ABSTRACT
ObjectivesThe effective exploitation of what are often called big data is increasingly important. They provide the evidence in evidence-based health care and underpin scientific progress in many domains including social/economic policy. Typically, an optimal analysis involves working directly with microdata; i.e. the detailed data relating to each individual in the dataset. But there are many ethico-legal and other governance restrictions on physically sharing microdata. Furthermore, researchers or institutions may have an extensive intellectual property investment in complex microdata and although keen for other researchers to analyse their data they may not wish to give them a physical copy. These restrictions can discourage the use of optimum approaches to analysing pivotal data and slow scientific progress. Data science groups across the world are exploring privacy-protected approaches to analysing microdata without having to physically share the data.
ApproachA two day international workshop was arranged focussing on privacy protected approaches to data analysis – particularly federated analysis where raw data remain at their original site of collection. The workshop considered the range of approaches that exist, and those that are currently being developed. It explored the strengths, weaknesses, opportunities and challenges associated with these methods and identified situations where specific approaches have a particularly important role. The workshop included a number of practical sessions where potential users could watch demonstrations of the various approaches in action and run analyses themselves.
ResultsThe Data Analysis with Privacy Protection for Epidemiological Research (DAPPER) workshop was held 22-23rd August 2016, Bristol. We report back to the broader community on the outcomes of this workshop that focussed on exploring current approaches, tools and technical solutions that facilitate sensitive data to be shared and analysed.
ConclusionsThe workshop has helped map out key opportunities and challenges and assisted potential users, developers and other stakeholders (e.g. funders/journals) to recognise the strengths and weaknesses of different privacy protected analytic approaches. The workshop will encourage further methodological work in this field and better informed application of existing methods.
IntroductionChildhood infection is a leading cause of morbidity and mortality worldwide, however, the epidemiology of infection-related hospitalisations (IRH) across high-income countries is not well described. Population-level data are valuable resources for studying the epidemiology of severe infections. Objectives and ApproachWe used data from our multi-country total population-based cohort study to describe the patterns of IRH across six populations in five countries. Our cohort study contains birth and hospitalisation data on all singleton live births from 1996-2015 from Australia (New South Wales and Western Australia), Denmark, Norway, Scotland, and England. Children were classified as having an IRH if they had an inpatient hospital admission that incurred at least one primary or secondary infectious disease discharge ICD code, at least one day after the birth-related discharge date and if they were less than five years of age at discharge. IRH were classified as overall and by clinical group. Here we present interim analyses from Denmark and Scotland, n=1,593,008 (further results will be presented at the conference). ResultsMore boys than girls had an IRH by 5 years of age (boys 26%, girls 21%). By 1 year, 12% of boys and 9% of girls experienced their first IRH, whereas between 1-5 years of age, 14% of boys and 12% of girls experienced their first IRH. Overall, 7% of children had >1 IRH. The majority of infections were lower and upper respiratory tract infections, followed by viral and gastrointestinal infections. Infection was commoner in the lowest socio-economic status groups. Conclusion / ImplicationsIRH remains a leading cause of hospitalisation in preschool children. Understanding the epidemiology of IRH in high-income countries is important for targeting appropriate interventions and reducing disease burden.