by Rebecca Hardy
Society has never quite come to terms with the change from imperial to metric measurements, particularly when it comes to weight and height. Ask people how tall they are or how much they weigh and you’re likely to get an answer in feet and inches, or stones and pounds. Ask again what that is in metric and more often than not you’ll get a blank look.
Longitudinal researchers are faced with a similar challenge. We have huge amounts of valuable data from the various birth cohort studies going back to 1946, but the data, and the way it’s been collected has changed over the years. This has made it difficult – if not impossible – to accurately compare data between the studies. This means it is hard to see how our society has changed over generations.
My team and I wanted to investigate how body mass index (BMI) – a measure of weight-for-height – overweight and obesity across life has changed across generations by comparing results from five of the UK’s longitudinal studies: the 1946 MRC National Survey of Health and Development, 1958 National Child Development Study, 1970 British Cohort Study, Millennium Cohort Study and the Avon Longitudinal Study of Parents and Children (Children of the 90s).
Height and weight had been recorded repeatedly over the years in all these studies but there was a mixture of imperial and metric measurements and a mixture of measurement by health professionals and that which has been self-reported (and we know that people are more likely to underestimate their weight and overestimate their height). There were also differences in the samples included in the different studies.
Supported by ESRC funding from CLOSER, we set about harmonising the data sets to enable us to conduct our research.
My colleague Dr Will Johnson, now a lecturer in Loughborough University’s School of Sport, Exercise and Health Sciences, did the really hard work – a lengthy and time consuming process taking many months. Will is a human biologist and was therefore an ideal person to do this as he has a good understanding of growth and development. The first stage was to document the relevant data from all the studies on a large spreadsheet – recording information such as the units of measurement and the range of dates over which the data collections took place. All measures were converted to kilograms for weight and metres for height and BMI was calculated as weight divided by height squared – this was the easy bit. Overweight and obesity can then be derived from BMI using age-specific cut-offs.
Will then applied a standardised “data cleaning” process across all studies to identify implausible measures – for example, identifying cases where height had decreased substantially between measurements. We had lengthy discussions within the team about whether we should “correct” the self-reported measures to allow for errors in reporting. We decided not to correct, as there is no obvious way of doing so and any correction would make strong assumptions – if anything our estimates of overweight and obesity levels are therefore conservative. However, in the harmonised data set we include an indicator of whether a height or weight was measured or self-reported to allow other researchers to make their own decisions about whether to adjust in their own analyses
After nearly two years of hard work we submitted our first manuscript using the harmonised data set to track how the age-related process of overweight and obesity development had changed across the five studies. We could now confidently show how overweight and obesity has become more prevalent at younger ages in more recently born generations. By linking our data set to harmonised measures of socioeconomic positions, we have also shown that three generations of less advantaged children faced a higher risk of overweight and obesity in middle age than their better-off peers.
Our harmonised data set is now deposited with the UK Data Service and freely available for others to use, through different licence arrangements.
Other CLOSER projects are creating a range of harmonised measures which will be deposited in the near future. These will enable the longitudinal research community to study change in society in areas such as class, income, educational qualifications, visual health and overcrowding.
And on a personal note, I now know how tall I am in both metres and feet and inches.
Rebecca Hardy is Professor of Epidemiology and Medical Statistics at the MRC Unit for Lifelong Health and Ageing at University College London.
Her research focuses on how prenatal development, childhood growth, adult BMI trajectories, and trajectories of physical function (such as blood pressure and lung function) influence later health and ageing. She also investigates the exposures that influence life course body size and the shape of the functional trajectories. Much of her research uses data from the MRC National Survey of Health and Development, a birth cohort born in 1946, with cross-cohort research being carried out through CLOSER, HALCyon and FALCon.
As part of CLOSER, Rebecca has carried out cross-cohort comparisons of BMI and overweight, tracking the rise in the obesity epidemic across generations. She is also interested in the methodology for the analysis of life course and longitudinal data and for cross-cohort comparisons, including data harmonisation.
Reblogged with kind permission from CLOSER. Read the original post.