Interpretation of findings

Our results suggest that there is a high prevalence of OSA risk factors among patients with dementia and MCI, which supports previous research indicating a higher prevalence of OSA in this population.^{Reference Shastri, Bangar and Holmes1–Reference Hahn, Wang, Andel and Fratiglioni4,Reference Guarnieri, Adorni, Musicco, Appollonio, Bonanni and Caffarra12} Although we were unable to compare these patients with the general population, these findings highlight that this group would benefit from routine OSA screening. At present, patients may be asked about their sleep at their initial consultation; however, this is not consistently carried out and recorded. The low frequency of the variables ‘snoring’ (0.9%) and ‘observed to stop breathing’ (0.3%) represent poor sleep history documentation for patients with memory impairment. This is in accordance with the results of an Alzheimer's Disease Neuroimaging Initiative study that measured sleep-disordered breathing using self-reporting (clinician documentation).^{Reference Osorio, Gumb, Pirraglia, Varga, Lu and Lim2} This cohort was found to have a prevalence of 7%, well below the general prevalence in older adults (50% in males and 25% in females), also highlighting an underreporting of sleep disturbances in this population.

Furthermore, many of the parameters, including neck circumference, blood pressure and BMI, are not routinely accounted for in memory services. As such, risk scores are likely to be higher than reported, as these factors are not considered in the scoring. It is possible that many high-risk patients are not being identified or referred for further sleep assessment. With routine use of tools such as the modified STOP BANG, patients will be more likely to receive the indicated treatment or investigations for OSA.

Of note, among Black, Asian and minority ethnic (BAME) individuals, there was a greater percentage of individuals in the high-risk group for OSA. However, the BAME population was severely underrepresented in this sample. It is likely that the results were confounded by selection bias due to the distinct characteristics of the individuals that did attend memory services. For example, it is possible that this group had more severe cognitive impairment than those that did not attend.

Those in the high-risk group had the highest prevalence of heavy smoking, which would support previous findings of this being a risk factor for OSA.^{Reference Lin ni, Li and Zhang13} The seemingly random distribution of the prevalence of EAC suggests that the findings regarding this factor may have been due to chance. Overall, the low prevalence of both factors in our sample suggests that these risk factors may not be as relevant as others in understanding the rates of OSA in MCI or dementia. It should be noted that the search terms used in our study for heavy smoking were limited. The term ‘heavy smoking’ is subjective; therefore, the definition may vary among healthcare professionals. This affects the validity of our findings. As heavy smoking and EAC have been previously identified as risk factors for OSA, we suggest that further research is required to outline their relevance in patients with MCI and dementia.

Using the modified STOP-BANG, we were able to identify a large proportion (96.64%) of the sample as being at intermediate to high risk of developing OSA. This is likely to be in accordance with previous studies showing that 60% of patients with dementia or MCI also have OSA. Although the presence of OSA was not determined, this study does highlight that this population has many factors suggestive of OSA. In addition, the use of different sleep disorder screening tools prevents direct comparisons among studies.

A systematic review concluded that patients with MCI do not have more risk factors than the general population, hypothesising that the pathophysiology of MCI may alter the strength of the association of some risk factors.^{Reference Mubashir, Abrahamyan, Niazi, Piyasena, Arif and Wong14} Another study suggested that although the prevalence of OSA is much higher in patients with MCI, the scores of these patients on sleep disorder subscales did not differ compared with controls.^{Reference Kim, Lee, Lee, Jhoo and Woo15} This may be the reason that other studies have failed to capture the increased risk. Although we did not compare the modified STOP BANG scores with controls, our findings contrasted with those of others as we did observe high scores on our risk-factor-based OSA scale. It may be that the modified STOP BANG contains parameters more appropriate to the measurement of OSA in this population.

The literature has also suggested that poor sleep overall is the true cause of the association with MCI.^{Reference Dlugaj, Weinreich, Weimar, Stang, Dragano and Wessendorf6} This may hold true, but our findings suggest that the poor sleep subtype OSA is likely to be highly prevalent among those with MCI and dementia. This is in line with other findings suggesting that sleep-disordered breathing is the most prevalent form of sleep disturbance in patients with MCI, Alzheimer's dementia, vascular dementia, frontotemporal dementia and Lewy body/Parkinson's disease dementia.^{Reference Guarnieri, Adorni, Musicco, Appollonio, Bonanni and Caffarra12}

A myriad of risk factors and associations have been proposed as explanations for this link.^{Reference Shastri, Bangar and Holmes1} However, to our knowledge, there has been limited research looking at several risk factors in the same sample. Our study found that most patients had three or more risk factors. There may be a synergistic relationship between the characteristics present in populations with cognitive impairment; thus, future research should investigate potential clustering patterns of risk factors to better understand their combined effect size.

Strengths and limitations

One limitation of using retrospective data from the CRIS database is that patient progress notes record clinical observations, which include elements of subjectivity. Variability in patient history collection and recording also made it difficult to confirm the presence of risk factors. Our keyword search attempted to capture commonly used descriptions of various risk factors. It is important to acknowledge that these lists were not exhaustive. In addition, we were looking for the presence of variables that are not routinely collected. Thus, an absence from our keyword search does not necessarily correlate with an absence of the variable. It could be that the variable was not assessed or was documented in a way that our search could not detect. Therefore, we believe a large percentage of patients that scored as intermediate risk were high risk. A positive word search also did not necessarily correlate with a positive finding of the STOP BANG variable. Keyword searches were not proofread for negative findings or context. A positive word search reflected the identification of that word in patient notes. For example, when searching for ‘snoring,’ the sentence could have been ‘does not report snoring’ in the progress notes text. However, given that these parameters are not routinely assessed, we believe that documentation is much more likely to indicate their presence. Therefore, any negative findings would have had minimal impact on our results.

In addition, the study did not use a comparator group. Therefore, we cannot conclude that this population's risk is significantly higher than that of the general population; however, we have found that this population is at high risk. OSA has deleterious effects on those with MCI and dementia, making its identification more urgent in this population compared with the general population. If this condition is untreated, patients with OSA will have increased rates of decline and symptom onset.^{Reference Osorio, Gumb, Pirraglia, Varga, Lu and Lim2}

Moving forward

Patients with cognitive impairment exhibit high frequencies of risk factors associated with OSA; however, further longitudinal research is needed to estimate the role of multiple risk factors and risk factor clusters in patients with MCI and dementia. We are not aware of any prospective studies that would help to quantify the extent and significance of specific risk factors in patients with cognitive impairment. Future studies could also explore the strength of the associations of smoking and alcohol use with increased risk in this population. In addition, increased BAME representation in studies could establish the true existence and degree of risk variation across ethnicities. The appropriateness of STOP BANG, and potential additional risk factors, should also be explored for this patient group.

Our findings agree with previous literature that indicate the need for special attention to patient sleep in memory clinics. We suggest that mental health memory clinics, such as those in South West London and St George's Mental Health NHS Trust, implement a modified STOP BANG assessment as part of a standardised cognitive impairment evaluation. The findings highlight the need for routine OSA screening for individuals with MCI and dementia. There is currently no systematic screening for OSA in UK memory services; therefore, this group of patients may not have equal access to referrals and treatment for OSA. Previous research has stated that OSA is commonly undiagnosed and untreated;^{Reference Osman, Carter, Carberry and Eckert5,Reference Chung, Abdullah and Liao7} our findings suggest that this is also true for this population. The delay in diagnosis and treatment for OSA emphasises the need for a practical and reliable method of identifying patients at high risk of OSA. The STOP BANG was specifically designed as a reliable and easy-to-use screening tool to quantify the risk factors predicting OSA.^{Reference Nagappa, Liao, Wong, Auckley, Ramachandran and Memtsoudis16} The STOP BANG has also been widely validated in various populations, which is why we would propose its introduction as a standardised screening tool for use at initial memory assessments.

Screening lends itself to false positives and unnecessary referrals to sleep services. To mitigate this, we recommend the following actions based on risk score: patients with moderate and high risk scores should discuss lifestyle interventions for OSA^{Reference Carneiro-Barrera, Díaz-Román, Guillén-Riquelme and Buela-Casal17} with their general practitioner. High-risk patients should be referred to sleep services to determine their eligibility for and tolerance of treatments such as continuous positive airway pressure (CPAP).