Lifetime Socioeconomic Inequalities in Physical and Cognitive Aging

Abstract

Objectives. We examined the relationship between childhood and adult socioeconomic position (SEP) and objectively assessed, later-life functioning.

Methods. We used the Medical Research Council’s National Survey of Health and Development data to examine performance at 60 to 64 years (obtained in 2006–2011) for a representative UK sample. We compared 9 physical and cognitive performance measures (forced expiratory volume, forced vital capacity, handgrip strength, chair rise time, standing balance time, timed get up and go speed, verbal memory score, processing speed, and simple reaction time) over the SEP distribution.

Results. Each performance measure was socially graded. Those at the top of the childhood SEP distribution had between 7% and 20% better performance than those at the bottom. Inequalities generally persisted after adjustment for adult SEP. When we combined the 9 performance measures, the relative difference was 66% (95% confidence interval = 53%, 78%).

Conclusions. Public health practice should monitor and target inequalities in functional performance, as well as risk of disease and death. Effective strategies will need to affect the social determinants of health in early life to influence inequalities into old age.

In 2012, the World Health Organization (WHO) made health and aging the theme for World Health Day,¹ a timely reminder that the striking increase in human longevity should give cause for celebration,² especially when “good health adds life to years.”¹ However, population aging is often framed in negative terms because of its anticipated societal effects, perhaps most prominently in terms of health and social care spending. The societal impacts of aging depend on health and functional capability. These vary considerably between individuals, but the extent to which inequalities in health and function in later life are driven by early life exposures is less well understood. There is growing interest in the impact of adverse earlier life experiences—already shown to be associated with inequalities in subsequent chronic disease risk³—on levels of functioning in later life.^4,5

Maintenance of physical, cognitive, psychological, and social function for the maximal period of time is a key feature of healthy aging.⁶ Higher levels of physical and cognitive functioning are associated with continued independent living,⁷ a higher quality of life,⁸ and lower rates of subsequent morbidity and mortality.^9–12 Lower adult socioeconomic position (SEP), typically captured by low levels of income, wealth, and educational attainment or low-skilled occupation, is associated with poorer functioning,¹³ lower age of onset of disabilities,¹⁴ and steeper trajectories of functional decline.¹⁵ The average difference in disability-free life expectancy is 17 years between people living in poor and rich areas of England.¹⁶ Prospective evidence suggests that childhood SEP, independently of adult SEP, is associated with subsequent disability and indicators of both physical and cognitive function at least up to middle age.^17,18 Childhood SEP, often captured retrospectively, is associated with function even into older age.^19–21

These findings suggest a wide temporal window within which socioeconomic factors might influence physical and cognitive function in later life, operating through a range of potential mediating biological and social pathways.²² They have important implications for public health interventions that target the factors across life that underlie social gradients in health. However, data to explore the relationship of prospectively assessed SEP from across life with a range of healthy aging indicators in older age through multiple mediating pathways have rarely been available. We examined the independent influence of SEP in childhood and adulthood on a range of markers of physical and cognitive performance, and an overall composite measure of these, in Britain’s oldest birth cohort study, now reaching retirement age. Although social function and psychological function are important components of healthy aging, we elected to focus on objectively assessed performance. We hypothesized that inequalities in physical and cognitive performance would be observed independently for childhood and adult SEP. A secondary aim was to assess the contribution of selected potential mediating factors to these inequalities.

METHODS

The Medical Research Council’s National Survey of Health and Development (NSHD) is a socially stratified sample of 5362 singleton births in Britain from 1 week of March 1946, with regular follow-up across life. Between 2006 and 2011, the NSHD scientific and data collection team invited all 2856 eligible study members (aged 60–64 years) who were known to be alive and with a known address in England, Scotland, or Wales for assessment at 1 of 6 clinical research facilities or by a research nurse at home. They did not send invitations to those who had died (n = 778), were living abroad (n = 570), had previously withdrawn from the study (n = 594), or had been lost to follow-up (n = 564). Of those invited, 2229 (78%) were assessed and 1690 (59.2%) attended a clinic; the remaining 539 were seen at home.²³ These form the analytical sample for the current study.

Aging Outcomes

We selected 9 objective assessments of physical and cognitive performance at age 60 to 64 years. Trained nurses recorded all of them using standardized protocols. They measured lung function twice by forced expiratory volume in 1 second and forced vital capacity, using a spirometer (Micro Medical Micro Plus, Rochester, UK). For this study, we used the maximum of 2 blows into the spirometer that agreed within 0.3 liters. Nurses measured handgrip strength²⁴ 3 times in each hand using an electronic handgrip dynamometer (custom made by Medical Physics and Clinical Engineering Department of Queen’s Medical Centre, Nottingham, UK), and the maximum value was used. Chair rise time²⁴ was the time taken to rise from a sitting to standing position and back to sitting 10 times, as fast as possible. Standing balance time²⁴ was the longest time that the study member was able to stand on 1 leg with eyes closed, up to a maximum of 30 seconds. We used natural log-transformed values in analyses. We assessed timed get up and go²⁵ as the time taken to get up from a chair, walk 3 meters at a normal pace, turn around, return to the chair, and sit back down, with or without the use of a walking aid (24 participants used walking aids). We then converted the recorded time to speed (meters per second). We assessed verbal memory with a 15-item word learning task devised by the NSHD, measured as the number of correct words recalled from a list of 15 over 3 learning trials (maximum score = 45). We measured processing speed as a score representing the total number of target letters searched within 1 minute. We defined simple reaction time as the speed with which participants pressed a key in response to a signal on 20 trials. In order that higher values represented better performance for all measures, we used reciprocal values of chair rise and reaction times in analyses.

Among clinic attendees, 75% provided performance data on all 9 outcomes and 98% on 7 or more outcomes. The corresponding numbers for those who had a home visit were 41% and 84%, the main reason for a drop in test completion being that it was not always possible to conduct the timed get up and go at home because of space constraints. We assumed that those study participants unable to perform the tests for health reasons (n = 22 for lung function, 49 for grip strength, 136 for chair rises, 89 for standing balance, 34 for timed get up and go, 10 for verbal memory, 11 for search speed, and 11 for reaction time) were likely to have poor performance and that their exclusion from analyses could introduce bias. We used the mean of the bottom quintile as the imputed value on the assumption that these participants were likely to have values within the lowest quintile of observed values. We ran sensitivity analyses excluding these individuals, and the results were essentially unchanged.

We derived a composite performance index by summing the z scores of all performance measures and dividing the sum by the number of tests. Those with 9 complete or imputed performance measures were included in this composite index (n = 1589).

Lifetime Socioeconomic Position

For ease of comparison, we used a single and well-recognized measure of SEP at both ages, namely, the occupational class of the head of the household. We based childhood occupational class on father’s occupation coded according to the UK Registrar General’s Social Class at age 4 years (or age 11 or 15 years if this was missing; n = 54). We based adult occupational class on head of household’s occupation at age 53 years (or ages 43 or 36 years if this was missing; n = 130). We did not use occupational class concurrent with the outcomes because by age 60 to 64 years approximately half of the men and three quarters of the women had retired from their main job.

Selected Potential Mediating Variables

We selected adult height (in centimeters) and childhood cognitive development as indicators that reflect physical and cognitive development, and as potential mediators linking childhood SEP and the aging outcomes. We measured cognitive development at age 8 years using the summed score from 4 tests: reading comprehension, word reading, vocabulary, and nonverbal reasoning. We selected current body mass index (BMI; defined as weight in kilograms divided by height in meters squared) and lifetime smoking status as indicators reflecting common health behaviors. Regarding smoking status, we distinguished “never smoker” (nonsmoker at ages 26, 33, 43, 53, and 60–64 years), “predominantly nonsmoker” (nonsmoker on at least 3 occasions), “long-term, ex-smoker” (smoker at 4 or more occasions but currently nonsmoker), and “long-term, current smoker” (currently smoking, smoked at 4 or more occasions). We selected adult occupation and education as 2 indicators that reflect social mediators. We grouped highest educational qualifications achieved by age 26 years as degree level (university degree and equivalent), advanced secondary qualifications (A levels and equivalent), ordinary secondary qualifications (O level and equivalent), and no formal qualifications.

Statistical Analysis

For each outcome separately, and with the maximum available sample, we used linear regression to model the outcome as a function of ridit scores (calculated as the proportion of the population with SEP higher than the midpoint for each SEP category) and gender. The ridit score coefficient is the slope index of inequality (SII), interpreted as the absolute difference in outcome between the top and bottom of the SEP distribution. We then divided these by mean outcome to obtain the relative index of inequality (i.e., the relative difference between the top and bottom of the SEP distribution). This technique takes account of the population distribution across SEP categories, enabling comparison across outcomes and over time, although it is only suitable for summarizing a linear relationship between SEP and outcome.²⁶ We included ridit-by-ridit interaction terms to assess nonlinearity.

Previous studies have shown large gender differences in several of these aging outcomes and some indication of differential socioeconomic gradients for men and women.²⁷ Hence, we present means by SEP separately for men and women in the descriptive tables; gender-by-ridit interaction terms were assessed in each model and are reported where statistically significant at the 5% level.

We assessed the SEP gradient in the outcomes separately for childhood and adult SEP, and then mutually adjusted. We tested for interaction between childhood and adulthood SEP for each aging outcome. A series of further models included the indicators for each potential mediator in turn to assess the percentage change in SII for childhood SEP in relation to each aging outcome. We weighted analyses to take account of the initial social stratification of the sample.

RESULTS

Those from more socioeconomically advantaged families (higher SEP in childhood) had a higher mean cognitive score in childhood and higher educational attainment at age 26 years, were less likely to be in manual SEP groups in adulthood, had lower current mean BMI, were taller as adults, and were less likely to be long-term cigarette smokers compared with those from lower SEP (P for linear trend = .015 for smoking, P < .001 for all others; Table A, available as a supplement to the online version of this article at http://www.ajph.org).

A clear gradient by childhood SEP was seen for all measures of physical and cognitive performance such that performance declined with decreasing SEP (Figure 1a and Table 1; P < .05 for all outcomes). The relative differences in outcome between the top and bottom of the SEP distribution were between 7% and 13%, with the exception of verbal memory (20%). A gender difference in the socioeconomic gradient was statistically significant only for reaction time, where the gender-by-SEP interaction was small in magnitude and indicated a slightly steeper gradient for men than for women. There was evidence of a leveling off of the gradient in standing balance and verbal memory in the manual SEP groups.

FIGURE 1—

Relative index of inequality (RII) for aging outcomes by (a) childhood SEP and (b) adulthood SEP: Medical Research Council National Survey of Health and Development, United Kingdom, 2006–2010.

Note. FEV1 = forced expiratory volume in 1 second; FVC = forced vital capacity; SEP = socioeconomic position; TUG = timed get up and go. Whiskers indicate 95% confidence intervals.

TABLE 1—

Aging Outcomes at Age 60 to 64 Years, by Childhood Socioeconomic Position (SEP), Adjusted for Gender Throughout: Medical Research Council National Survey of Health and Development, United Kingdom, 2006–2010

		Childhood SEP, Mean
Physical and Cognitive Performance at Age 60–64 Years	No.	All	I	II	IIINM	IIIM	IV	V	Slope Index of Inequality (95% CI)	Relative Index of Inequality (95% CI)
Lung function
FEV1, L									−0.32 (−0.59, −0.25)	1.13 (1.08, 1.17)
Men	928	2.80	3.00	3.00	2.88	2.69	2.67	2.66
Women	1012	1.98	2.10	2.06	2.09	1.93	1.90	1.73
FVC, L									−0.51 (−0.73, −0.30)	1.12 (1.07, 1.17)
Men	839	3.54	3.79	3.86	3.62	3.35	3.42	3.39
Women	975	2.48	2.60	2.57	2.62	2.43	2.38	2.13
Muscle strength
Grip strength, kg									−4.08 (−5.97, −2.20)	1.07 (1.04, 1.10)
Men	973	45.5	46.3	47.2	45.5	45.1	44.4	44.6
Women	1038	26.6	27.9	27.2	27.0	26.3	25.8	25.1
Physical performance
Chair rise time reciprocal, s−1 × 100									−0.42 (−0.66, −0.19)	1.09 (1.04, 1.14)
Men	1002	4.30	4.39	4.49	4.54	4.21	3.99	4.38
Women	1085	4.16	4.45	4.29	4.23	4.04	4.10	3.98
Standing balance, loge s ab									−0.16 (−0.29, −0.03)	1.11 (1.03, 1.19)
Men	1006	1.33	1.44	1.40	1.38	1.26	1.28	1.30
Women	1083	1.21	1.44	1.24	1.24	1.14	1.17	1.14
TUG, m/s									−0.06 (−0.10, −0.03)	1.08 (1.04, 1.12)
Men	935	0.71	0.75	0.72	0.73	0.72	0.68	0.65
Women	1017	0.68	0.72	0.70	0.69	0.67	0.38	0.64
Cognitive performance
Verbal memory,b no. correct words recalled									−4.72 (−5.84, −3.61)	1.20 (1.15, 1.24)
Men	981	23.0	26.6	24.0	24.6	21.7	21.1	22.1
Women	1069	25.2	28.8	27.0	26.5	23.9	23.8	21.7
Processing speed, score for no. letters processed									−28.87 (−57.74, −15.30)	1.11 (1.06, 1.15)
Men	993	260	273	272	266	255	247	255
Women	1081	272	296	278	280	261	268	254
Simple reaction time reciprocal, s−1 × 100c
Men	1006	36.5	38.2	38.3	38.1	25.7	34.2	34.8	−0.05 (−0.07, −0.02)	1.12 (1.07, 1.17)
Women	1081	36.5	28.4	36.8	37.7	35.9	36.2	33.3	−0.04 (−0.05, −0.02)	1.09 (1.05, 1.14)

Note. CI = confidence interval; FEV1 = forced expiratory volume in 1 second; FVC = forced vital capacity; TUG = timed get up and go. Childhood SEP is based on the father’s occupation under the UK Registrar General’s classification system, as follows: I, professional occupations; II, managerial and lower professional occupations; IIINM, nonmanual skilled occupations; IIIM, manual skilled occupations; IV, semiskilled occupations; V, unskilled occupations.

^aGeometric means presented but regression performed with log-transformed data.

^bUnderlying relationship nonlinear.

^cGender-by-childhood SEP interaction.

A clear gradient by adult SEP was also seen for all measures of physical and cognitive performance (Figure 1b and Table 2; P < .05 for all outcomes). Relative differences were between 6% and 13%, with the exception of forced expiratory volume (16%), standing balance (18%), and verbal memory (27%). There was evidence of a slightly steeper SEP gradient in processing speed among men compared with women.

TABLE 2—

Aging Outcomes at Age 60 to 64, by Adult Socioeconomic Position (SEP), Adjusted for Gender Throughout: Medical Research Council National Survey of Health and Development, United Kingdom, 2006–2010

		Adult SEP, Mean
Physical and Cognitive Performance at Age 60–64 Years	No.	All	I	II	IIINM	IIIM	IV	V	Slope Index of Inequality (95% CI)	Relative Index of Inequality (95% CI)
Lung function
FEV1, L									−0.55 (−0.72, −0.38)	1.16 (1.11, 1.21)
Men	958	2.81	3.04	2.90	2.78	2.63	2.68	2.05
Women	1065	1.98	2.15	2.07	1.91	1.92	1.85	1.62
FVC, L									−0.55 (−0.77, −0.34)	1.13 (1.08, 1.18)
Men	870	3.55	3.80	3.66	3.55	3.32	3.52	2.65
Women	1026	2.48	2.62	2.57	2.41	2.39	2.38	2.22
Muscle strength
Grip strength, kg									−3·57 (−5.51, −1.63)	1.06 (1.03, 1.10)
Men	1004	45.59	46.09	46.56	43.54	45.54	44.15	38.67
Women	1091	26.57	27.87	27.31	25.89	25.68	25.73	24.61
Physical performance
Chair rise time reciprocal, s−1 × 100									−0.50 (−0.73, −0.27)	1.11 (1.06, 1.16)
Men	1035	4.31	4.67	4.38	4.23	4.13	3.98	4.22
Women	1139	4.16	4.25	4.27	4.10	4.17	3.86	3.53
Standing balance, loge s a									−0.27 (−0.40, −0.15)	1.18 (1.10, 1.25)
Men	1038	1.33	1.48	1.38	1.30	1.20	1.34	1.13
Women	1137	1.21	1.32	1.26	1.24	1.12	1.11	0.98
TUG, m/s									−0.05 (−0.09, −0.02)	1.07 (1.03, 1.11)
Men	966	0.71	0.75	0.72	0.73	0.69	0.66	0.68
Women	1069	0.68	0.70	0.69	0.69	0.67	0.66	0.60
Cognitive performance
Verbal memory, no. correct words recalled									−6.67 (−7.73, −5.61)	1.27 (1.23, 1.31)
Men	1016	23.0	26.2	24.3	22.9	19.8	20.3	20.4
Women	1122	25.3	28.0	26.8	24.9	23.0	22.5	22.4
Processing speed, score for no. letters processedb
Men	1027	261	274	269	265	244	241	235	−33.74 (−51.50, −15.98)	1.13 (1.06, 1.18)
Women	1136	271	278	276	274	263	268	250	−23.28 (−41.16, −5.39)	1.08 (1.02, 1.15)
Simple reaction time reciprocal, s−1  × 100									−0.05 (−0.06, −0.03)	1.12 (1.09, 1.15)
Men	1040	36.5	37.6	37.9	35.7	34.7	34.1	32.7
Women	1136	36.5	28.4	37.1	36.6	35.4	35.5	33.8

Note. CI = confidence interval; FEV1 = forced expiratory volume in 1 second; FVC = forced vital capacity; TUG = timed get up and go. Adult SEP is based on the UK Registrar General’s classification of occupations, as follows: I, professional occupations; II, managerial and lower professional occupations; IIINM, nonmanual skilled occupations; IIIM, manual skilled occupations; IV, semiskilled occupations; V, unskilled occupations.

^aGeometric means presented but regression performed with log-transformed data.

^bGender-by-childhood SEP interaction.

The relative index of inequality for the composite score indicated a 66% difference (95% confidence interval [CI] = 53%, 78%) between the top and bottom of the childhood SEP distribution, and a relative difference of 74% (95% CI = 61%, 87%) based on adult SEP.

Table 3 is limited to study members who had complete covariate data. It shows the gender-adjusted slope index of inequality and the percentage reduction in this with adjustment for each possible mediating variable separately. For example, FEV1 was 0.46 liters lower at the top compared with the bottom of the childhood SEP distribution (SII = −0.46), and a 27% reduction in this SII was observed when adult SEP was included in the model. The association between childhood SEP and each aging outcome was attenuated when adjusted for adult SEP, although sizable gradients remained for all outcomes, with the exception of standing balance. When we tested for an interaction between child and adult SEP with each outcome, the associations were additive. We found no evidence of multiplicative associations.

TABLE 3—

Change in Slope Index of Inequality (SII) in Relation to Childhood Socioeconomic Position (SEP), With Inclusion of Selected Social, Behavioral, and Developmental Covariates: Medical Research Council National Survey of Health and Development, United Kingdom, 2006–2010

			Percentage Change From Base Model Gender-Adjusted SII
Clinical Measure at Age 60–64 Years (Variable Added to Base Model)	No.	Base Model Gender-Adjusted, SII (95% CI)	Adult SEP	Educational Attainment	Adult BMI	Lifetime Smoking	Adult Height	Cognitive Ability Age 8 Years
Lung function
FEV1	1660	−0.46 (−0.65, −0.27)	−27	−34	−4	−15	−29	−33
FVCa	1550	−0.55 (−0.78, −0.31)	−22	−27	−7	−12	−30	−31
Muscle strength: grip strength	1710	−4.52 (−6.54, −2.50)	−17	−15	−1	−5	−35	−12
Physical performance
Chair rise time reciprocal	1781	−0.44 (−0.70, −0.18)	−17	−32	−22	−14	12	−9
Standing balance, loge	1784	−0.16 (−0.29, −0.02)	−35	−68	−40	−6	14	−53
Time to get up and go	1658	−0.07 (−0.10, −0.03)	−14	−15	−19	−8	−4	−9
Cognitive performance
Verbal memory	1756	−4.60 (−5.83, −3.37)	−32	−67	−7	−8	−3	−64
Processing speed	1773	−30.10 (−44.47, −15.73)	−18	−41	−5	−9	−9	−30
Simple reaction time reciprocala	1784
Men	837	−0.04 (−0.06, −0.02)	−24	−32	−1	−1	+3	−35
Women	947	−0.03 (−0.05, −0.01)	−25	−50	−9	−0	−10	−33

Note. BMI = body mass index; CI = confidence interval; FEV1 = forced expiratory volume in 1 second; FVC = forced vital capacity.

^aGender-by-childhood SEP interaction.

Estimates suggested that childhood cognitive ability and educational attainment explained a considerable amount of the childhood SEP gradient across most physical and cognitive outcomes. Height was an important mediator of the gradient in lung function and muscle strength, and BMI was an important mediator of the gradient for measures of physical performance. Inclusion of lifetime smoking history did not substantially attenuate the childhood SEP gradient.

DISCUSSION

This study demonstrates that in a representative British cohort born in 1946, the legacy of childhood SEP persists at least up to conventional retirement age, affecting both physical and cognitive performance. These relative inequalities were strikingly consistent and persisted after adjustment for adult SEP for all but 1 functional outcome. Whereas the mortality, disease, and health care burden of socioeconomic inequality has been documented,²⁸ the early origins of inequalities in physical and cognitive aging, which considerably affect independent living and quality of life, have not so far been comprehensively described. The consistency of the gradient in performance across physical and cognitive outcomes may reflect shared risk factors.

Our findings are supported by evidence from this cohort at a younger age, and in international data, showing that childhood and adult SEP are independently associated with adult lung function,¹⁹ muscle strength and physical performance,²⁰ and various aspects of cognitive function.^17,18 However, our study is the first to document socioeconomic inequalities that can be compared across a range of objectively measured physical and cognitive measures around retirement age, using prospectively obtained data from birth. Our findings, using a composite index of performance, illustrate the considerable burden of poorer functioning at each point below the most advantaged SEP backgrounds. They add to evidence suggesting that rising levels of education among older generations and their parents have contributed to recent declines in late-life disability.²⁹

Although childhood SEP maintained an independent association with aging outcomes after adjustment for adult SEP, several other factors across the life course contributed to explaining the childhood SEP gradient. This is consistent with previous work indicating that multiple mediating pathways link childhood SEP to health in later life.³⁰ For example, adult height was a shared mediator for the social gradient in physical performance measures, and is strongly associated with early environmental factors.³¹ Cognitive development was an important mediator of the social gradients in cognitive performance, lung function, and standing balance. Educational achievement was also a strong mediator of the childhood SEP gradients in both physical and cognitive performance. Childhood SEP predicts early cognitive ability and educational achievement, and both have major impacts on occupational choice and opportunities in adult life. Early cognition and educational achievement may also mediate the childhood SEP gradient through their influence on psychosocial factors and behaviors.³² Behaviors, such as physical activity, may be important, both for physical³³ and cognitive³⁴ performance. Our findings suggest that BMI mediated the gradient for physical performance, but they do not indicate a strong role for smoking in this cohort. Neurodevelopmental processes governing peak attainment and reserve and neurodegenerative aging processes governing rate of decline, which affect both physical and cognitive performance,³⁵ also need to be considered as explanations for these consistent relative inequalities. Further mediators warrant investigation (e.g., personality development).³⁶ Our intention was not to attempt to isolate the influence of different causal pathways, and we have not examined interactions between mediators or quantified the precise contribution of each variable to the social gradients in performance. Nevertheless, our findings are consistent with the operation of both accumulation of disadvantage and the biological embedding of disadvantage at sensitive developmental periods.^3,22

These findings concur with previous research in the NSHD; for example, in showing that parental SEP is associated with midlife fluid cognition and is mediated by education and cognitive development,³⁷ and that a range of childhood socioeconomic indicators are related to lung function, chair rising, and standing balance (but not grip strength) at earlier ages.^27,38 The timing of the emergence of socioeconomic differences depends on the extent of socioeconomic influences on development, maintenance, and decline in performance. This plausibly varies over physical performance measures—for example, because of occupational physical activity. Differences between these and earlier findings are unlikely to be attributable to the inclusion in the current analyses of imputed values for those unable to do the tests, because we undertook a sensitivity analysis by excluding this group and found similar results.

A previous systematic review recommended that researchers should investigate the differences between individual and aggregate measures of healthy aging within the same population, having found little evidence of an effect of socioeconomic factors when using composite outcomes.³⁹ This contrasts with our finding that relative inequalities in functional outcomes were stronger when we used a composite measure than when we used single measures. It is possible that these effects may not be as strong when other healthy aging indicators, such as social and psychological well-being, are used.

Limitations

The study benefits from the availability of prospectively obtained data across the life course and objective measures of functioning in early old age. The objective measures of physical and cognitive function allowed us to examine variation across the full spectrum of performance, which self-reported measures that primarily aim to capture loss of function are not able to distinguish. However, we acknowledge that these objective assessments are subject to measurement error and exclude some people who are unable to perform the tests, and that caution is required in using performance on objective tests to make direct inferences about individuals’ abilities to undertake the tasks of daily living.⁴⁰ Lung function had been measured in previous data collections in this population, so we used the same measure to provide continuity, although techniques have been more recently revised. The measures we used had more error than if recent guidelines had been followed, but this should not introduce any bias to our findings. There are other areas of function for which such objective measures are less well defined—for example, social and psychological function. A limitation of our study is that these aspects of healthy aging were not included.

Some attrition is unavoidable, although this study remains broadly representative of the British population born in 1946. An analysis of response and attrition in the latest data collection demonstrated that certain factors independently predicted overall response and others predicted participation in clinic rather than home visit.²⁴ Nonresponders tended to have a poorer socioeconomic profile, health behaviors, and cognition, suggesting that the present findings may underestimate the socioeconomic gradient in function.

In this study, we used occupational class to enable comparison both with past research in this cohort and with the greatest volume of other work undertaken in the United Kingdom. Limitations with this measure are documented⁴¹; however, classification structures that have a clearer theoretical basis have not demonstrated the same graded socioeconomic hierarchy.⁴² Adult SEP was indicated by the occupation of the household male because, in this cohort, it is likely to accurately reflect household SEP. However, we acknowledge it does not reflect occupational class for female study members. The relative index of inequality method enabled us to make comparisons across outcomes, but linearity of association is assumed. For 2 outcomes (standing balance and verbal memory), a small leveling off of the gradient was observed in manual groups, and so caution is required in interpreting these results.

This cohort benefited from postwar opportunities for social mobility, which gave us sufficient power to separate statistically the relationship of aging outcomes to child and adult SEP. The generalizability to younger cohorts who may not have had such opportunities for social mobility is unclear.

Conclusions

Global evidence of socioeconomic gradients in mortality and morbidity are well documented. Our findings support a widening of emphasis in public health research and practice toward reducing inequalities in physical and cognitive function in aging populations, as well as risk of disease and death, as life expectancy increases. They suggest that effective strategies will need to be long term and to target mediating pathways that link socioeconomic circumstances in early life (such as education) to health in old age. The graded association between childhood SEP and performance supports population-wide interventions, using a “universal proportionalism” approach,¹⁶ rather than a focus on disadvantaged sections of society.⁴³ Our findings indicate that in addition to public health initiatives targeted in later life,⁴⁴ to celebrate equitable “life to years” in the future, a life course approach to healthy aging¹⁶ is required.