Our study investigated the prevalence of HW at low, moderate, and high rOSA and the feasibility of a screening program targeted at the detection of undiagnosed OSA in a population of over 5,000 HW belonging to a large Italian University Hospital.

The proportion of response to the online questionnaire was similar in the present study and in a previous one, simultaneously carried out in the same University Hospital (31 vs 34%), and higher than others targeting HW populations (Spiteri et al. 2023). In agreement with the previous study, also the percentage of response among women and nurses was significantly higher than among men and other professional groups. No significant association was detected between the percentage of response and night shifts.

Our study showed that 13.6% of Verona HW were at high rOSA. Interestingly, this value falls in the lower range of diagnosed OSA prevalence (9–38%), reported for the general population aged > 18 years. It should be reminded, however, that the prevalence estimated for the working age classes (20–65 years old) is lower than the prevalence reported for elderly people (Senaratna et al. 2017).

The prevalence of rOSA, found in studies involving only HW, varied widely as a consequence of different characteristics of the enrolled populations and different methods for identifying at-risk individuals. Pascoe et al. (Pascoe et al. 2020) found a very high proportion of individuals at high rOSA (37%) in a population of 2851 American caregivers using the STOP questionnaire. Accordingly, 7% of this population even had a prior diagnosis of OSA. However, as acknowledged by the authors themselves, the percentage of respondents was very low (about 6%) in that study, so that nonresponse bias cannot be excluded (Verlato et al. 2010). In particular, the lack of sociodemographic data did not allow determining the age characteristics of the sample. Hence, considering the high number of subjects with a prior diagnosis of OSA (7%), it could be assumed that the HW who participated in this study were older than non-responders.

Three studies, dealing with OSA risk among HW, adopted the Berlin Questionnaire rather than SBQ (Seyedmehdi et al. 2016; Aydın Güçlü et al. 2019; Alexandropoulou et al. 2019). Seyedmehdi et al. (Seyedmehdi et al. 2016) reported a prevalence of 6.9% among a population of 715 HW. The lower prevalence as compared to our study could be explained by the younger age (mean = 33.5 years) beside our study (48.7 years). However, another study performed on 604 HW of similar age (median = 34.9 years) found a prevalence of rOSA (15.1%) comparable to our results (13.6%) (Aydın Güçlü et al. 2019). Hence, age alone cannot fully explain variability across studies, which is likely influenced by several other factors. For instance, Alexandropoulou et al. (Alexandropoulou et al. 2019) found a very high prevalence of rOSA in a Greek nursing staff population (20.5% of 444 participants), mostly overweight or obese. Indeed, the median BMI among participants was 27.2 kg/m2, significantly higher than the value of 23.4 kg/m2, found in our study population, and the higher BMI could possibly explain the higher prevalence in rOSA.

In our study, night shifts were associated with a lower rOSA. Paciorek et al. (Paciorek et al. 2011) investigated the effect of the night shifts in a group of 10 workers affected by OSAS, and found that the severity of the disease worsened after working at night. Other studies confirmed that sleep disorders can be triggered by night shifts (Yazdi et al. 2014; Pascoe et al. 2020). On the other hand, a systematic review by Yang et al. (Yang et al. 2021), exploring the association between shift work and OSA, reported a small, non-significant increase in the rOSA in night shift workers (RR = 1.05; 95% CI 0.85–1.30), so they concluded that the results were inconclusive. In our opinion, the negative association between night shifts and rOSA could be attributed to reverse causation, i.e., an healthy worker effect, as subjects with symptoms related to OSA are often excluded from night shifts.

During the clinical phase, we performed 64 polygraphies. Of these, 40 were positive, showing that SQB achieved a rather good PPV (68.8%) for mild to severe OSA. An even higher value was estimated by meta-analysis performed by Abrishami et al. (2010), which yielded a PPV of 84% for AHI ≥ 5 in patients without a history of sleep disorders. Restricting the outcome definition to moderate/severe or only severe OSA the value raised to 93% and 100%, respectively. PPV is largely influenced by the research setting, as it increases with the prevalence of the disease. However, also sensitivity, which is independent of disease prevalence from a theoretical point of view (Altman 1991) but not in clinical practice (Mark 2005), was quite good. Sensitivity was 73% and 97.6% in meta-analyses of studies on the general population or sleep clinic patients, respectively (Amra et al. 2018; Chen et al. 2021). High sensitivity (83.6%) was reported also in a study involving only sleep clinic patients (Luo et al. 2014; Amra et al. 2018). These data show that the PPV and the sensitivity of SBQ improve when tests are administered to high-prevalence populations, but they remain reliable even for healthy working populations.

Our study showed that the PPV increased from 68% to 96.9% when the questionnaire was administered by trained medical staff. This finding seems to suggest that screening for high rOSA in the workplace should be performed by occupational physicians during the periodic health surveillance, leading to an improvement of the test effectiveness, as well as health promotion in work settings.

This study has some limitations. First of all, the low response rate to the on-line questionnaire (31%). However, as discussed before, it is similar or even higher than in previous studies among HW, suggesting that greater adhesion rates are difficult to achieve in this population. Moreover, only a low number of HW underwent polygraphy (64 out of 212). Indeed, it should be underlined that adherence to the clinical phase was influenced by the COVID pandemic during 2020. Finally, sociodemographic and clinical data are self-reported, affecting their reliability.

Our study has also several strengths. To the best of our knowledge, our research is the second largest by the number of HW involved and it is the largest that has clinical and sociodemographic data available for all the HW included. Furthermore, using of a standardized questionnaire let us to achieve comparable and repeatable results. Moreover, although most of the studies that investigated the rOSA among HW used the Berlin Questionnaire, we chose the SBQ because it was validated in work settings, such as commercial drivers, had the highest sensitivity, and was the easiest and fastest to fill in (Abrishami et al. 2010; Amra et al. 2018; Lonia et al. 2020; Chen et al. 2021). A further strength regards the instrumental confirmation performed in HW classified at high rOSA by SBQ; this approach allowed us to calculate PPV of both self and trained medical staff administered SBQ.