Severe obstructive sleep apnoea can be accurately diagnosed in primary care centres

Abstract

We aimed to assess an obstructive sleep apnoea (OSA) diagnostic approach performed solely in primary care centres (PCC) with the support of an autoscoring home sleep apnoea testing (aHSAT, ApneaLink^TM Air) device and compare the diagnoses with those undertaken by the manual analysis of home sleep apnoea testing (mHSAT), and polysomnography (PSG) if necessary, of a certified sleep specialist. This multicentre, cross-sectional study was undertaken between April 2016 and November 2020. We randomly selected patients aged 30–70 years with a high probability of OSA (≥ 3 points on the STOP-Bang questionnaire) who were visiting any of the four PCCs assigned for referral to the University Hospital Doctor Josep Trueta, Girona, Spain. 2599 patients were assessed for eligibility; 403 provided a high probability of OSA and 329 could be compared between PCC and hospital. 210 (63.8%) patients were male and the mean age was 56.5 (SD: 9.2) years. The global agreement between PCC and hospital diagnoses was 41.6% and severe OSA showed the highest level of agreement (96.2%). The Kappa index for severe OSA was 0.46 (95% CI: 0.37, 0.55) and the specificity was 0.99 (95% CI: 0.97, 1.00). The ApneaLink^TM Air device showed high specificity for severe OSA in a high-risk primary care population with a high pre-test probability of OSA. When aHSAT indicates severe OSA, hospital confirmation may be unnecessary, whereas negative or moderate findings may still require specialist assessment.

Introduction

Obstructive sleep apnoea (OSA) is the consequence of repetitive periods of breathing cessation, causing disrupted sleep¹. Men are more prone to suffer from this disorder; the estimated prevalence of OSA in the 30–70-year-old North American population is 33.9% in men and 17.4% in women². In addition, the rising obesity levels in the western population could, partially, explain the increasing trend in OSA prevalence³. Notably, OSA is largely underdiagnosed and undertreated^1,4, mainly because of the low level of suspicion among primary care physicians (PCP)⁵ and the elevated number of paucisymptomatic⁶ and asymptomatic patients³. Also, undiagnosed OSA is associated with high cardiovascular morbidity and mortality and poor quality of life if untreated^7,8. Besides, this lack and delay of diagnosis and treatment entails increased direct and indirect health costs⁹. On the contrary, early management of OSA patients has been proven to be much more cost-effective¹⁰ and could positively impact the consequences of this disorder¹¹.

The gold standard method for the objective testing of OSA is in-hospital polysomnography (PSG), but manual analysis of home sleep apnoea testing (mHSAT) is also an accepted diagnostic method^12,13. However, both require certified sleep specialists to perform the scoring¹³. Interestingly, some autoscoring home sleep apnoea testing (aHSAT) devices have been demonstrated to correctly diagnose OSA, especially in highly symptomatic patients^14–16. The rising obesity rates¹⁷ in the general population and the limited resources of sleep laboratories¹⁸ have sparked an interest in involving PCP and other healthcare providers in the early diagnosis of this disorder¹⁹. Indeed, some studies have pointed out that PCP could screen OSA patients to prevent unnecessary referrals²⁰, a management model that could be more cost-effective than that performed by hospital sleep units^21–24. However, only a few studies^25,26, as far as we know, have evaluated the diagnosis of OSA carried out exclusively by PCP, i.e., without any intervention from the in-hospital specialist in the diagnostic process.

Therefore, the objective of our study was to assess an entire OSA diagnostic approach performed by a PCP with support from an aHSAT device and to compare the resulting diagnoses with those undertaken by a hospital sleep unit specialist using the mHSAT and, if necessary, the PSG.

Methods

Current prevalent practice

In Catalonia, patients presenting to a PCP with clinically suspected OSA are referred to a pulmonary physician, usually working at the hospital. After a clinical interview, the pulmonary physician may refer the patient to a sleep specialist in the same hospital, who may order an HSAT or a hospital PSG. In the first case, patients have to collect the device from the hospital and return it the next day so the sleep specialist can manually analyse the records. When a PSG is required, the limited capacity to meet its demand may lead to a delay in OSA diagnosis^27,28. In both cases, after an HSAT or PSG, the sleep specialist makes the diagnosis, notifies the patient about it, and proposes a treatment.

Study design and participants

In this multicentre, cross-sectional study, we evaluated the diagnostic accuracy of a novel pragmatic method of OSA diagnosis, undertaken entirely in primary care centres (PCC), compared with that performed in a hospital sleep unit. Four PCCs of Girona, Spain, participated in the study, two of which were located in urban areas (Santa Clara and Salt, Alfons Moré i Paretas) and the remaining two in rural areas (Sarrià de Ter and Sils) (Supplementary Table S1). The hospital sleep unit at University Hospital Josep Trueta (Girona, Spain), which was involved in the study, served as the reference hospital for those four PCC.

We randomly selected people attending PCC visits for any reason between April 2016 and November 2020. Randomisation was conducted by selecting a number indicating the position of a patient on the consultation schedule. A random number generator (range: 1–15) was used, based on the maximum number of patients that could be seen per day at each participating PCC.

We included in the study all patients between 30 and 70 years old who presented with an intermediate or high risk of OSA (≥ 3 points on the STOP-Bang questionnaire²⁹). We excluded those with severe chronic heart failure NYHA class III/ IV, severe renal failure, chronic obstructive pulmonary disease, neuromuscular diseases, neoplasm, poorly controlled psychiatric disorders, oxygen support needs, continuous positive airway pressure or OSA diagnosis or some modality of non-invasive positive pressure ventilation, pregnancy status, alcohol addiction, use of opioids or other agents that might alter sleep, and other serious illnesses, such as patients under palliative care and individuals with cognitive or motor limitations. We also excluded those unable to set up the aHSAT device and those who declined to participate in the study. We reported the results of this study in agreement with the STARD (Standards for Reporting of Diagnostic Accuracy Studies) guidelines. We anonymised personal patients’ information to ensure data confidentiality and obtained written informed consent from all study participants.

Materials

We used the ApneaLink^TM Air device (ResMed Corporation, Sydney, Australia; Firmware: SX566-0302; Sofware Airview: 4.49.0-24.0.0) as the aHSAT and mHSAT.

No algorithm updates occurred during the data collection period that affected the analysis.

The ApneaLink^TM Air device is an ambulatory respiratory polygraph that records up to five channels of information: respiratory effort, pulse, oxygen saturation, nasal flow, and snoring. We used the Airview^TM diagnosis platform (ResMed Corporation, Sydney, Australia) to visualize the aHSAT results. The mHSAT was analysed according to the criteria of the American Academy of Sleep Medicine (AASM)³⁰.

We recorded PSG with the E-Series Sleep System (Compumedics Ltd., Melbourne, Australia). The recording setting comprised an electroencephalogram, left and right electrooculograms, a chin electromyogram, left and right anterior tibial electromyograms, and an electrocardiogram. We analysed the PSG according to AASM criteria³¹. We obtained respiratory recordings from a nasal pressure cannula, an oronasal flow thermistor, a thoracoabdominal band, and the sum of band amplitudes. We measured oxygen saturation with a Nonin 8000 J (Nonin Medical, Plymouth, Minnesota) pulse oximetry sensor. We monitored the position with a V7 TP sensor and the snoring with a V4 TP-SCR microphone. We used the 2007 AASM recommendations regarding configuration, filters, and signal sampling rates³⁰.

Procedures

The sleep unit specialist performed a two-hour theoretical training session on OSA and its diagnosis for participating personnel at the centres. Besides, he conducted a six-hour practical training session on the assembly, disassembly, software use, and data download of the aHSAT device for a nurse from each participating PCC. He delivered a one-hour practical training session on interpreting the automatic analysis of the aHSAT device to PCP. The latter and nurses measured simulated people to verify their acquired skills and knowledge and clarify doubts.

The nurse in each PCC instructed the study participants on installing the aHSAT device. Written instructions for this procedure were also handed to them. The people installed the aHSAT device at home and returned it to their PCC the following day. Participants were instructed to manually start and stop the device at bedtime and upon waking and to record these times in a written register. Each PCP and nurse had access to the aHSAT diagnosis platform (Airview^TM). The PCP interpreted the aHSAT records and diagnosed the study participants without intervention from the hospital’s sleep unit. The PCP considered as valid a recorded period and flow or band and oximetry measurements of at least four hours, excluding periods of non-interpretable signals or where artefacts were detected. In addition, participants had to report having slept at least four hours. An invalid recording could be repeated only once.

The record was sent through the Airview^TM diagnosis platform to the participating hospital’s sleep unit. Then, the sleep medicine specialist performed the mHSAT without knowing the participant’s identity, the aHSAT analysis, or the PCP’s diagnosis. Participants with a respiratory event index (REI) < 30 in their mHSAT underwent a PSG study at the hospital. In these cases, the patient was informed by the hospital sleep unit of the date the PSG would be performed.

Definitions

Apnoea was scored when there was a drop in the peak signal excursion by ≥ 90% of the pre-event baseline using an oronasal thermal sensor for ≥ 10 s. Hypopnoea was scored when the peak signal excursions dropped by ≥ 30% of the pre-event baseline using nasal pressure for ≥ 10 s in association with ≥ 3% arterial oxygen desaturation³¹. The total number of apnoea and hypopnoeas divided by the recording time was referred to as the REI, and the percentage of time spent below 90% of oxygen saturation (SaO₂) was determined for each participant.

We defined OSA as REI ≥ 5 per hour and considered no respiratory sleep disturbances (NRSD) for REI < 5 per hour. For the aHSAT, mHSAT, and PSG, we considered REI ≥ 5 and < 15 per hour as mild OSA, REI ≥ 15 and < 30 per hour as moderate OSA, and REI ≥ 30 per hour as severe OSA. We diagnosed OSA for mHSAT and PSG according to international guidelines on the diagnosis and treatment of sleep apnoea-hypopnoea syndrome^32,33.

Statistical analysis

We described categorical variables as absolute frequencies and percentages and continuous variables as mean and standard deviation (SD) if normally distributed, and as median and range if non-normally distributed. We used the Student’s t-test or the U-Mann-Whitney for independent samples and the Fisher’s exact test and chi-square test to analyse continuous or discrete variables, respectively.

We determined the ratio of correct OSA diagnosis performed by the aHSAT device and interpreted by the PCP. We evaluated the sensitivity, specificity, predictive values, and likelihood ratio of the aHSAT. Agreement between the REI of the mHSAT and that of the aHSAT was assessed using Cohen’s kappa statistic. We regarded a kappa statistic ≤ 0.2 as poor agreement, > 0.2–0.4 as fair, > 0.4–0.6 as moderate, > 0.6–0.8 as substantial, and > 0.8 as almost perfect agreement. All statistical analyses were performed using R software (version 4.5.0; R Foundation for Statistical Computing, Vienna, Austria).

Ethics approval

We drafted the study according to the Declaration of Helsinki. The study was approved by the Clinical Research Ethics Committees of Hospital Universitari de Girona Doctor Josep Trueta (code: 2021.049). We anonymised personal patients’ information to ensure data confidentiality and obtained written informed consent from all study participants.

Results

Baseline characteristics

A total of 2599 patients were assessed in PCC for study eligibility; of these, 2196 met at least one exclusion criterion (Supplementary Table S2), and 403 were enrolled and screened with aHSAT. A total of 355 (88.1%) participants provided a valid aHSAT record in the PCC setting. Finally, 290 patients with a mHSAT and REI < 30 were admitted to the hospital, 264 of whom had a valid video-PSG, while 65 patients with a mHSAT and REI > 30 had only their recordings sent. The average time interval between the HSAT and PSG recording was 101 days. Therefore, 329 (81.6%) patients who also completed the diagnosis assessment were included in the comparative analysis (Fig. 1). Most of these study participants were male (210, 63.8%) and the mean age was 56.5 (SD: 9.2). The most common comorbidities in our cohort were hypertension (181, 55%), dyslipidaemia (107, 32.6%), and diabetes (55, 16.8%) (Table 1).

Fig. 1. Patient screening flowchart.

aHSAT, autoscoring home sleep apnoea testing; mHSAT, manual analysis of home sleep apnoea testing; OSA, obstructive sleep apnoea; PCC, primary care centres; PSG, polysomnography; REI, respiratory event index.

Table 1.

Baseline characteristics of participants included in the comparative analysis between the diagnosis made in the PCC and the hospital.

	Overall (n = 329)
Sex, No. (%)
Male	210 (63.8)
Age, mean (SD), years	56.5 (9.2)
Ethnicity, No. (%)
Caucasian	297 (90.8)
BMI, mean (SD), kg/m2	30.1 (5.2)
Neck circumference, mean (SD), cm	40.3 (4.4)
Epworth sleepiness scale, median [IQR]	7.0 [4.0–10.0]
STOP-Bang result, mean (SD)	4.3 (1.1)
Alcohol intake (g)a, median [IQR]	16.0 [8.0–24.0]
Smoking historyb, No. (%)
Current or former	209 (63.7)
Never	119 (36.3)
Prior use of benzodiazepinesb, No. (%)	62 (18.9)
Comorbidities, No. (%)
Hypertension	181 (55)
Dyslipidaemia	107 (32.6)
Diabetes	55 (16.8)
Ischemic heart disease	17 (5.2)
Arrhythmia	11 (3.4)
Stroke	3 (0.9)

^aInformation available only for 104 participants.

^bInformation available only for 328 participants.

BMI body mass index, IQR interquartile range (Q1–Q3), PCC primary care centres, SD standard deviation, STOP-Bang: questions on snoring, tiredness during daytime, observed apnoea, high blood pressure, body mass index, age, neck circumference, and gender.

aHSAT, mHSAT, and PSG results and concordance of primary care and hospital diagnoses

The results of the analysis of aHSAT, mHSAT, and PSG records are shown in Table 2. In PCC, 62 (18.8%) diagnoses of NRSD and 267 (81.2%) of any OSA were made. In the hospital, 18 (5.5%) patients were diagnosed with NRSD, 309 (93.9%) with any OSA, and two (0.6%) with central sleep apnoea. The global agreement of hospital diagnoses with those made at the PCC was 41.6%. Agreement in the diagnosis between PCC and hospital was found in 11 (17.7%) patients with NRSD and in 258 (78.4%) with any OSA. Additionally, 126 (38.3%) OSA diagnoses made at the PCC matched the disease severity with hospital findings. The highest level of agreement in OSA severity was found for severe OSA (96.2%), indicating that nearly all cases diagnosed at the PCC as severe OSA were confirmed by hospital diagnoses. Of the 73 cases diagnosed by PCC as moderate OSA, 63 (86.3%) were confirmed by the hospital as moderate or severe OSA. Thus, this diagnostic strategy allowed the diagnosis of 51 severe OSA cases by PCC of the 122 cases identified by the hospital sleep unit (representing 41.8%) (Table 3).

Table 2.

Summary of the results of the aHSAT, mHSAT, and PSG tests, median [IQR].

	aHSAT (n = 355)	mHSAT (n = 355)	PSG (n = 264)
Total recording time (min)	460.0 [256.0–729.0]	459.0 [256.0–729.0]	-
Total sleep time (min)	-	-	365.5 [180.0–499.5]
Total wakefulness time (min)	-	-	51.0 [1.5–301.5]
Sleep efficiency	-	-	81.8 [32.4–97.5]
NREM stages	-	-	83.0 [19.9–309.5]
REM stage	-	-	17.1 [0.0–161.0]
Total number of apnoea	7.0 [0.0–413.0]	11.0 [0.0–416.0]	4.0 [0.0–248.0]
Total number of hypopnoea	53.0 [0.0–367.0]	55.0 [2.0–482.0]	86.0 [4.0–397.0]
REI	10.6 [0.0–71.0]	10.5 [0.4–77.4]	18.1 [0.6–83.0]
AI	1.1 [0.0–59.1]	1.6 [0.0–60.9]	0.7 [0.0–32.7]
OAI	0.6 [0.0–49.0]	1.1 [0.0–45.5]	0.4 [0.0–27.0]
CAI	0.2 [0.0–24.5]	0.1 [0.0–21.4]	0.0 [0.0–29.3]
MAI	0.0 [0.0–5.2]	0.0 [0.0–28.1]	0.0 [0.0–3.0]
HI	8.5 [0.0–51.7]	7.7 [0.0–65.4]	15.5 [0.5–79.0]
Supine REI	13.1 [0.0–78.6]	13.4 [0.0–79.0]	15.7 [0.0–120.0]
REM REI	-	-	18.4 [0.0–82.5]
Percentage of sleep or recorded time spent below 90%SaO2	8.0 [0.0–100]	8.0 [0.0–100]	0.7 [0.0–90.3]
Percentage of sleep or recorded time spent below 80%SaO2	0.0 [0.0–18.0]	0.0 [0.0–18.0]	0.0 [0.0–9.0]
Desaturation index	14.9 [0.0–74.8]	10.8 [0.5–74.8]	8.2 [0.0–74.4]
Arousal index	-	-	19.0 [0.5–85.0]

aHSAT autoscoring home sleep apnoea testing, AI apnoea index, CAI central apnoea index, HI hypopnoea index, IQR interquartile range (Q1–Q3), MAI mixt apnoea index, mHSAT manual analysis of home sleep apnoea testing, NREM non-rapid eye movement, OAI obstructive apnoea index, PSG polysomnography, REI respiratory event index, REM rapid eye movement; SaO₂, oxygen saturation.

Table 3.

Diagnostic concordance between primary care centres and the hospital, considering mHSAT and PSG if necessary, No (%).

Hospital (n = 329)	PCC (n = 329)
	NRSD (n = 62)	Mild OSA (n = 141)	Moderate OSA (n = 73)	Severe OSA (n = 53)	CSA (n = 0)
NRSD (n = 18)	11 (17.7)	6 (4.3)	1 (1.4)	0 (0.0)	0 (0.0)
Mild OSA (n = 88)	30 (48.4)	49 (34.8)	9 (12.3)	0 (0.0)	0 (0.0)
Moderate OSA (n = 99)	17 (27.4)	54 (38.3)	26 (35.6)	2 (3.8)	0 (0.0)
Severe OSA (n = 122)	4 (6.5)	30 (21.3)	37 (50.7)	51 (96.2)	0 (0.0)
CSA (n = 2)	0 (0.0)	2 (1.4)	0 (0.0)	0 (0.0)	0 (0.0)
Agreement	11 (17.7)	49 (34.8)	26 (35.6)	51 (96.2)	0 (0.0)
Disagreement	51 (82.3)	92 (62.2)	47 (64.4)	2 (3.8)	2 (100)

CSA central sleep apnoea, mHSAT manual analysis of home sleep apnoea testing, NRSD no respiratory sleep disturbances, OSA obstructive sleep apnoea, PCC primary care centre, PSG polysomnography.

Comparative assessment of primary care and hospital diagnoses

The agreement between hospital and PCC diagnoses for NRSD subjects was fair, as was the case for overall OSA, with a Kappa index of 0.21 (95% CI: 0.12, 0.29) for both of them. When analysing by severity, the best concordance was found in severe OSA (Kappa index = 0.46; 95% CI: 0.37, 0.55), with a sensitivity of 0.42 (95% CI: 0.33, 0.51) and a specificity of 0.99 (95% CI: 0.97, 1.00). The positive likelihood ratio for severe OSA was very high (42.8, 95% CI: 10.6, 172.9). For the remaining OSA, concordance was poor, with a Kappa index of 0.15 (95% CI: 0.05, 0.25) for mild and of 0.06 (95% CI: –0.05, 0.16) for moderate OSA (Table 4).

Table 4.

Comparative assessment of diagnoses at the primary care centre and the hospital, (95% CI).

	NRSD	OSA	Mild OSA	Moderate OSA	Severe OSA
Sensitivity	0.61 (0.36, 0.83)	0.83 (0.79, 0.87)	0.56 (0.45, 0.66)	0.26 (0.18, 0.36)	0.42 (0.33, 0.51)
Specificity	0.83 (0.79, 0.87)	0.61 (0.36, 0.83)	0.62 (0.56, 0.69)	0.79 (0.74, 0.84)	0.99 (0.97, 1.00)
LR +	3.70 (2.37, 5.78)	2.15 (1.20, 3.84)	1.48 (1.15, 1.89)	1.27 (0.84, 1.93)	42.8 (10.6, 172.9)
LR–	0.47 (0.26, 0.83)	0.27 (0.17, 0.42)	0.71 (0.55, 0.92)	0.93 (0.81, 1.06)	0.59 (0.51, 0.68)
Kappa index	0.21 (0.12, 0.29)	0.21 (0.12, 0.29)	0.15 (0.05, 0.25)	0.06 (–0.05, 0.16)	0.46 (0.37, 0.55)

95% CI: 95% confidence interval; LR + positive likelihood ratio, LR–, negative likelihood ratio, NRSD no respiratory sleep disturbances, OSA obstructive sleep apnoea.

Autoscoring home sleep apnoea testing at PCC versus manual analysis at the hospital

Supplementary Table S3 shows the diagnostic concordance between aHSAT conducted at PCC and the manual analysis of the same records by the sleep specialist at the hospital (without considering PSG). The interpretation of mHSAT by the sleep specialist classified 131 patients (36.9%) as having moderate or severe OSA.

Discussion

Our study demonstrated the efficacy of a diagnostic strategy for OSA in PCC, supported by aHSAT and without the intervention of a sleep specialist. Almost all cases of severe OSA diagnosed at the PCC were confirmed by a sleep specialist. Given the extremely high specificity of aHSAT at PCC for identifying severe OSA and the fact that this simple approach enabled the reliable diagnosis of nearly half of all patients with severe OSA, this strategy could lead to early treatment initiation in primary care centres for patients at the highest risk of complications. However, mild-to-moderate OSA, with poor agreement between hospital and PCC in our study, would require further hospital investigations to avoid delays in diagnostic assessment and subsequent treatment.

OSA is known to be significantly underdiagnosed, mainly because of the heterogeneity and lack of symptoms in this condition⁴. To illustrate this, most studies evaluating the diagnostic efficacy of OSA have included only individuals clinically suspected of OSA^34,35. Our study included all individuals attending PCC visits for any reason who showed a high probability of OSA based on the STOP-Bang questionnaire. This questionnaire, validated in both the general population³⁶ and in PCC³⁷, is quick and simple and has demonstrated high sensitivity for diagnosing OSA³⁸. Comparative studies with other screening methods, such as the Berlin Questionnaire, the Epworth Sleepiness Scale, and even home respiratory polygraphy with five channels, concluded that the STOP-Bang questionnaire was the most reliable method for identifying OSA³⁹. In our study, most patients with a STOP-Bang score ≥ 3 were confirmed as having OSA by mHSAT and/or PSG, supporting its use as a screening tool for this condition in PCC, given its high prevalence and healthcare demand. The higher percentage of OSA cases, especially severe cases relative to other general population cohorts^2,40, may well reflect the different selection criteria. Not only did we include individuals attending the PCC for any reason, which is more representative of the broad clinical spectrum of the disease^41,42, but also a high-risk OSA population using the STOP-Bang questionnaire, a highly sensitive questionnaire for diagnosing OSA^38,39. From another perspective, we decided that patients with REI < 30 would undergo PSG to identify those who would have been underestimated by HSAT if we had not carried out PSG in patients with REI ≥ 15^32,33. However, performing PSG would not be necessary in routine clinical practice in cases with REI ≥ 15 on mHSAT as analysed by a sleep specialist^32,33,43.

Another advantage of our two-step diagnostic approach, the STOP-Bang questionnaire followed by aHSAT, is that not only does it require minimal primary care training of doctors and nurses, but it also enables carrying out later during clinical work. aHSAT does not require specialist experience, and PCC personnel alone can make the diagnosis, conduct the assessment, and make the decision. The invalid aHSAT percentage, mainly due to mispositioned sensors or insufficient recording time, was acceptable in our study and was similar to that of other reports^44,45. The simplicity and the possibility of early diagnosis in severe OSA patients in PCC could be suitable in remote geographical zones, where communication with a specialist sleep centre is difficult, or in overwhelmed sleep centres that cannot accommodate demand, given the high global prevalence of OSA⁴⁶. However, our diagnostic approach should be adapted to available clinical practice resources and guidelines recommendations, suggesting the use of automatic HSAT in selected patients³³.

As far as we know, only two more reports^25,26 have assessed the efficacy of a diagnostic strategy for OSA in PCC without any intervention of a sleep specialist. A study evaluating a simplified method for diagnosing OSA observed that aHSAT could be a practical tool, but its sample size was small, subjected to selection bias, a nurse had to undertake home visits to install the devices, and no comparison with the gold standard was established²⁵. The authors also used the Chicago Criteria, which overestimated the diagnosis of OSA. Another relevant report concluded that OSA could be diagnosed, and severe OSA could be confidently ruled out, in the primary care setting²⁶. However, in contrast to our study, patients were included after clinical suspicion of OSA and a home sleep testing was performed on three consecutive nights with an automatic single-channel nasal flow monitor using the mean of the individual night REIs in the analysis.

Another study found that automatic analysis was useful in moderate and severe OSA cases⁴⁷. However, the population in that study was evaluated by a sleep pulmonologist in PCC, and the aHSAT setup instructions were provided by a hospital technician. On the contrary, in our approach, the diagnosis was made in PCC, and the same polygraphy recording was used without the need to conduct a new test or refer patients to the hospital. The manual analysis of polygraphy by a specialist or sleep unit is known to improve the underdiagnosis efficacy of the aHSAT in diagnosing OSA^8,15,16. Peñacoba et al. evaluated the diagnostic decision agreement between the PCP and the sleep specialist by conducting a screening with a 2-channel aHSAT and compared it to the gold standard. The authors found a poor concordance, which improved upon the mHSAT analysis by a sleep specialist²⁰. Another clinical trial, undertaken in patients attending pulmonology consultations with intermediate to high clinical suspicion of OSA, determined that a sequential analysis (aHSAT followed by mHSAT for inconclusive cases) was the most cost-effective approach⁴⁸. Other strategies that did not attempt to provide improved diagnosis for OSA in primary care settings^22,23,49 have shown that once diagnosed, OSA can be treated cost-effectively compared to a sleep hospital unit. Communication between the patient and the two healthcare services, with a virtual sleep unit, for instance, could enable the diagnosis with the same aHSAT recording and the early treatment of the identified OSA patients⁵⁰. This would mean that, with minimal intervention, many hospital visits and PSG could be avoided⁵¹.

The generalisability of our findings may be subject to certain limitations. First, we may not have included the entire population needing a sleep study. Other studies with older populations and cardiovascular comorbidities could be necessary, although the guidelines of the American College of Physicians do not recommend home testing in this population⁷. Second, differences were found in various tests, but they could be explained by differences in evaluation properties, in device algorithms, and nightly variability⁵². Besides, these discrepancies are common in clinical practice, no clinical differences in the patients during the time between tests were observed, and the differences found were not clinically relevant. Third, errors in aHSAT analysis were also detected, invalidating valid recordings when analysed manually. Fourth, we did not undertake a comprehensive OSA management model with a virtual sleep unit’s support, which could not only provide data on treatment and outcomes but also on cost-effectiveness in PCC. Fifth, PSG was not performed in all patients, but mHSAT by a sleep specialist is an accepted method in patients with a high probability of OSA, according to pre-test^12,13. Finally, in severe OSA, HSAT rarely underestimates the disorder. The probability of false negatives is very low; PSG would not be indicated, as it would delay treatment and generate higher costs, as pointed out by some authors and guidelines^48,53,54.

Conclusion

The ApneaLink^TM Air device showed high specificity for severe OSA in a high-risk primary care population defined by a STOP-Bang questionnaire score ≥ 3. These results suggest that when aHSAT indicates severe OSA, hospital confirmation may be unnecessary, whereas negative or moderate findings may still require specialist assessment.

Author contributions

All authors had full access to the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: R.O., E.R., and A.O. Analysis and interpretation of the data: R.O., E.R., R.R., I.C.D., X. E., M.S., E.M., and M.B.-V. Statistical analysis: M.C.-C. Supervision: R.O. Drafting of the manuscript: R.O. and E.R. Critical revision of the manuscript for important intellectual content: R.O., E.R, A.O., M.C.-C., R.R., I. C. D., X. E., M. S., E.M. and M.B.-V.

Data Availability

The datasets generated and/or analysed during the current study are not publicly available because of confidentiality issues but are available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41533-026-00496-4.