This analysis of German SHI data found that just over half of all eligible individuals with OSA were prescribed PAP. In the absence of data on variables used to phenotype individuals with OSA in clinical practice, predictors of PAP prescription in this insured population included the presence of a variety of comorbidities (especially metabolic and cardiovascular diseases) and use of several classes of medication. However, clinical variables might play a more important role in determining PAP prescription.

One potential explanation for the low rate of PAP prescription in the study population could be that a substantial proportion of individuals with mild sleep apnoea on cardiorespiratory polygraphy may not have been referred to a sleep laboratory for further investigation, and were instead managed using options other than PAP, such as weight reduction and avoiding the supine position during sleep. Overall, the current findings provide valuable real-world data that are representative of the individuals and activities in clinical practice, and can therefore help to inform the overall care of individuals with OSA.

There is currently a lack of data on rates of PAP prescription in patients with OSA, and about factors that predict the prescription of PAP therapy. A recent meta-analysis found significant associations between six factors and the purchase of a PAP device by patients with OSA [23]. These were older age, more years of education, higher income, current smoking, higher Epworth Sleepiness Scale score and higher apnoea–hypopnoea/respiratory disturbance index. Interestingly, the association between the presence of hypertension/cardiovascular disease and PAP device purchase did not reach statistical significance in that analysis [23]. In contrast, cardiovascular diseases, including hypertension and heart failure, were significant predictors of PAP initiation in treatment-naïve patients with OSA in the current study.

Overall, the number of comorbidities in our study population was high, especially coronary artery disease, obesity, chronic obstructive pulmonary disease, diabetes mellitus and arterial hypertension, which is consistent with database studies from other countries [24, 25]. Obesity is a well-known risk factor for the development of OSA [26], and overweight/obese patients were more likely to start PAP therapy in the current analysis.

The burden of mental health-related comorbidities in the study population was relatively high, with nearly a quarter of all participants (22.8%) having depression, similar to rates reported in other studies of individuals with OSA [27, 28]. Relatively high rates of mental health issues in patients with OSA may be due to a variety of factors. OSA symptoms such as EDS, fatigue, poor concentration, irritability, psychomotor issues and weight gain can overlap with symptoms of depression [29]. Alternatively, depression might also be caused by residual EDS, or depression could coexist with OSA symptoms (such as EDS) [30]. Data from a small study (n = 50) indicated a possible connection between depression and residual EDS in OSA, and suggested that PAP therapy might be able to improve both [31].

The German SHI claims data used for this analysis provide a comprehensive, patient-level picture of all reimbursed health-related services, allowing determination of epidemiological estimates and healthcare resource utilization. The large dataset is representative of the German population [16] and, with approximately 90% of the population enrolled in the SHI system and no opt-out possibility for this anonymized analysis, Germany offers a near-ideal setting to analyse population-based epidemiological estimates. Another strength of the current analysis are the objective and consistent definitions of PAP usage or non-usage (control group).

There are also several limitations that need to be considered when interpreting the findings of this study. Firstly, the maximum observational period of 6 years may not reflect the entire diagnostic process and some patients with later treatment of OSA might have been missed due to the specific timeframes applied. Secondly, the strict inclusion criteria applied to simulate the diagnostic scheme could have biased the study results. It is likely that these criteria have high specificity but poor sensitivity for identifying treatment-naïve OSA patients in the population. Thirdly, specific PSG and PG results were not available in the claims database used, and it was not possible to distinguish between diagnostic and titration PSGs. This means that there are no data on the severity of OSA, and therefore the specific indication for PAP therapy (which is primarily recommended for individuals with moderate to severe OSA). In addition, the control group did not receive treatment for OSA, not even alternatives to PAP such as mandibular repositioning devices. One potential limitation of the large population included in this study is that small between-group differences that achieved statistical significance may not be clinically significant, so this needs to be considered when interpreting the study data.