Use of Biomeasures in Social Surveys 2 |
|
| Session Organisers | Ms Anne Conolly (NatCen Social Research) Dr Shaun Scholes (Research Department of Epidemiology and Public Health, University College London) Mr Matt Brown (Centre for Longitudinal Studies, UCL Institute of Education) Dr Emily Gilbert (Centre for Longitudinal Studies, UCL Institute of Education) |
| Time | Thursday 18th July, 14:00 - 15:30 |
| Room | D26 |
Many social surveys, both cross-sectional and longitudinal, include the collection of biomeasures. While traditional questionnaires can obtain self-reported assessments of general health, diagnosed disease and health behaviours, it is known that they can be prone to misreporting. Objective health measures can add to survey data considerably, enabling researchers to discover things that cannot be captured through self-reported measures. The inclusion of objective measurements within social surveys allows us to assess health with significantly greater accuracy and therefore to deepen our understanding of the interplay between social and biological factors in explaining human behaviour.
Such measurements encompass a range of anthropometric (e.g. height, weight, waist), functional (e.g. grip strength, balance), and sensory measurements (e.g. hearing), as well as biological samples (e.g. blood, saliva, urine), other physiological health measurements (e.g. blood pressure, lung function), and device-based measurement of physical activity.
Typically these data are collected either in clinics or at participants’ homes and may be carried out by trained field interviewers or by those with medical training and expertise (such as nurses). Technological advances and the development of minimally invasive techniques of data collection have increased the feasibility of collecting biomeasures at home and by fieldworkers with no medical training. Respondent-led collection of their own biomedical data is also now emerging as a data collection method – for example, some studies now ask respondents to self-collect dried blood spots. Additionally, the increase in the use of smartphone apps (e.g. activity tracking, food logs) and wearable technology (e.g. fitness trackers, smart watches, smart eyewear) has led to a growing interest in using such technology for data collection in survey research.
This session invites survey practitioners to share their experiences of incorporating the collection of biomeasures into social surveys. We welcome submissions relating to:
• Innovative approaches to the collection of biomeasures
• Comparisons of objective measures with self-reported data
• Analyses to assess the diagnostic ability of biomarkers
• Training of fieldworkers to collect biomeasures
• Respondent-led collection of biomeasures
• Methods to maximise response to and/or representativeness of biomeasures
• Collecting biomeasures in special populations (e.g. older people)
• Ethical challenges in the collection of biomeasures (e.g. relating to feedback of results, consent for ongoing use of biological samples)
Papers need not be restricted to these specific examples.
Keywords: biomeasures, health data, biosocial, biomarkers
Dr Erin Ware (University of Michigan) - Presenting Author
Dr Jessica Faul (University of Michigan)
Dr Colter Mitchell (University of Michigan)
Polygenic scores (PGSs), for better or worse, have become pervasive in social science due to their relative ease of computation and implementation. Many complex health outcomes/behaviors have been shown to be polygenic in nature and single genetic variants or candidate genes may not truly capture the dynamic nature of these traits.
Building on previous work examining the different methods by which PGSs can be constructed (Ware, et al. 2017), one recommendation from our analysis was to use genotyped data only. This was due to our empirical findings of no added benefit to using imputed data in the percent variance explained of the analogous phenotype and the computation benefit to using a smaller set of variants.
However, conclusions were based on the Health and Retirement Study (HRS) which uses lllumina HumanOmni2.5 BeadChips (HumanOmni2.5-4v1, HumanOmni2.5-8v1), measuring ~2.4 million SNPs. To leverage coverage, cost, and customizability, population-based studies invest in a number of different arrays. Less dense genome-wide arrays such as Infinium PsychArray-24 Kit (~500k variants), Illumina Human Omni1-Quad BeadChip (~1 million variants), Infinium Multi-Ethnic Global “MEGA” (1.7 million variants), and arrays with more coverage such as Illumina Infinium Omni5-4 Kit (4.3 million variants) have not been evaluated based on the use of genotyped data in polygenic scores to the extent that the lllumina HumanOmni2.5 BeadChips have in the HRS.
This project will use the Health and Retirement Study imputation to the HRC and Illumina manifests from multiple arrays (Infinium PsychArray-24 Kit, Illumina Human Omni1-Quad BeadChip, Infinium Multi-Ethnic Global “MEGA”, and Illumina Infinium Omni5-4 Kit) to mimic genotyped data from these specific chips. We will then assess the predictive capacity and correlation of PGS using all overlapping SNPs between the simulated genotyped data and GWAS summary statistics from height, educational attainment, and BMI.
Dr Jessica Faul (University of Michigan)
Dr Colter Mitchell (University of Michigan) - Presenting Author
Over the last decade molecular biomarkers such as telomere length (TL), the repeated DNA that create protective caps at the ends of chromosomes, and DNA methylation (DNAm), chemical tags that attach to DNA modifying gene expression, have become more common in social science and health research. Studies have shown associations between these molecular measures and various health outcomes, as well as significant social and environmental exposures. This might suggest they could be potential biomarkers of health &/or exposures, or maybe even mechanisms linking social contexts and health/behavior. The ability to detect small differences in TL or DNAm, and to do so reliably, may be influenced by several factors including those related to collection in the field as well as handling and storage preceding analysis. These are issues that have not been systematically investigated and are especially relevant to population-based studies, which have collected or are considering adding the collection of biological material. This study documents the effects of storage and handling conditions on the measurement of TL and DNAm using observational and experimental data. Using two representative panel studies, the Health and Retirement Study and the Fragile Families and Child Wellbeing Study, we examine relationships between time in shipping, and storage at room temperature and DNA quality, TL, and multiple DNAm measures (i.e. epigenetic age and methylation distribution). In addition, we conducted two simple experiments that allow for more causal associations. In one we collected saliva in Oragene kits and kept it at room temperature extracting DNA at 0, 1, 2, 3, 6, and 12 months. We found that DNA quality and TL measures decline significantly after 6 months. In another experiment we collected whole blood samples, aliquoted the samples, left one aliquot frozen in the lab and shipped the other across the country—arriving a different times between 24 and 76 hours. Those results are forthcoming.
Ms Jill Darling (USC)
Dr Arie Kapteyn (USC) - Presenting Author
Commercially available activity trackers are gaining in reliability and functionality, which makes them increasingly attractive for use in population based research. One concern with the use of the commercial activity trackers is that these provide feedback to the wearers, which is part of their commercial appeal. In observational studies (as opposed to interventions [1]), however, this potentially compromises validity of measurement.
We will present results of a randomized controlled trial that aims to gauge the extent to which feedback affects behavior. We designed a two-arm experiment in which participants (1) wear an activity monitor that provides feedback (Fitbit Charge 3) or (2) an activity monitor that does not provide feedback (GENEActiv). Participants wore the devices for 7 days, and answered survey questions relevant to analysis of the results, including about their sleep cycle, and levels of activity. One week later, participants in arm (1) wore the non-feedback device and those in arm (2) wore the feedback device for seven days. Participants were members of the USC Center for Economic and Social Research’s Understanding America Study probability-based internet panel who consented to participate. To control for the effect of comparing two different devices, in a separate experiment, students wore both devices simultaneously for a seven-day period and the correlation between the simultaneously worn devices was included in our analysis. We report on the results of these experiments, including participation rates, compliance, correlation rates, and the impact of feedback on respondent activity levels.
Reference
1. Van der Walt, N., et al., Feedback From Activity Trackers Improves Daily Step Count After Knee and Hip Arthroplasty: A Randomized Controlled Trial. The Journal of arthroplasty, 2018. 33(11): p. 3422-3428.
Miss Fiona Pashazadeh (University of Manchester) - Presenting Author
Dr Alexandru Cernat (University of Manchester)
The collection of biological data as part of large-scale longitudinal surveys has increased in recent years, creating new possibilities for research into the interactions between physical and social phenomena in the general population. Whilst the possibilities are undoubtedly exciting, these data can create additional challenges from the viewpoints of both collection and analysis. In particular, representative samples often suffer from missing data caused by sample members not responding to a survey request. The additional burden of biological data collection can lead to increased incidences of this nonresponse, potentially affecting the quality of the data and the robustness of results from subsequent analysis. There is a large body of research on the effects of interviewers on the likelihood of sample members to respond and some evidence that these effects may carry over from previous survey waves. Given the key role of nurses in collecting biological data through home visits in surveys such as UKHLS and ELSA, nurse effects may also be present.
This paper uses data from waves 2 and 3 of UKHLS to investigate how nonresponse to biological data collection may be related to the survey employees who carry out the relevant fieldwork. The methods centre on building response propensity models for the stages of biological data collection and testing for both interviewer and nurse effects on the likelihood to respond. This includes estimating cross-classified logistic multilevel models to account for the complex data structure.
The results suggest that both the interviewer and nurse assigned to a sample member can indeed affect their participation in biological data collection, adjusting for other explanatory variables. This has potential implications for the future collection of biological data, as survey organisations may be able to use this information to adapt their fieldwork procedures, and also for secondary data analysis as improvements could be made to missing data techniques to incorporate these effects.