The impact of community-acquired pneumonia on the health-related quality-of-life in elderly

The current study, “Costs, Health status and Outcomes of CAP (Community-Acquired Pneumonia)” (CHO-CAP), was executed in parallel to the “Community-Acquired Pneumonia Immunization trial in Adults” (CAPiTA) trial, a placebo-controlled double-blinded RCT evaluating the effectiveness of a 13-valent pneumococcal conjugate vaccine in 84,496 community-dwelling elderly in the Netherlands [13, 14]. The CHO-CAP-study used a nested matched-cohort study design when recruiting patients hospitalized with a clinical suspicion of a pneumonia episode from the CAPITA-study population and prospectively followed them, along with non-diseased subjects (i.e. no pneumonia), for a one-year post-discharge period (Fig. 1) [15]. CAPiTA-participants were approached for study participation in CHO-CAP at the time of vaccination (November-2008-January-2010). Overall, 72,074 CAPiTA-participants¹ received the invitation to return the CHO-CAP-baseline questionnaire together with a signed informed consent form. Those who returned both questionnaire and informed consent formed the CHO-CAP source population (n = 47,476²). This source population was a priori eligible for the nested matched-cohort study. Within the CAPiTA-trial, 3225 patients with a suspected pneumonia were identified in 56 Dutch sentinel hospitals. Potential cases for the “diseased”-cohort were subjects with a first-time suspected pneumonia episode, participating in the CHO-CAP source population, without recently diagnosed malignancy and able to complete questionnaires. After hospital discharge, these subjects were invited for participation in the diseased cohort. For feasibility reasons, we included all patients hospitalized with a clinical suspicion, which were later stratified upon either or not receiving a radiological confirmation of CAP by the blinded chest X-ray adjudication committee. Those who consented were visited at home within 2 weeks of hospital discharge by trained interviewers. They were asked to complete self-administered questionnaires during the home visit (day 0), and at 1, 6 and 12 months after the home visit through postal questionnaires. For each subject in the diseased cohort, two matched non-diseased subjects (i.e. no pneumonia) were identified from the CHO-CAP source population (referred hereafter as non-diseased subjects). Matching was based on age, sex, and EQ-5D-3L-index collected at vaccination, implying that health status of people with suspected pneumonia and their matched non-diseased subjects was similar at the time of vaccination. Non-diseased subjects adhered to the same inclusion criteria as suspected pneumonia patients. They were asked to complete self-administered questionnaires during the home visit (day 0). Further questionnaires were sent by regular mail 1, 6 and 12 months after the home visit. For full details see Mangen et al. [15].

Additionally, we did distinguish between “radiologically confirmed” CAP cases and the matched non-diseased subjects and “radiologically non-confirmed” CAP cases and their matched non-diseased subjects. A “radiologically confirmed” CAP was defined as the presence of two or more clinical signs of pneumonia together with a chest x-ray consistent with pneumonia, identical to the definition used within the CAPiTA-trial [13].

Date of birth, sex, place of residence, loss-to-follow-up due to death during the follow-up period and causes of death were extracted from the CAPiTA-study files [14]. Health status (EQ5D) and socio-demographic status (living situation and education) were collected at the time of vaccination with the CHO-CAP-baseline questionnaire. Full details of data collection of nested matched-cohort study are provided in Mangen et al. [15]. In short, comorbidity details were collected during the home visit. Current living situation was collected at all four contact moments. Health status (EQ5D) was collected thrice for suspected pneumonia cases during the home-visit interview, reflecting health status (1) at day of interview, (2) at the worst moment during the recent pneumonia episode, and (3) previous to the recent pneumonia episode. Health status was also collected at month 1, month 6 and month 12 after initial visit using both EQ5D and SF-36. For non-diseased subjects, both EQ5D and SF-36 were administered at all four contact moments. For suspected pneumonia cases, clinical information on hospital admission (e.g. X-ray result; clinical symptoms; length of stay) was extracted from the CAPiTA-study files [14].

The SF-36 is composed of 36-items measuring health across eight domains (physical functioning, social functioning, role limitations with respect to physical activities, role limitations with respect to emotional activities, pain, mental health, vitality and general health perception). Responses to each item within a domain are combined to generate a score from 0 to 100, where 100 indicates best health [11]. Because of the elderly population - some potentially in poor health - the usual order of items in question 3 was inversed [16]. Applying the scoring-method developed by Brazier et al. [17], we further derived the SF-6D health-index from the SF-36 survey, a numerical index between 0 (“death”) and 1 (“full health”) [17]. The EQ5D consists of two parts, the EQ-5D-descriptive system and the EQ-visual analogue scale (VAS) [18]. The EQ-VAS records the participants’ self-reported health on a VAS from 0 (“Worst imaginable health state”) to 100 (“Best imaginable health state”) [18]. The EQ5D-descriptive system consists of five domains (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) and three functioning levels (no problems, some problems or severe problems) [12, 18]. The EQ5D health states were scored with the Dutch value-set [19], to obtain EQ-5D-3L summary indexes (EQ-index) ranging from 0 (“death”) to 1 (“full health”) [12, 19].

Quality Adjusted Life Years (QALY) is a concept used to reflect a year in full quality of life. Hence, QALYs combine both length of life and quality of life. Quality of life is represented in a value between 0 (death) and 1 (optimal quality of life). QALYs are estimated by multiplying length of life with the indicator value for quality of life. One-year QALY estimates, with and without pneumonia episode included, were calculated for both cohorts, using the self-reported EQ5D health states and its associated index values at the different contact moments based on recorded date of contact moment. An area under the curve approach was followed by interpolating between the observations provided by the patients. For patients who died, we calculated QALY by using the date of death and a utility score of 0 from that date onwards. For missing EQ-indexes, ten imputations were performed. QALYs were calculated in each imputed dataset and averaged over the ten imputated datasets. The observed utility difference between both cohorts was attributed to the CAP episode. Excess QALY loss was calculated for the one-year post-discharge period, excluding and including the CAP episode, respectively (see Additional file 1: Figure S1A for illustration). Additionally, we estimated the QALY and the excess QALY loss for one-year survivors. Pairs of pneumonia cases and non-diseased subjects in which one of the three died during the follow-up were excluded from these estimates.

Health status and QALY estimates in the suspected pneumonia cases and non-diseased subjects are presented for the different follow-up moments, in a decomposed manner (i.e. at the level of the different domains of quality of life) and as summary index. Causes of death were categorized into five categories: (1) infectious diseases; (2) chronic lung diseases; (3) cancer; (4) cardiovascular events and stroke and (5) others. Depending on the nature and distribution of data, we used Chi-square test for categorical variables (e.g. sex, education) and non-parametric tests for non-normally distributed variables (e.g. scores) to test for differences between both cohorts (i.e. diseased and non-diseased cohort) and for differences between patients with and without confirmed CAP. Correlation between the different instruments was studied using Spearman’s rank correlation coefficient. All analyses were performed using IBM SPSS Statistics version 21.

Of the 3225 identified suspected pneumonia episodes in the CAPiTA-trial, 1750 (54%) belonged to the CHO-CAP source population, of which 562 (32%) participated; 341 (61%) had radiologically confirmed CAP and 221 (39%) had radiologically non-confirmed CAP (Fig. 1). Reasons for non-participation/exclusion are provided in in the Additional file 1: Table S1A.

Baseline characteristics of the suspected pneumonia cases and their non-diseased subjects are summarized in Table 1. Despite adequate matching on three criteria, non-diseased subjects had fewer comorbidities, were higher educated and were more often from the south of the Netherlands than pneumonia cases (Table 1).

Compared to the non-diseased subjects, suspected pneumonia cases more frequently died during the one-year follow-up period (8.4% vs 1.2%;p < 0.001 (in the Additional file 1: Figure S2A.); attributable risk:0.059(95% CI:0.058–0.060)) or withdrew from the study (16.7% vs 9.9%; p < 0.001), mainly because of bad health (in the Additional file 1: Table S2A). Chronic lung diseases, cardiovascular events and stroke were more frequently reported as being the cause of death for suspected pneumonia cases than for non-diseased subjects (p = 0.054; in the Additional file 1: Table S2A)).

Figure 2 shows the distribution of suspected pneumonia cases and non-diseased subjects reporting problems in the five domains of the EQ5D-instrument at all observation moments. By definition, non-diseased subjects and suspected pneumonia cases had a similar health status at the moment of vaccination. However, during follow-up suspected pneumonia cases more frequently reported problems in the five domains of the EQ5D-instrument than non-diseased subjects at all contact moments and for all domains (Fig. 2). This finding was confirmed by the SF36-questionnaire. Suspected pneumonia cases had persistently lower SF-36 mean scale scores on all domains and during all contact moments, compared to their non-diseased subjects (Fig. 3). Furthermore, HrQol, as expressed in EQ5D and SF-6D index value and EQ-VAS score of suspected pneumonia cases was constantly lower than HrQol of non-diseased subjects during the one-year follow-up (Table 2, in the Additional file 1: Figure S3A). At month-12, HrQol-scores were significantly lower for survivors in the suspected pneumonia cohort compared to the non-diseased cohort (EQ5D-index:0.74 vs 0.82(p < 0.001) and SF6D-index:0.68 vs 0.75(p < 0.001)).

One-year QALY estimates, excluding the CAP episode and using the self-reported EQ5D health status and its associated-index values, were 0.68 and 0.81 for suspected pneumonia cases and non-diseased subjects (p < 0.001), yielding a utility difference between both cohorts of −0.13 attributable to suspected pneumonia. One-year QALY estimates, including the CAP episode, were 0.67 and 0.81 for suspected pneumonia cases and non-diseased subjects (p < 0.001), yielding an excess QALY loss of 0.15. Slightly smaller QALY differences were obtained when considering only survivors (Table 3), resulting in excess QALY loss of 0.10, if excluding the CAP episode, and 0.11, if including the CAP episode, respectively.

EQ5D-index, EQ-VAS and SF6D-index were positively correlated at all contact moments for suspected pneumonia cases and non-diseased subjects (rho > 0.45), in the Additional file 1: Table S3A and Table S4A. The highest correlation was found between EQ5D-index and SF6D-index for both suspected pneumonia cases and non-diseased subjects (rho >0.67 for all contact moments).

There were apparent differences between patients with radiologically confirmed and non-confirmed CAP, but when repeating analyses in a stratified manner interpretation did not change. Radiologically confirmed and non-confirmed CAP cases did not differ in baseline characteristics and were adequately matched to their non-diseased subjects (in the Additional file 1: Table S5A). Radiologically confirmed CAP cases stayed significantly longer in hospital than radiologically non-confirmed CAP cases, but had comparable median numbers of clinical symptoms, pneumonia severity index (PSI) scores [20], admissions to ICU and readmissions within 30 days, Table 4. Mortality was higher for patients with radiologically confirmed CAP (10.3%) compared to non-confirmed CAP (5.4%; p = 0.043), but the causes of death were comparable (in the Additional file 1: Table S6A). The attributable risks of dying, compared to non-diseased subjects, were 0.083 (95%CI:0.082–0.084) for those with radiologically confirmed CAP and 0.047 (95%CI:0.047–0.048) for those with non-confirmed CAP. As compared to their non-diseased subjects, both radiologically confirmed and non-confirmed CAP cases had a significantly persistent lower HrQol, independent of the instrument used (see in the Additional file 1: Figure S4A-S6A and Table S7A.). The excess QALY loss attributable to CAP was with 0.14 for radiologically confirmed CAP cases and 0.12 for radiologically non-confirmed CAP cases, of the same magnitude, independent of radiological confirmation of CAP (in the Additional file 1: Table S8A.).