We found that HRQoL, as measured by the SRI questionnaire, was inversely associated with mortality in LTMV patients before and after adjustment for covariates. In the total group of LTMV patients, the adjusted analyses showed significant inverse associations between mortality and the SRI sum score and the SRI subscales, ‘physical functioning’, ‘social functioning’ and ‘psychological well-being’. Furthermore, mortality varied considerably between the disease groups during the six-year period. The highest mortality was among COPD patients with established CHRF receiving LTMV. The majority of mortality in COPD is related to cardiac disease and the requirement of LTMV in COPD might be understood as a marker of overall frailty and multi-system disease severity. The lowest mortality was in the CWD group, reflecting the non-progressive nature of the disease in these patients.
As shown in previous studies [
3,
10‐
13,
23], mortality in patients treated with LTMV is associated with underlying disease categories. Previous studies have shown large variations in the attending patient categories, severity of disease, and follow-up times. Thus, a direct comparison of mortality between studies on patients with LTMV is challenging and might lead to an oversimplification.
HRQoL as a prognostic factor
The association between poor HRQoL and increased mortality in the total group of LTMV patients is consistent with the main findings of other similar studies on LTMV patients [
30,
31]. In line with Budweiser (2007a), crude analyses of SRI were significantly associated with mortality in all SRI subscales, with the exception of the ‘attendant symptoms and sleep’ scale.
The adjusted analyses among NMD patients showed that SRI ‘physical functioning’, ‘psychological well-being’ and ‘social functioning’ continued to be significant factors for mortality, which was consistent with the study by Budweiser [
30], but with different adjustment variables than those in our study. We also found associations between SRI and mortality among COPD patients in the adjusted analyses in the ‘attendant symptoms and sleep’ and ‘psychological well-being’ SRI subscales.
The initial choice of the adjustment variables in the present study was based on previous work that evaluated age [
10,
15,
26,
45], sex [
18,
48], education level [
44], marital status [
45], disease categories (NMD, COPD, OHS and CWD) [
2,
3,
10‐
13,
23] and comorbidity [
13,
18,
22,
46]. The variables ventilator dependency and time since LTMV was initiated were chosen a priori. Marital status was not associated with neither mortality nor the SRI sum score and was therefore excluded as adjustment variable. There were no sex differences between survivors and those who deceased, thus sex also was excluded as an adjustment variable.
However, we have considered the possibility that HRQoL could be influenced by other confounding covariates that might also pose a risk of death, such as PaCO
2. Reduced PaCO
2 levels have been related to lower one-year mortality and improved SRI scores in COPD patients treated with LTMV [
7,
8]. On the other hand, exploratory analyses did not identify any significant correlations between changes in hypercapnia status or baseline hypercapnia status and mortality in this group [
5]. However, in the present study, PaCO
2 values were normalized at baseline as a result of ongoing LTMV and were therefore not included in the analyses. The results from studies on lung function and survival in LTMV patients are not conclusive. Some studies [
19,
23,
30] reported associations between low FEV
1 and FVC and mortality, whereas another study [
11] found no differences in baseline lung function between the survivors and deceased patients. When FVC and FEV
1 were added to the Cox regression analysis in the current study, the result was altered only slightly; however, this result might also be influenced by missing lung function data (FVC baseline numbers did not sum to 112 due to 23 missing data points, FEV
1 baseline numbers did not sum to 112 due to 22 missing data points), some of the missing data might be explained due to patients having difficulties performing the spirometry test.
We also considered to include ventilation mode as a covariate as longer survival were reported in patients with DMD using non-invasive LTMV compared to those receiving LTMV via a tracheostomy [
14,
49]. However, another study concluded that the risk of death was not associated with use of invasive versus non-invasive LTMV in patients with DMD [
16], No significant difference in one year mortality was found between patients receiving LTMV via a tracheostomy and those weaned after discharged from the Intensive Care Unit (ICU) and no significant difference in HRQoL measured by SRI at discharge from ICU were found between the two groups [
50]. However, HRQoL tended to be lower, in the SRI ‘physical functioning’, while scores for ‘anxieties’ tended to be better in patients receiving LTMV via tracheostomy compared to those treated with non-invasive LTMV [
51].
Although the analyses in the present study were adjusted for education level, other economic confounding variables, such as income, might also have an impact on HRQoL and mortality. On the other hand, Norwegian society and health care services probably represent one of the most equitable systems worldwide, where all citizens have equal access to health care services. Nevertheless, the number of covariates that could be included in the analyses in this study was limited by the sample size at baseline, and we can never exhaustively cover all variables of minor importance among LTMV patients.
Why and how SRI predicts mortality
Previous studies using patient-reported measures other than SRI have also reported an association between self-reported health and mortality in patients treated with LTMV [
11,
26]. However, these studies did not adjust for the same covariates as the present study, and they lacked important variables, such as comorbidity and education level. There is a large body of evidence on the association between self-reported health measures and mortality in other settings and disorders, such as in communities [
24], in patients with cancer [
25] and idiopathic pulmonary fibrosis [
27]. Explanations of these consistent findings are complex and imply that survey respondents’ perceptions of health status are holistic; they include information on medical status but that information might be evaluated differently by men and women in different social positions, with different reference groups providing different social comparisons [
24]. Further, the accuracy of self-reported health as a predictor of mortality depends on the comprehensiveness and accuracy of the information that the person incorporates into the self-rating [
52]. This hypothesis corresponds with SRI as a multidimensional comprehensive questionnaire that captures the symptoms of CHRF and covers essential aspects of LTMV patients’ daily life [
28].
Clinical implication of the associations between SRI and mortality
Individuals suffering from CHRF treated with LTMV often have an incurable disease [
2‐
4]. Health care professionals and relatives tend to behave differently depending on whether the disease is perceived as a chronic or terminal condition. However, the distinction between the patient’s condition as chronic or terminal might become vague and can sometimes be ambiguous and difficult to interpret [
53]. Prognostic information from the SRI questionnaire might provide valuable knowledge on how to cope with these situations, improving treatment plans and communication between involved professionals, family members, and the LTMV patient. Our study demonstrates that the risk of death decreases by each unit increase in the SRI score. This result suggests that LTMV patients with low SRI should be identified, initiating thorough considerations on how to improve HRQoL. However, whether the relationship between mortality and quality of life is causal and changes in HRQoL status in some way influences mortality cannot be confirmed in this study design.
The minimal clinically important difference of the SRI questionnaire has not been defined [
41]. However, the great numerical difference in SRI score at baseline between the surviving LTMV patients and those who died during the follow-up, support the clinical relevance of the study.
Strengths and limitations
As far as we are aware, this study is among the very first to examine SRI scores as a predictor for mortality in LTMV patients with a follow-up time as long as 80 months. The strengths of the study include the use of standardized data collection [
32‐
34], including relevant confounders, such as comorbidity, which is often lacking in study of this type, and the prospective study design. Another strength is the use of the specific and validated SRI questionnaire, which can capture HRQoL related to symptoms and the experience of having CHRF and LTMV [
2,
28‐
30,
34‐
43].
The study has some limitations. First, its small sample size may decrease the statistical power to detect clinically relevant associations in multivariate Cox analyses. Second, comorbidity modeled simply as the number of somatic diagnoses. Charlson Comorbidity Index [
54] is a common index to measure comorbidity using ICD-10 codes. However, as complete ICD-10 codes were not available in our data, we chose to measure comorbidity as the number of somatic diagnoses. Thirdly, some of the LTMV patients answered that they received help to complete the questionnaire, which might introduce some information bias in SRI scores. However, it is of great importance to include the SRI scores from patients who needed help to fill out the questionnaire.
In addition, because of the observational study design, we cannot exclude the possibility of residual or unknown confounding. Whether HRQoL score reflects a perception by the LTMV patient of progression of her or his condition or whether change in HRQoL status in some way also influences the course of the condition is an interesting question. However, the research design cannot confirm causality between improvement in HRQoL and survival in this study. To address this question a randomized interventional study aiming to improve HRQoL with a control group receiving standard treatment would be more suitable.