Economic evaluation of telemonitoring as a follow‐up approach for patients with obstructive sleep apnea syndrome starting treatment with continuous positive airway pressure

Telemonitoring of obstructive sleep apnea patients is increasingly being adopted though its cost‐effectiveness evidence base is scanty. This study investigated whether telemonitoring is a cost‐effective strategy compared with the standard follow‐up in patients with obstructive sleep apnea who are starting continuous positive airway pressure treatment. In total, 167 obstructive sleep apnea patients were randomised into telemonitoring (n = 79) or standard follow‐up (n = 88), initiated continuous positive airway pressure treatment, and were followed up for 6 months. The frequencies of healthcare contacts, related costs (in USD 2021 prices), treatment effect and compliance were compared between the follow‐up approaches using generalised linear models. The cost effectiveness analysis was conducted from a healthcare perspective and the results presented as cost per avoided extra clinic visit. Additionally, patient satisfaction between the two approaches was explored. The analysis showed no baseline differences. At follow‐up, there was no significant difference in treatment compliance, and the mean residual apnea–hypoapnea index. There was no difference in total visits, adjusted incidence rate ratio 0.87 (0.72–1.06). Participants in the telemonitoring arm made eight times more telephone visits, 8.10 (5.04–13.84), and about 73% fewer physical healthcare visits 0.27 (0.20–0.36). This translated into significantly lower total costs for the telemonitoring approach compared with standard follow‐up, −192 USD (−346 to −41). The form of follow‐up seemed to have no impact on the extent of patient satisfaction. These results demonstrate the telemonitoring of patients with obstructive sleep apnea initiating continuous positive airway pressure treatment as a cost saving strategy and can be argued as a potential worthy investment.

The most common and established treatment for OSA, is continuous positive airway pressure (CPAP) (Council on Health Technology Assessment, 2007;Patil et al., 2019;Qaseem et al., 2013) and the use of mandibular advancement devices (Randerath et al., 2021).Of the current treatment approaches, CPAP usage is the most researched and recommended as the cost-effective treatment alternative for moderate to severe OSA (McMillan et al., 2015;Pachito et al., 2022;Patil Susheel et al., 2019;Toraldo et al., 2017).However, there is continued difficulty of compliance to CPAP treatment which in itself involves a cost if not addressed (O'Reilly, 2022).The non-compliance shortly after initiation of treatment may be attributed to side-effects of treatment including mouth dryness, mask leakage, facial pain due to the mask, gastrointestinal disturbances and disturbed sleep (Ghadiri & Grunstein, 2020).Thus, patients with OSA starting CPAP treatment require tight and close follow-up when initiating treatment.
During the CPAP initiation period, the healthcare personnel get to address the immediate adverse effects of the treatment and titrate the pressure settings to fit the patient's needs.This period is cost intensive to the healthcare provider and thus information and communications technologies (ICT) based e-health interventions have been implemented and tested as an alternative to the usual face to face follow-up approach (Isetta et al., 2015;Lugo et al., 2019;Pachito et al., 2022;Toraldo et al., 2017;Turino et al., 2017).Other strategies have been to delegate the follow-up to primary healthcare physicians, nurses, and physiotherapists other than the physicians at the specialist clinics (Sánchez-de-la-Torre et al., 2015;Sánchez-Quiroga et al., 2018).
Of the current approaches, e-health interventions such as telemonitoring in OSA care are promising (Alsaif et al., 2022;Dusart et al., 2022;Kosky et al., 2022;Pachito et al., 2022;Verbraecken, 2021).Telemonitoring is a form of e-health monitoring that involves the ability to remotely access the patient's CPAP apparatus data via an electronic platform.In a number of research studies, e-health interventions in regard of care for OSA patients using CPAP, have been deemed cost-effective, free health personnel's time to focus on other duties, and not to compromise patient satisfaction and compliance (Anttalainen et al., 2016;Contal et al., 2021;Pépin et al., 2017;Suarez-Giron et al., 2020;Thong et al., 2022).Although there are studies presenting conflicting results with regard to patient satisfaction and treatment adherence (Dusart et al., 2022;Turino et al., 2017) and the cost effectiveness literature of e-health monitoring interventions is still limited (Dusart et al., 2022).
Thus far, most studies have investigated cost effectiveness, satisfaction, and compliance over varying periods of follow-up (Dusart et al., 2022;Thong et al., 2022).They have also focused on e-health applications where the clinician has the ability to communicate with the patient through a digital application, adjust treatment, give feedback, and make assessments based on the patient's sleep details from the application database or a secure cloud (Isetta et al., 2015).However, this still involves frequent clinician contact, and is thus costly.
Presently, no such investigation has been done in a Swedish setting where much of the follow-up of OSA patients has been digitalised and delegated to other healthcare personnel other than specialist physicians in sleep medicine.We hypothesised that the telemonitoring approach would involve fewer physical visits to the OSA clinic and thus potentially less cost.Therefore, the work in this study aimed at addressing the following research questions: 1. How cost-effective is telemonitoring compared with standard follow-up of OSA patients starting CPAP treatment?2. How does telemonitoring impact patient satisfaction?

Study design
This study was a within trial health economic evaluation from a healthcare perspective.The work is based on data from a randomised controlled study (RCT) comparing two follow-up approaches, telemonitoring (TM) and standard follow-up (SF) of patients with OSA initiating CPAP treatment at home.The study was carried out at an academic hospital sleep apnea clinic in Uppsala, Sweden.Adults with newly diagnosed OSA, who had also accepted initiating CPAP treatment were randomly assigned to either the TM or a SF approach of follow-up.All participants provided informed consent, were assessed at baseline using questionnaires and thereafter followed up for 6 months through telephone-or physical (face-to face) contacts with healthcare personnel.This study received ethical approval from ethical review board in Uppsala (Dnr 2017/014, approval date 2017-04-05).

Participants and procedure
Adults (≥18 years) referred to the sleep apnea clinic at Uppsala university hospital with suspected OSA, underwent a home sleep apnea test (HSAT).The OSA diagnosis was set by a physician and treatment was discussed at a visit at the clinic.Participants with diagnosed OSA who were prescribed treatment with CPAP were eligible for the study.At the CPAP education and testing visit to initiate treatment, individuals were attended to in groups of six to eight participants.Individuals that offered their informed consent were thereafter randomised to the two different follow-up approaches.Individuals requiring increased care needs for example, need of a personal assistant, and those that had language difficulties were therefore booked in on an individual basis and excluded from the study.Both in the TM and SF approach of follow-up, the visits were carried out by specialist sleep nurses, physiotherapists, and technicians.
Data on healthcare consumption at the 6 months follow-up was retrieved from the patient electronic medical records at the sleep apnea clinic.Data on CPAP compliance was collected either by downloading data from the memory card in the CPAP machine (ResScan™, ResMed, USA; standard care follow-up) or via the telemonitoring platform (AirView™; ResMed, USA; TM group).Data on patient satisfaction was gathered using questionnaires.

Randomisation
Block randomisation using a ratio 1:1 was applied, where groups of patients that attended the CPAP education and testing session before initiation of treatment were allocated to either the TM arm or the SF arm.No blinding of patients or assessors was done.

Telemonitoring follow-up
In this study, the ResMed AirView platform was used (AirView™; ResMed, USA; TM group).The telemonitoring platform, provides information on settings, given CPAP pressures, air leakage, apnea-hypoapnea index (AHI), and duration of usage.Based on these measures, the clinician can adjust treatment or initiate contact with the patient.
In this study, participants randomised to the telemonitoring group had a telephone contact with a CPAP nurse after 2-5 days of initiating treatment.At the contact, the patient's compliance to treatment (measured in hours/night and nights/month) were assessed via the telemonitoring system, side-effects, and symptoms were assessed, and if needed adjustments to the treatment were made or the patient was called in for a face-to face appointment.Thereafter, continued telephone or physical visits took place until the treatment was deemed well established and functional.Any further visits or contacts after established CPAP treatment were made at the patient's initiative.

Standard follow-up
Under the standard follow-up approach, the patient came for a clinic visit after 2 weeks of initiating treatment.During the visit, compliance to treatment (data downloaded from the memory card of the CPAP machine), side-effects and symptoms were assessed and treatment adjusted as needed.Thereafter, continued physical visits took place until when the treatment was deemed well established and functional.
Just like in the telemonitoring approach, any further visits/contacts after established CPAP treatment were made at the patient's initiative.

Baseline characteristics and socioeconomic
At baseline, data on demographic characteristics, lifestyle habits, and baseline home sleep apnea testing (HSAT)/polygraphy for example, AHI and oxygen desaturation index (ODI) results were collected.

Comorbidities
In the baseline questionnaire, participants declared which other diseases they were being treated for and what kind of treatment they were receiving.This variable was treated as categorical based on the number of comorbidities a participant had.That is, 0 for participants with <3, 1 for those with 3-5, and 2 for those with >5 comorbid conditions.

Quality from the patient's perspective (patient satisfaction)
Data on Quality from the Patient's Perspective (QPP) as a measure of patient satisfaction, was collected at the baseline and at 6 months follow-up.A modified version of the QPP questionnaire (Wilde et al., 1994) was used to capture information about patient satisfaction with the care received.The questionnaire entails questions regarding information, participation, healthcare personnel, organisation of care, trust, and atmosphere among others.In the QPP questionnaire, the items are evaluated in two ways by the respondent; assessment of perceived reality and evaluation of subjective importance (Likert scales).A quality of care index is computed on the basis of the relationship between these two scores, and results are presented in the form of a measure index from the concepts: Deficient quality, Some deficient quality, Balanced low quality, Balanced acceptable quality, Balanced

Healthcare usage and related costs
The frequencies of healthcare contacts over the 6 months follow-up, that is, both telephone and physical contacts with the personnel at the sleep apnea clinic were retrieved from the hospital's patients' electronic database based on journal entries.Thereafter, related total cost was estimated as the product of frequencies and the unit cost per healthcare contact.No discounting was applied as all the costs occurred in less than a year.All cost data were collected in Swedish kronor (SEK) and adjusted to US dollars, in 2021 prices using purchasing power parities (Shemilt et al., 2010).

Descriptives
Baseline characteristics and observed outcomes for the whole sample in the respective defined groups were summarised as counts and percentages (%) for categorical variables, mean counts and standard deviations (SD) for the numerical variables.Baseline differences in numerical variables were assessed using Student's t-tests and Mann-Whitney tests depending on the distribution properties of the variables.For categorical variables, chi-square tests were used to assess baseline differences.

Analysis of cost and outcome data
Using data from participants that had complete data on the outcomes of interest at both baseline and follow-up, the relationship between the follow-up approaches and differences in mean count of healthcare contacts was examined using generalised linear models.To accommodate the distributional properties of the data, based on theory and visual inspection using histograms, a quasi-Poisson distribution and a log link function were used for the count data and a gamma distribution with an identity link for the cost data.Cumulative counts and costs of resources over time (between baseline and 6 months follow-up) were compared between trial arms.All analyses were adjusted for age, gender, treatment compliance in hours, baseline AHI, and comorbidities.Only the adjusted estimates are reported in this work.The effect measure of the relationship for count data was expressed as an incidence rate ratio (IRR) and cost data as a mean difference.Using non-parametric bootstrapping, we run 5000 iterations of the regression analysis estimates to generate mean cost and mean effect differences, as well as confidence intervals to characterise the uncertainty around them.Mean effect differences were estimated as a subtraction of the intercept (mean visits in the SF group) from the product of the intercept and the IRR.The joint distribution of costs and effects were represented visually on a cost effectiveness plane (Drummond et al., 2015), and the probability of telemonitoring being cost effective at different willingness to pay thresholds was represented visually on a cost effectiveness acceptability curve (Fenwick et al., 2001).

Sensitivity analysis
The delegation of follow-up of OSA patients to other clinical personnel than specialist physicians in sleep medicine, results in a different charge per patient visit, hence a possibility of affecting the results.
Therefore, a deterministic sensitivity analysis was conducted where the unit costs of the healthcare visits were increased by 50% percent from the baseline value, thus reflecting the cost of a specialist sleep physician's visit, and changes in the cost-effectiveness were studied.
All analyses were conducted in R version 4.0.0 with statistical significance set at p < 0.05.

Descriptives
The study population consisted of 216 participants, predominately male (70%), and of a mean age of 57 years.A total of 35 individuals discontinued from the study and 14 had missing values on the primary outcome.Therefore, a final sample of 167 OSA patients with both baseline and follow-up data on the outcomes of interest was used in the analysis.There were no baseline differences between the TM and SF groups on the observed covariates.See Figure 1 and Table 1 for details.

Telephone contacts
Individuals in the TM group had significantly higher mean telephone contacts with healthcare personnel, compared with the SF group over the 6 months follow-up period, that is, 1.84 (1.29) and 0.24 (0.59) contacts, respectively.They made eight times more telephone contacts than the SF group with an adjusted incidence rate ratio, aIRR 8.10 (5.04-13.84)p < 0.001.

Physical face to face contacts
The standard follow-up group had, on average, significantly more physical face-to-face healthcare contacts compared with the TM group, 2.77 (1.57) and 0.76 (0.99) visits respectively.The TM group made 73% fewer physical visits compared with the SF group, aIRR 0.27, 95% CI (0.20 to 0.36) p < 0.001.

Overall contacts
There were no significant differences between the mean total healthcare contacts in the TM, 2.59 (1.90) and the SF, 3.01 (1.54).Although the TM group made 13% fewer visits than the SF group, aIRR 0.87, 95% CI (0.72-1.06) p = 0.1627.See Table 2 for details.

Telephone contacts
Participants in the TM group had a significantly higher telephone contacts related cost compared with the SF group amounting to a mean difference of 364 USD, 95% CI (200 to 528) USD p < 0.001.

Physical face to face contacts
Patients in the TM group had significantly lower physical face-to-face healthcare contacts related costs compared with the SF arm, À637 USD, 95% CI (À837 to À437) USD p < 0.001.

Overall contacts
Overall, the TM group had significantly lower total healthcare related costs, that is telephone and physical visits related costs altogether compared with the SF group.Mean cost difference, À192 USD, (À346 to À41) USD p < 0.0108.See Table 2 for details.

Base case analysis
The base case analysis demonstrated the TM approach to be cost saving from a healthcare perspective, with an overall bootstrapped mean cost difference of À215 USD, 95% CI (À352 to À63) USD.
The mean difference in overall healthcare contacts avoided was estimated as 0.33 visits, 95% CI (À0.15-0.86)visits, though statistically non-significant.See Table 3 below   difference) and better effects (a positive difference in avoided visits) thus overall, a cost saving intervention.However, it also shows a wide spread of the data pairings along the x-axis showing the possibly large uncertainty in the effect difference estimates (see Figure 2).The cost effectiveness acceptability curve showed the probability of the intervention being cost-saving of 99% at a willingness to pay (WTP) of 0 USD per avoided visit.It also demonstrates a slow decline in the extent of cost savings with increasing willingness to pay per avoided healthcare contact/visit, physical or telephone (see Figure 3).Cost effectiveness plane for the sensitivity analysis 92.4% 7.6% F I G U R E 2 Cost effectiveness analysis plane for the base case, with follow up by specialist sleep nurses, physiotherapists or technicians, and a sensitivity analysis, increasing healthcare contacts unit costs by 50%, to reflect a visit to a specialist physician in sleep medicine

Sensitivity analysis
Increasing healthcare contacts unit costs by 50%, to reflect a visit to a specialist physician in sleep medicine, made the TM approach even more cost saving.See Table 3 as well as Figures 2 and 3 below.
Quality from the patient's perspective (patient satisfaction) Among patients who responded to the QPP questionnaire (n = 163), there was a high level of satisfaction despite not answering all the

Dropout analysis
A comparison of the baseline characteristics of the main sample and the dropout participants is shown in Table S1.Significant differences were noted in OSA severity, with AHI 33.4 (19.1) in the main sample and 27.0 (15.7) in the dropout participants, sick leave where there were more individuals on sick leave in the dropout sample, and nationality with more of the participants in the main sample being Swedish born (see Table S1).

Treatment effect and compliance
No significant differences were noted in measures of treatment effect (change in AHI), and compliance among participants, in the different follow-up approaches (see Table S2).

Summary of results
In this work, we explored whether telemonitoring of OSA patients starting CPAP treatment was a cost-effective approach compared with the standard follow-up approach.The work was carried out at an academic hospital sleep apnea clinic where the follow-up was run by dedicated sleep apnea specialised nurses.Our findings demonstrate that telemonitoring is a cost saving approach with acceptable satisfaction ratings.The observed cost-savings were mainly due to the switch from physical face-to-face visits to more telephone contacts in the TM group, which were relatively cheaper.However, no net avoidance in healthcare contacts between the two approaches was observed, that is, no significant differences were noted in the overall healthcare contacts, telephone and physical visits combined.In a sensitivity analysis an increase in the unit cost per physical visit was investigated, and the telemonitoring approach remained cost saving.

Comparison with other studies
The results from our work corroborate the current literature on e-interventions in the management and follow-up of OSA patients initiating treatment.(Lugo et al., 2019) studied the cost-effectiveness of a virtual sleep unit run outside the hospital compared with hospitalbased management, and also found a significant difference in the OSA-related direct medical costs, even though in their study, it was the physicians meeting and communicating with the patients virtually.
In our work, the follow-up was mainly driven by specialised nurses with the help of technicians and physiotherapists, thereby freeing the physicians' time for other patients having their first-time contact.
There were no significant differences noted in the total visits made by the telemonitoring group compared with the standard follow-up group in our study.On the contrary, work from Spain by Isetta et al. (2015), found that the telemonitoring group made significantly more visits compared with the traditional face-to-face followup group.This could be attributed to the differences in the structure of the e-treatment used.In our study, it was mainly remotely reading off sleep data from the CPAP machine and, thus making adjustments accordingly as well as telephone contacts or physical visits when deemed fit.

Strengths and limitations
This work was conducted in a day to day already established and running academic hospital sleep apnea clinic, thus closely representing what would happen in routine practice.That is, patient demographic, patterns of dropout, and comorbidity profiles.No stringent inclusion criteria regarding the patient population studied was considered, for example patients did not need to have information communications technology (ICT) skills, as would be required in situations of treatment platforms for a patient to log in.This resulted in including even older people with possibly limited ICT-skills, thus a more representative case mix of patients encountered in routine practice.This study also uses a relatively large and well-balanced sample that is followed up over a long time.However, there are considerations when interpreting our results.Firstly, the drop out population had a lower mean AHI compared with the main sample.Thus, there may be an element of self-selection for the individuals with moderate to severe OSA within the study.However, the dropout rate was similar in each group, thus any effect that would arise from such a self-selection problem would cancel out, and the present observed results would still apply.
Secondly, not all information on the reasons for dropout was provided by the participants.Such information is relevant as it provides insight into what could have been done differently, especially if the dropout was due to the type of follow-up approach.Such information also provides a gateway to speculate on how these patients could be helped in the future.
Thirdly, it was difficult to ascertain which visits were planned by healthcare personnel and which visits were guided by the patient's own initiative during the follow-up period after established treatment adherence.This has economic implications as visits that could have been avoided would consume resources that could be utilised elsewhere.
Fourth, a conservative approach was taken when billing telephone contacts, where all phone calls were charged the same fee irrespective of duration.This could mean over-or under estimation of the total cost for the telephone visits compared with a setting where the telephone visits are billed based on the doctor's cost per unit time.
Further, the study did not include a tool to capture health related Lastly, we only examined OSA related direct medical costs as no further data could be collected during the trial.A broader societal costing perspective would have been more informative and relevant given the impacts of OSA on other aspects of a patient's life.

Implications for clinical practice and decision making
The findings in this work support the current evidence demonstrating telemonitoring of OSA patients starting CPAP treatment as a costsaving alternative compared with the standard face-to-face follow-up.
It further demonstrates and strengthens the evidence that telemonitoring is cost-saving also when the follow up is carried out by specialised sleep medicine nurses (Anttalainen et al., 2016).
Taken together, current evidence supports that telemonitoring can be cost-saving, without compromising patient clinical outcomes, in different clinical and technological settings.That is, both in situations where there is a possibility for direct real time conversation with the patient and treatment adjustments as well as remotely reading off

CONCLUSION
Telemonitoring is a potential cost saving approach of follow-up for OSA patients initiating treatment with CPAP at home.Therefore, a careful roll out of telemonitoring may be financially beneficial to the healthcare system.
good quality, Somewhat excessive quality, and Excessive quality.A balanced quality indicates met quality need for patients in items that are of importance to the patients, and is what should be strived for.Low quality index indicates inadequately met quality needs in items of importance to the patient, whereas excess quality index indicates quality met in items/issues of low or no importance to the patient.Measures of treatment effect and adherence CPAP data at 6 months follow-up in comparison with the baseline measures were used to depict treatment benefit, for example differences in AHI.Compliance was measured as average time of use of the CPAP per night in hours, where usage of ≥4 h/night was considered good or established compliancy.Detailed results on treatment adherence are reported elsewhere (Delijaj et al., 2023), since the focus of this work was on cost-effectiveness.Unit costs for healthcare contacts Unit costs for the different healthcare contacts were retrieved from administrative data from the Uppsala university hospital accounts office.Patients were primarily followed up by specialist sleep nurses, physiotherapists, and technicians.A mean unit cost was applied, 2564 SEK (290 USD) per physical visit and 2071 SEK (234 USD) for a telephone contact, in 2021 prices.
for details.A visual presentation of the distribution of cost and effect differences pairs on a cost effectiveness plane shows 92.4% of the estimates in the southern east quadrant demonstrating savings (a negative cost Study sample at baseline and follow-up T A B L E 1 Descriptive statistics of the study sample Cost effectiveness acceptability curves (CEAC) for the base case and sensitivity analyses questions, showing balanced high quality index for the items in the questionnaire.This was shown both in the TM and SC group.Further details are available in Figure S1a,b.
quality of life with a possibility to allow the estimation of estimation of quality adjusted life years (QALYs), which are an important outcome in economic evaluation due to its intrinsic pragmatic economic credentials.QALYs can be benchmarked against willingness to pay values, which can help determine whether the cost per QALY estimated in a study is acceptable in terms of cost-effectiveness.It would also allow us to compare our study results with other studies which have used QALYs as outcome measure.Nevertheless, the outcome used in this work has intrinsic value to healthcare in terms of planning and allocation of resources.
CPAP results, followed by contacting the patient to have a telephone or physical visit, as deemed fit.Upscale of such e-interventions is at the forefront of healthcare systems' goals in Europe and has the ability to be financially beneficial without compromising patient clinical outcomes.
However, in the work byIsetta et al.there was a special designed platform where the patients and clinicians would have regular contact through a messaging tool.Hence there was a possibility of initiating even more extra visits that may or may not have been necessary.Isetta's work also demonstrates telemonitoring to be cheaper and that the probability of cost effectiveness decreases with increasing willingness to pay per QALY.A similar observation was made in our work, though using willingness to pay per avoided extra visit.