[Skip to Content]
[Skip to Content Landing]
Table 1.  Characteristics of the Analytic Sample (n = 7850)
Characteristics of the Analytic Sample (n = 7850)
Table 2.  Association of Combined Noncognitive Life Skills With Mortality
Association of Combined Noncognitive Life Skills With Mortality
Table 3.  Association of Combined Noncognitive Life Skills With Mortality After Removing Each Component
Association of Combined Noncognitive Life Skills With Mortality After Removing Each Component
Table 4.  Association of Combined Noncognitive Life Skills and Potential Mediators and Confounders With Mortality
Association of Combined Noncognitive Life Skills and Potential Mediators and Confounders With Mortality
1.
Heckman  JJ, Humphries  JE, Kautz  T, eds.  The Myth of Achievement Tests: The GED and the Role of Character in American Life. University of Chicago Press; 2014.
2.
Humphries  JE, Kosse  F.  On the interpretation of non-cognitive skills—what is being measured and why it matters.   J Econ Behav Organ. 2017;136:174-185. doi:10.1016/j.jebo.2017.02.001Google ScholarCrossref
3.
Thiel  H, Thomsen  SL.  Noncognitive skills in economics: models, measurement, and empirical evidence.   Res Econ. 2013;67:189-214. doi:10.1016/j.rie.2013.03.002Google ScholarCrossref
4.
Gutman  LM, Schoon  I.  The Impact of Non-cognitive Skills on Outcomes for Young People. Education Endowment Fund and Cabinet Office; 2013.
5.
Steptoe  A, Wardle  J.  Life skills, wealth, health, and wellbeing in later life.   Proc Natl Acad Sci U S A. 2017;114(17):4354-4359. doi:10.1073/pnas.1616011114PubMedGoogle ScholarCrossref
6.
Steptoe  A, Jackson  SE.  The life skills of older Americans: association with economic, psychological, social, and health outcomes.   Sci Rep. 2018;8(1):9669. doi:10.1038/s41598-018-27909-wPubMedGoogle ScholarCrossref
7.
Graham  EK, Rutsohn  JP, Turiano  NA,  et al.  Personality predicts mortality risk: an integrative data analysis of 15 international longitudinal studies.   J Res Pers. 2017;70:174-186. doi:10.1016/j.jrp.2017.07.005PubMedGoogle ScholarCrossref
8.
Jokela  M, Batty  GD, Nyberg  ST,  et al.  Personality and all-cause mortality: individual-participant meta-analysis of 3,947 deaths in 76,150 adults.   Am J Epidemiol. 2013;178(5):667-675. doi:10.1093/aje/kwt170PubMedGoogle ScholarCrossref
9.
Kozela  M, Pająk  A, Micek  A,  et al.  Impact of perceived control on all-cause and cardiovascular disease mortality in three urban populations of Central and Eastern Europe: the HAPIEE study.   J Epidemiol Community Health. 2017;71(8):771-778. doi:10.1136/jech-2017-208992PubMedGoogle ScholarCrossref
10.
Rasmussen  HN, Scheier  MF, Greenhouse  JB.  Optimism and physical health: a meta-analytic review.   Ann Behav Med. 2009;37(3):239-256. doi:10.1007/s12160-009-9111-xPubMedGoogle ScholarCrossref
11.
Steptoe  A, Breeze  E, Banks  J, Nazroo  J.  Cohort profile: the English longitudinal study of ageing.   Int J Epidemiol. 2013;42(6):1640-1648. doi:10.1093/ije/dys168PubMedGoogle ScholarCrossref
12.
English Longitudinal Study of Ageing. About. Accessed April 9, 2020. http://www.elsa-project.ac.uk
13.
UK Economic and Social Research Council. UK Data Service. Accessed April 9, 2020. https://www.ukdataservice.ac.uk/
14.
World Medical Association.  World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects.   JAMA. 2013;310(20):2191-2194. doi:10.1001/jama.2013.281053.Google ScholarCrossref
15.
Lachman  ME, Weaver  SL.  Midlife Development Inventory (MIDI) Personality Scales: Scale Construction and Scoring. Brandeis University; 1997.
16.
Demakakos  P, Biddulph  JP, Bobak  M, Marmot  MG.  Wealth and mortality at older ages: a prospective cohort study.   J Epidemiol Community Health. 2016;70(4):346-353. doi:10.1136/jech-2015-206173PubMedGoogle ScholarCrossref
17.
Steptoe  A, Shankar  A, Demakakos  P, Wardle  J.  Social isolation, loneliness, and all-cause mortality in older men and women.   Proc Natl Acad Sci U S A. 2013;110(15):5797-5801. doi:10.1073/pnas.1219686110PubMedGoogle ScholarCrossref
18.
Marmot  M, Banks  J, Blundell  R, Lessof  C, Nazroo  J, eds.  Health, Wealth and Lifestyles of the Older Population in England. Institute of Fiscal Studies; 2003.
19.
VanderWeele  TJ, Ding  P.  Sensitivity analysis in observational research: introducing the E-Value.   Ann Intern Med. 2017;167(4):268-274. doi:10.7326/M16-2607PubMedGoogle ScholarCrossref
20.
Haneuse  S, VanderWeele  TJ, Arterburn  D.  Using the E-Value to assess the potential effect of unmeasured confounding in observational studies.   JAMA. 2019;321(6):602-603. doi:10.1001/jama.2018.21554PubMedGoogle ScholarCrossref
21.
Lin  DY, Fleming  TR, De Gruttola  V.  Estimating the proportion of treatment effect explained by a surrogate marker.   Stat Med. 1997;16(13):1515-1527. doi:10.1002/(SICI)1097-0258(19970715)16:13<1515::AID-SIM572>3.0.CO;2-1PubMedGoogle ScholarCrossref
22.
Elovainio  M, Hakulinen  C, Pulkki-Råback  L,  et al.  Contribution of risk factors to excess mortality in isolated and lonely individuals: an analysis of data from the UK Biobank cohort study.   Lancet Public Health. 2017;2(6):e260-e266. doi:10.1016/S2468-2667(17)30075-0PubMedGoogle ScholarCrossref
23.
Zhou  K; United Nations Educational, Scientific and Cultural Organization. Non-cognitive skills: definitions, measurement and malleability. Accessed April 8, 2020. https://unesdoc.unesco.org/images/0024/002455/245576E.pdf
24.
Thomas  CB, McCabe  OL.  Precursors of premature disease and death: habits of nervous tension.   Johns Hopkins Med J. 1980;147(4):137-145.PubMedGoogle Scholar
25.
Vaillant  GE.  Natural history of male psychologic health: effects of mental health on physical health.   N Engl J Med. 1979;301(23):1249-1254. doi:10.1056/NEJM197912063012302PubMedGoogle ScholarCrossref
26.
Chapman  BP, Fiscella  K, Kawachi  I, Duberstein  PR.  Personality, socioeconomic status, and all-cause mortality in the United States.   Am J Epidemiol. 2010;171(1):83-92. doi:10.1093/aje/kwp323PubMedGoogle ScholarCrossref
27.
Terracciano  A, Löckenhoff  CE, Zonderman  AB, Ferrucci  L, Costa  PT  Jr.  Personality predictors of longevity: activity, emotional stability, and conscientiousness.   Psychosom Med. 2008;70(6):621-627. doi:10.1097/PSY.0b013e31817b9371PubMedGoogle ScholarCrossref
28.
Kim  ES, Hagan  KA, Grodstein  F, DeMeo  DL, De Vivo  I, Kubzansky  LD.  Optimism and cause-specific mortality: a prospective cohort study.   Am J Epidemiol. 2017;185(1):21-29. doi:10.1093/aje/kww182PubMedGoogle ScholarCrossref
29.
Turiano  NA, Chapman  BP, Agrigoroaei  S, Infurna  FJ, Lachman  M.  Perceived control reduces mortality risk at low, not high, education levels.   Health Psychol. 2014;33(8):883-890. doi:10.1037/hea0000022PubMedGoogle ScholarCrossref
30.
Kim  ES, Strecher  VJ, Ryff  CD.  Purpose in life and use of preventive health care services.   Proc Natl Acad Sci U S A. 2014;111(46):16331-16336. doi:10.1073/pnas.1414826111PubMedGoogle ScholarCrossref
31.
Steptoe  A, Easterlin  E, Kirschbaum  C.  Conscientiousness, hair cortisol concentration, and health behaviour in older men and women.   Psychoneuroendocrinology. 2017;86:122-127. doi:10.1016/j.psyneuen.2017.09.016PubMedGoogle ScholarCrossref
32.
Roy  B, Diez-Roux  AV, Seeman  T, Ranjit  N, Shea  S, Cushman  M.  Association of optimism and pessimism with inflammation and hemostasis in the Multi-Ethnic Study of Atherosclerosis (MESA).   Psychosom Med. 2010;72(2):134-140. doi:10.1097/PSY.0b013e3181cb981bPubMedGoogle ScholarCrossref
33.
Nater  UM, Hoppmann  C, Klumb  PL.  Neuroticism and conscientiousness are associated with cortisol diurnal profiles in adults—role of positive and negative affect.   Psychoneuroendocrinology. 2010;35(10):1573-1577. doi:10.1016/j.psyneuen.2010.02.017PubMedGoogle ScholarCrossref
34.
Pitsavos  C, Panagiotakos  DB, Papageorgiou  C, Tsetsekou  E, Soldatos  C, Stefanadis  C.  Anxiety in relation to inflammation and coagulation markers, among healthy adults: the ATTICA study.   Atherosclerosis. 2006;185(2):320-326. doi:10.1016/j.atherosclerosis.2005.06.001PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Original Investigation
    Public Health
    May 14, 2020

    Association of Noncognitive Life Skills With Mortality at Middle and Older Ages in England

    Author Affiliations
    • 1Department of Behavioural Science and Health, University College London, London, United Kingdom
    JAMA Netw Open. 2020;3(5):e204808. doi:10.1001/jamanetworkopen.2020.4808
    Key Points español 中文 (chinese)

    Question  Are greater noncognitive life skills associated with reduced mortality in older adults?

    Findings  In this cohort study of 7850 adults aged 52 years and older followed up for approximately 7 years, the combination of conscientiousness, emotional stability, persistence, optimism, and sense of control was associated with reduced mortality independently of sociodemographic, health, and behavioral factors, but no single life skill explained the association.

    Meaning  These findings suggest that higher levels of noncognitive life skills are associated with longer survival, suggesting that maintenance of these skills in later life is relevant to health.

    Abstract

    Importance  Noncognitive life skills are patterns of behavior, thoughts, and feelings that complement cognitive ability in promoting positive economic and educational outcomes. These positive attributes have been associated with favorable social and health outcomes at older ages, but their combined association with survival is not known.

    Objectives  To evaluate the association of the combination of 5 noncognitive life skills with mortality, and to explore the role of sociodemographic, health, cognitive, and behavioral factors in explaining associations.

    Design, Setting, and Participants  This cohort study used data from wave 5 of the English Longitudinal Study of Ageing, conducted in 2010. Participants included adults 52 years and older. Associations of scores on 5 noncognitive life skills, including conscientiousness, perseverance, emotional stability, optimism, and control, with all-cause mortality were analyzed for a mean (SD) of 7.2 (1.3) years. Data analyses were completed in November 2019.

    Exposures  Response to wave 5 of the English Longitudinal Study of Ageing.

    Main Outcomes and Measures  Noncognitive life skills scores, including conscientiousness, perseverance, emotional stability, optimism, and control, were measured by questionnaire. The main outcome was all-cause mortality, and the associations with noncognitive life skills scores were analyzed using Cox proportional hazards regressions models to estimate hazard ratios per 1-SD increase in score.

    Results  A total of 7850 participants (mean [SD] age, 66.5 [9.0] years; 4333 [55.2%] women) were included. Combined life skill score was positively associated with survival, with a hazard ratio of 0.81 (95% CI, 0.72-0.90) per 1-SD increase in positive attributes after adjustment for age, sex, race/ethnicity, childhood socioeconomic status, educational attainment, baseline chronic disease, depressive symptoms, cognitive function, mobility impairment, social isolation, smoking, physical activity, alcohol intake, and fruit and vegetable consumption (P < .001). Excluding deaths during 24 months after baseline as a check for reverse causation showed the same pattern (adjusted hazard ratio, 0.79; 95% CI, 0.70-0.89; P < .001). Associations were maintained after each life skill was omitted in turn from the aggregate score, indicating that no single positive attribute accounted for the protective association.

    Conclusions and Relevance  These findings suggest that noncognitive life skills are associated with survival at older ages. Whether training and education programs could enhance these attributes and influence mortality risk is not known, but fostering and maintaining life skills may be important in later life.

    Introduction

    Noncognitive life skills are patterns of behavior, thoughts, and feelings, such as perseverance, self-control, conscientiousness, social skills, and emotional stability. They have come into prominence in economic and policy research as positive psychological attributes that contrast with cognitive skills or ability. There is substantial evidence that they are important in promoting future economic, educational, and prosocial outcomes in early life.1-4 Some of these characteristics can be regarded as aspects of personality, although the term skill has been used to highlight the notion that they are malleable rather than fixed traits. The accumulation of noncognitive life skills in later life seems to be relevant as well. An analysis of more than 8000 participants in the English Longitudinal Study of Ageing (ELSA) measured 5 positive attributes, including conscientiousness, emotional stability, persistence or determination, optimism, and sense of control, and demonstrated that aggregate life skills were associated with favorable economic, social, health, and biological outcomes.5 The combination of life skills was also associated with less depression and loneliness, a lower incidence of chronic disease and disability, and better physical capability 4 years later, even after taking baseline function into account. These findings from older people in England were replicated in the Health and Retirement Study in the US, where longitudinal associations were also observed between number of life skills and less anxiety and chronic stress and more prosocial behavior.6 No single life skill was responsible for these findings, and associations were independent of cognitive ability, education, and childhood socioeconomic advantage.

    The association of combined noncognitive life skills with survival has not been investigated extensively, to our knowledge. Individual components, such as conscientiousness, emotional stability, optimism, and sense of control, have been associated with reduced mortality,7-10 but it is not known whether the combination of these positive attributes is associated with longer survival. Accordingly, we assessed all-cause mortality during an approximately 7-year follow-up period in the ELSA sample in which other outcomes have previously been investigated.5 In addition to evaluating the combination of noncognitive life skills, we tested whether any particular skill was especially associated with survival. We also assessed the extent to which socioeconomic factors, chronic physical illness, depression, mobility impairment, cognitive ability, social isolation, and health behaviors explained the associations of noncognitive life skills with survival.

    Methods
    Study Sample

    The ELSA is a longitudinal panel study of men and women 50 years and older living in England that started in 2002.11 The methods of data collection and questionnaires have been published previously.12 The data used for our analyses were collected in wave 5 of ELSA in 2010, since this was the first occasion on which data for several of the noncognitive life skills were obtained. Data from ELSA can be accessed from the UK Data Service.13 The study was conducted in accordance with the World Medical Association Declaration of Helsinki.14 The ELSA study was approved by the UK’s National Research Ethics Service, and all participants provided informed consent in person. This report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. Data analyses were completed in November 2019.

    Measures
    Noncognitive Life Skills

    Scales from the Midlife Development Inventory Personality Scales15 were used to measure conscientiousness and emotional stability. These scales have been used in several previous analyses of the Midlife in the United States study and the Health and Retirement Study. Participants were asked the extent to which each of 26 adjectives described themselves on a scale ranging from 1, indicating not at all, to 4, a lot. Four items (eg, organized, responsible) contributed to the conscientiousness scale, and 6 items (eg, moody, worrying) to the emotional stability scale. Persistence was assessed with a single rating of the extent to which participants had felt determined during the past 30 days (responses ranged from not at all to very much). Optimism was measured with ratings of 2 statements “I feel that life is full of opportunities” and “I feel that the future looks good for me,” while sense of control was indexed by the single item “At home, I feel I have control over what happens in most situations.” An aggregated index of noncognitive life skills was created by standardizing ratings for each skill (mean [SD], 0 [1.0]), and deriving the mean of standardized ratings. We also created aggregated life skill scores omitting each of the 5 capabilities in turn to assess whether any single skill accounted for associations with mortality.

    Mortality

    The main outcome, all-cause mortality, was assessed up to February 2018. Mortality data were obtained through linkage with the National Health Service central data registry for all participants who consented to mortality follow-up, indicating death by year and month.

    Covariates

    We included as covariates demographic variables together with factors that might be associated with noncognitive life skills and mortality risk. Age at baseline (ie, in 2010) was classified into 4 categories: 52 to 59 years, 60 to 69 years, 70 to 79 years, and 80 years or older. Race/ethnicity was categorized as white or other and was self-reported. Childhood socioeconomic status (SES) was indexed by the occupation of the participant’s father or main caregiver when they were aged 14 years, divided into routine, intermediate, and managerial or professional. Education was measured as the participant’s highest educational qualification and divided into 4 categories: basic education (ie, no qualifications), some junior high school qualifications (ie, O levels, approximately equivalent to US 7th or 8th grades), high school graduation (ie, A levels, approximately equivalent to US 12th grade), and college or university education. Chronic physical illness at baseline was indicated by summing the number of diagnosed serious diseases, including coronary heart disease, cancer, stroke, diabetes, arthritis, and chronic lung disease. Depression was assessed with symptoms measured using the 8-item Center for Epidemiologic Studies Depression Scale, with a score of 4 or higher being used to indicate significant symptoms, as used previously in ELSA analyses.16 Cognitive capacity was measured by aggregating performance on 5 objective tests administered in face-to-face interviews. These were immediate recall, delayed recall, verbal fluency, and speed and accuracy on a letter cancellation task. We z-transformed scores on the 5 tests and found the mean to generate an index of cognitive function. Mobility impairment was assessed by summing the presence or absence of 10 mobility issues (eg, walking 100 yards, reaching or extending arms above shoulder level). Social isolation was assessed using a measure of the extent of contact with children, other family members, and friends and participation in organizations and clubs.17 Four activities were included in the measure of health behavior: smoking, physical activity (a 5-level categorization detailed elsewhere18), number of units of alcohol consumed per week, and number of portions of fruit and vegetables eaten daily.

    Statistical Analysis

    Associations between the index of noncognitive life skills and mortality were analyzed using Cox proportional hazards regression models, estimating hazard ratios (HRs) and 95% CIs per 1-SD increase in score. Checking Schoenfeld residuals indicated that proportional hazards assumptions were met in the Cox regression models. Survival time was measured in months from baseline (ie, the date of the wave 5 interview) to death or March 2018, whichever came first. We computed 2 models: a basic model adjusting for age and sex, and a fully adjusted model adjusting for age, sex, race/ethnicity, childhood SES, educational attainment, chronic disease, depressive symptoms, cognitive function, mobility impairment, social isolation, smoking, physical activity, fruit and vegetable consumption, and alcohol intake. Observational studies are liable to confounding by unmeasured variables. As a check for this possibility, E-values were calculated for HRs to estimate the strength of association an unmeasured confounder would need to have with both life skills and mortality to explain the positive attribute–mortality associations.19,20 The possibility of reverse causality (ie, terminal illness leading to reduced noncognitive life skills) was addressed by excluding deaths occurring within 24 months of baseline. The importance of individual life skills in association with mortality was tested by removing each in turn from the aggregated index and calculating the proportion of the protective association that remained after that component was removed.

    We also tested the extent to which different factors explained the associations between noncognitive life skills and mortality by testing 7 sets of potential mediators and confounders: sociodemographic factors (ie, race/ethnicity, childhood SES, and educational attainment), baseline physical health (ie, number of chronic diseases), depressive symptoms, cognitive function, mobility impairment, social isolation, and health behavior (ie, smoking, physical activity, fruit and vegetable consumption, and alcohol intake). In each case, we computed HRs for noncognitive life skills when including the potential mediator in the model, and computed the percentage of the protective association explained (PPAE) using the formula PPAE = (HR [adjusted for noncognitive life skills, age, sex, race/ethnicity, childhood SES, education, and mediator] – HR [adjusted for noncognitive life skills, age, sex, race/ethnicity, childhood SES, and education]) / (1 − HR [adjusted for noncognitive life skills, age, sex, race/ethnicity, childhood SES, and education]) × 100.21 This method has been widely used in epidemiological studies.22

    Results

    Of 8119 adults aged 52 years and older assessed in 2010 and analyzed in our previous study,5 8117 (99.9%) were tracked until March 2018, a mean follow-up of 7.2 [1.3] years. Data were missing on some covariates for 267 individuals, so the analytic sample included 7850 individuals (mean [SD] age, 66.5 [9.0] years; 4333 [55.2%] women). Table 1 summarizes the characteristics of the study sample. Participants were predominantly white (7673 individuals [97.7%]). There were 1030 deaths (13.1%) during the follow-up period. The univariable associations between sample characteristics and mortality are detailed in eTable 1 in the Supplement. Death was more common among men than women (549 men [15.6%] vs 481 women [11.1%]; P < .001), and was associated with greater age (age 52-59 years: 63 deaths among 2003 individuals [3.1%]; age 60-69 years: 205 deaths among 3125 individuals [6.6%]; age 70-79 years: 405 deaths among 1990 individuals [20.4%]; age ≥80 years: 357 deaths among 732 individuals [48.8%]; P < .001), lower childhood SES (routine: 371 deaths among 2484 individuals [14.9%]; intermediate: 435 deaths among 3308 individuals [13.6%]; managerial or professional: 224 deaths among 2158 individuals [10.4%]; P < .001), less education (no qualifications: 384 deaths among 1827 individuals [21.0%]; up to O level: 234 deaths among 875 individuals [12.5%]; A level or equivalent: 146 deaths among 1249 individuals [11.7%]; college or university: 266 deaths among 2899 individuals [9.2%]; P < .001), more chronic diseases (none: 246 deaths among 3445 individuals [7.1%]; 1 chronic disease: 393 deaths among 2948 individuals [13.3%]; 2 chronic diseases: 253 deaths among 1069 individuals [23.7%]; 3 chronic diseases: 111 deaths among 320 individuals [34.7%]; 4 chronic diseases: 19 deaths among 55 individuals [34.5%]; ≥5 chronic diseases: 8 deaths among 13 individuals [61.5%]; P < .001), depressive symptoms (no symptoms: 802 deaths among 6822 individuals [11.8%]; significant symptoms: 228 deaths among 1028 individuals [22.2%]; P < .001), social isolation (not isolated: 536 deaths among 4450 individuals [12.0%]; isolated: 494 deaths among 3400 individuals [14.5%]; P < .001), and smoking (no smoking: 866 deaths among 6907 individuals [12.5%]; current smoking: 164 deaths among 943 individuals [17.4%]; P < .001). Mortality risk was reduced in participants with higher cognitive scores (r = −0.224; P < .001), more physical activity (r = −0.234; P < .001), and greater fruit and vegetable consumption (r = −0.059; P < .001) and increased for individuals with mobility impairment (r = 0.215; P < .001).

    Results of the Cox proportional hazards regression analyses of the association of aggregate noncognitive life skills with all-cause mortality are shown in Table 2. There was a significant protective association in the basic age-and sex-adjusted model, with an HR of 0.58 (95% CI, 0.53-0.63; P < .001). The association was reduced in the fully-adjusted model but remained statistically significant, with an HR of 0.81 (95% CI, 0.72-0.90; P < .001), indicating a 19% reduction of hazard of dying for every 1-SD increase in noncognitive life skills, after controlling for sociodemographic, health, social, and lifestyle factors. In the full regression model, mortality was independently associated with sex (adjusted HR for women compared with men, 0.59; 95% CI, 0.51-0.67; P < .001), greater age (adjusted HR compared with age 52-59 years: age 60-69 years, 1.97; 95 % CI, 1.49-2.62; age 70-79 years, 5.43; 95 % CI, 4.13-7.15; age ≥80 years, 12.80; 95 % CI, 9.61-17.04; P < .001), baseline chronic disease (adjusted HR, 1.16; 95% CI, 1.08-1.23; P < .001), lower cognition (adjusted HR, 0.76; 95% CI, 0.69-0.84; P < .001), smoking (adjusted HR, 1.64; 95% CI, 1.37-1.96; P < .001) and less physical activity (adjusted HR, 0.81; 95% CI, 0.77-0.86; P < .001) (eTable 2 in the Supplement). The E-value for the fully adjusted model was 1.77 (lower 95% CI, 1.46) (Table 2).

    We tested for reverse causality by excluding all deaths occurring within 24 months of the measurement of noncognitive life skills, reasoning that participants might have had serious illnesses leading to death soon after baseline that reduced life skill scores, resulting in reverse causation among those experiencing early death. However, the association between noncognitive life skills and mortality was maintained (Table 2), with a 21% reduction in hazard for every 1-SD increase in life skills in the fully adjusted model (adjusted HR, 0.79; 95% CI, 0.70-0.89; P < .001).

    Since several of noncognitive life skills have previously been shown to be associated with survival, we tested whether a particular factor was responsible for the association of the aggregate life skill index with survival. Each component was therefore omitted from the combined index in turn, and the associations of the remaining factors with survival were calculated. The results indicate that the associations of noncognitive life skills with survival were slightly reduced after each component was removed in turn, but nonetheless remained significant in each case (conscientiousness: adjusted HR, 0.86; 95% CI, 0.77-0.95; P < .001; emotional stability: adjusted HR, 0.83; 95% CI, 0.76-0.90; P < .001; persistence: adjusted HR, 0.81; 95% CI, 0.73-0.90; P < .001; optimism: adjusted HR, 0.83; 95% CI, 0.75-0.92; P < .001; control: adjusted HR, 0.80; 95% CI, 0.72-0.90; P < .001) (Table 3).

    Table 4 summarizes the role of 7 sets of potential mediators and confounders in explaining the association of combined noncognitive life skills with survival. The associations of life skills with survival remained significant after each potential mediator or confounder was added to the model. The importance of each set of factors ranged from social isolation, which explained only 2% of the association of noncognitive life skills with survival, to sociodemographic factors (PPAE, 26%), health behaviors (PPAE, 26%), and mobility impairment (PPAE, 24%). Nonetheless, all factors together explained only approximately 55% of the association of noncognitive life skills with survival, implying that almost half of the association was not explained by these factors.

    Discussion

    The findings of this cohort study suggest that the combination of noncognitive life skills at older ages was associated with reduced all-cause mortality in a representative population sample. No single life skill explained this association, since the association with survival remained significant after each component was omitted in turn. The results indicated that approximately half the association of noncognitive life skills and survival was explained by the various mediators studied, including sociodemographic factors (eg, early life SES and educational attainment), baseline health, depressive symptoms, mobility, cognition, social isolation, and health behaviors. Nevertheless, even when all these factors were taken into account, individuals with high scores in noncognitive life skills had a 19% reduction in risk of dying during the 7-year follow-up period compared with those with lower noncognitive life skill scores. The E-values suggest that the associations are probably robust to unmeasured confounders. There is no absolute level for E-values above which confounding is definitely ruled out.20 However, the E-value of 1.77 in the primary analysis indicates that an unmeasured confounder would have to have a relative risk association with both noncognitive life skills and mortality of at least 1.77 to explain the association. Considering that smoking had an HR of 1.64 (95% CI, 1.37-1.96) for mortality, the unmeasured confounder would need to have a greater association even than smoking, and this seems unlikely.

    Nomenclature in this field of research is problematic. Some, but not all, of the noncognitive life skills are personality factors. The term noncognitive life skill was introduced by researchers from economic and public policy disciplines to distinguish these attributes from cognitive skills and to emphasize that they are modifiable. Other authorities have labeled these factors as aspects of personality or character.23 We used the term noncognitive life skill to encompass the range of phenomena under investigation and to distinguish these factors from features of positive well-being, such as happiness or sense of purpose. There is increasing evidence from epidemiology and personality psychology that individual positive attributes are associated with survival, following from the pioneering work of Thomas and McCabe,24 Vaillant,25 and others. Thus, large population studies have demonstrated that individuals high in conscientiousness and emotional stability are at reduced risk of mortality.7,8,26,27 Greater optimism is associated with reduced all-cause and cardiovascular mortality,10,28 while greater control also has protective associations.9,29 Few studies have examined combinations of these factors, although some investigators have assessed whether associations are independent of each other.7,8 Our study adds to the literature by showing the relevance of noncognitive life skills in combination. It is interesting that no single component drove the findings, supporting the argument that the accumulation of these life skills is relevant to healthy aging. Noncognitive life skills are important not only for survival but also for social, economic, and mental health outcomes.5,6

    We evaluated the processes potentially responsible for explaining the associations of noncognitive life skills with survival by studying 7 sets of factors, some of which could potentially act as confounders (eg, education, chronic disease) whereas others could be mediators (eg, social isolation, health behaviors). In addition to sociodemographic factors, such as education and childhood SES, 2 factors were particularly important. First, mobility impairment accounted for 24% of the protective association of noncognitive life skills with mortality. We do not know whether lower life skills earlier in life determined in part the development of impaired mobility, or whether mobility deficits led to reductions in noncognitive skills, such as optimism and persistence. Second, health behaviors, including smoking, physical inactivity, excessive alcohol consumption, and poor diet, explained 26% of the association. Previous studies have shown that the association of emotional stability (but not conscientiousness) with mortality is partly mediated by smoking,7 and our results suggest that a broader set of health behaviors is relevant.

    Nevertheless, a substantial proportion of the association of noncognitive life skills with mortality was not explained by mobility, health behavior, demographic factors, cognitive function, the presence of serious illness at baseline, depressive symptoms, or social isolation. Additional pathways must therefore be in part responsible. Unmeasured behaviors may be relevant, including adherence to medical advice or medication, taking opportunities for screening for serious diseases, or responding promptly to medical symptoms.30 Direct influences on biological processes contributing to health risk, such as cardiovascular and metabolic regulation, inflammatory responses, and neuroendocrine pathways, may also contribute. For example, high levels of conscientiousness, emotional stability, and greater optimism have been associated with reduced cortisol output and inflammation.31-34

    Limitations

    This study had some limitations. First, we were limited by the noncognitive life skills assessed in ELSA. Therefore, we were not able to include measures of social competence and social skill, coping, or other elements that are prominent in the noncognitive life skill literature.2,4 Some measures of life skills were suboptimal. For example, optimism was assessed with 2 items, although similar results were obtained in the Health and Retirement Study when optimism was quantified with a standard questionnaire.6 We do not know whether the life skills assessed in this study reflect capabilities maintained since early in life or were developed more recently. Life course studies are needed to investigate this issue, since the youngest participants in ELSA were in their 50s. Research involving more diverse cohorts in terms of age, race/ethnicity, and geography are desirable to help gauge the generalizability of the findings.

    Conclusions

    The results of this cohort study add weight to the importance of noncognitive life skills at older ages by documenting associations with survival. Training and education programs may be relevant even in middle and older age, but evaluating the effects of intervention on longevity is a major challenge.

    Back to top
    Article Information

    Accepted for Publication: March 8, 2020.

    Published: May 14, 2020. doi:10.1001/jamanetworkopen.2020.4808

    Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Steptoe A et al. JAMA Network Open.

    Corresponding Author: Andrew Steptoe, DSc, Department of Behavioural Science and Health, University College London, 1-19, Torrington Place, London WC1E 6BT (a.steptoe@ucl.ac.uk).

    Author Contributions: Drs Steptoe and Jackson had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Steptoe.

    Acquisition, analysis, or interpretation of data: Both authors.

    Drafting of the manuscript: Steptoe.

    Critical revision of the manuscript for important intellectual content: Both authors.

    Statistical analysis: Both authors.

    Obtained funding: Steptoe.

    Conflict of Interest Disclosures: None reported.

    Funding/Support: The English Longitudinal Study of Ageing is administered by a team of researchers based at the University College London, NatCen Social Research, the Institute for Fiscal Studies, and the University of Manchester. Funding is provided by National Institute on Aging (R01AG017644; principal investigator: Dr Steptoe) and by a consortium of UK government departments coordinated by the National Institute for Health Research.

    Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    References
    1.
    Heckman  JJ, Humphries  JE, Kautz  T, eds.  The Myth of Achievement Tests: The GED and the Role of Character in American Life. University of Chicago Press; 2014.
    2.
    Humphries  JE, Kosse  F.  On the interpretation of non-cognitive skills—what is being measured and why it matters.   J Econ Behav Organ. 2017;136:174-185. doi:10.1016/j.jebo.2017.02.001Google ScholarCrossref
    3.
    Thiel  H, Thomsen  SL.  Noncognitive skills in economics: models, measurement, and empirical evidence.   Res Econ. 2013;67:189-214. doi:10.1016/j.rie.2013.03.002Google ScholarCrossref
    4.
    Gutman  LM, Schoon  I.  The Impact of Non-cognitive Skills on Outcomes for Young People. Education Endowment Fund and Cabinet Office; 2013.
    5.
    Steptoe  A, Wardle  J.  Life skills, wealth, health, and wellbeing in later life.   Proc Natl Acad Sci U S A. 2017;114(17):4354-4359. doi:10.1073/pnas.1616011114PubMedGoogle ScholarCrossref
    6.
    Steptoe  A, Jackson  SE.  The life skills of older Americans: association with economic, psychological, social, and health outcomes.   Sci Rep. 2018;8(1):9669. doi:10.1038/s41598-018-27909-wPubMedGoogle ScholarCrossref
    7.
    Graham  EK, Rutsohn  JP, Turiano  NA,  et al.  Personality predicts mortality risk: an integrative data analysis of 15 international longitudinal studies.   J Res Pers. 2017;70:174-186. doi:10.1016/j.jrp.2017.07.005PubMedGoogle ScholarCrossref
    8.
    Jokela  M, Batty  GD, Nyberg  ST,  et al.  Personality and all-cause mortality: individual-participant meta-analysis of 3,947 deaths in 76,150 adults.   Am J Epidemiol. 2013;178(5):667-675. doi:10.1093/aje/kwt170PubMedGoogle ScholarCrossref
    9.
    Kozela  M, Pająk  A, Micek  A,  et al.  Impact of perceived control on all-cause and cardiovascular disease mortality in three urban populations of Central and Eastern Europe: the HAPIEE study.   J Epidemiol Community Health. 2017;71(8):771-778. doi:10.1136/jech-2017-208992PubMedGoogle ScholarCrossref
    10.
    Rasmussen  HN, Scheier  MF, Greenhouse  JB.  Optimism and physical health: a meta-analytic review.   Ann Behav Med. 2009;37(3):239-256. doi:10.1007/s12160-009-9111-xPubMedGoogle ScholarCrossref
    11.
    Steptoe  A, Breeze  E, Banks  J, Nazroo  J.  Cohort profile: the English longitudinal study of ageing.   Int J Epidemiol. 2013;42(6):1640-1648. doi:10.1093/ije/dys168PubMedGoogle ScholarCrossref
    12.
    English Longitudinal Study of Ageing. About. Accessed April 9, 2020. http://www.elsa-project.ac.uk
    13.
    UK Economic and Social Research Council. UK Data Service. Accessed April 9, 2020. https://www.ukdataservice.ac.uk/
    14.
    World Medical Association.  World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects.   JAMA. 2013;310(20):2191-2194. doi:10.1001/jama.2013.281053.Google ScholarCrossref
    15.
    Lachman  ME, Weaver  SL.  Midlife Development Inventory (MIDI) Personality Scales: Scale Construction and Scoring. Brandeis University; 1997.
    16.
    Demakakos  P, Biddulph  JP, Bobak  M, Marmot  MG.  Wealth and mortality at older ages: a prospective cohort study.   J Epidemiol Community Health. 2016;70(4):346-353. doi:10.1136/jech-2015-206173PubMedGoogle ScholarCrossref
    17.
    Steptoe  A, Shankar  A, Demakakos  P, Wardle  J.  Social isolation, loneliness, and all-cause mortality in older men and women.   Proc Natl Acad Sci U S A. 2013;110(15):5797-5801. doi:10.1073/pnas.1219686110PubMedGoogle ScholarCrossref
    18.
    Marmot  M, Banks  J, Blundell  R, Lessof  C, Nazroo  J, eds.  Health, Wealth and Lifestyles of the Older Population in England. Institute of Fiscal Studies; 2003.
    19.
    VanderWeele  TJ, Ding  P.  Sensitivity analysis in observational research: introducing the E-Value.   Ann Intern Med. 2017;167(4):268-274. doi:10.7326/M16-2607PubMedGoogle ScholarCrossref
    20.
    Haneuse  S, VanderWeele  TJ, Arterburn  D.  Using the E-Value to assess the potential effect of unmeasured confounding in observational studies.   JAMA. 2019;321(6):602-603. doi:10.1001/jama.2018.21554PubMedGoogle ScholarCrossref
    21.
    Lin  DY, Fleming  TR, De Gruttola  V.  Estimating the proportion of treatment effect explained by a surrogate marker.   Stat Med. 1997;16(13):1515-1527. doi:10.1002/(SICI)1097-0258(19970715)16:13<1515::AID-SIM572>3.0.CO;2-1PubMedGoogle ScholarCrossref
    22.
    Elovainio  M, Hakulinen  C, Pulkki-Råback  L,  et al.  Contribution of risk factors to excess mortality in isolated and lonely individuals: an analysis of data from the UK Biobank cohort study.   Lancet Public Health. 2017;2(6):e260-e266. doi:10.1016/S2468-2667(17)30075-0PubMedGoogle ScholarCrossref
    23.
    Zhou  K; United Nations Educational, Scientific and Cultural Organization. Non-cognitive skills: definitions, measurement and malleability. Accessed April 8, 2020. https://unesdoc.unesco.org/images/0024/002455/245576E.pdf
    24.
    Thomas  CB, McCabe  OL.  Precursors of premature disease and death: habits of nervous tension.   Johns Hopkins Med J. 1980;147(4):137-145.PubMedGoogle Scholar
    25.
    Vaillant  GE.  Natural history of male psychologic health: effects of mental health on physical health.   N Engl J Med. 1979;301(23):1249-1254. doi:10.1056/NEJM197912063012302PubMedGoogle ScholarCrossref
    26.
    Chapman  BP, Fiscella  K, Kawachi  I, Duberstein  PR.  Personality, socioeconomic status, and all-cause mortality in the United States.   Am J Epidemiol. 2010;171(1):83-92. doi:10.1093/aje/kwp323PubMedGoogle ScholarCrossref
    27.
    Terracciano  A, Löckenhoff  CE, Zonderman  AB, Ferrucci  L, Costa  PT  Jr.  Personality predictors of longevity: activity, emotional stability, and conscientiousness.   Psychosom Med. 2008;70(6):621-627. doi:10.1097/PSY.0b013e31817b9371PubMedGoogle ScholarCrossref
    28.
    Kim  ES, Hagan  KA, Grodstein  F, DeMeo  DL, De Vivo  I, Kubzansky  LD.  Optimism and cause-specific mortality: a prospective cohort study.   Am J Epidemiol. 2017;185(1):21-29. doi:10.1093/aje/kww182PubMedGoogle ScholarCrossref
    29.
    Turiano  NA, Chapman  BP, Agrigoroaei  S, Infurna  FJ, Lachman  M.  Perceived control reduces mortality risk at low, not high, education levels.   Health Psychol. 2014;33(8):883-890. doi:10.1037/hea0000022PubMedGoogle ScholarCrossref
    30.
    Kim  ES, Strecher  VJ, Ryff  CD.  Purpose in life and use of preventive health care services.   Proc Natl Acad Sci U S A. 2014;111(46):16331-16336. doi:10.1073/pnas.1414826111PubMedGoogle ScholarCrossref
    31.
    Steptoe  A, Easterlin  E, Kirschbaum  C.  Conscientiousness, hair cortisol concentration, and health behaviour in older men and women.   Psychoneuroendocrinology. 2017;86:122-127. doi:10.1016/j.psyneuen.2017.09.016PubMedGoogle ScholarCrossref
    32.
    Roy  B, Diez-Roux  AV, Seeman  T, Ranjit  N, Shea  S, Cushman  M.  Association of optimism and pessimism with inflammation and hemostasis in the Multi-Ethnic Study of Atherosclerosis (MESA).   Psychosom Med. 2010;72(2):134-140. doi:10.1097/PSY.0b013e3181cb981bPubMedGoogle ScholarCrossref
    33.
    Nater  UM, Hoppmann  C, Klumb  PL.  Neuroticism and conscientiousness are associated with cortisol diurnal profiles in adults—role of positive and negative affect.   Psychoneuroendocrinology. 2010;35(10):1573-1577. doi:10.1016/j.psyneuen.2010.02.017PubMedGoogle ScholarCrossref
    34.
    Pitsavos  C, Panagiotakos  DB, Papageorgiou  C, Tsetsekou  E, Soldatos  C, Stefanadis  C.  Anxiety in relation to inflammation and coagulation markers, among healthy adults: the ATTICA study.   Atherosclerosis. 2006;185(2):320-326. doi:10.1016/j.atherosclerosis.2005.06.001PubMedGoogle ScholarCrossref
    ×