Study and population
A cross-sectional study was conducted between June and August 2019, which enrolled people aged above 18 using a proportionate random sample from all Lebanese governorates (Beirut, Mount Lebanon, North, South, and Beqaa). Each governorate is divided into Caza (stratum), which in turn is divided into multiple villages. Two villages per casa were selected randomly from the list of villages provided by the Central Agency of Statistics in Lebanon. Households were randomly selected from each village using an online software (Research Randomizer)26. The interviews were conducted inside the household and all adults living there were eligible to participate. Excluded were people unable to understand Arabic (n=1) -the national language- and people with cognitive impairment (n=5) (trouble remembering or concentrating)27 as reported by a family member.
Sample size calculation
According to the Epi info, sample size calculations assuming a 31% frequency of OSA among the general population in the absence of similar studies in the country, a 95% confidence level, a power of 80%, and a sample of 329 contestants was required to fulfill the objectives and allow for adequate power for bivariate and multivariable analyses. We conducted the questionnaire on a total of 600 individuals. Eighty-seven were excluded from the study (14.5%), eighty-one refused to participate, one was unable to understand Arabic and five people had cognitive impairment; twelve terminated the interview before completion (2%). A total of 472 (83.5%) completed the interview and was included in the final analysis.
Data collection
A standardized method of a face-to-face interview was adopted by a trained and study-independent personnel. The questionnaire was divided into four parts:
The first part (Part 1: Socio-demographic characteristics) was collected through a multiple-choice format of 19 questions. The gender, the age, the weight, the height, the educational level (low (illiterate/primary), intermediate (complementary/ secondary) or high (university level) and the health insurance were mentioned. The governorate and the lifestyle (smoking, alcohol and coffee consumption) were also included, in addition to the number of traffic accidents per year. The monthly income was divided into 4 categories, as follows, based on the salary: none, low (<1000 USD), intermediate (1000–2000 USD), and high (>2000 USD). We asked about the last medical visit and its yearly frequency. The BMI was calculated from the measured weight and height of the individual.
The second part (Part 2: Personal diseases) evaluated whether the respondent had ever heard about OSA and if he/she had a prior physician diagnosis of OSA. If yes, the respondent was asked if he/she was currently on any sort of treatment. Questions about the history of personal diseases included HTN, diabetes (DM), cerebrovascular accident (CVA), arrhythmia, and myocardial infarction (MI). Nocturia was also mentioned, by citing the number of times a participant gets out of bed, at night, to urinate.
The third part (Part 3: The Knowledge Scale) intended to get information concerning the knowledge of OSA, subdivided into two main focuses: suggestive characteristics and possible complications. There were 13 items concerning suggestive OSA symptoms and characteristics (10 correct, plus 3 distractors) and 13 possible complications (7 correct, plus 6 distractors), as mentioned in the Appendix1. This yielded an overall score ranging between 0 and 26.
Currently, the Obstructive Sleep Apnea and Attitudes Questionnaire (OSAKA) is a valid tool to assess OSA knowledge among physicians28. However, few scales consisted of questions addressed to the general population22,23. To collect data related to knowledge and to investigate disease-related beliefs among the general population, we used a questionnaire based on the previously published “Guidelines for clinical practice in OSAHS in adults”29. It has been translated from the Loraine’s questionnaire23 into Arabic. Forward and back method was adopted for the translation from French to Arabic then from Arabic to French by two different translators. The two French versions were compared; discrepancies were resolved by consensus between the authors and the translators.
We considered the following answers as acceptable for the symptoms and characteristics: Snoring, respiratory breaks, daytime fatigue, night suffocating sensation, non-restorative sleep, daytime somnolence, concentration disorder, morning headache, nocturia, and obesity. For health consequences, those were the right answers: stoke, DM, HTN, dementia, cardiac arrythmia, MI, and road accidents. Score details are mentioned in the Appendix2.
The fourth part (Part 4: Screening): A thorough literature review highlighted the presence of well-validated scales used in research studies to diagnose OSA: the Epworth Sleepiness Scale (ESS)30 and the STOP-BANG questionnaire (SBQ)31. These two scales were chosen since the ESS is recommended to be included in screening evaluations32,33, and the SBQ for being superior in detecting OSA in the general population34,35.
ESS inquires about falling asleep in some circumstances, presenting a subjective measure of daytime sleepiness. A scale of 0–3 (0 means “would never doze” and 3 means “high chance of dozing”). The total score ranges between 0 and 24. Higher scores indicate more daytime sleepiness; the cutoff for normal daytime sleepiness is 1030. The Arabic form has been validated as an authentic tool36. The SBQ, also valid in Arabic, included the STOP questions: Snoring behavior, Tiredness, Obstruction (gasping) and Pressure (hypertension). From the BANG questions we included the Neck circumference37. BMI, Age, and Gender were included in the score based on previous answers to avoid repetition and redundancy. It is scaled as low, intermediate and high OSA Risk based on the number of positive answers31 – Appendix3-.
A pilot study was run on about 20 subjects -not included in the study- to ensure the understanding and acceptability of the questions in the general population. Few linguistic modifications improved the response rate in the final questionnaire.
Statistical analysis
Statistical Package for Social Science (SPSS) version 23 was used for the statistical analyses. Descriptive statistics were presented using mean and standard deviation for continuous measures, frequencies, and percentages for categorical variables.
The Student t-test and ANOVA test were used to assess the association between each continuous independent variable (ESS total score, SBQ total score, and knowledge score) and the sociodemographic and other variables. To calculate the p-value of the statistical significance, the Bonferroni correction compensates for that increase by testing each hypothesis at a significance level of α/m, where α is the desired overall alpha level and m is the number of hypotheses/tests conducted (23). Concerning the different scores, we tested 19 hypotheses/variables in each model, with a desired error α of 0.05; therefore, the Bonferroni correction would test each hypothesis at a p-value of 0.05/19=0.002.
Multivariable linear regression models were done to explore factors associated with the three scores as dependent variables and taking all variables that showed a p≤0.002 in the bivariate analysis as independent variables. A p<0.05 in the multivariable model was considered significant. Moreover, Cronbach’s alpha was recorded for reliability analysis for each scale.