This prospective cohort study demonstrated that among community-dwelling older outpatients in Italy poor QOL and HRQOL, as described by the lowest score-based quartiles of the OPQOL total score and health-related OPQOL sub-score respectively, were independent predictors of several adverse health outcomes: falls and ED admissions for overall QOL as well as falls, nursing home placement and death for HRQOL. Our findings lend support to the prognostic value of QOL measures in older people and grant further insight into the association between QOL and adverse health events. As far as the novelty of the study is concerned, some points deserve particular mention. First, to the best of our knowledge, our study provides the first evidence of the predictive value of a poor HRQOL on the occurrence not only of death but also of nursing home placement at one year, after statistical correction for a number of variables including the frailty syndrome. Indeed the latter is an acknowledged predictor of adverse health outcomes, as illustrated in the Background section, and has recently been shown to be the main condition leading community-dwelling older people to death [38
]. The choice of the SOF criteria to diagnose frailty is justified by their having been recently validated in large population studies in the U.S. [27
] and successfully applied to a sample of older subjects attending the same geriatric clinic [30
Second, the finding that a poor QOL and HRQOL are independently associated with a greater risk of falls at one year is also a novel one. A possible explanation could be that a poor QOL at the baseline visit actually selected a subset of participants who had already experienced falls in the previous year. In fact it is widely recognised that patients who have fallen are at greater risk of further falls [39
] and it is equally well known that falls worsen the QOL. This latter effect is mediated by the "fear of falling" syndrome by which older adults who have fallen develop psychological distress and unnecessarily restrict their activity [40
]; indeed fall prevention programmes have improved several dimensions of the HRQOL (i.e. physical function, social function, vitality, mental health and environmental domains) in elders living in the community [41
]. Yet, the hypothesis of a selection bias does not hold since this association persisted after correction for previous falls at multivariate analysis. An alternative explanation could be that a poor QOL and HRQOL may derive from a number of factors - such as dissatisfaction with one's health, lower social participation or support, negative feelings about the neighbourhood - which reduce the individual's confidence and lead to a constriction of his/her life space. The latter is a measure of spatial mobility, defined as the size of the spatial area people purposely move through in their daily life [42
]. Constriction of the life-space is a condition known to decrease physical activity, accelerate physical deconditioning and the decline in physiological reserves [43
]: it can be thus speculated that it may increase the risk of falls through a pathophysiological mechanism resembling that of the "fear of falling" syndrome. It can also be supposed that constriction of the life-space contributed to our finding of a correlation between HRQOL and death even after correction for disability and the frailty syndrome: in a population study involving older women, not frail at baseline, it emerged as an independent predictor of both frailty and frailty-free mortality [43
]. Of course all hypotheses concerning the relationship between the QOL, life space constriction and adverse health outcomes should be verified by appropriate studies.
Third, another element of novelty of the study resides in the fact that we considered both HRQOL and generic QOL. It is interesting to note that HRQOL and QOL were found to have an impact on different adverse health outcomes. Death and nursing home placement were predicted only by a poor HRQOL, probably because they are mainly due to poor health and poor functional status. ED admissions were instead predicted only by a poor generic QOL. This latter finding suggests that a greater use of the ED by elders is associated with dimensions of the QOL other than the HRQOL, such as dissatisfaction with social support, personal relationships and living environment as well as with a negative perception of one's independence and control over life. In other words, it seems that the subjective distress which makes older people seek help from the ED may be caused not only by physical dysfunction but also by purely social/psychological factors. In keeping with this hypothesis, it has been shown that in older patients discharged from an emergency department in Italy, a multidimensional intervention, based on a CGA performed after discharge, was able to reduce the rate of ED readmissions at a three-month follow-up and was also able to improve not only morale and nutritional status but also generic QOL [44
]. It must be emphasised that a poor QOL is associated with several acknowledged predictors of ED admissions such as depressive symptoms, lack of social support, loneliness, larger use of ED visits [45
]. However, it is noteworthy that in our study this correlation persisted after adjustment for living conditions, depression and previous admissions to the ED.
Finally, some discussion must be devoted to a few methodological issues. When taking falls, ED admissions and hospitalisation as adverse health outcomes we decided for a qualitative rather than a quantitative approach - i.e. we chose to assess the occurrence of any such event in the year after the baseline visit and not the number of events. The latter would in fact have introduced a greater recall bias since it is reasonable to suppose that after a relatively long period of time participants would be able to more accurately report on the absence/presence of adverse events than on the specific number of intervening events. Indeed the reliability of the data so collected is testified by the rate of falls within our sample: we found a 40% prevalence of any fall during one year which appears consistent with figures in the literature - 27% (95% CI 19-36%) according to a review of 18 studies on older community-dwelling subjects [39
] - considering the outpatient nature of our population. In fact older subjects referred to a geriatric clinic for health care are likely to be selected for greater comorbidity and risk of adverse events. This same explanation can apply to the high prevalence of frailty, dementia and depression observed in the sample and is supported by the fact that in other recent studies on older outpatients with a disability referred to the same geriatric service the rates of depressive disorders and cognitive impairment were found to be even greater [50
]. Moreover, it must be noted that frail subjects make larger use of health and community services than subjects who are not frail [52
]. Another methodological issue deserving discussion is that we decided to include in the study even subjects suffering from mild or moderate dementia if they were able to understand and reliably answer the OPQOL questionnaire. Such choice was based on the fact that a large proportion of older people can reliably answer questions about their QOL even if they are affected by mild or moderate cognitive deficits. This notion has generally been reported by the literature [53
] and is consistent with the baseline data of the study, which has specifically shown that the OPQOL questionnaire is applicable to subjects with cognitive impairment [10
With reference to the limitations of the study, it must be remarked that in the statistical models we found a rather large 95% confidence interval for the odds ratio of nursing home placement and death in relation to the OPQOL health-related sub-score. Although this is certainly not due to multi-collinearity between variables, as previously explained in the Methods, the predictive value of the OPQOL on these two health outcomes needs to be confirmed by further studies conducted on larger samples of community-dwelling older people. Moreover, since the sample analysed consisted of outpatients referred to a geriatric clinic by their general practitioners, our findings cannot be automatically extended to the entire population of older people living at home in Italy. Although we cannot exclude that we might have selected a group of community-dwelling older adults with better social and health assistance, a selection based on economic status can certainly be ruled out since in the specific Italian setting all citizens are granted free access to outpatient services. However, the possible occurrence of a selection bias does not invalidate the clinical relevance of our results and indeed may enhance it. First, the predictive value of the OPQOL score was established in what could be a "best scenario" population. In fact, among the subjects recruited at baseline we lost to follow-up the older and sicker ones who were likely to exhibit greater vulnerability. Moreover - and foremost - all the subjects considered had undergone a CGA and had received individually-tailored therapeutic advice focused on improving their health and QOL, which is the standard approach of geriatric outpatient visits. This highlights the fact that, within the CGA, the administration of the OPQOL questionnaire to evaluate the QOL - particularly in its health-related domain - could better identify those high-risk subjects to whom additional measures should be targeted. Even though specific treatments for frail and vulnerable older patients are yet to be developed and clinically tested [15
], and although QOL has seldom been shown to be improved in the very few randomised controlled trials targeting even QOL in frail older people [55
], our findings underscore the need for research along this line employing also QOL measures such as the OPQOL.