Using linked educational attainment data to reduce bias due to missing outcome data in estimates of the association between the duration of breastfeeding and IQ at 15 years

An exploration into the causal relationships between educational attainment, intelligence, and wellbeing: an observational and two-sample Mendelian randomisation study

Educational attainment is associated with a range of positive outcomes, yet its impact on wellbeing is unclear, and complicated by high correlations with intelligence. We use genetic and observational data to investigate for the first time, whether educational attainment and intelligence are causally and independently related to wellbeing. Results from our multivariable Mendelian randomisation demonstrated a positive causal impact of a genetic predisposition to higher educational attainment on wellbeing that remained after accounting for intelligence, and a negative impact of intelligence that was independent of educational attainment. Observational analyses suggested that these associations may be subject to sex differences, with benefits to wellbeing greater for females who attend higher education compared to males. For intelligence, males scoring more highly on measures related to happiness were those with lower intelligence. Our findings demonstrate a unique benefit for wellbeing of staying in school, over and above improving cognitive abilities, with benefits likely to be greater for females compared to males.

Using linked administrative data to aid the handling of non-response and restore sample representativeness in cohort studies: the 1958 national child development study and hospital episode statistics data

BACKGROUND

There is growing interest in whether linked administrative data have the potential to aid analyses subject to missing data in cohort studies.

METHODS

Using linked 1958 National Child Development Study (NCDS; British cohort born in 1958, n = 18,558) and Hospital Episode Statistics (HES) data, we applied a LASSO variable selection approach to identify HES variables which are predictive of non-response at the age 55 sweep of NCDS. We then included these variables as auxiliary variables in multiple imputation (MI) analyses to explore the extent to which they helped restore sample representativeness of the respondents together with the imputed non-respondents in terms of early life variables (father's social class at birth, cognitive ability at age 7) and relative to external population benchmarks (educational qualifications and marital status at age 55).

RESULTS

We identified 10 HES variables that were predictive of non-response at age 55 in NCDS. For example, cohort members who had been treated for adult mental illness had more than 70% greater odds of bring non-respondents (odds ratio 1.73; 95% confidence interval 1.17, 2.51). Inclusion of these HES variables in MI analyses only helped to restore sample representativeness to a limited extent. Furthermore, there was essentially no additional gain in sample representativeness relative to analyses using only previously identified survey predictors of non-response (i.e. NCDS rather than HES variables).

CONCLUSIONS

Inclusion of HES variables only aided missing data handling in NCDS to a limited extent. However, these findings may not generalise to other analyses, cohorts or linked administrative datasets. This work provides a demonstration of the use of linked administrative data for the handling of missing cohort data which we hope will act as template for others.

UK research data resources based on primary care electronic health records: review and summary for potential users

BACKGROUND

The range and scope of electronic health record (EHR) data assets in the UK has recently increased, which has been mainly in response to the COVID-19 pandemic. Summarising and comparing the large primary care resources will help researchers to choose the data resources most suited to their needs.

AIM

To describe the current landscape of UK EHR databases and considerations of access and use of these resources relevant to researchers.

DESIGN & SETTING

Narrative review of EHR databases in the UK.

METHOD

Information was collected from the Health Data Research Innovation Gateway, publicly available websites and other published data, and from key informants. The eligibility criteria were population-based open-access databases sampling EHRs across the whole population of one or more countries in the UK. Published database characteristics were extracted and summarised, and these were corroborated with resource providers. Results were synthesised narratively.

RESULTS

Nine large national primary care EHR data resources were identified and summarised. These resources are enhanced by linkage to other administrative data to a varying extent. Resources are mainly intended to support observational research, although some can support experimental studies. There is considerable overlap of populations covered. While all resources are accessible to bona fide researchers, access mechanisms, costs, timescales, and other considerations vary across databases.

CONCLUSION

Researchers are currently able to access primary care EHR data from several sources. Choice of data resource is likely to be driven by project needs and access considerations. The landscape of data resources based on primary care EHRs in the UK continues to evolve.

Does depression in mid-life predispose to greater cognitive decline in later life in the Whitehall II cohort?

BACKGROUND

Later-life depression appears to have different symptomatology and possibly underlying pathology to younger adults. Depression is linked to dementia but whether it is a risk factor or an early sign of dementia remains unclear. Neuroinflammation is increasingly recognised in both conditions.

AIMS

To investigate the link between depression, inflammation and dementia. We hypothesised that recurrent depression increases the rate of cognitive decline in older adults and that this effect is modified by anti-inflammatory medication.

METHODS

We used data from Whitehall II including cognitive test results and reliable measures to assess depression. Depression was defined as a self-reported diagnosis or a score of ≥20 on the CESD. The presence/absence of inflammatory illness was assessed using a standardised list of inflammatory conditions. Individuals with dementia, chronic neurological and psychotic conditions were excluded. Logistic and linear regression was used to examine the effect of depression on cognitive test performance and the effect of chronic inflammation.

LIMITATIONS

Lack of clinical diagnoses of depression.

RESULTS

There were 1063 individuals with and 2572 without depression. Depression did not affect deterioration in episodic memory, verbal fluency or the AH4 test at 15-year follow up. We found no evidence of an effect of anti-inflammatory medication. Depressed individuals had worse cross-sectional performance on the Mill Hill test and tests of abstract reasoning and verbal fluency at both baseline and 15-year follow-up.

CONCLUSIONS

Using a UK based study with a long follow-up interval we have shown that depression in individuals aged >50 is not associated with increased cognitive decline.

Exploring causality from observational data: An example assessing whether religiosity promotes cooperation

Causal inference from observational data is notoriously difficult, and relies upon many unverifiable assumptions, including no confounding or selection bias. Here, we demonstrate how to apply a range of sensitivity analyses to examine whether a causal interpretation from observational data may be justified. These methods include: testing different confounding structures (as the assumed confounding model may be incorrect), exploring potential residual confounding and assessing the impact of selection bias due to missing data. We aim to answer the causal question 'Does religiosity promote cooperative behaviour?' as a motivating example of how these methods can be applied. We use data from the parental generation of a large-scale (n = approximately 14,000) prospective UK birth cohort (the Avon Longitudinal Study of Parents and Children), which has detailed information on religiosity and potential confounding variables, while cooperation was measured via self-reported history of blood donation. In this study, there was no association between religious belief or affiliation and blood donation. Religious attendance was positively associated with blood donation, but could plausibly be explained by unmeasured confounding. In this population, evidence that religiosity causes blood donation is suggestive, but rather weak. These analyses illustrate how sensitivity analyses can aid causal inference from observational research.

Bias from questionnaire invitation and response in COVID-19 research: an example using ALSPAC

Background: Longitudinal studies are crucial for identifying potential risk factors for infection with, and consequences of, COVID-19, but relationships can be biased if they are associated with invitation and response to data collection. We describe factors relating to questionnaire invitation and response in COVID-19 questionnaire data collection in a multigenerational birth cohort (the Avon Longitudinal Study of Parents and Children, ALSPAC). Methods: We analysed online questionnaires completed between the beginning of the pandemic and easing of the first UK lockdown by participants with valid email addresses who had not actively disengaged from the study. We assessed associations of pre-pandemic sociodemographic, behavioural, anthropometric and health-related factors with: i) being sent a questionnaire; ii) returning a questionnaire; and iii) item response (for specific questions). Analyses were conducted in three cohorts: the index children born in the early 1990s (now young adults; 41 variables assessed), their mothers (35 variables) and the mothers' partners (27 variables). Results: Of 14,849 young adults, 41% were sent a questionnaire, of whom 57% returned one. Item response was >95%. In this cohort, 78% of factors were associated with being sent a questionnaire, 56% with returning one, and, as an example of item response, 20% with keyworker status response. For instance, children from mothers educated to degree-level had greater odds of being sent a questionnaire (OR=5.59; 95% CI=4.87-6.41), returning one (OR=1.60; 95% CI=1.31-1.95), and responding to items (e.g., keyworker status OR=1.65; 95% CI=0.88-3.04), relative to children from mothers with fewer qualifications. Invitation and response rates and associations were similar in all cohorts. Conclusions: These results highlight the importance of considering potential biases due to non-response when using longitudinal studies in COVID-19 research and interpreting results. We recommend researchers report response rates and factors associated with invitation and response in all COVID-19 observational research studies, which can inform sensitivity analyses.

Recurrent depression has persistent effects on cognition but this does not appear to be mediated by neuroinflammation

BACKGROUND

Later-life depression appears to be different to depression in younger adults. The underlying pathology may also differ. Depression is linked to dementia but whether it is a risk factor or an early sign of a developing dementia remains unclear. Neuroinflammation is increasingly recognised in both depression and Alzheimer's Disease.

AIMS

To investigate the link between depression, inflammation and dementia. We hypothesised that recurrent depression has adverse effects on performance in cognitive tests in middle to older age and that this effect is modified by anti-inflammatory medication.

METHODS

We identified UK based cohort studies which included individuals aged >50, had medical information, results from detailed cognitive testing and had used reliable measures to assess depression. Individuals with recurrent depression had ≥ 2 episodes of depression. Controls had no history of depression. The presence/absence of inflammatory illness was assessed using a standardised list of inflammatory conditions. Individuals with dementia, chronic neurological and psychotic conditions were excluded. Logistic and linear regression were used to examine the effect of depression on cognitive test performance and the mediating effect of chronic inflammation.

RESULTS

Unexpectedly in both studies there was evidence that those with recurrent depression performed better in some cognitive tasks (e.g Mill Hill vocabulary) but worse in others (e.g. reaction time). In UK Biobank there was no evidence that anti-inflammatories moderated this effect.

LIMITATIONS

Cross-sectional assessment of cognition.

CONCLUSIONS

Although previous recurrent depression has small effects on cognitive test performance this does not appear to be mediated by chronic inflammatory disease.

Imputing cognitive impairment in SPARK, a large autism cohort

Diverse large cohorts are necessary for dissecting subtypes of autism, and intellectual disability is one of the most robust endophenotypes for analysis. However, current cognitive assessment methods are not feasible at scale. We developed five commonly used machine learning models to predict cognitive impairment (FSIQ<80 and FSIQ<70) and FSIQ scores among 521 children with autism using parent-reported online surveys in SPARK, and evaluated them in an independent set (n = 1346) with a missing data rate up to 70%. We assessed accuracy, sensitivity, and specificity by comparing predicted cognitive levels against clinical IQ data. The elastic-net model has good performance (AUC = 0.876, sensitivity = 0.772, specificity = 0.803) using 129 predictive features to impute cognitive impairment (FSIQ<80). Top-ranked predictive features included parent-reported language and cognitive levels, age at autism diagnosis, and history of services. Prediction of FSIQ<70 and FSIQ scores also showed good performance. We show cognitive levels can be imputed with high accuracy for children with autism, using commonly collected parent-reported data and standardized surveys. The current model offers a method for large-scale autism studies seeking estimates of cognitive ability when standardized psychometric testing is not feasible. LAY SUMMARY: Children with autism who have more severe learning challenges or cognitive impairment have different needs that are important to consider in research studies. When children in our study were missing standardized cognitive testing scores, we were able to use machine learning with other information to correctly "guess" when they have cognitive impairment about 80% of the time. We can use this information in research in the future to develop more appropriate treatments for children with autism and cognitive impairment.

Bias from questionnaire invitation and response in COVID-19 research: an example using ALSPAC

Background: Longitudinal studies are crucial for identifying potential risk factors for infection with, and consequences of, COVID-19, but relationships can be biased if they are associated with invitation and response to data collection. We describe factors relating to questionnaire invitation and response in COVID-19 questionnaire data collection in a multigenerational birth cohort (the Avon Longitudinal Study of Parents and Children, ALSPAC). Methods: We analysed online questionnaires completed between the beginning of the pandemic and easing of the first UK lockdown by participants with valid email addresses who had not actively disengaged from the study. We assessed associations of pre-pandemic sociodemographic, behavioural, anthropometric and health-related factors with: i) being sent a questionnaire; ii) returning a questionnaire; and iii) item response (for specific questions). Analyses were conducted in three cohorts: the index children born in the early 1990s (now young adults; 41 variables assessed), their mothers (35 variables) and the mothers' partners (27 variables). Results: Of 14,849 young adults, 41% were sent a questionnaire, of whom 57% returned one. Item response was >95%. In this cohort, 78% of factors were associated with being sent a questionnaire, 56% with returning one, and, as an example of item response, 20% with keyworker status response. For instance, children from mothers educated to degree-level had greater odds of being sent a questionnaire (OR=5.59; 95% CI=4.87-6.41), returning one (OR=1.60; 95% CI=1.31-1.95), and responding to items (e.g., keyworker status OR=1.65; 95% CI=0.88-3.04), relative to children from mothers with fewer qualifications. Invitation and response rates and associations were similar in all cohorts. Conclusions: These results highlight the importance of considering potential biases due to non-response when using longitudinal studies in COVID-19 research and interpreting results. We recommend researchers report response rates and factors associated with invitation and response in all COVID-19 observational research studies, which can inform sensitivity analyses.

Factors associated with participation over time in the Avon Longitudinal Study of Parents and Children: a study using linked education and primary care data

Background

In observational research, choosing an optimal analysis strategy when variables are incomplete requires an understanding of the factors associated with ongoing participation and non-response, but this cannot be fully examined with incomplete data. Linkage to external datasets provides additional information on those with incomplete data, allowing examination of factors related to missingness.

Methods

We examined the association between baseline sociodemographic factors and ongoing participation in the Avon Longitudinal Study of Parents and Children. We investigated whether child and adolescent outcomes measured in linked education and primary care data were associated with participation, after accounting for baseline factors. To demonstrate the potential for bias, we examined whether the association between maternal smoking and these outcomes differed in the subsample who completed the 19-year questionnaire.

Results

Lower levels of school attainment, lower general practitioner (GP) consultation and prescription rates, higher body mass index (BMI), special educational needs (SEN) status, not having an asthma diagnosis, depression and being a smoker were associated with lower participation after adjustment for baseline factors. For example, the adjusted odds ratio (OR) for participation comparing ever smokers (by 18 years) with non-smokers was: 0.65, 95% CI (0.56, 0.75). The associations with maternal smoking differed between the subsample of participants at 19 years and the entire sample, although differences were small and confidence intervals overlapped. For example: for SEN status, OR = 1.19 (1.06, 1.33) (all participants); OR = 1.03 (0.79, 1.45) (subsample).

Conclusions

A range of health-related and educational factors are associated with ongoing participation in ALSPAC; this is likely to be the case in other cohort studies. Researchers need to be aware of this when planning their analysis. Cohort studies can use linkage to routine data to explore predictors of ongoing participation and conduct sensitivity analyses to assess potential bias.

Accounting for missing data in statistical analyses: multiple imputation is not always the answer

Background

Missing data are unavoidable in epidemiological research, potentially leading to bias and loss of precision. Multiple imputation (MI) is widely advocated as an improvement over complete case analysis (CCA). However, contrary to widespread belief, CCA is preferable to MI in some situations.

Methods

We provide guidance on choice of analysis when data are incomplete. Using causal diagrams to depict missingness mechanisms, we describe when CCA will not be biased by missing data and compare MI and CCA, with respect to bias and efficiency, in a range of missing data situations. We illustrate selection of an appropriate method in practice.

Results

For most regression models, CCA gives unbiased results when the chance of being a complete case does not depend on the outcome after taking the covariates into consideration, which includes situations where data are missing not at random. Consequently, there are situations in which CCA analyses are unbiased while MI analyses, assuming missing at random (MAR), are biased. By contrast MI, unlike CCA, is valid for all MAR situations and has the potential to use information contained in the incomplete cases and auxiliary variables to reduce bias and/or improve precision. For this reason, MI was preferred over CCA in our real data example.

Conclusions

Choice of method for dealing with missing data is crucial for validity of conclusions, and should be based on careful consideration of the reasons for the missing data, missing data patterns and the availability of auxiliary information.

Preventing bias from selective non-response in population-based survey studies: findings from a Monte Carlo simulation study

Background

Health researchers often use survey studies to examine associations between risk factors at one time point and health outcomes later in life. Previous studies have shown that missing not at random (MNAR) may produce biased estimates in such studies. Medical researchers typically do not employ statistical methods for treating MNAR. Hence, there is a need to increase knowledge about how to prevent occurrence of such bias in the first place.

Methods

Monte Carlo simulations were used to examine the degree to which selective non-response leads to biased estimates of associations between risk factors and health outcomes when persons with the highest levels of health problems are under-represented or totally missing from the sample. This was examined under different response rates and different degrees of dependency between non-response and study variables.

Results

Response rate per se had little effect on bias. When extreme values on the health outcome were completely missing, rather than under-represented, results were heavily biased even at a 70% response rate. In most situations, 50–100% of this bias could be prevented by including some persons with extreme scores on the health outcome in the sample, even when these persons were under-represented. When some extreme scores were present, estimates of associations were unbiased in several situations, only mildly biased in other situations, and became biased only when non-response was related to both risk factor and health outcome to substantial degrees.

Conclusions

The potential for preventing bias by including some extreme scorers in the sample is high (50–100% in many scenarios). Estimates may then be relatively unbiased in many situations, also at low response rates. Hence, researchers should prioritize to spend their resources on recruiting and retaining at least some individuals with extreme levels of health problems, rather than to obtain very high response rates from people who typically respond to survey studies. This may contribute to preventing bias due to selective non-response in longitudinal studies of risk factors and health outcomes.

Personal health information in research: Perceived risk, trustworthiness and opinions from patients attending a tertiary healthcare facility

BACKGROUND

Personal health information is a valuable resource to the advancement of research. In order to achieve a comprehensive reform of data infrastructure in Australia, both public engagement and building social trust is vital. In light of this, we conducted a study to explore the opinions, perceived risks and trustworthiness regarding the use of personal health information for research, in a sample of the public attending a tertiary healthcare facility.

METHODS

The Consumer Opinions of Research Data Sharing (CORDS) study was a questionnaire-based design with 249 participants who were attending a public tertiary healthcare facility located on the Gold Coast, Australia. The questionnaire was designed to explore opinions and evaluate trust and perceived risk in research that uses personal health information. Concept analysis was used to identify key dimensions of perceived risk.

RESULTS

Overall participants were supportive of research, highly likely to participate and mostly willing to share their personal health information. However, where the perceived risk of data misuse was high and trust in others was low, participants expressed hesitation to share particular types of information. Performance, physical and privacy risks were identified as key dimensions of perceived risk.

CONCLUSION

This study highlights that while participant views on the use of personal health information in research is mostly positive, where there is perceived risk in an environment of low trust, support for research decreases. The three key findings of this research are that willingness to share data is contingent upon: (i) data type; (ii) risk perception; and (iii) trust in who is accessing the data. Understanding which factors play a key role in a person's decision to share their personal health information for research is vital to securing a social license.

Data Resource: the National Pupil Database (NPD)

Introduction

The National Pupil Database (NPD) is a record-level administrative data resource curated by the UK government’s Department for Education that is used for funding purposes, school performance tables, policy making, and research.

Processes

Data are sourced from schools, exam awarding bodies, and local authorities who collect data on an on-going basis and submit to the Department for Education either termly or yearly.

Data contents

NPD contains child-level and school-level data on all pupils in state schools in England (6.6 million in the 2016/17 academic year). The primary module is the census, which has information on characteristics and school enrolment. Other modules include alternative provision, exam attainment, absence and exclusions. Data from children’s social care are also available on children referred for support and those who become looked after. Children’s records are linkable across different modules and across time using a nationally unique, anonymised child-level identifier. Linkage to external datasets has also been accomplished using child-level identifiers.

Conclusions

The NPD is an especially valuable data resource for researchers interested in the educational experience and outcomes of children and young people in England. Although limited by the fact that children in private schools or who are home schooled are not included, it provides a near-complete picture of school trajectories and outcomes for the majority of children. Linkage to other datasets can enhance analyses and provide answers to questions that would otherwise be costly, time consuming and difficult to find

The second generation of The Avon Longitudinal Study of Parents and Children (ALSPAC-G2): a cohort profile

Background: The Avon Longitudinal Study of Parents and Children-Generation 2 (ALSPAC-G2) was set up to provide a unique multi-generational cohort. It builds on the existing ALSPAC resource, which recruited 14,541 pregnancies to women resident in the South West of England who were expected to deliver between 01/04/1991 and 31/12/1992. Those women and their partners (Generation 0; ALSPAC-G0) and their offspring (ALSPAC-G1) have been followed for the last 26 years. This profile describes recruitment and data collection on the next generation (ALSPAC-G2)-the grandchildren of ALSPAC-G0 and children of ALSPAC-G1. Recruitment: Recruitment began on the 6 th of June 2012 and we present details of recruitment and participants up to 30 th June 2018 (~6 years). We knew at the start of recruitment that some ALSPAC-G1 participants had already become parents and ALSPAC-G2 is an open cohort; we recruit at any age. We hope to continue recruiting until all ALSPAC-G1 participants have completed their families. Up to 30 th June 2018 we recruited 810 ALSPAC-G2 participants from 548 families. Of these 810, 389 (48%) were recruited during their mother's pregnancy, 287 (35%) before age 3 years, 104 (13%) between 3-6 years and 30 (4%) after 6 years. Over 70% of those invited to early pregnancy, late pregnancy, second week of life, 6-, 12- and 24-month assessments (whether for their recruitment, or a follow-up, visit) have attended, with attendance being over 60% for subsequent visits up to 7 years (to few are eligible for the 9- and 11-year assessments to analyse). Data collection: We collect a wide-range of social, lifestyle, clinical, anthropometric and biological data on all family members repeatedly. Biological samples include blood (including cord-blood), urine, meconium and faeces, and placental tissue. In subgroups detailed data collection, such as continuous glucose monitoring and videos of parent-child interactions, are being collected.

The second generation of The Avon Longitudinal Study of Parents and Children (ALSPAC-G2): a cohort profile

Background: The Avon Longitudinal Study of Parents and Children-Generation 2 (ALSPAC-G2) was set up to provide a unique multi-generational cohort. It builds on the existing ALSPAC resource, which recruited 14,541 pregnancies to women resident in the South West of England who were expected to deliver between 01/04/1991 and 31/12/1992. Those women and their partners (Generation 0; ALSPAC-G0) and their offspring (ALSPAC-G1) have been followed for the last 26 years. This profile describes recruitment and data collection on the next generation (ALSPAC-G2)-the grandchildren of ALSPAC-G0 and children of ALSPAC-G1. Recruitment: Recruitment began on the 6 ^th of June 2012 and we present details of recruitment and participants up to 30 ^th June 2018 (~6 years). We knew at the start of recruitment that some ALSPAC-G1 participants had already become parents and ALSPAC-G2 is an open cohort; we recruit at any age. We hope to continue recruiting until all ALSPAC-G1 participants have completed their families. Up to 30 ^th June 2018 we recruited 810 ALSPAC-G2 participants from 548 families. Of these 810, 389 (48%) were recruited during their mother's pregnancy, 287 (35%) before age 3 years, 104 (13%) between 3-6 years and 30 (4%) after 6 years. Over 70% of those invited to early pregnancy, late pregnancy, second week of life, 6-, 12- and 24-month assessments (whether for their recruitment, or a follow-up, visit) have attended, with attendance being over 60% for subsequent visits up to 7 years (to few are eligible for the 9- and 11-year assessments to analyse). Data collection: We collect a wide-range of social, lifestyle, clinical, anthropometric and biological data on all family members repeatedly. Biological samples include blood (including cord-blood), urine, meconium and faeces, and placental tissue. In subgroups detailed data collection, such as continuous glucose monitoring and videos of parent-child interactions, are being collected.

Comparison of statistical approaches for analyzing incomplete longitudinal patient-reported outcome data in randomized controlled trials

Purpose

Missing data are a potential source of bias in the results of RCTs, but are often unavoidable in clinical research, particularly in patient-reported outcome measures (PROMs). Maximum likelihood (ML), multiple imputation (MI), and inverse probability weighting (IPW) can be used to handle incomplete longitudinal data. This paper compares their performance when analyzing PROMs, using a simulation study based on an RCT data set.

Methods

Realistic missing-at-random data were simulated based on patterns observed during the follow-up of the knee arthroscopy trial (ISRCTN45837371). Simulation scenarios covered different sample sizes, with missing PROM data in 10%–60% of participants. Monotone and nonmonotone missing data patterns were considered. Missing data were addressed by using ML, MI, and IPW and analyzed via multilevel mixed-effects linear regression models. Root mean square errors in the treatment effects were used as performance parameters across 1,000 simulations.

Results

Nonconvergence issues were observed for IPW at small sample sizes. The performance of all three approaches worsened with decreasing sample size and increasing proportions of missing data. MI and ML performed similarly when the MI model was restricted to baseline variables, but MI performed better when using postrandomization data in the imputation model and also in nonmonotone versus monotone missing data scenarios. IPW performed worse than ML and MI in all simulation scenarios.

Conclusion

When additional postrandomization information is available, MI can be beneficial over ML for handling incomplete longitudinal PROM data. IPW is not recommended for handling missing PROM data in the simulated scenarios.

On the use of the not-at-random fully conditional specification (NARFCS) procedure in practice

The not‐at‐random fully conditional specification (NARFCS) procedure provides a flexible means for the imputation of multivariable missing data under missing‐not‐at‐random conditions. Recent work has outlined difficulties with eliciting the sensitivity parameters of the procedure from expert opinion due to their conditional nature. Failure to adequately account for this conditioning will generate imputations that are inconsistent with the assumptions of the user.

In this paper, we clarify the importance of correct conditioning of NARFCS sensitivity parameters and develop procedures to calibrate these sensitivity parameters by relating them to more easily elicited quantities, in particular, the sensitivity parameters from simpler pattern mixture models. Additionally, we consider how to include the missingness indicators as part of the imputation models of NARFCS, recommending including all of them in each model as default practice.

Algorithms are developed to perform the calibration procedure and demonstrated on data from the Avon Longitudinal Study of Parents and Children, as well as with simulation studies.

Continually improving standards of care: The UK Renal Registry as a translational public health tool

A disease registry uses observational study methods to collect defined data on patients with a particular condition for a predetermined purpose. By providing comprehensive standardised data on patients with kidney disease, renal registries aim to provide a 'real world' representation of practice patterns, treatment and patient outcomes that may not be captured accurately by other methods, including randomised controlled trials. Additionally, using registries to measure variations in outcomes and audit care against standards is crucial to understanding how to improve quality of care for patients in an efficacious and cost-effective manner. Registries also have the potential to be a powerful scientific tool that can monitor and support the translational process between research and routine clinical practice, although their limitations must be borne in mind. In this review, we describe the role of the UK Renal Registry as a tool to support translational research. We describe its involvement across each stage of the translational pathway: from hypothesis generation, study design and data collection, to reporting of long-term outcomes and quality improvement initiatives. Furthermore we explore how this role may bring about improvements in care for adults and children with kidney disease.

Multiple imputation using linked proxy outcome data resulted in important bias reduction and efficiency gains: a simulation study

Background

When an outcome variable is missing not at random (MNAR: probability of missingness depends on outcome values), estimates of the effect of an exposure on this outcome are often biased. We investigated the extent of this bias and examined whether the bias can be reduced through incorporating proxy outcomes obtained through linkage to administrative data as auxiliary variables in multiple imputation (MI).

Methods

Using data from the Avon Longitudinal Study of Parents and Children (ALSPAC) we estimated the association between breastfeeding and IQ (continuous outcome), incorporating linked attainment data (proxies for IQ) as auxiliary variables in MI models. Simulation studies explored the impact of varying the proportion of missing data (from 20 to 80%), the correlation between the outcome and its proxy (0.1-0.9), the strength of the missing data mechanism, and having a proxy variable that was incomplete.

Results

Incorporating a linked proxy for the missing outcome as an auxiliary variable reduced bias and increased efficiency in all scenarios, even when 80% of the outcome was missing. Using an incomplete proxy was similarly beneficial. High correlations (> 0.5) between the outcome and its proxy substantially reduced the missing information. Consistent with this, ALSPAC analysis showed inclusion of a proxy reduced bias and improved efficiency. Gains with additional proxies were modest.

Conclusions

In longitudinal studies with loss to follow-up, incorporating proxies for this study outcome obtained via linkage to external sources of data as auxiliary variables in MI models can give practically important bias reduction and efficiency gains when the study outcome is MNAR.

The Cumulative Effect of Health Adversities on Children's Later Academic Achievement

OBJECTIVE

We aimed to determine whether the accumulation of physical, psychosocial, and combined health adversities measured at age 8 to 9 years predicts worsening of academic scores cross-sectionally at 8 to 9 and longitudinally at 10 to 11 years.

METHODS

Design: Longitudinal data from Waves 3 and 4 in the Longitudinal Study of Australian Children (83% of 4983 retained). Exposures (8-9 years): Physical health adversities (yes/no; summed range, 0-5): overweight, special health care needs, chronic illness, PedsQL Physical, and global health. Psychosocial health adversities (yes/no; summed range, 0-4): parent- and teacher-reported behavior, PedsQL Psychosocial, sleep problems. Combined health adversities (range 0-9). Outcomes (8-9, and 10-11 years): National academic standardized test scores.

ANALYSIS

Generalized estimating equations, accounting for multiple academic domains in each year and socioeconomic position and cognition.

RESULTS

At 8 to 9 years, 23.9%, 9.9%, and 5.3% had 1, 2, or ≥3 physical health adversities, respectively, while 27.2%, 9.5%, and 4.9% had 1, 2, or ≥3 psychosocial health adversities. For each additional health adversity at 8 to 9 years, academic scores fell incrementally in year 3 and year 5 (both P < .001), with reductions of at least 0.4 SDs for ≥3 health adversities. Number was more important than type (physical, psychosocial) of adversity.

CONCLUSIONS

The accumulation of health adversities predicts poorer academic achievement up to 2 years later. Interventions might need to address multiple domains to improve child academic outcomes and be delivered across the health-education interface.

Young people's views about the purpose and composition of research ethics committees: findings from the PEARL qualitative study

BACKGROUND

Avon Longitudinal Study of Parents and Children (ALSPAC) is a birth cohort study within which the Project to Enhance ALSPAC through Record Linkage (PEARL) was established to enrich the ALSPAC resource through linkage between ALSPAC participants and routine sources of health and social data. PEARL incorporated qualitative research to seek the views of young people about data linkage, including their opinions about appropriate safeguards and research governance. In this paper we focus on views expressed about the purpose and composition of research ethics committees.

METHODS

Digitally recorded interviews were conducted with 48 participants aged 17-19 years. Participants were asked about whether medical research should be monitored and controlled, their knowledge of research ethics committees, who should sit on these committees and what their role should be. Interview recordings were fully transcribed and anonymised. Thematic analysis was undertaken, assisted by the Framework approach to data management.

RESULTS

The majority of interviewees had little or no specific knowledge of ethics committees. Once given basic information about research ethics committees, only three respondents suggested there was no need for such bodies to scrutinise research. The key tasks of ethics committees were identified as monitoring the research process and protecting research participants. The difficulty of balancing the potential to inhibit research against the need to protect research participants was acknowledged. The importance of relevant research and professional expertise was identified but it was also considered important to represent wider public opinion, and to counter the bias potentially associated with self-selection possibly through a selection process similar to 'jury duty'.

CONCLUSIONS

There is a need for more education and public awareness about the role and composition of research ethics committees. Despite an initial lack of knowledge, interviewees were able to contribute their ideas and balance the rights of individuals with the wider benefits from research. The suggestion that public opinion should be represented through random selection similar to jury duty may be worth pursuing in the light of the need to ensure diversity of opinion and establish trust amongst the general public about the use of 'big data' for the wider public good.

Young people's views about consenting to data linkage: findings from the PEARL qualitative study

BACKGROUND

Electronic administrative data exist in several domains which, if linked, are potentially useful for research. However, benefits from data linkage should be considered alongside risks such as the threat to privacy. Avon Longitudinal Study of Parents and Children (ALSPAC) is a birth cohort study. The Project to Enhance ALSPAC through Record Linkage (PEARL) was established to enrich the ALSPAC resource through linkage between ALSPAC participants and routine sources of health and social data. Qualitative research was incorporated in the PEARL study to examine participants' views about data linkage and inform approaches to information sharing. This paper focusses on issues of consent.

METHODS

Digitally recorded interviews were conducted with 55 participants aged 17-19 years. Terms and processes relating to consent, anonymization and data linkage were explained to interviewees. Scenarios were used to prompt consideration of linking different sources of data, and whether consent should be requested. Interview recordings were fully transcribed. Thematic analysis was undertaken using the Framework approach.

RESULTS

Participant views on data linkage appeared to be most influenced by: considerations around the social sensitivity of the research question, and; the possibility of tangible health benefits in the public interest. Some participants appeared unsure about the effectiveness of anonymization, or did not always view effective anonymization as making consent unnecessary. This was related to notions of ownership of personal information and etiquette around asking permission for secondary use. Despite different consent procedures being explained, participants tended to equate consent with 'opt-in' consent through which participants are 'asked' if their data can be used for a specific study. Participants raising similar concerns came to differing conclusions about whether consent was needed. Views changed when presented with different scenarios, and were sometimes inconsistent.

CONCLUSIONS

Findings from this study question the validity of 'informed consent' as a cornerstone of good governance, and the extent to which potential research participants understand different types of consent and what they are consenting, or not consenting, to. Pragmatic, imaginative and flexible approaches are needed if research using data linkage is to successfully realise its potential for public good without undermining public trust in the research process.