Research methods and evidence use
Over its seven centres, SRI carries out research that is regularly used to inform decisions in policy and practice. Here we describe the methods we use to do this, and what ‘evidence use’ means.

Research methods
SRI is a transdisciplinary department utilising a wide range of research methods to achieve its objectives. It uses established methods from qualitative and quantitative traditions but also develops new approaches when necessary. Its research methods include: evidence synthesis; evidence use; access to syntheses of research evidence; guidance, supporting and studying research use; survey methodology; and statistical methodology.
Evidence synthesis
An area of active development for more than two decades now has been in the methods used to synthesise research evidence – i.e., to understand what multiple research studies are saying about a given issue. Evidence synthesis is most often used for informing decisions in policy and practice and often involves combining research from different traditions and using different methods. Our methods developments have included:
- thematic synthesis for combining qualitative research studies
- ‘mixed methods’ synthesis for combining qualitative and quantitative studies in the same analysis
- involving stakeholders in research and in setting the research agenda
- using machine learning and other information technologies to make evidence synthesis more efficient.
Evidence use
When most people think about research use, they think about a linear process where research evidence informs decision-making directly. However, there is growing recognition that research use is a complex social process consisting of several multi-directional elements. A further insight is that research evidence can ‘inform’ policy and practice in many ways. It can be used in everyday practice alongside other forms of knowledge, such as that from professional or personal experience. It can be used directly to inform policy and/or practice, where it is seen as having a direct, instrumental influence on behaviour change. In addition, it is understood that there are indirect, more conceptual or enlightened, uses of research in shaping knowledge, understanding and attitudes. There is also the crucial issue of how research users are engaged in informing the setting of research agendas and approaches to research. Although not exhaustive, an overview of the different types of research use is shown below:
SRI is interested in all types of research use. We are working in the following areas.
Access to syntheses of research evidence
Reviews published by the EPPI-Centre are freely available through our Evidence Library. When conducting or supporting the conduct of reviews, we invite outsiders to focus our efforts, in order to make the reviews relevant to those asking the review questions, those well placed to make use of the findings, and those whose lives may be affected by any subsequent decisions.
Guidance on research use
As part of our commitment to capacity-building activities to help ensure that those who want and need research can find, understand and use it effectively, we have produced a number of introductory guides for postgraduate researchers and academic staff to assist the understanding of terms such as ‘research impact’ and ‘knowledge exchange’ and the key issues within them.
Supporting research use
Building on our coordination of two European Commission-funded projects in this area, our work includes supporting those who wish to use research by providing direct support services. We currently offer a Research Advisory Service (RSA) to enable users of research to consider whether research might be helpful to their decision-making, and, if so, in what way and what type of research would meet that need.
Studying research use
A growing area of interest is the study of research use in policy and practice, including the processes, structures and systems (both formal and informal) that shape this use. We believe that research can assist decision-making; in which case, how best to do this is itself an important research question. We need to be ‘evidence-informed’ about how we use research evidence!
Survey methodology
The Centre for Longitudinal Studies is home to four national longitudinal cohort studies, which follow the lives of tens of thousands of people. The design of our studies is underpinned by the latest evidence and our research in longitudinal survey methods research is internationally recognised.
Our key research areas are summarised below:
- Data collection mode – The mode of data collection can impact how participants respond. Our studies have primarily been conducted face-to-face but we have used the web and telephone – and in two current surveys have used video interviewing for the first time. We conduct research into the implications of mode choice in surveys on measurement and data quality and how we can make the most effective use of the web and maximise participation.
- New technologies and innovations in data collection – New technologies such as apps and wearables can be used to enhance surveys with new types of data. Our studies have, for example, used apps to collect ‘time-use’ data and accelerometers to measure physical activity. We research how to make optimal use of new technologies for research while understanding their limitations and overcoming their challenges.
- Questionnaire design – Questionnaire design choices impact upon measurement and data quality. Our research explores how questionnaires for longitudinal studies should be optimally designed.
- Record linkage – Enhancing survey data through linkage to administrative records (e.g. health records, economic records) is hugely beneficial for research but how should surveys be designed to maximise consent and which factors determine whether cohort members agree to these permissions?
- Participant tracking and engagement – Keeping track of study members who move, and keeping them engaged over time, is a crucial aim for all longitudinal studies. We conduct research into how to optimally do this.
- Combining social and biomedical data collection – Combining objective health measures with survey data enables researchers to further understand the interplay between social and biological factors in explaining human behaviour. Our research explores how to optimally include biomedical data collection in longitudinal surveys, and the impact of different approaches on cost, response rates and measurement.
- Surveying children and young people – Our work on the MCS in particular has pushed forward best practices in surveying children and young people – including questionnaire design, engagement and ethical approaches.
Statistical methodology
The Centre for Longitudinal Studies "applied statistical methods" research programme supports and enables users of the four national longitudinal cohort studies to tackle some of the important challenges in using longitudinal data. We bring together ideas and methods from a number of disciplines, such as statistics, econometrics, psychometrics, epidemiology and computer science.
Our main areas of interest are summarised below:
- Non-response and missing data – Some non-response is inevitable in longitudinal studies. We know different types of people tend to drop out of our studies at different times, depending on their individual circumstances and characteristics, and we can take account of this in the methods we use. We have developed approaches which capitalise on the rich data cohort members provided over the years, before they left the study, in order to deal with missing data and reduce bias using well-known methods such as multiple imputation, inverse probability weighting and full information maximum likelihood. We have implemented systematic data-driven approaches to identify predictors of non-response which can be used in such analyses.
- Causal inference – Making causal inferences using observational data is not straightforward. However, the wealth of information we have collected about cohort members across the life course gives data users the opportunity to select rich controls for multivariable adjustment. There are also circumstances in which causal inferences may be made using approaches such as instrumental variable modelling/Mendelian randomisation, regression discontinuity, and fixed effects/correlated random effects methods. Sensitivity analysis methods such as negative controls and falsification tests also allow quantification of potential bias due to unobserved heterogeneity and residual confounding. We are developing a programme of translational work to demonstrate the application of these methods within the context of the CLS cohorts and provide detailed guidance to users in applying these in their own research.
- Measurement error – Data from self-reported measures can be biased due to processes driven by cohort members’ personalities and circumstances. Data from objective measures may also be affected by instrumental errors, for example, the precision of the blood pressure device used by the nurse. Additional sources of error arise when comparing data from multiple studies as there can be variation in how different groups interpret the same question and in response tendencies. In our work on measurement error, we use the latest extensions of the generalised latent variable modelling framework to specify complex error structures. This allows us to investigate the properties of some key areas of measurement of our cohort data, including physical health, mental health and cognition, and to establish within and between cohort equivalent measures.
- Mode effects – Data are increasingly being collected using different modes (for example, face-to-face, telephone, Web), either during the same sweep of data collection (mixed mode) or across different sweeps of data collection in longitudinal studies (mode discontinuity). Mode effects – differences in observed responses which are due solely to the mode of data collection – may limit the ability to compare data from different modes in a meaningful way. We are working on methods to aid in the detection and handling of mode effects in longitudinal data.