Self-reported health (SRH) is one of the most frequently used measures of health perception in survey research studies [1]. It is based on a single question (e.g., “Overall, how would you rate your health?”) with a four- or five-point answering scale ranging from “very poor” to “very good.” SRH has consistently been associated with both lifestyle factors, e.g., physical activity, overweight/obesity, smoking, diet, alcohol consumption [1–3], and with the occurrence of diseases, e.g., cardiovascular disease (CVD), lung disease, arthritis, metabolic disease [4–7]. Furthermore, numerous studies have found SRH to be a strong predictor of all-cause mortality [8–10]; and in search for a causal pathway, studies have explored underlying factors that may influence both objectively measured health and a person’s subjective health rating including gender, age, socioeconomic status, and ethnicity [11–13]. Kaplan et al. were the first to examine how the correlation between SRH and mortality was attenuated when controlling for subclinical conditions and disease. They concluded that SRH levels mainly reflect underlying disease burden, and accordingly predict mortality [14]. Later studies have confirmed these findings and additionally reported that SRH influences health-related behaviour that affects outcome, reflects coping resources, and encompass a health awareness not captured by the specific health indicators but relevant to the overall health rating [10, 15–17].

Several studies have also found a strong correlation between SRH and objective health measures [18–22]. Kananen et al. investigated the association of SRH with 150 biomarkers from blood and urine and found 57 of these biomarkers to be associated with SRH, independently of disease and physical functioning in half of the cases [18].

The impact of psychosocial health on SRH including adverse exposures throughout the life course on long-term adult health outcomes have been reported and conceptualized by Erikson and in social determinants of health models [23]. In order to comprehend the potential of SRH as a measure of health in research and clinical practice, it is important to further explore how SRH reflects the condition of the human body [24–26].

Links have been revealed between SRH and allostatic load (AL), defined by cumulative strain of the body due to chronic fluctuation of neural and neuroendocrine responses [26]. AL derives from the concept of allostasis, which is a dynamic regulatory process where homeostasis is maintained through adaptation in the presence of physical and behavioral stressors. The concept was introduced by McEwan and Stellar in 1993 [26]. AL cannot be directly measured but it can be estimated using an AL index, a composite of biomarkers deriving most commonly from the endocrine, metabolic, cardiovascular, and inflammatory systems [27]. AL has been associated with various health outcomes also related to SRH, including CVD, metabolic disease, and all-cause mortality [27, 28]. In our previous study we found a strong correlation between high AL and all-cause mortality; this association was stronger for the AL index than the association between any individual biomarker and all-cause mortality indicating that incorporating risks at the entire range of biomarkers may yield the best prediction of health in general [29].

The association between AL and SRH has been explored in a few studies with methodological limitations including small sample sizes and use of data from only elderly people [20–22]. Therefore, to explore whether SRH has a biologic foundation, we investigated the relation between SRH and AL, and all-cause mortality in the following years in the general adult population from the Lolland-Falster region of Denmark.

A total of 14,104 participants were included (54.1% women and 45.9% men) (Table 1). Poor/very poor SRH was reported by 4.2% (4.5% in women and 3.8% in men) (Supplementary Table S3). One-fourth of the participants had a higher education, and one-fifth were current smokers. Presence of diabetes at the time of LOFUS participation was reported by 5% (3.7% women and 6.8% men), cancer by 4% (3.4% women and 4.2% men), and CVD by 28% (25.6% women and 31.0% men). For BMI, 63% of the participants were overweight or obese (56.6% women and 70.4% men). High AL was observed in 28.1% (29.0% women and 27.1% men), mid AL in 39.8% (39.5% women and 40.2% men) and low AL in 32.1% (31.5% women and 32.7% men). During a median follow-up time of 4.6 years, 456 participants died (189 women and 267 men); 126 were excluded due to missing data (Supplementary Table S4).

The association between SRH and all-cause mortality was investigated in a multivariate Cox proportional hazard regression, Table 2. Compared with the mortality of persons who had reported very good health, the HRs for death was for women statistically significantly elevated for all other levels of SRH; good SRH, HR 2.23 (95% CI 1.12–4.42); fair SRH, HR 2.42 (95% CI 1.20–4.90; and poor/very poor SRH, HR 6.31 (95% CI 2.84–13.9). For men a statistically significant increase in mortality was seen for fair SRH, HR 2.20 (95% CI 1.32–3.66) and for poor/very poor SRH, HR 3.92 (95% CI 2.12–7.25).

Compared with the mortality of persons with low AL, women with high AL had increased mortality, HR 2.40 (95% CI 1.61–3.57) and so had men, HR 1.78 (95% CI 1.27–2.40) (Supplementary Figure S1). Mutually adjusted models for SRH, AL and mortality are presented in Supplementary Table S5.

Multivariate Cox proportional hazard regression of all-cause mortality by SRH and individual biomarkers, mutually adjusted are presented in Supplementary Table S6. The impact of the biomarkers were weak for most biomarkers when adjusting for SRH; except for LDL and WHR in women HR 1.70 (95% CI 1.25–2.32) and HR 1.59 (1.15–2.20), respectively, and for albumin in men HR 1.83 (95% CI 1.44–2.33).

Multinomial regression models for SRH as predictor of AL for women and men are presented in Table 3. In women the risk of having medium as compared with low AL increased with decreasing level of reported SRH. In the multivariate adjusted model, RRR of medium vs. low AL was 1.23 (95% CI 1.04–1.44) in women reporting good as compared with very good SRH; RRR 1.52 (95% CI 1.26–1.84) in fair compared with very good SRH; and RRR 1.58 (95% CI 1.11–2.23) in poor/very poor compared with very good SRH. The pattern was less clear for men, and only the RRR for medium vs. low AL in men reporting poor/very poor as compared very good SRH reached statistical significance, RRR 1.84 (95% CI 1.20–2.81). For high vs. low AL, the RRR increased more sharply across levels of SRH. In the multivariate adjusted model for women, the RRR for good vs. very good SRH was 1.34 (95% CI 1.08–1.67); for fair vs. very good SRH it was RRR 2.03 (95% CI 1.60–2.58); and for poor/very poor vs. very good SRH it was RRR 2.43 (95% CI 1.66–3.56). In this model for men, only the two latter comparisons reached statistical significance, for fair vs. very good SRH, RRR 1.58 (95% CI 1.23–2.03); and for poor/very poor vs. very good, RRR 2.96 (95% CI 1.87–4.70). For each level of SRH, the probability of having high AL increased by increasing age, Figure 2.

A notable difference was found between the age-adjusted and fully-adjusted model for RRR of high vs. low AL for poor/very poor SRH in women and men. We therefore further explored the intermediate variables in a multinomial model and found that when adjusting for BMI the association between AL and SRH attenuated more than when adjusting for the other variables (Supplementary Table S7). Finally, we made sensitivity analysis by investigating the association between AL and mortality and SRH and AL by using clinical cut-off values to define AL. The AL categories were slightly changed to low AL: 0–2, medium AL: 3, and high AL: 4–9, as no participants had AL10 (Supplementary Tables S8, S9). We found mostly a weaker association between AL and SRH, and also AL and mortality when comparing with AL based on the distribution of our population.

The distribution of biomarkers according to self-reported health are presented in (Supplementary Table S10).

In the general population study from Lolland-Falster in Denmark, SRH was strongly associated with AL. Despite SRH being an instant and subjective assessment of overall health; the relative risk of harbouring high vs. low AL at the same time was more than doubled among both women and men reporting poor/very poor health as compared with those reporting very good health, even when controlling for several known risk factors. Additionally, after follow-up of the study population for a median of 4.6 years, SRH was found to be strongly associated with all-cause mortality even after control for known risk factors. Women reporting poor/very poor SRH had 6-fold the death rate of those reporting very good SRH and for men, an almost 4-fold ratio was seen.

Among the 14,104 LOFUS-participants included in the study, 4.2% reported poor/very poor SRH. This was higher than the 1.4% reported in the Copenhagen Aging and Midlife Biobank study, comprising of participants from the Metropolit 1953 Danish Male Birth Cohort, the Copenhagen Perinatal Cohort, and the Danish Longitudinal Study on Work, Unemployment and Health cohort [41]. In Europe, reporting of poor SRH in the adult population has decreased during the last decades, and poor SRH has been most prevalent in lower socioeconomic classes [2, 42, 43]. The higher rate of poor SRH in LOFUS than in other Danish cohorts may be linked to the fact that Lolland-Falster is a socio-economically disadvantaged area of Denmark, with health problems reported more frequently, and with shorter life expectancy than the rest of Denmark [44].

We found that the association between SRH and AL was stronger when the AL was calculated using our own cut-off values, compared to AL based on clinical defined cut-off values. This might be a consequence of the fact that in our population clinical cut-offs tend to produce a generally lower AL, smoothing down the differences between the AL categories. The number of studies exploring the relationship between SRH and AL is sparse. Hasson et al. found in a cross-sectional study, that poor SRH, along with older age, low education, and work in healthcare as compared with work in information technology was associated with higher AL in Swedish women [20]. These findings were supported by the Norwegian HUNT study further indicating level of SRH in adolescence to be a predictor of AL in young adulthood [21]. Among elderly persons in Taiwan, Hu et al. found SRH to be related to AL [22], contributing valuable evidence of construct validity of AL in a non-Western population as most studies on AL are of Western orientation. A more recent study from the United States used cross-sectional data from the National Health and Nutrition Examination Survey (NHANES) and also found higher levels of AL to be associated with higher odds of reporting poor/fair SRH [45]. This association was found to differ by race/ethnicity.

When evaluating the intermediate variables separately, we found that the association between SRH and AL was primarily attenuated by BMI, a well-established risk factor for numerous diseases and all-cause mortality [32].

Our findings on the association between SRH and mortality is also consistent with previous studies [8–10]; in a systematic review, Idler et al. found a strong correlation between SRH and all-cause mortality [9]. Additionally, in a subsequent meta-analysis DeSalvo et al. found that people with poor SRH had a two-fold mortality risk when comparing with those reporting very good SRH [46]. Studies from Europe have reported a similar association between SRH and all-cause mortality in both adults and in older populations [47–50]. In a comprehensive analysis conducted by Parker et al., encompassing a systematic review and meta-analysis of 17 studies, it was concluded that despite significant variations in the ages of the participants, AL indices, and follow-up time, still a notable association existed between high AL and elevated risk of mortality [28]. The pooled estimates indicated that individuals with high AL had a 22% higher risk of all-cause mortality and a 31% higher risk of cardiovascular mortality compared to those with low AL. In the majority of studies, the most substantial correlation with mortality was observed when incorporating biomarkers that captured all four organ systems. In our data, we found an almost two-fold mortality in persons with high AL as compared with those with low AL when using our own AL construct. It should however be taken into consideration that the construct of our AL index was slightly different than in other studies due to the fact that both the upper and lower quartiles for some biomarkers defined high risk values [27]. This may also explain the stronger association found when using cut off-values based on the population to define AL when comparing with AL based on cut-off values.

SRH is a subjective summary of the understanding and interpretation of “health,” and it might be influenced by several factors; 1) direct information of one’s disease, severity, and prognosis; 2) lifestyle and behavioral risk factors including unhealthy diet and smoking; and 3) perceptions and sensations about the body and mind, such as pain and tiredness [19]. It is known that the level of expectations regarding “good” health decreases with increasing age, where more health problems are tolerated at a given level of stated SRH [48].

The present study contributed by studying SRH, AL and mortality in a large population with a wide age span. There are several ways in which AL can have an impact on SRH. First, biomarkers may act as surrogate measures for medical conditions that were not directly measured in this study [47]. Second, biomarkers may reflect an unhealthy lifestyle known to effect SRH [7]. Third, the state of physiologic system may be accessible to individuals through interoception, a system of feelings that represents a sense of the physiologic condition of the body [48–50]. Most studies on the interoceptive signalling of humoral processes and changes in biomarker levels concern inflammation, in which higher circulating levels of inflammatory biomarkers cause fatigue, poor appetite, low mood and general malaise. However, the empirical evidence on interoceptive signalling of humoral processes is still lacking and almost non-existent for biomarkers of other organ systems [48].

Our results contributed to the ongoing debate on the role of SRH in predicting mortality. According to our findings SRH has a biologic basis and may reflect strain on the body in terms of AL. Future research are needed to support our findings and to investigate both the physiological processes that underlie sensations and the reason that governs ones perception about health. Further studies should focus on the pathways that mediate information from the human organism to the consciousness incorporating information into rating ones health.