Self-Rated Health as a Predictor of Mortality in Older Adults: A Systematic Review

The aim of this study was to investigate the link between self-reported health (SRH) and mortality in older adults. In total, 505 studies were found in PubMed and Scopus, of which 26 were included in this review. In total, 6 of the 26 studies included did not find any evidence of an association between SRH and mortality. Of the 21 studies that included community dwellers, 16 found a significant relationship between SRH and mortality. In total, 17 studies involved patients with no specific medical conditions; among these, 12 found a significant link between SRH and mortality. Among the studies in adults with specific medical conditions, eight showed a significant association between SRH and mortality. Among the 20 studies that definitely included people younger than 80 years, 14 found a significant association between SRH and mortality. Of the twenty-six studies, four examined short-term mortality; seven, medium-term mortality; and eighteen, long-term mortality. Among these, a significant association between SRH and mortality was found in 3, 7, and 12 studies, respectively. This study supports the existence of a significant relation between SRH and mortality. A better understanding of the components of SRH might help guide preventive health policies aimed at delaying mortality in the long term.


Introduction
Numerous studies have investigated the predictive value of self-reported health (SRH) on mortality or adverse health outcomes in both young and old adults [1][2][3]. Overall, the results of these studies, particularly regarding the link between SRH and mortality, widely vary according to the age and sex of the population studied, the length of follow-up, or the presence or absence of specific diseases [4]. It is, therefore, difficult to know with any certainty what weight should be given to patients' SRH. This difficulty is particularly marked among older adults, who are often frail and multimorbid, and who may have a life expectancy that is limited by one or more chronic diseases. SRH is a valuable assessment, because it covers multiple components and is easy to collect. Several authors [5,6] have shown the multiple domains are encompassed by the term self-reported health. However, the contribution of each individual component to the overall evaluation remains to be determined and seems to vary according to the context (gender, socio-economic or educational level, age category, religion, etc.). The evaluation of SRH yields a more comprehensive view of an individual's health and may be more accurate than a purely medical evaluation. Moreover, it allows physicians to understand complex predictive factors of health, such as chronic inflammatory status [7,8]. Finally, SRH can be evaluated by asking a single, simple question [9].
In this systematic review, we aimed to determine whether there is a significant link between SRH and mortality in older adults.

Study Selection Criteria
Study eligibility criteria were defined prior to performing the literature search by two senior researchers (L.G. and M.D.) according to the PICOS framework. Studies were eligible for inclusion if they reported data on self-rated health. The population of the studies included people aged 65 years or older, of any sex, ethnicity, or living place. The groups to be compared were defined according to their levels of self-rated health (SRH). The outcome was death, whatever the timepoint. Basic science articles, reviews, case reports and case series, editorials, and correspondence were excluded.

Data Extraction
Data analysis was performed using Covidence systematic review software© (Veritas Health Innovation, Melbourne, Australia), available at www.covidence.org (accessed on 23 March 2022). After eliminating duplicates, two senior researchers (L.G. and M.D.) independently reviewed the titles and abstracts of all articles. In case of disagreement about whether or not to include an article, the case was discussed until consensus was reached. Overlap between studies in the results reported was checked. We independently extracted the data, using the same data extraction form. For descriptive analyses, the following data were extracted: publication year, country where the study was conducted, study design, study setting, medical condition (if any), sample size, and age (mean or median and their statistical dispersion parameters, when available). To analyse the relation between SRH and mortality, the following information was collected: outcome (death or survival), type of analysis (whether adjusted or not), SRH levels, statistical estimates (hazard ratios, odds ratios, rate ratios, and rates) and their respective 95% confidence intervals (95% CIs), and the level of significance (p-values).

Quality Assessment
The quality of the included studies was assessed independently by two researchers (L.G. and M.D.) using the Newcastle-Ottawa Scale (NOS) [10]. The NOS consists of three quality parameters: selection, comparability, and outcome assessment. The "selection" criterion is scored between 0 and 4 points; the "comparability" criterion is scored between 0 and 2 points; and the "outcome" criterion is scored between 0 and 3 points. The sum of the scores of these three criteria gives an NOS total score between 0 and 9 points. NOS scores of 7 or over were considered to be of high quality, while 5-6 indicated moderate quality, and scores under 5 indicated low quality. Disagreement was resolved by means of a joint review of the manuscript to reach consensus, and the opinion of a third researcher was requested when necessary. When appropriate and possible, certain parameters were calculated from available data (e.g., pooled mean age and/or standard deviations, odds ratios, rate ratios, etc.).

Results
In total, 505 studies were identified during the literature search ( Figure 1). Among these, 195 duplicates were excluded. After examination of the titles and abstracts of the remaining 310 studies, 98 articles were retained for full-text assessment. After reading the full text of these 98 studies, 72 were excluded for one or more of the following reasons: inappropriate age of the study population, wrong study design, or wrong outcome. Thus, 26 studies were included in the final review. criterion is scored between 0 and 4 points; the "comparability" criterion is scored between 0 and 2 points; and the "outcome" criterion is scored between 0 and 3 points. The sum of the scores of these three criteria gives an NOS total score between 0 and 9 points. NOS scores of 7 or over were considered to be of high quality, while 5-6 indicated moderate quality, and scores under 5 indicated low quality. Disagreement was resolved by means of a joint review of the manuscript to reach consensus, and the opinion of a third researcher was requested when necessary. When appropriate and possible, certain parameters were calculated from available data (e.g., pooled mean age and/or standard deviations, odds ratios, rate ratios, etc.).
The quality of the included studies, as assessed using the NOS, is summarized in Table 3. The quality was considered high for all 26 studies. Table 3. Quality assessment of the different studies included in this systematic review performed using the Newcastle-Ottawa scale (NOS).
For the two studies that included people aged 80 years or over, the authors showed a significant relationship between SRH and all-cause mortality at each timepoint (6 weeks; 6 months; and 1, 2 and 3 years). However, it seems difficult to extrapolate these results, as they all concern the same population, hospitalised via the emergency department for an acute condition.
SRH is a composite concept that encompasses medical, social, cultural, religious, ethnical, and individual dimensions. Several authors have attempted to characterise the different dimensions of health under the term "SRH" [1,5,6,[38][39][40]. The share of each dimension in the overall subjective feeling varies from one individual to another, explaining the variable strength of the link between SRH and mortality according to gender, culture, ethnicity, socio-economic level. and even age group [33,36,[41][42][43]. Zajacova et al. [44] showed that the individual criteria taken into account when assessing SRH varied from one sex to another as well as according to the period of life. Younger women tended to assess their SRH more unfavourably than men of the same age, while older women had a more favourable view of their SRH than men of the same age. This trend is even more salient if socio-economic factors (such as education, marital status, or income) are taken into account. As people age, the SRH is generally poorer, in both sexes, and this worsens as health problems and loss of autonomy increase. This illustrates the likely important role of medical criteria and functional status in the assessment of SRH with advancing age. Zajacova et al. [44] pointed out that all health indicators (physical health such as functioning or pain, mental health such as depressive symptoms, and health behaviours) are significantly associated with SRH, regardless of age or sex. Cott et al. [45] made the same observation in adult populations with one or more chronic diseases.
SRH is also associated with other factors known to predict outcome in the elderly population, such as interkeukin-6 (IL-6) [7]. Arnberg et al. [7] found that good or very good SRH was associated with low levels of systemic markers of inflammation in a population with a median age of 74 years (range of 60-93 at inclusion). Christian et al. [8] reported similar findings. Taken together, these data confirm that the collection of SRH in routine practice would be a simple and effective way of complementing the usual medical assessment to extrapolate an individual's life trajectory.
Throughout life, including in the older population, the SRH seems to be a fairly accurate assessment of an individual's functional capacities and even functional reserves for coping with the hazards of life, as evidenced by the predictive link with all-cause mortality demonstrated at all ages of adult life and at all timepoints. SRH is easily collected [9], even in people with mild-to-moderate cognitive impairment [31,46].
The methods used to collect SRH are variable. In our systematic review, some authors chose to assess the SRH on a value scale from excellent to very poor (excellent, very good, good, fair, poor, and very poor). Others chose to class SRH on a binary scale (SRH (excellent, very good or good) versus (fair, poor or very poor)). Of the 11 authors who evaluated SRH on a binary scale, 9 (i.e., 81.8%) found a predictive link between SRH and mortality in the short, medium, or long term [11,12,18,19,23,24,28,30,31]. Among the authors who treated the SRH according to a multiple choice scale, 12 (i.e., 70.6%) [13][14][15][16][17][18][20][21][22]26,34,35] showed a predictive link between SRH and mortality for at least one time point. The predictive capacity of the SRH with respect to mortality seems to be better when SRH is treated as a binary variable, most likely because there is greater statistical power with a dichotomous variable than with a non-binary, categorical one.
The predictive link between SRH and specific mortality in specific medical conditions seems to be more difficult to establish, because it is less well documented. In this systematic review, three articles investigated mortality linked to cancer [13,15,16], and two of them found a significant predictive relationship between SRH and cancer-related death in the medium [16] and long term [13]. Five articles investigated cardiovascular mortality [13,15,16,23,26], of which four [13,15,23,26] found that SRH significantly predicted cardiovascular death in the long term in a population with a mean age of <80 years.

Conclusions
SRH seems to be a good criterion for assessing the risk of mortality in the short, medium, or long term in a population of elderly subjects living at home according to the articles studied in this systematic review. SRH assessment is complementary to so-called objective medical measures. SRH is simple to collect, which makes it easy to use for health professionals and acceptable to the population. Its composite nature makes it possible to take into account an individual's health in a global manner.
A better understanding of the components of SRH and their respective weight at each age might help to guide preventive health policies aimed at delaying mortality in the long term. However, there are currently no studies that have established that improving the criteria comprising SRH would make it possible to reduce mortality.
Moreover, as the weight of each criterion seems to vary according to the individual and the age considered, targeted interventions may not be very effective. The composite nature of the SRH concept should encourage us to implement comprehensive prevention strategies from the outset, individualised and variable over time for greater effectiveness.
Prevention strategies should be implemented early in the life of the individual and continue throughout life. The identification of poor SRH in a patient should prompt healthcare providers to promptly look for associated modifiable factors in an attempt to improve them.  Data Availability Statement: Data can be make available at moustapha.drame@chu-martinique.fr.