The NSHD is based on a social class stratified sample of 5,362 births of all singleton births that occurred within marriage in a week in March 1946 in England, Scotland and Wales. Previous follow-ups occurred approximately every 2 years in childhood, and the previous main data collections in adult life were at 26, 36, 43 and 53 years (Wadsworth et al. 2006).
The 60–64 year data collection consisted of a postal questionnaire to eligible study members, followed up (between 2 months and 2 years later) by invitation to attend one of six CRFs across the UK for assessments or, if they were unable or unwilling to travel, to have a research nurse visit the study member at home (Kuh et al. 2011). A reduced set of clinical measures was carried out for those visited at home. All participants were also asked to complete a pre-assessment questionnaire before the visit.
Ethical approval for the study was obtained from the Greater Manchester Local Research Ethics Committee and the Scotland A Research Ethics Committee. Written, informed consent was obtained from the study member for each component of data collection.
Defining response and cooperation rates
Study members were defined as eligible for the 60–64 year follow-up if study records indicated that they were living in England, Scotland or Wales, and had not previously withdrawn from the study or remained untraced since the previous follow-up at 53 years. Of the original cohort of 5,362 study members, 3,163 (59.0 %) were included in the target sample. Contact was not attempted with the remaining 2,198 of the original cohort who were considered ineligible for inclusion at 60–64 years: 718 (13.4 %) had died, 567 (10.6 %) lived abroad, 594 (11.1 %) were prior refusals and 320 remained permanently untraced since the last contact in 1999. Drop-outs due to death were not investigated here but previous analyses show that low childhood socioeconomic position (indicated by father being in a manual occupation, mother attaining primary level education or below, or poor housing quality) among women and low adult socioeconomic position (indicated by head of household manual occupation, not owning one’s home or low household income) among men and women predicted premature all-cause mortality (Kuh et al. 2009). Low childhood cognition, being a smoker, or having psychiatric disorder in early adulthood also predicted premature mortality (Kuh et al. 2009; Henderson et al. 2011). Of the 3,163 target sample invited to complete the postal questionnaire, between the postal questionnaire and CRF invitation, 60 died, 17 emigrated or moved out of the catchment area, and 230 were found to have an unknown address. This group of 307 were considered ineligible for inclusion in the clinical assessment mode of data collection because of the cost and participant burden of travelling, or because their whereabouts was unknown, yielding a target sample of 2,856 (Figure 1).
The primary aim was to investigate health and socio-demographic differences in response to the different elements of the study. We calculated the overall response rate by dividing the number who provided any information at the 60–64 year follow-up by the total eligible target sample. We calculated the visit cooperation rate by dividing the number who completed a CRF or home visit by the number known to be eligible for the clinical assessment (Table 2). We calculated the CRF cooperation rate by dividing the number who attended a CRF by the number who completed either a CRF or home visit.
Explanatory measures and analysis methods
Associations between overall response rate and the visit and CRF cooperation rates and (a) socioeconomic characteristics in childhood and adulthood, and (b) adult health and health-related behaviours were examined bivariately using Wald tests. Exposures were selected to capture childhood, early adult and midlife characteristics. These were childhood cognitive ability, father’s social class in childhood, educational attainment by 26 years and housing tenure at 26 years. Cognitive ability was measured at age 8 (or at ages 11 or 15 if this was missing) using the summed score from four tests: reading comprehension, word reading, vocabulary and nonverbal reasoning (Richards et al. 2004). Father’s occupational social class at age 4 was coded according to the UK Registrar General’s Standard’s Occupation Classification. Midlife socioeconomic factors considered were economic activity, occupation-based social class (also coded to the UK Registrar General’s Standard’s Occupation Classification), housing tenure and marital status, all at 53 years. Midlife health-related factors considered were physical and cognitive performance, mental health profiles, health conditions, cardiovascular disease, obesity, smoking, physical activity and alcohol problems, at 53 years with the exception of mental health and smoking. Physical performance was evaluated utilising measures of grip strength, balance and time to rise from a chair ten times. These three indicators were summed to create an aggregate physical performance score (Guralnik et al. 2006). Cognitive performance was captured by verbal memory, measured as the number of correct words recalled from a list of 15 over three learning trials (Richards et al. 2004). Accepted thresholds indicating suboptimal cognitive and physical performance have not been defined. The 10 and 90 % cut-offs were used here in line with a previous study which aimed to evaluate functional outcomes that ‘were potentially meaningful in this middle-aged cohort’ (Guralnik et al. 2006, pp. 696). Health conditions (including cardiovascular, respiratory, cancer and other conditions) were self-reported. A further list of cardiovascular disease indicators were also self-reported (angina, leg claudication, doctor-diagnosed stroke, valvular disease, aortic stenosis). Study members were classified as obese if they had a body mass index of ≥30 kg/m2 based on measured height and weight. Self-reported leisure-time physical activity was assessed as the number of occasions in which study members participated in sport, vigorous leisure activities or exercises in leisure time, not including getting to and from work, in the past 4 weeks (Cooper et al. 2011). Alcohol problems were captured by the CAGE screen for potential alcohol abuse (Hatch et al. 2007). Life time smoking behaviour was derived from smoking status at 26, 31, 36, 43 and 53 years (Clennell et al. 2008). Mental health profiles based on latent classes of measures of affective signs and symptoms at 13, 15, 36, 43 and 53 years were used (Colman et al. 2007). Study members were assigned to one of four latent classes summarising their symptoms in adolescence and adulthood which can be broadly described as having symptoms in adolescence which were not present in adulthood, having adult onset symptoms, having symptoms in adolescence and adulthood and not having symptoms. Multivariable logistic regression was used to identify independent predictors of response from childhood and adult life in three steps: (i) including all socioeconomic characteristics found to be associated bivariately with response at the 20 % level of significance, (ii) including all health-related characteristics found to be bivariately associated at the 20 % level of significance and (iii) all socioeconomic and health-related characteristics identified as statistically significant predictors at the 5 % level in the two preceding steps.
Health and socioeconomic characteristics of NSHD study members were compared with those of the general population aged 60–64 years using 2001 England Census data (accessed through the Census Dissemination Unit, MIMAS (University of Manchester)). Since it was not possible to distinguish those born in and outside mainland Britain from routinely available census statistics, we additionally used data for 60–64 year olds of white ethnic origin living in England, Wales or Scotland in 2010 extracted from the Integrated Household Survey (Office for National Statistics 2010). Available sample size varies by item but is more than 28,000 for all characteristics tabled.