Symptom content analysis of OSA questionnaires: time to identify and improve relevance of diversity of OSA symptoms?

STUDY OBJECTIVES: Obstructive sleep apnea (OSA) is a heterogeneous condition covering many clinical phenotypes in terms of the diversity of symptoms. Patient-based OSA screening questionnaires used in routine practice contain significantly varying contents that can impact the reliability and validity of the screening. We investigated to what extent common patient-based OSA screening questionnaires differ or overlap in their item content by conducting a rigorous, methodical, and quantified content overlap analysis. METHODS: We conducted an item content analysis of 11 OSA screening questionnaires validated in adult populations and characterized their overlap using a 4-step approach: (1) selection of OSA screening questionnaires, (2) item extraction and selection, (3) extraction of symptoms from items, and (4) assessment of content overlap with the Jaccard index (from 0, no overlap, to 1, full overlap). RESULTS: We extracted 72 items that provided 25 distinct symptoms from 11 selected OSA questionnaires. The overlap between them was weak (mean Jaccard index: 0.224; ranging from 0.138 to 0.329). All questionnaires contained symptoms of the "OSA symptom" dimension (eg, snoring or witnessed apneas). The STOP-BANG (0.329) and the Berlin (0.280) questionnaires exhibited the highest overlap content. Ten symptoms (40%) were investigated in only 1 questionnaire. CONCLUSIONS: The heterogeneity of content and the low overlap across these questionnaires reflect the challenges of screening OSA. The different OSA questionnaires potentially capture varying aspects of the disorder, with the risk of biased results in studies. Suggestions are made for better OSA screening and refinement of clinical OSA phenotypes. CITATION: Gauld C, Baillieul S, Martin VP, et al. Symptom content analysis of OSA questionnaires: time to identify and improve relevance of diversity of OSA symptoms? J Clin Sleep Med. 2024;20(7):1105-1117.

Sleepiness should be reinvestigated through the lens of clinical neurophysiology: A mixed expertal and big-data Natural Language Processing approach.

Historically, the field of sleep medicine has revolved around electrophysiological tools. However, the use of these tools as a neurophysiological method of investigation seems to be underrepresented today, from both international recommendations and sleep centers, in contrast to behavioral and psychometric tools. The aim of this article is to combine a data-driven approach and neurophysiological and sleep medicine expertise to confirm or refute the hypothesis that neurophysiology has declined in favor of behavioral or self-reported dimensions in sleep medicine for the investigation of sleepiness, despite the use of electrophysiological tools. Using Natural Language Processing methods, we analyzed the abstracts of the 18,370 articles indexed by PubMed containing the terms 'sleepiness' or 'sleepy' in the title, abstract, or keywords. For this purpose, we examined these abstracts using two methods: a lexical network, enabling the identification of concepts (neurophysiological or clinical) related to sleepiness in these articles and their interconnections; furthermore, we analyzed the temporal evolution of these concepts to extract historical trends. These results confirm the hypothesis that neurophysiology has declined in favor of behavioral or self-reported dimensions in sleep medicine for the investigation of sleepiness. In order to bring sleepiness measurements closer to brain functioning and to reintroduce neurophysiology into sleep medicine, we discuss two strategies: the first is reanalyzing electrophysiological signals collected during the standard sleep electrophysiological test; the second takes advantage of the current trend towards dimensional models of sleepiness to situate clinical neurophysiology at the heart of the redefinition of sleepiness.

Objective evaluation of excessive daytime sleepiness.

Excessive daytime sleepiness (EDS) is multifactorial. It combines, among other things, an excessive propensity to fall asleep ("physiological sleepiness") and a continuous non-imperative sleepiness (or drowsiness/hypo-arousal) leading to difficulties remaining awake and maintaining sustained attention and vigilance over the long term ("manifest sleepiness"). There is no stand-alone biological measure of EDS. EDS measures can either capture the severity of physiological sleepiness, which corresponds to the propensity to fall asleep, or the severity of manifest sleepiness, which corresponds to behavioral consequences of sleepiness and reduced vigilance. Neuropsychological tests (The psychomotor vigilance task (PVT), Oxford Sleep Resistance Test (OSLeR), Sustained Attention to Response Task (SART)) explore manifest sleepiness through several sustained attention tests but the lack of normative values and standardized protocols make the results difficult to interpret and use in clinical practice. Neurophysiological tests explore the two main aspects of EDS, i.e. the propensity to fall asleep (Multiple sleep latency test, MSLT) and the capacity to remain awake (Maintenance of wakefulness test, MWT). The MSLT and the MWT are widely used in clinical practice. The MSLT is recognized as the "gold standard" test for measuring the severity of the propensity to fall asleep and it is a diagnostic criterion for narcolepsy. The MWT measures the ability to stay awake. The MWT is not a diagnostic test as it is recommended only to evaluate the evolution of EDS and efficacy of EDS treatment. Even if some efforts to standardize the protocols for administration of these tests have been ongoing, MSLT and MWT have numerous limitations: age effect, floor or ceiling effects, binding protocol, no normal or cutoff value (or determined in small samples), and no or low test-retest values in some pathologies. Moreover, the recommended electrophysiological set-up and the determination of sleep onset using the 30­sec epochs scoring rule show some limitations. New, more precise neurophysiological techniques should aim to detect very brief periods of physiological sleepiness and, in the future, the brain local phenomenon of sleepiness likely to underpin drowsiness, which could be called "physiological drowsiness".

The 68 symptoms of the clinical high risk for psychosis: Low similarity among fourteen screening questionnaires.

The Clinical High Risk for psychosis (CHR) is a heterogeneous condition with multiple symptoms. CHR screening is challenging in routine care, as a wide variety of questionnaires exists. We propose to explore the extent to which these questionnaires differ or overlap in item content. We performed a systematic and quantitative analysis of item content in a set of widely-used CHR screening questionnaires. Items were extracted from questionnaires and reworded according to the Structured Interview for Psychosis-Risk Syndromes (SIPS). Then, symptoms were generated from individual items. The Jaccard Index was calculated to assess content overlap. The 14 analysed questionnaires were composed of 347 items, from which 198 symptoms were generated and, in turn, collapsed into 68 distinct symptoms. Positive symptoms were the most commonly represented. The overall overlap across questionnaires showed weak similarity (Jaccard = 0.19±0.50). CHR screening questionnaires might evaluate the same broad clinical construct, but have different scopes within that construct, and may be more or less comprehensive than one another. Clinicians and researchers should be mindful of the specific features of each instrument for optimal CHR screening.

Content analysis of insomnia questionnaires: A step to better evaluate the complex and multifaceted construct of insomnia disorder.

Insomnia disorder is a mental disorder that includes various types of symptoms (e.g., insomnia initiating, worries, mood disturbances) and impairments (e.g., distress related to sleep alterations). Self-report questionnaires are the most common method for assessing insomnia but no systematic quantified analysis of their content and overlap has been carried out. We used content analysis and a visualization method to better identify the different types of clinical manifestations that are investigated by nine commonly used insomnia questionnaires for adults and the Jaccard index to quantify the degree to which they overlap. Content analysis found and visualized 16 different clinical manifestations classified into five dimensions ("Insomnia symptoms", "Insomnia-related symptoms", "Daytime symptoms", "Insomnia-related impairments", "Sleep behaviors"). The average Jaccard Index was 0.409 (moderate overlap in content). There is a lack of distinction between symptoms and impairments, and the assessment of sleep duration and hyperarousal symptoms remains overlooked. This preliminary analysis makes it possible to visualize the content of each of the nine questionnaires and to select the most appropriate questionnaire based on the issue to be addressed. Suggestions are made regarding the development of future questionnaires to better distinguish symptoms and impairments, and the different phenotypes of insomnia disorder.

Systematic Item Content and Overlap Analysis of Self-Reported Multiple Sleep Disorder Screening Questionnaires in Adults.

Sleep disorders are defined on the basis of diagnostic criteria presented in medical classifications. However, no consensus has emerged on the exact list of operational symptoms that should be systematically investigated in the field of sleep medicine. We propose a systematic analysis of sleep symptoms that figure in a set of self-reported multiple sleep disorder screening questionnaires for adult populations, to identify the content overlap of symptoms that probe the presence of central sleep symptoms, and to highlight the potential level of heterogeneity among sleep disorder questionnaires. The method comprises three steps: (i) the selection of self-reported multiple sleep disorder screening questionnaires; (ii) item extraction and selection; (iii) the extraction of symptoms from items. Frequency of sleep symptoms and content overlap (Jaccard Index) are analyzed. We extracted 469 items that provide 60 different symptoms from 12 questionnaires. Insomnia, somnolence, and sleep-related breathing symptoms were found in all the questionnaires. The mean overlap among all questionnaires evaluated with the Jaccard Index is 0.44, i.e., moderate similarity. Despite limitations related to the selection of questionnaires and the symptom extraction and harmonization, this study underlines the need to standardize sleep symptom contents for sleep medicine in order to enhance the practicability, reliability, and validity of sleep disorder diagnoses.

Sleepiness in adults: An umbrella review of a complex construct.

Sleepiness involves many dimensions that require investigation. Since sleepiness is often defined operationally, we exhaustively inventoried all the assessment tools designed to measure it in an umbrella review, without any preconceptions, i.e. a review of reviews. We included all reviews and systematic reviews related to sleepiness assessment tools published up to March 2021. Three investigators independently assessed the eligibility of studies for inclusion and identified 36 relevant reviews. In total, 99 tools were identified and classified into 8 categories. We classified them depending on their category, their publication year and the number of mentions in the 36 included reviews. The 6 most frequently cited were the Epworth sleepiness scale, the multiple sleep latency test, the maintenance of wakefulness test, the Stanford sleepiness scale, the Karolinska sleepiness scale, and the psychomotor vigilance task. Despite the limitation that we may have missed some recently developed tools, this historical perspective on sleepiness measurement is a first step toward a better delineation of the different dimensions underlying the constructs of sleepiness, and will serve as a basis for further discussion in the clinical and research sleep community.

How Does Comparison With Artificial Intelligence Shed Light on the Way Clinicians Reason? A Cross-Talk Perspective.

In order to create a dynamic for the psychiatry of the future, bringing together digital technology and clinical practice, we propose in this paper a cross-teaching translational roadmap comparing clinical reasoning with computational reasoning. Based on the relevant literature on clinical ways of thinking, we differentiate the process of clinical judgment into four main stages: collection of variables, theoretical background, construction of the model, and use of the model. We detail, for each step, parallels between: i) clinical reasoning; ii) the ML engineer methodology to build a ML model; iii) and the ML model itself. Such analysis supports the understanding of the empirical practice of each of the disciplines (psychiatry and ML engineering). Thus, ML does not only bring methods to the clinician, but also supports educational issues for clinical practice. Psychiatry can rely on developments in ML reasoning to shed light on its own practice in a clever way. In return, this analysis highlights the importance of subjectivity of the ML engineers and their methodologies.

How to Design a Relevant Corpus for Sleepiness Detection Through Voice?

This article presents research on the detection of pathologies affecting speech through automatic analysis. Voice processing has indeed been used for evaluating several diseases such as Parkinson, Alzheimer, or depression. If some studies present results that seem sufficient for clinical applications, this is not the case for the detection of sleepiness. Even two international challenges and the recent advent of deep learning techniques have still not managed to change this situation. This article explores the hypothesis that the observed average performances of automatic processing find their cause in the design of the corpora. To this aim, we first discuss and refine the concept of sleepiness related to the ground-truth labels. Second, we present an in-depth study of four corpora, bringing to light the methodological choices that have been made and the underlying biases they may have induced. Finally, in light of this information, we propose guidelines for the design of new corpora.