These areas can create a vicious cycle: emotional exhaustion triggers detachment from patients, which in turn erodes professional self-worth, particularly in high-stakes environments like intensive care units (ICUs)⁴.

Maslach and Leiter’s multidimensional model of burnout identifies five behavioral profiles across the burnout–engagement continuum: Engaged, Ineffective, Overextended, Disengaged, and Burnout. The Engaged profile reflects low exhaustion and depersonalization with a strong sense of efficacy, representing a positive and energized relationship with work. The Ineffective profile shows reduced personal accomplishment without significant exhaustion or depersonalization, suggesting low self-efficacy but preserved motivation. The Overextended profile presents high exhaustion alone, often linked to workload stress but with sustained involvement. The Disengaged profile is characterized by elevated depersonalization and detachment while maintaining energy and competence, reflecting emotional withdrawal from work. Finally, the Burnout profile encompasses high exhaustion, high depersonalization, and low personal accomplishment, indicating severe strain and loss of psychological connection to work⁴.

Global data show that pre-pandemic burnout in pediatric ICU (PICU) staff from high-income countries ranged approximately from 42% to 77%, according to a systematic review of six studies using validated tools like the Maslach Burnout Inventory⁵. In Brazil, a 2014 observational study comparing pediatric intensivists and general pediatricians in two hospitals found burnout in 71% of intensivists versus 29% of generalists⁶. A broader 2016 survey involving intensivists across five Brazilian capitals reported 61.7% prevalence of burnout (defined by high score in at least one MBI dimension)⁷. During the COVID-19 pandemic, a multicenter Brazilian study across 29 PICUs documented a 24% prevalence of burnout, along with high emotional exhaustion scores and elevated anxiety, depression, and post-traumatic stress disorder (PTSD) symptoms among staff⁸. In the U.S., longitudinal research at a children’s hospital revealed an increase in pediatric healthcare worker burnout from 18.5% in April 2020 to 28.4% by March 2021⁹. Structural risk factors identified in Brazil included extended work shifts—such as ≥ 12-hour shifts—which were linked to higher burnout rates¹⁰, as well as inadequate institutional psychological support in PICUs (inferred from high distress and mental health symptom prevalence)⁸.

Healthcare professionals have long faced distressingly high burnout rates. A 2023 systematic review and meta-analysis examining 20,723 adult ICU physicians and nurses found that 41% of physicians and 44% of nurses experienced high-level burnout — defined by the Maslach Burnout Inventory — prior to and during the COVID-19 era¹¹. Burnout in intensive care has repeatedly been linked to depression, increased medical errors, and impaired clinical decision-making, particularly when intuitive judgment is needed^12,13.

The COVID-19 pandemic intensified this crisis. Frontline healthcare workers faced burnout prevalence exceeding 50%, with physicians and nurses disproportionately affected¹⁴. In adult ICUs, nurse burnout jumped from 37% before COVID-19 to 61% during the pandemic¹¹. Pediatric units, while seeing fewer COVID-19 cases, encountered significant stressors like redeployment and symptoms suggestive of Burnout Syndrome, as suggested by a longitudinal study conducted at a U.S. pediatric hospital that found that the proportion of healthcare workers experiencing high levels of emotional exhaustion and/or depersonalization increased significantly from 18.5% in April 2020 to 28.4% in March 2021⁹.

Post-pandemic, the healthcare environment remains highly challenging. The ongoing management of long COVID symptoms, persistent staffing shortages, and increasing patient backlogs continue to exert substantial stress on healthcare workers¹⁵. In PICUs, elevated rates of anxiety, depression, and PTSD among staff are strongly associated with increased intentions to leave their roles, posing a significant threat to the quality and continuity of care¹⁵.

Emerging artificial intelligence technologies offer promising tools for early detection of burnout risk. Machine-learning models trained on validated burnout inventories, combined with analysis of electronic health record usage patterns, have demonstrated an accuracy of 60.4% for self-reported burnout and 68.5% in emotional exhaustion¹⁶. Although ethical considerations around data privacy and informed consent persist, these tools have the potential to facilitate timely and proactive mental health interventions.

Evidence-based organizational interventions have demonstrated significant impact in reducing burnout among PICU staff. A pilot study implementing brief, on-shift mindfulness sessions for PICU nurses—consisting of 5-minute meditations before each shift—found a notable decrease in perceived stress, suggesting feasibility for stress-reduction in high-intensity settings¹⁷. Similarly, a quasi-experimental study involving a 6-week mindfulness-based intervention for PICU nurses showed statistically significant reductions in burnout and secondary traumatic stress, alongside increased compassion satisfaction¹⁸.

Yet, implementation remains inconsistent across regions. A European survey found that 78% of PICUs provided formal psychological support programs, reflecting an institutional commitment to clinician mental health¹⁹.

Cultural contexts shape burnout dynamics among healthcare workers. A cross-sectional Brazilian study found that 67% of PICU nurses reported difficult interactions with families as a major stressor—higher than comparable findings from U.S. units²⁰.

This study focuses on PICUs to address critical gaps in understanding post-pandemic burnout. We combine traditional MBI assessment with analysis of pandemic-specific stressors, structural determinants, and cultural influences. The primary objective is to quantify burnout prevalence among PICU staff, while secondary aims include: (1) identifying modifiable risk factors in Brazil’s unique context, (2) evaluating AI-assisted screening feasibility, and (3) proposing targeted interventions for LMIC PICU teams.

Methods

This is a cross-sectional study that collected data via an online questionnaire distributed by email to healthcare workers in the PICU of the Instituto da Criança at Hospital das Clínicas da Universidade de São Paulo. This research was approved by the National Committee of Research Ethics led by the National Health Council (Conselho Nacional de Saúde) on February 2nd 2021 under the number 36466520.1.1001.0068, in accordance with its guidelines, Resolution 466/2012, which regulates research involving human participants. All participants agreed to participate on this research after signing an online consent form.

The survey was administered three times (2021, 2022, and 2023) to assess burnout trends during and after the COVID-19 pandemic. A convenience sampling strategy was employed, where all healthcare workers currently employed within the participating PICU were invited to participate. This non-probabilistic approach was selected because it was practical, feasible, and appropriate for a workforce-based study conducted during a period of considerable operational pressure and staffing instability. Inclusion criteria required participants to be active, permanent healthcare staff in the PICU at the time of analysis. Exclusion criteria applied to non-healthcare workers, temporary staff, and those no longer employed in the unit. Unfortunately, due to high staff turnover rates and an erratic response rate, it was not possible to individually follow up participants during the three periods of assessment.

The questionnaire, distributed securely via Google Forms using the university’s email system, included the Maslach Burnout Inventory - Human Services Survey (MBI-HSS) in Portuguese, a validated tool for assessing burnout in healthcare professionals^4,21. Responses were frequency-based, evaluating three domains: emotional exhaustion, depersonalization, and personal accomplishment. Based on these scores, participants were categorized into behavioral profiles: engaged, ineffective, overextended, disengaged, or burnout. Additional demographic and subjective questions covered age, gender, profession, family proximity, financial concerns, team dynamics, and perceived pandemic impacts. The online questionnaire blocks duplicate entries and multiple responses as it only allows one email registration per person with one link to have access to the questionnaire.

Part of the data processing was conducted using AI-powered tools, including the Transform™ Platform by Mind Garden—a validated international publisher— and a license was acquired for MBI-HSS administration and profile analysis. The Transform™ system is a secure online platform for administering and scoring validated tools such as the Maslach Burnout Inventory–Human Services Survey (MBI-HSS). It automatically calculates scores for emotional exhaustion, depersonalization, and personal accomplishment, providing individual and group reports benchmarked against normative data. The system allows repeated assessments to track changes over time and provides a practical and scalable method for reliable burnout evaluation and follow-up in both research and institutional settings.

Data analysis

Categorical variables were described using absolute and relative frequencies, with associations assessed via chi-square or likelihood ratio tests²². Continuous burnout domain scores were summarized using means, standard deviations, medians, and quartiles. Differences across assessment periods were evaluated using analysis of variance (ANOVA) with Bonferroni multiple comparison tests to identify specific time points where significant variations occurred²³. The same approach was applied to compare scores across behavioral profiles.

Data analyses were conducted in IBM SPSS Statistics (v22.0), with data tabulation performed in Microsoft Excel 2013. A 5% significance threshold (α = 0.05) was applied for all tests.

Due to natural healthcare workforce turnover, the participant cohort varied across assessments. While the PICU’s operational structure and patient demographics remained consistent, the primary distinction was the exposure to patients with COVID-19 and its challenges in 2020–2022 versus post-pandemic conditions in 2023. The sample size was constrained by fixed staffing; to maximize participation, email reminders were sent, but no modifications to the total eligible population were possible.

Results

The study flow diagram is presented in Fig. 1. The initial assessment conducted in 2020 was sent to 105 healthcare workers from PICU and achieved a response rate of 38%, followed by a subsequent evaluation in 2021 that was sent to 88 staff members with a 30% response rate. Lastly, the 2023 assessment was sent to 90 potential participants and had a 27.7% response rate. As shown in Tables 1 and 2, the demographic characteristics of participants remained consistent across all three assessment periods, with no statistically significant differences observed in gender distribution, age, race, marital status, parental status, professional category, or years of professional experience.

Fig. 1.

Study flow diagram.

Table 1.

Demographic characteristics of the population.

Variable	Year			Total (N = 92)	p
	2020 (N = 40)	2021 (N = 27)	2023 (N = 25)
Gender					0.508
Female	34 (85)	20 (74.1)	21 (84)	75 (81.5)
Male	6 (15)	7 (25.9)	4 (16)	17 (18.5)
Age (years)					0.920
20–29	13 (32.5)	6 (22.2)	6 (24)	25 (27.2)
30–39	15 (37.5)	14 (51.9)	12 (48)	41 (44.6)
40–49	9 (22.5)	5 (18.5)	6 (24)	20 (21.7)
50–59	2 (5)	1 (3.7)	1 (4)	4 (4.3)
60+	1 (2.5)	1 (3.7)	0 (0)	2 (2.2)
Race					0.698
White	29 (72.5)	21 (77.8)	18 (72)	68 (73.9)
Mixed	7 (17.5)	6 (22.2)	5 (20)	18 (19.6)
Black	2 (5)	0 (0)	1 (4)	3 (3.3)
Asian	1 (2.5)	0 (0)	1 (4)	2 (2.2)
Other	1 (2.5)	0 (0)	0 (0)	1 (1.1)
Marital status					0.229*
Single	16 (40)	10 (37)	5 (20)	31 (33.7)
With partner	24 (60)	17 (63)	20 (80)	61 (66.3)
Has children					0.461*
No	26 (65)	15 (55.6)	18 (72)	59 (64.1)
Yes	14 (35)	12 (44.4)	7 (28)	33 (35.9)

*Chi-square test; **ANOVA; Likelihood ratio test used where unspecified% in brackets

Table 2.

Professional characteristics of the population.

Variable	Year			Total (N = 92)	p
	2020 (N = 40)	2021 (N = 27)	2023 (N = 25)
Profession					0.180
Nursing technician	4 (10)	4 (14.8)	6 (24)	14 (15.2)
Nurse	8 (20)	7 (25.9)	1 (4)	16 (17.4)
Physiotherapist	7 (17.5)	4 (14.8)	7 (28)	18 (19.6)
Nutritionist	2 (5)	0 (0)	0 (0)	2 (2.2)
Doctor	19 (47.5)	12 (44.4)	11 (44)	42 (45.7)
Years of professional experience					0.875
0–5	14 (35)	6 (22.2)	8 (32)	28 (30.4)
5–10	11 (27.5)	8 (29.6)	4 (16)	23 (25)
10–15	8 (20)	7 (25.9)	7 (28)	22 (23.9)
15–20	3 (7.5)	3 (11.1)	4 (16)	10 (10.9)
20 +	4 (10)	3 (11.1)	2 (8)	9 (9.8)
Financial instability caused by the pandemic?					0.040*
No	16 (40)	19 (70.4)	15 (60)	50 (54.3)
Yes	24 (60)	8 (29.6)	10 (40)	42 (45.7)
Do you think you could be affected by rising unemployment rates during/after the pandemic?					0.010*
No	11 (27.5)	9 (33.3)	16 (64)	36 (39.1)
Yes	29 (72.5)	18 (66.7)	9 (36)	56 (60.9)
How are you affected by media news?					0.138
Negatively	22 (55)	19 (70.4)	19 (76)	60 (65.2)
Not influenced	18 (45)	8 (29.6)	5 (20)	31 (33.7)
Positively	0 (0)	0 (0)	1 (4)	1 (1.1)

*Chi-square test; **ANOVA; Likelihood ratio test used where unspecified% in brackets; results with statistical significance are in bold

Furthermore, participants’ subjective perceptions of pandemic-related impacts also showed significant variation as presented in Table 2. Concerns about financial instability and unemployment effects were most pronounced in the 2020 assessment, with a majority of respondents reporting negative impacts. However, these concerns diminished significantly in subsequent years.

While behavioral profiles did not demonstrate statistically significant variation over time, as demonstrated in Table 3, notable patterns emerged in their distribution. The 2020 assessment revealed a predominance of the committed profile, while 2021 showed a more balanced distribution across engaged, ineffective, and overextended profiles. By 2023, the overextended profile had become most prevalent. Importantly, all Maslach Burnout Inventory subscales showed statistically significant changes over time. Emotional exhaustion scores increased from 2.47 ± 1.17 in 2020 to 3.33 ± 1.3 in 2023 (p = 0.027), while depersonalization rose from 1.13 ± 1.07 to 1.86 ± 0.98 (p = 0.014) during the same period. Personal accomplishment scores declined from 4.62 ± 0.79 to 4.11 ± 0.84 (p = 0.048), indicating progressive deterioration across all burnout dimensions.

Table 3.

MBI-HSS results of the population.

Variable	Year			Total (N = 92)	p
	2020 (N = 40)	2021 (N = 27)	2023 (N = 25)
Behavioral profile					0.195
Engaged	18 (45)	7 (25.9)	4 (16)	29 (31.5)
Ineffective	9 (22.5)	9 (33.3)	5 (20)	23 (25)
Overextended	9 (22.5)	8 (29.6)	11 (44)	28 (30.4)
Disengaged	0 (0)	1 (3.7)	1 (4)	2 (2.2)
Burnout	4 (10)	2 (7.4)	4 (16)	10 (10.9)
Emotional exhaustion					0.030**
Mean ± SD	2.47 ± 1.17	2.7 ± 1.36	3.33 ± 1.3	2.77 ± 1.3
Median (p25; p75)	2.2 (1.7; 3.38)	2.4 (1.8; 3.8)	3.3 (2.25; 4.35)	2.6 (1.8; 3.8)
Depersonalization					0.014**
Mean ± SD	1.13 ± 1.07	1.25 ± 0.9	1.86 ± 0.98	1.36 ± 1.04
Median (p25; p75)	0.6 (0.25; 1.6)	1.2 (0.6; 1.6)	1.6 (1; 2.6)	1.2 (0.6; 2)
Personal accomplishment					0.048**
Mean ± SD	4.62 ± 0.79	4.3 ± 0.86	4.11 ± 0.84	4.39 ± 0.84
Median (p25; p75)	4.8 (4.3; 5.1)	4.4 (3.8; 4.8)	4 (3.45; 4.9)	4.5 (3.8; 5)

*Chi-square test; **ANOVA; Likelihood ratio test used where unspecified% in brackets, except where centile stated; results with statistical significance are in bold

Table 4 provides detailed comparisons of MBI domain scores across assessment periods, confirming that emotional exhaustion and depersonalization were significantly higher in 2023 compared to 2020 (p = 0.027 and p = 0.014, respectively), while the difference in personal accomplishment approached but did not reach statistical significance (p = 0.052).

Table 4.

Multiple comparison results of scores in the 3 domains of MBI-HSS across time points.

Variable	Comparison		Mean difference	Standard error	p	CI (95%)
						Inferior	Superior
Emotional exhaustion	2020	2021	−0.22	0.31	> 0.999	−0.99	0.54
	2020	2023	−0.86	0.32	0.027	−1.65	−0.07
	2021	2023	−0.64	0.35	0.219	−1.49	0.22
Depersonalization	2020	2021	−0.13	0.25	> 0.999	−0.73	0.48
	2020	2023	−0.74	0.25	0.014	−1.36	−0.12
	2021	2023	−0.61	0.28	0.089	−1.29	0.06
Personal accomplishment	2020	2021	0.32	0.21	0.375	−0.18	0.82
	2020	2023	0.51	0.21	0.052	−0.00	1.02
	2021	2023	0.19	0.23	> 0.999	−0.37	0.75

Bonferroni multiple comparisons. Results with statistical significance are in bold

Table 5 demonstrates statistically significant differences in MBI domain scores across behavioral profiles (p < 0.001). Participants classified as engaged exhibited the most favorable pattern, with the lowest scores for emotional exhaustion and depersonalization coupled with the highest personal accomplishment. In contrast, those with ineffective profiles showed elevated emotional exhaustion and depersonalization alongside reduced personal accomplishment. The overextended profile was characterized by high emotional exhaustion but maintained moderate personal accomplishment and relatively lower depersonalization compared to disengaged participants, who displayed reduced emotional exhaustion but elevated depersonalization and diminished personal accomplishment. As expected, respondents meeting criteria for burnout demonstrated the most severe pattern, with the highest levels of emotional exhaustion and depersonalization combined with the lowest personal accomplishment scores.

Table 5.

Description of scores in the 3 domains by classification profile and results of comparative analyses.

Variable	Behavioral profile					p
	Engaged	Ineffective	Overextended	Disengaged	Burnout
Emotional exhaustion						< 0.001
Mean ± SD	1.59 ± 0.61	2.09 ± 0.52	4.03 ± 0.71	2.85 ± 0.07	4.24 ± 0.87
Median (p25; p75)	1.7 (1.05; 2.1)	2.2 (1.8; 2.6)	3.95 (3.33; 4.55)	2.85 (2.8; &)	3.8 (3.55; 4.95)
Depersonalization						< 0.001
Mean ± SD	0.72 ± 0.56	1.17 ± 0.62	1.37 ± 0.81	3 ± 0	3.34 ± 0.76
Median (p25; p75)	0.6 (0.4; 1.1)	1.2 (0.6; 1.4)	1.3 (0.65; 2.15)	3 (3; 3)	3.1 (2.8; 3.65)
Personal accomplishment						< 0.001
Mean ± SD	5.23 ± 0.35	3.86 ± 0.69	4.13 ± 0.66	3.85 ± 1.06	4 ± 0.87
Median (p25; p75)	5.1 (5; 5.5)	4 (3.4; 4.4)	4.05 (3.53; 4.75)	3.85 (3.1; &)	4.2 (3.38; 4.65)

ANOVA; & Cannot be estimated; results with statistical significance are in bold

Discussion

The longitudinal data from our study reveal a statistically significant worsening of burnout metrics in this Brazilian public hospital’s PICU. Emotional exhaustion scores escalated from 2.47 ± 1.17 in 2020 to 3.33 ± 1.3 in 2023 (p = 0.027), while depersonalization increased from 1.13 ± 1.07 to 1.86 ± 0.98 (p = 0.014) over the same period. These findings align with the concept of “delayed burnout,” where prolonged stress gradually depletes coping strategies²⁴. The 44% increase in emotional exhaustion among overextended staff (mean EE = 4.03 ± 0.71) is particularly alarming, as this group now represents nearly half of the workforce.

Behavioral profile patterns suggest important shifts in workforce well-being. In this study, the Overextended group was most common (30.4%), followed by Engaged (31.5%) and Ineffective (25%), closely resembling findings among Irish hospital doctors, where Overextended (30%) and Ineffective (22%) were similarly prominent²⁵. This alignment underscores emotional exhaustion as an early and widespread marker of occupational strain. The substantial proportion of overextended staff (~ 44%) in Brazilian healthcare settings also reflects international trends. For example, a U.S. national survey found that 69% of general surgery residents experienced burnout linked to workload and stress²⁶, while a European multicenter study of ICU intensivists during COVID-19 reported 51% severe burnout, again emphasizing emotional exhaustion as a key element²⁷.

Across the three survey years, the decline in Engaged staff (45% in 2020 to 16% in 2023) paralleled an important rise in Overextended profiles (22.5% to 44%). Compared with engaged colleagues, overextended individuals showed significantly higher emotional exhaustion (by 2.44 points, p < 0.001) and depersonalization (by 0.65 points, p = 0.005). This pattern reflects a transitional stage preceding full burnout, which increased from 10% to 16% over the study period. As demonstrated by Shanafelt et al., such shifts often foreshadow workforce instability, with burned-out clinicians showing a three times higher likelihood of leaving their profession²⁸. Altogether, these trends suggest that the Overextended state represents a critical point in the burnout trajectory characterized by heavy workload, cumulative fatigue, and escalating stress, but still preceding the disengagement seen in complete burnout.

However, the maintenance of moderate personal accomplishment in this sample, contrasts with the European pattern where personal accomplishment typically declines alongside emotional exhaustion²⁹. This could suggest that despite systemic pressures, the participants were able to maintain professional identity longer than their peers abroad—though without sustained support, they remain at risk of eventual breakdown.

The progressive rise in emotional exhaustion (p = 0.030) and decline in Personal Accomplishment (p = 0.048) over time in the current cohort suggest a gradual shift from engagement toward fatigue and reduced efficacy, likely reflecting increased workload and sustained stress. Comparable findings were also observed among medical and nursing students in a lower-middle-income country, where only one-third were Engaged and one-third Ineffective, with low personal efficacy as the main burnout feature³⁰. Similarly, the present data indicate that although full burnout remains limited, the predominance of Overextended and Ineffective profiles represents an early warning of deteriorating wellbeing³⁰.

Over the course of the study, the sample size decreased progressively from 38% to 27.7%, a limitation that mirrors the declining response rates described in similar burnout research among healthcare workers. Reduced participation over time can introduce selection bias, as those most affected by burnout are often less likely to engage in research activities, leading to an underestimation of its true prevalence. Hodkinson et al. demonstrated that physicians experiencing burnout are more likely to exhibit career disengagement and intentions to leave their posts, directly affecting participation and continuity in research³¹. Similarly, Chen et al. highlighted that high workload and perceived organizational stress significantly reduce engagement in institutional surveys, as healthcare professionals facing emotional exhaustion or depersonalization often deprioritize non-mandatory academic or research activities³². Thus, the differences observed in the present sample may reflect both systemic staff turnover and the reduced willingness of emotionally exhausted personnel to engage in voluntary research efforts—a well-recognized bias in burnout studies that may underestimate the true prevalence of the syndrome. In our Brazilian context, this phenomenon was also seen in others studies, such as noted by Villela et al.³³ in their article on Brazil’s healthcare workers.

In Brazil, the lack of a unified pediatric intensive care unit (PICU) research network remains a structural barrier to multicenter collaboration. Currently, burnout data are fragmented and limited to isolated institutions. This study, conducted in the largest hospital in Latin America, may serve as an important starting point for future multicenter research once a national PICU network is established—enabling standardized data collection, improved monitoring of staff wellbeing, and more comprehensive evaluation of interventions across the country.

While financial instability concerns decreased from 60% to 40%, media impact persisted, with 65% reporting negative psychological effects from coverage. This aligns with Dyrbye et al.‘s³⁴ findings that media coverage of healthcare crises exacerbates anxiety even as material conditions stabilize. In Brazil’s public health system, where staff often face criticism in mainstream media, this effect may be particularly pronounced³⁵.

Artificial intelligence is the branch of computer science that enables machines to learn from data, recognize complex patterns and make adaptive decisions that emulate human reasoning³⁶ and represents a new approach for assessing burnout among healthcare professionals. In this study it was used to analyze the answer to the MBI-HSS and categorize them into the behavioral patterns established by Dr Christina Maslach⁴.

In the future it could allow the analysis of multidimensional information such as responses to validated instruments like the Maslach Burnout Inventory, workload patterns, and contextual variables. Recent advances such as the Hierarchical burnout Prediction based on Activity Logs (HiPAL) model, which seeks to predict burnout based on electronic health records and activity logs, have demonstrated the capacity of machine learning algorithms and could predict burnout risk³⁷, providing continuous and automated surveillance of wellbeing rather than relying solely on periodic self-reported surveys. This integration of AI-based tools into burnout assessment could facilitate more frequent, objective, and scalable monitoring, improving diagnostic accuracy and enabling earlier, data-driven interventions aimed at preventing progression and supporting workforce resilience.

The preservation of moderate personal accomplishment (4.13 ± 0.66) in overextended workers suggests targeted support could yield meaningful recovery. This finding might be particularly relevant for Brazilian public hospitals, where resources for comprehensive mental health programs are limited but peer support networks show strong cultural acceptance³⁸. Profile-specific interventions adapted to our workforce composition could maximize impact within existing constraints.

However, the general decline in personal accomplishment from 4.62 ± 0.79 to 4.11 ± 0.84 (p = 0.048) indicates a workforce experiencing a loss of professional efficacy. This decline could be likely driven by chronic understaffing, with over half (52%) of ICU staff reporting increased workload, and moral distress stemming from pediatric COVID-19 outcomes³⁹. Similar effects have been observed in Brazilian intensive care units, where elevated patient-to-staff ratios correlate with higher burnout and diminished job satisfaction, a recognized marker of professional efficacy decline²⁹.

Our cohort of participants was predominantly female (81.5%), partnered (66.3%), and with children (35.9%). It reflects demographic patterns frequently observed in healthcare burnout research. A meta-analysis by Gómez-Urquiza et al. showed that gender, marital status, and parenthood modestly influence burnout risk, with female and married professionals often reporting higher emotional exhaustion but lower depersonalization compared to single or childless peers, likely due to differing social and cultural coping mechanisms⁴⁰. Despite no significant associations in our data currently, the progressive rise in emotional exhaustion might suggest that traditional protective factors, such as family support, may be insufficient in mitigating chronic occupational stress, emphasizing the need for institutional strategies that account for cultural and familial dimensions of burnout.

The next steps in our research would be implementing changes to improve staff wellbeing, followed by a new assessment with MBI-HSS. The following suggestions are the ones likely to be carried out: First, apply “protected time” protocols ensuring uninterrupted breaks during shifts that could reduce emotional exhaustion and general burnout levels, as demonstrated in a British pilot study showing a significant improvement these areas⁴¹. Second, establishing peer support circles facilitated by senior staff which would harness Brazil’s collectivist work culture, require minimal resources, and effectively reduce burnout and stress—as demonstrated by Meredith et al. in similar settings³⁸. Third, implementing media literacy workshops to equip staff to better process negative coverage and misinformation, enhancing their psychological resilience—mirroring the approach described by Huang et al.⁴². Fourth, putting into practice a rotating low-acuity assignment system tailored to individual staff needs which could offer periodic relief for overextended workers, as demonstrated to be effective by Uhde et al.⁴³. Fifth, implementing anonymous real-time feedback systems to help identify burnout triggers while overcoming cultural reluctance to report distress, modeled after a US system that showed reduction in burnout levels⁴⁴.

The Brazilian Unified Health System is publicly funded and provides free healthcare to all Brazilian citizens (SUS). It faces significant challenges in implementing burnout interventions due to rigid personnel policies. Nonetheless, addressing burnout among healthcare workers remains critical. A 2019 study estimated that physician burnout costs the U.S. healthcare system approximately $4.6 billion annually⁴⁵. Although comparable data for Brazil are lacking, it is hypothesized that the financial burden of burnout is substantial within the SUS as well.

PICU staff encounter intense emotional burdens, as they care for critically ill children while simultaneously supporting distressed families—creating compounded psychological stress. This challenge is amplified in Brazil’s family-centered care model which promotes high levels of family involvement in care and indirect emotional impact on staff⁴⁶. As reported in 2023 by Lima-Setta et al., PICU workers from a different Brazilian unit report notably elevated burnout levels, with 24% meeting high burnout criteria amid pandemic peaks, underscoring the need for targeted support systems⁸.

The study’s strengths include its longitudinal design in a representative Brazilian public hospital, providing rare documentation of burnout progression through the pandemic’s acute and chronic phases. The integration of quantitative MBI data with behavioral profiling offers nuanced understanding beyond conventional surveys. However, limitations include potential attrition bias, as the 18% dropout rate may underrepresent the most severely affected staff. The small disengaged sample (N = 2) also warrants cautious interpretation, suggesting need for qualitative follow-up studies to capture this group’s experiences. Lastly, the inability to follow up employes across the years made analysis more cautious.

Conclusion

This longitudinal study highlights a concerning escalation of burnout among PICU staff in a Brazilian public hospital, marked by significant increases in emotional exhaustion and depersonalization, alongside a sharp decline in engaged workers. The rise of “overextended” staff—now nearly half the workforce—signals a critical intervention window before irreversible attrition occurs. While cultural and systemic barriers complicate mitigation efforts, targeted strategies such as protected break times, peer support networks, and AI-driven early detection could help curb burnout progression. The preservation of moderate personal accomplishment suggests resilience remains, but without urgent, context-specific interventions which includes policy reforms to address rigid SUS regulations, the healthcare system risks further workforce depletion. These findings underscore the need for PICU-tailored solutions that address both global burnout trends and Brazil’s unique healthcare challenges, ensuring staff well-being and sustained quality of care.

Author contributions

IR: Conceptualization, methodology, investigation, data curation, writing—original draft, writing—reviewing and editing. APC: Conceptualization, methodology, investigation. PFG: Investigation. CJ: Conceptualization, writing—reviewing and editing. WB: Conceptualization, writing—reviewing and editing. AFD: Conceptualization, methodology, investigation, writing—reviewing and editing.

Funding

This study did not receive any external financial support.

Data availability

The datasets used and/or analyzed during this study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval and consent to participate

This research was approved by the National Committee of Research Ethics led by the National Health Council (Conselho Nacional de Saúde) under the number 36466520.1.1001.0068, in accordance with its guidelines, Resolution 466/2012, which regulates research involving human participants. All participants consented to participate prior to answering the questionnaire.