Introduction

Physicians usually initiate consultations with open-ended questions such as, “How are you?” or, “How are you feeling?” These questions address patients’ perception of their overall health status, a concept also known as self-rated health (SRH). However, the validity of SRH has been a matter of debate in epidemiological and clinical research [1], due to its uncertain biological significance [2]. Some aspects of SRH remain open to debate, notably the biological significance of rating oneself with poor or very poor health. A number of questions arise nowadays regarding this issue, such as: In terms of biological correlates, what is the significance of answers to global questions about health status, such as ‘How would you rate your health’? How could these global health assessments be used in daily clinical work? How are these subjective judgments influenced by personal bias as opposed to objective health criteria?

SRH has become a well-established predictor of mortality and other adverse outcomes [3–9]. The predictive value of SRH appears to be independent of other well-known factors associated with mortality, such as cardiovascular disease and smoking, among others. In spite of its widespread use in different health surveys, there is a lack of information concerning the consistency of perceptions about SRH across populations in different cultural and socioeconomic contexts [10–13].

Physical performance tests also have good predictive value in terms of mortality, disability, and rate of hospitalization, and appear to be consistent across populations, due to their objectivity and the possibility of reliable third-party measurement [14]. These tests include a variety of measures, including grip strength, gait speed, endurance, balance, and others. In particular, the Short Physical Performance Battery (SPPB) consists in a set of three tests that assess gait speed, balance, and strength. It accurately predicts adverse outcomes such as mobility disability [15, 16], ADL disability, hospitalization, nursing home admission, frailty, and mortality [17–20]. In addition, the validity of the SPPB has been confirmed in studies conducted in Brazil and Colombia [21–23]. Among the SPPB tests, gait speed is the most widely used and studied for older adults, and has a clear association with mortality and other adverse outcomes [24, 25]. Moreover, SRH and gait speed seem to have an additive prognostic significance, as shown by Cesari et al. who found that Mexican-American older adults with both low SRH and low gait speed had the strongest association with mortality [26].

Since both SRH and SPPB have a graded prediction of mortality (the higher the score, the lower the probability of dying), we would expect physical performance to be associated with SRH. In fact, slow gait speed and difficulty in walking were associated with poor SRH in cross-sectional analyses of the Women's Health and Aging Study, conducted on a sample of mobility-disabled women in the United States [27]. However, to our knowledge no study has shown the association between SRH and the SPPB in diverse populations of older men and women. Such a study could demonstrate the validity of SRH as a measure of global health status that aligns with objective indicators of physical function among older adults in different cultural settings. In addition, it could provide evidence of the significance of SRH in clinical settings.

Therefore, the aim of this study is to explore the validity of SRH across different populations of older men and women, by comparing it to the SPPB.

Materials and Methods

Results

Of a total of 1,995 older adults assessed, 52.1% (n = 1,040) were women (without significant differences between sites). Mean age was 69.1 (±2.6 standard deviations [SD]) overall and was significantly higher in non-Canadian sites (p < .001) (see Table 1). Regarding years in school, 65.7% of participants had 7 years or more; in non-Canadian sites, women were less educated than men. The majority of participants were married (67.5%), and the mean number of chronic conditions was not significantly different across sites (p = .73) (overall mean 1.8 [±1.32 SD]). LCT scores were not different between sites, 27.9 (±3.8 SD); and the overall mean for CES D was 11.02 points (±11.02 SD). The overall mean height of men was 1.67m (±.08 SD), with no differences between sites; the mean height of women was 1.54m (±.07 SD), with significant difference between sites (p < .001), the lowest height being observed in women of Natal and Manizales (1.5m).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. General Characteristics of the Sample According to Site and Stratified by Sex.

https://doi.org/10.1371/journal.pone.0153855.t001

There was a significant difference in SPPB score across sites: the overall mean was 9.5 (±2.4 SD), with the lowest score in Manizales (9.04, ±2.9 SD), and the highest in Kingston (10.2 ±1.8 SD) (p < .0001). With the exception of Kingston and Manizales, there were also significant differences in SPPB total scores between men and women at each site.

Poor physical performance was defined as a SPPB score below eight. Poor physical performance was observed in 14.8% (n = 296) of the full sample, with the highest proportion of older adults below this threshold in Tirana (30.4%, n = 90) and the lowest in Saint-Hyacinthe (9.1%, n = 27) (p < .0001).

SRH was distributed as follows: 383 (19.2%) participants reported their health as ‘very good’, 744 (37.2%) as ‘good’, 705 (35.3%) as ‘fair’, and 163 (8.1%) as ‘poor/very poor’, with significant differences between all the sites. In Manizales, Natal, and Tirana, poor/very poor perceptions of health were reported by 20% of the population. The site with the highest proportion of very good/good health was Kingston, and the site with the highest proportion of poor/very poor health was Manizales. In addition, women reported poor and very poor health more frequently than men, but only in the non-Canadian sites (p = .33). At the Canadian sites, there was no significant difference in self-rated health between men and women (p = .24).

A significant inverse gradient in physical performance was observed with respect to categories of self-rated health in all analyses by sex and city (p < .0001) (Table 2). Among men, a steep gradient was observed at each site, going from a mean SPPB score of 11.1 (±.9 SD) for SRH ‘very good’ in Tirana to 10.3 (±1.9) in Natal. The mean SPPB scores of women with ‘very good’ SRH ranged from 11.2 (±.9 SD) in Tirana to 9.2 (±3.1 SD) in Manizales. The lowest mean SPPB score for men in the SRH category of ‘poor/very poor’ was 5.5 (±.7 SD) in Saint-Hyacinthe, while the highest score in the same category was 8 (±2.7 SD) in Natal. For women, the lowest mean SPPB score for the SRH category of ‘poor/very poor’ was 5.7 (±3.1 SD) in Kingston, and the highest was 7.6 (±2.3 SD) in Natal.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Mean (SD) of SPPB^† According to Self-Rated Health (Four Categories) at Each Site and Stratified by Sex^†.

https://doi.org/10.1371/journal.pone.0153855.t002

The results of the multiple logistic regression models are shown in Table 3. Among men, reporting poor/very poor health was significantly associated with poor SPPB at all sites except for Manizales, after controlling for age (Model 1). After extensive adjustment for socio-demographic variables, chronic conditions, body mass index, cognitive function, and depressive symptoms, these associations were attenuated and lost significance at all sites except at Saint-Hyacinthe (Model 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multiple Logistic Regression Models for SPPB as Dependent Variable (<8 points) and Self-rated Health in Three Categories (Good/Very Good, Fair [reference], and Poor/Very Poor) as Independent Variable Stratified by Sex.

https://doi.org/10.1371/journal.pone.0153855.t003

Among women, large significant associations were observed in Kingston, Tirana, and Natal between poor/very poor SRH and poor physical performance; these associations were also strong in Saint-Hyacinthe and Manizales, but did not reach statistical significance. After adjustment for all covariates, associations became non-significant for women living in Manizales and the Canadian cities, but they remained strong and significant for women in Tirana and Natal.

Discussion

These findings further confirm the validity of SRH as a measure of overall health status in older adults. According to our results, the prevalence of poor physical function increased as perceptions of health worsened, for both men and women and across all five research sites. Associations remained when controlling for age, but became attenuated or disappeared after extensive adjustments for education and objective indicators of physical and mental health. Even across different cultural backgrounds, when compared with a robust reference standard such as the SPPB, SRH proved to be an effective and valid global measurement tool for poor health and physical function.

The fact that associations tended to disappear when adjusting for socioeconomic and objective physical and mental health indicators further strengthens the validity of the concept, since these adjustments show that SRH is capturing important aspects of health and socioeconomic status that explain the pathway from SRH to physical performance. Although adverse conditions can exacerbate the perception of a poor SRH, the association persists when these conditions are not present. However, women living in Natal and Tirana who stated their health was poor or very poor continued to be more likely to have poor physical function, indicating that additional aspects influencing SRH have not been taken into account among women of these two cities. Importantly, it is these same women who had the worst perceptions of health and the poorest function among all sex and city comparison groups. However, adding more variables to the models might have resulted in loss of power.

Thus, when a clinician asks a patient to rate his/her health and the patient indicates it is ‘poor’ or ‘very poor’, the clinician can be fairly certain that the likelihood of poor physical function is high; that potential adverse outcomes such as mobility disability, hospitalization, and death are likely outcomes for that particular older adult; and that the meaning of this self-reported status is fairly consistent across countries. This is very clear, as it was true for all groups of men and women at each of the five research sites. Admittedly, some of the differences when rating own health could be solely attributed to country of origin (sociocultural background) and education, or to rating different constructs of health (e.g., a specific disease versus overall physical health) [35]. Nevertheless, low SRH should prompt the clinician to search for chronic conditions that could be affecting the patient’s perception.

Another contribution of this study is its focus on the age span of 65 through 74. This is a very dynamic period in disability transitions and a key age range for prevention of mobility loss and disability. Many people in this age group are also going through retirement from the work force. It is a very opportune time for health behaviour changes that can delay further physical decline. However, this strength of the study could also represent a limitation, since some literature suggests that the predictive value of SRH for mortality diminishes as people become older, and therefore the same might be true of the predictive value of SRH for physical function [36].

In clinical settings, the ease of obtaining SRH may diminish the need to run unnecessary tests that could be harmful to older adults, and in particular to those in the frail spectrum of health. We provide evidence that a single question on health status perception could have enormous implications for the care of older adults. Even where SRH cannot replace other diagnostic tools, it could help in decision making, especially in lower and middle-income countries, where resources are scarce [37].

The weakest associations of poor SRH with poor function, found among men and women in Manizales, could be due to cultural lack of acceptance or to lower recognition of being in poor health. We have already reported that older adults in Manizales tended to report better health than older adults of Natal within each level of mobility, as assessed by the videos of the Mobility Assessment Tool-Short Form [33]. However, even with a lack of recognition of poor health in Manizales, the reports of poor health had validity in the unadjusted analyses (Table 2). Social inequality has been shown to impact physical performance, particularly in early old age, an aspect that also needs to be taken into account when interpreting our results [27].

This study has limitations: the sample size for each strata by sex and city is one of the most important–as shown by the large confidence intervals of the associations, particularly in Model 2. We found no substantial differences when we carried out sensitivity analyses using SPPB cut-off points of less than 9 and less than 10, as appropriate–as was done in the recently published Lifestyle Interventions and Independence for Elders LIFE [38] clinical trial. However, this also points to the need for cut-off points tailored to each population, which has been demonstrated in the case of other physical performance tests [39]. Notwithstanding, with a larger sample size we would have been able to estimate the associations of poor physical function across the five levels of SRH precisely, to construct ROC curves at each site, and to validate the cut-off points for poor physical function for each sex and city group, resulting in a fuller investigation of the association under study. In addition, SRH and SPPB should be compared with other tools that assess physical function, are widely accepted, and also have standardized methodology in their application and interpretation (e.g., the World Health Organization’s International Classification of Function). Finally, while one might think that having to collapse the categories of SRH due to limited sample size might have reduced the accuracy of our construct, in actuality patients in clinical practice used few categories when answering this type of question. Therefore, having three categories in our analyses might in fact better represent the real clinical situation and the actual answers of older adults. Further research will help confirm and expand on the results obtained in this manuscript.

Acknowledgments

We thank the IMIAS team for their contributions.