Design and study sample

For the current paper, we used data from the first wave of the Leiden 85-plus Study. All inhabitants of the Dutch municipality of Leiden who turned 85 between 1 September 1997 and 1 September 1999 (n = 705) were identified from the municipal’s register. No exclusion criteria were employed at inclusion, but for the current analysis, participants were excluded if they had severe cognitive dysfunction, operationalized as MMSE score < 19 points, as self-reported questionnaires on mental health characteristics were not administered to these subjects [11]. Out of the eligible population, 14 died before enrolment, and 92 declined participation. The remaining 599 participants (87%) agreed to participate [11]. 98 Participants had an MMSE score below 19 points, setting the final study sample at 501.

Measures

Participants received home visits of a trained researcher three or four times in their 85^th year during which standardised interviews and physiological measurements were conducted [11, 12]. All measurements were performed only once per year. Included measures are hereafter divided in demographics, physical health, mental health characteristics, and life satisfaction.

Data on demographics, sex and marital status was collected from the municipality. Participants provided information on their income (state pension or state pension and additional income), education (primary school or primary school and additional education) and living situation (independent or institutionalized) during one of the interviews.

For mental health characteristics, depressive symptoms were measured with the Geriatric Depression Scale 15 (GDS-15) [12]: a 15-item screening instrument for depression developed for older populations. For analysis, the outcome score that can range from 1 to 15 was divided into score ≤ 4 for less depressive symptoms and ≥ 5 for more depressive symptoms [15] and for figures, the outcomes were divided into scores ≤ 4, 5–9, and ≥ 10 [16]. Perceived loneliness was measured with the De Jong-Gierveld Loneliness scale [12]: an 11-item questionnaire designed to measure loneliness with scoring range 0 to 11. For analysis, the total scores were categorized into ≤ 2 for those who did not experience loneliness and ≥ 3 for those who did, and for figures scores were categorized into ≤ 2, 3–9, and ≥ 10 [17].

Life satisfaction was measured with Cantril’s 10-point self-anchoring ladder [18]. Participants were presented with a depiction of a ladder, the steps numbered one to ten. Participants were told that the bottom step is the worst and the top step is the best imaginable life for them and were asked to indicate where on the ladder they felt they stood at that time, resulting in a life satisfaction score ranging from 1 to 10.

Data analysis

Description of participants

The original sample of the Leiden 85-plus Study included 599 individuals and was representative of the entire source population that was intended for recruitment and the general Dutch population in terms of physical health, mortality hazard and demographics [11, 20]. For present analyses we excluded 98 participants with severe cognitive impairment, which was operationalized as an MMSE score below 19 points. Table 1 shows characteristics of the 501 participants included in the analyses. Life satisfaction was rated highly by most participants (median 8 out of 10 points (IQR 7 to 9)). Note the additional peak in the distribution at score 10 points (n = 88). Six participants rated their life satisfaction with 1 point and 21 participants rated it with less than 5 points. Ratings of life satisfaction were available for 43 of the 98 participants who were excluded due to severe cognitive impairment; the median score in this group was 8 out of 10 points (IQR 6 to 8).

Determinants of life satisfaction

There were no differences in life satisfaction between the sexes (0.1 points, 95% CI −0.4 to 0.3), nor between types of marital status (0.0 points, 95% CI −0.4 to 0.4), nor between those who had a state pension and those with additional income (0.0 points, 95% CI −0.4 to 0.5), nor between those who went to primary school and those with additional education (0.1 points, 95% CI −0.2 to 0.4). Life satisfaction among those who were institutionalized was 1.0 point lower (p < 0.001, 95% CI −1.5 to −0.5) than among those who were not institutionalized. This difference became smaller and lost statistical significance after adjustment for mental health characteristics (−0.5 points, 95% CI −1.0 to 0.1).

To provide further information, we performed various trend analyses (Figs. 1, 2 and 3). Figure 1 shows mean life satisfaction estimated by various characteristics of physical health: comorbidity, physical performance score, cognitive function, and functional status. Trend analysis showed that poorer physical performance had a significant but small association with lower life satisfaction in the crude model (Fig. 1b), but lost significance after adjustment for characteristics of mental health. Poorer functional status showed a significant but small association with lower life satisfaction in both the crude and the adjusted model corrected for unequal distributions in mental health characteristics. Other physical health characteristics were not related to life satisfaction.

Figure 2 shows the mean life satisfaction dependent on characteristics of mental health: depressive symptoms and perceived loneliness. Trend analysis showed that having more depressive symptoms and perceiving greater loneliness were strongly and significantly associated with lower life satisfaction (both p < 0.001). Further adjustment for unequal distributions of physical health did not substantially change these results.

All participants were followed up for vital status. Figure 3 shows ratings of life satisfaction at age 85 years dependent on their residual lifespan. No significant relation was observed, neither for those who died within one year after assessment, nor for those who died later in time. Participants who lived up to centenarian age, thus had a residual lifespan of 15 years and over, had reported a particularly high life satisfaction at age 85. More specifically, out of this small subgroup of sixteen participants, only two participants reported life satisfaction below the median of 8 points.

Finally, we examined effect sizes of the associations between physical and mental health characteristics and life satisfaction with squared Pearson correlation coefficients using the crude, uncategorised, full range variables. For physical health, the explained variance ranged from 0.1% for comorbidity to 3.6% for functional status. The explained variance of life satisfaction for all physical health variables together was 2.8%. For mental health, the explained variance was 25.9% for depressive symptoms and 13.0% for perceived loneliness, and 28.8% combined. When adjusting for demographic and mental health characteristics, the partially explained variance ranged from 0.0 to 2.0% for the various physical health characteristics. The partially explained variance was 23.1 and 11.1% for depressive symptoms and perceived loneliness respectively when adjusting for demographic and physical health characteristics.

Limitations

The original study sample can be considered characteristic for the Dutch population at 85 years, as there were no exclusion criteria. However, in the present study, we had to exclude those with severe cognitive impairments as we did not have (reliable) estimates for, amongst others, mental health for these subjects. The data can therefore not be extrapolated to participants with severe cognitive impairment. We did, however, obtain ratings on life satisfaction for almost half of them; these ratings were in the same range as for those individuals who were included in the analysis. Furthermore, since we showed that residual lifespan was not associated with life satisfaction.

The presented association between mental health and life satisfaction could be inflated by shared personality factors, such as neuroticism or by environmental factors, such as stressful life events. However, the focus of the current paper was on the relationship between physical health and life satisfaction, and mental health was merely included to find an explanation for the differences in life satisfaction between older people, since we showed that these were not due to differences in physical health. Therefore, in this paper we did not delineate specific affective disorders, nor did we investigate the nature of their association with life satisfaction (as long as it was not due to confounding by physical health). An inflated association between life satisfaction and mental health due to self-reporting could be suggested, but this is rather unlikely, because functional status was also self-reported but failed to show an association with life satisfaction in the same magnitude as mental health.

A weakness of the current study is that mental health characteristics were measured less extensively than physical health characteristics. However, this makes it all the more striking that the level of life satisfaction was so strongly explained by the few mental health measures that were available.

The objective of this study was to uncover the impact of physical and mental health on life satisfaction at advanced age. We found that in general, older people in the Netherlands were highly satisfied with their life, even in the presence of disease and disability. Poor physical health, including cognitive impairment, was not related to lower life satisfaction, and functional status was only weakly related to lower life satisfaction, especially when depressive symptoms and perceived loneliness were taken into account. In contrast, poor mental health was strongly related to lower life satisfaction, and characteristics of physical health did not alter these relationships.

The findings of the present study relate to international criticism and discussion on the use of Gross Domestic Product as a measure of a society’s ‘success’ and ‘well-being’ [34]. As is now widely argued, the relationship between Gross Domestic Product and well-being is not straightforward, and using these constructs interchangeably can lead to a narrow and biased view on well-being and its determinants. Our somewhat unexpected finding that physical health is hardly related to life satisfaction in the oldest old indicates that a similar warning is in order here.

The present data compel refraining altogether from measures that are composed of external determinants of wellbeing. Instead, well-being should be measured by asking people directly [28]. Improving well-being is a priority in public health [35]. and healthcare professionals and policymakers now increasingly make use of self-reported life satisfaction and its correlates to decide on the development and implementation of clinical interventions for individuals or on policies to improve well-being of the population [34]. In line with this, we believe that the present study contributes an important message: it suggests that more success in optimising life satisfaction for older people will come from interventions that prevent depressive symptoms and loneliness than from those that target physical health.