The present work is a retrospective longitudinal analysis of data from a cohort of people with chronic respiratory failure treated at home with long-term, nocturnal, noninvasive ventilation (NIV) who underwent regular assessment between February 2019 and May 2023.

Potential participants were identified through a regional home-care provider database (AGIR à dom). AGIR à dom provides NIV treatment to over 3000 individuals in France, which corresponds to 3 out of every 100 people treated with NIV nationwide.21,22 The inclusion criteria were being aged≥18 years and having completed the S3-NIV questionnaire during a home follow-up visit. Non Inclusion criteria were: (1) NIV prescription for >12h/day (to exclude diurnal use of NIV at the beginning of the study); (2) use of the volumetric mode for the small size of this group (0.16% of the cohort); (3) use of adaptive servo-ventilation (ASV) mode, mainly reserved for patients with central or mixed sleep apnoea syndromes, but not necessarily chronic respiratory failures; and (4) no agreement for the use of their personal data.

The study was carried out in accordance with the European General Protection Regulation (EGPR) and the reference methodology (MR004) of the French information technology and personal data protection authority (Commission Nationale Informatique et Liberté (C.N.I.L)). Before starting the study, a research protocol was drawn up and submitted to the Health-Data-Hub.23 An information letter was mailed to all potential participants to provide them with the opportunity to object to the use of their data. This study is reported according to the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) recommendations.24

The following data were retrospectively collected from the homecare provider's database: age, sex, anthropometrics, type of respiratory condition, time elapsed since NIV initiation, NIV settings, supplemental oxygen therapy, adherence (from the device's built-in time counters), and type of interface and S3-NIV scores. For each participant, the first measurement time-point was the date of completion of the first S3-NIV questionnaire. In France, the health authority requires that the home care provider's nurses or technicians visit the individual every three to four months until death or NIV discontinuation. Data for each participant were therefore updated and the S3-NIV questionnaire was completed at each visit.

The S3-NIV questionnaire has 11 items. Each item is scored using a 5-point Likert scale from “always true” to “never true”. The overall S3-NIV score is the average of all items multiplied by 2.5. The S3-NIV evaluates two domains: “respiratory symptoms” and “sleep & side effects”. The score for the first domain is the average score of items 1, 4, 5, 6 and 7 multiplied by 2.5; the score for the latter domain is the average score of items 2, 3, 8, 9, 10 and 11 multiplied by 2.5. The total score and the two sub scores range from 0 to 10 points. Lower scores represent a higher, negative impact of the disease and NIV on the individual.

Quantitative variables are described by the median and quartiles, and categorical variables by the frequency and percentage.

The time-course of the S3-NIV score and its two domains were modeled using multivariate linear mixed effect models. Before creating the models, we assessed the distribution of scores graphically and using the Kolmogorov–Smirnov test. All three models included Individual random effects for both the intercept and the slope. A likelihood ratio test showed that the models with a random-effect of time fitted better than the non-random-effect models (p<0.0001). Time since initiation of treatment, number of provider visits between two visits during which the S3-NIV was completed, sex, age, BMI, type of respiratory condition, type of Interface, adherence (hours/day), oxygen therapy, pressure support and back-up rate were included as fixed effects. Fig. E1 shows an example of the longitudinal follow up of a participant. We chose to include all variables, even those with non-significant effects, to compare their effect on the total score and the sub-scores under the same conditions. No collinearity was detected among the variables that could potentially influence the results. Assessment of residuals confirmed the quality of the three models. Although the sample size has not been calculated, we conducted a post hoc calculation of the statistical power of the study for the equivalence test between the initial and the final S3-NIV scores.it had been estimated to be 0.99 for 2135 participants, considering a minimal clinically significant difference in S3-NIV score from 0.6 to 1.02.20

The change in S3-NIV item scores over time was modeled using a mixed effect multivariable multinomial ordinal regression because of the discrete, ordinal, non-normal distribution of the scores. All item models also included Individual random effects for both the intercept and the slope.

Before statistical analysis, we used intra class correlation (ICC) to check that there were no significant intra-participant changes for the predictor variables: BMI, pressure support, PEP, and back-up rate. Missing data for those variables were imputed from the last records (last observation carried forward). Following the imputation, records with remaining missing values were excluded (Appendix A, Supplementary results).

Statistical computations were performed with SAS for Windows, Version 9.4. p-Values<0.05 were considered significant.

Study flow chart is depicted in Fig. 1. The median (IQR) follow-up duration was 2 years (10 months–3 years) with a median of 5 (3–7) completed S3-NIV questionnaires per individual. Median time between completions of each S3-NIV questionnaire was 4 months (3–6 months). 595 (28%) participants died during the follow-up; 153 (7%) participants were of lost to follow-up. The reason for their drop-out was not reported.

Fig. 1.

Study flow.

(0.19MB).

Participants’ baseline characteristics are presented in Table 1. The two most frequent respiratory conditions for which NIV was prescribed were obesity hypoventilation syndrome (OHS) and chronic obstructive pulmonary disease (COPD). At inclusion, these participants had been treated with NIV for a median (IQR) duration of 3 years (7 months–7 years) with a median NIV adherence of 7.9h (5.6–10h/day). Median S3-NIV score was 7.9 (6.4–8.6), with a median respiratory domain score of 6.5 (5.0–8.5) and a median sleep and side effects domain score of 8.3 (7.1–9.2).

Analysis of change in total S3-NIV, respiratory symptoms domain and sleep and side effects domain scores over time

Fig. 2 (and Table S1) shows the variables independently associated with changes in the total score and the two S3-NIV domains. S3-NIV total score was lower for female than male participants over time (β=−0.238, 95% CI: −0.344; −0.132), this was the case for both domain scores (respiratory symptoms: β=−0.176, sleep and side effects: β=−0.268). Older age was associated with lower respiratory symptom scores over time (for each additional 10 years, β=−0.045 95% CI: −0.085; −0.005). BMI was not significantly associated with changes in S3-NIV total score or the domain scores.

Fig. 2.

Effects Coefficients and Confidence Intervals of predictor variables on the S3-NIV Scores and the two domain scores: respiratory symptoms domain and sleep and side effects domain in multivariate Linear Mixed-Effect Models. COPD: Chronic obstructive pulmonary disease, OHS: Obesity hypoventilation syndrome, PEP: positive expiratory pressure.

(0.22MB).

S3-NIV total score improved over time for the whole group (for each additional 4 months, β=0.041 95% CI: 0.032; 0.050). The S3-NIV improvement was mainly associated with positive changes in the sleep and NIV side effects domain (p<.0001) but not the respiratory symptoms domain (p=0.132) (Fig. 3). Time elapsed since initiation of NIV at baseline was also associated with improvement in S3-NIV total score (β=0.035 95% CI: 0.024; 0.046) and the scores of both domains (respiratory symptoms: β=0.039, sleep and side effects: β=0.026) over time. Daily adherence was not significantly associated with change in total S3-NIV score over time (p=0.309), but was associated with a better sleep and NIV side effects score (β=0.017 95% CI: 0.008; 0.025) and a poorer respiratory symptoms score (β=−0.027 95% CI: −0.038; −0.017).

Fig. 3.

Estimated changes in S3-NIV, respiratory symptoms domain and sleep and side effects domain scores over time from the multivariate linear mixed effect models.

(0.14MB).

In comparison to participants with OHS, participants with COPD or other respiratory diseases had significantly lower S3-NIV scores over time (COPD vs OHS: β=−0.294 95% CI: −0.441; −0.146, COPD vs OHS: β=−0.256 95% CI: −0.458; −0.054). These differences were associated with a lower respiratory symptoms score, since sleep and NIV related side effects scores did not differ significantly between pathologies.

Participants who used face masks had lower scores over time than those who used nasal/nasal pillow masks (β=−0.210, 95% CI: −0.097; −0.324). Participants who used additional oxygen therapy also had lower scores over time than those who did not (β=−0.337 95% CI: −0.431; −0.242). With regard to NIV settings, higher positive end-expiratory pressure was positively associated with change in S3-NIV total score (for each additional 2cmH2O β=0.06 95% CI: 0.013; 0.01) and with change in the respiratory symptoms domain score (for each additional 2cmH2O β=0.114 95% CI: 0.049; 0.179) over time. Conversely, higher levels of pressure support were associated with lower S3-NIV total scores (for each additional 2cmH2O β=−0.036 95% CI: −0.070; −0.002) and lower sleep and NIV related side effect domain scores (meaning more adverse effects and poorer sleep) over time. Participants provided with humidifier had lower S3-NIV scores over time (β=−0.072 95% CI: −0.137; −0.008), which is mainly due to having more side effects (β=−0.119 95% CI: −0.187; −0.050). The later the changes in NIV settings occurred in the participants’ follow-up, the lower the evolution of the S3-NIV total score over time. This was mainly due to poorer sleep quality and greater adverse effects.

Although the main linear model of the S3-NIV total-score-change over time was adjusted for time since NIV initiation, we performed a sensitivity analysis for participants who had been on NIV for less than a year. The model (Figure Supplement 01) was fairly similar to the main model (including all participants) except for the effect of type of interface. The nasal mask did not appear to be associated with a more positive change of the S3-NIV score than the face mask.

Fig. 4 (and Table S2 supplement) shows the impact of the main NIV settings, type of interface, adherence, the presence of humidifier and additional oxygen therapy on changes in the scores of the 11 items of the S3-NIV questionnaire.

Fig. 4.

Odds ratio estimates of the variables in the mixed effect multivariable multinomial ordinal models of the 11 items of the S3-NIV questionnaire. (A) Variables Interface, supplemental oxygen and humidifier (B) Variables Adherence, pressure support, back up rate and positive expiratory pressure.

(0.21MB).

After controlling for confounding variables, perceiving too much air, or experiencing discomfort with the mask was less likely to occur over time in participants with higher levels of adherence. Participants with higher adherence levels were more likely to experience more respiratory symptoms over time, except for mucus in the airways.

In contrast, participants who required higher levels of pressure support were more likely to report NIV related side effects, including nasal and oral dryness, perceiving too much air, being disturbed by leaks, and waking up at night with breathing difficulties over time. However, participants with a higher NIV back up rate were less likely to experience mucus in their airways, have trouble breathing when they speak and shortness of breath. They were also less likely to report receiving too much air or waking up at night with breathing difficulties. Additionally, participants who used higher values of PEP were less likely to experience respiratory symptoms and sleep problems, however they were more likely to report being disturbed by leaks.

Dryness, mask discomfort, disturbing leaks, and waking-up during the night were less likely side-effects for participants who used a nasal mask or nasal pillows than for those who used a face mask. They were also less likely to report shortness of breath or problems breathing when speaking or eating.

Participants who had a humidifier were more likely to report complaints of dryness, mask discomfort, disturbing leaks, and mucus in airways. Participants who had additional oxygen therapy were more likely to report more respiratory symptoms (all those included in S3-NIV), poorer sleep and more NIV side-effects, except for the perception of disturbing leaks.