International Journal of Population Data Science最新文献

Introduction: "Big data" - including linked administrative data - can be exploited to evaluate interventions for maternal and child health, providing time- and cost-effective alternatives to randomised controlled trials. However, using these data to evaluate population-level interventions can be challenging.

Objectives: We aimed to inform future evaluations of complex interventions by describing sources of bias, lessons learned, and suggestions for improvements, based on two observational studies using linked administrative data from health, education and social care sectors to evaluate the Family Nurse Partnership (FNP) in England and Scotland.

Methods: We first considered how different sources of potential bias within the administrative data could affect results of the evaluations. We explored how each study design addressed these sources of bias using maternal confounders captured in the data. We then determined what additional information could be captured at each step of the complex intervention to enable analysts to minimise bias and maximise comparability between intervention and usual care groups, so that any observed differences can be attributed to the intervention.

Results: Lessons learned include the need for i) detailed data on intervention activity (dates/geography) and usual care; ii) improved information on data linkage quality to accurately characterise control groups; iii) more efficient provision of linked data to ensure timeliness of results; iv) better measurement of confounding characteristics affecting who is eligible, approached and enrolled.

Conclusions: Linked administrative data are a valuable resource for evaluations of the FNP national programme and other complex population-level interventions. However, information on local programme delivery and usual care are required to account for biases that characterise those who receive the intervention, and to inform understanding of mechanisms of effect. National, ongoing, robust evaluations of complex public health evaluations would be more achievable if programme implementation was integrated with improved national and local data collection, and robust quasi-experimental designs.

Introduction: Research to date has established that the COVID-19 pandemic has not impacted everyone equitably. Whether this unequitable impact was seen educationally with regards to educator reported barriers to distance learning, concerns and mental health is less clear.

Objective: The objective of this study was to explore the association between the neighbourhood composition of the school and kindergarten educator-reported barriers and concerns regarding children's learning during the first wave of COVID-19 related school closures in Ontario, Canada.

Methods: In the spring of 2020, we collected data from Ontario kindergarten educators (n = 2569; 74.2% kindergarten teachers, 25.8% early childhood educators; 97.6% female) using an online survey asking them about their experiences and challenges with online learning during the first round of school closures. We linked the educator responses to 2016 Canadian Census variables based on schools' postal codes. Bivariate correlations and Poisson regression analyses were used to determine if there was an association between neighbourhood composition and educator mental health, and the number of barriers and concerns reported by kindergarten educators.

Results: There were no significant findings with educator mental health and school neighbourhood characteristics. Educators who taught at schools in neighbourhoods with lower median income reported a greater number of barriers to online learning (e.g., parents/guardians not submitting assignments/providing updates on their child's learning) and concerns regarding the return to school in the fall of 2020 (e.g., students' readjustment to routines). There were no significant associations with educator reported barriers or concerns and any of the other Census neighbourhood variables (proportion of lone parent families, average household size, proportion of population that do no speak official language, proportion of population that are recent immigrants, or proportion of population ages 0-4).

Conclusions: Overall, our study suggests that the neighbourhood composition of the children's school location did not exacerbate the potential negative learning experiences of kindergarten students and educators during the COVID-19 pandemic, although we did find that educators teaching in schools in lower-SES neighbourhoods reported more barriers to online learning during this time. Taken together, our study suggests that remediation efforts should be focused on individual kindergarten children and their families as opposed to school location.

Introduction: Population wide educational attainment registers are necessary for educational planning and research. Regular linking of databases is needed to build and update such a register. Without availability of unique national identification numbers, record linkage must be based on quasi-identifiers such as name, date of birth and sex. However, the data protection principle of data minimization aims to minimize the set of identifiers in databases.

Objectives: Therefore, the German Federal Ministry of Research and Education commissioned a study to inform legislation on the minimum set of identifiers required for a national educational register.

Methods: To justify our recommendations empirically, we implemented a microsimulation of about 20 million people. The simulated register accumulates changes and errors in identifiers due to migration, regional mobility, marriage, school career and mortality, thereby allowing the study of errors on longitudinal datasets. Updated records were linked yearly to the simulated register using several linkage methods. Clear-text methods as well as privacy-preserving (PPRL) methods were compared.

Results: The results indicate linkage bias if only the primary identifiers are available in the register. More detailed identifiers, including place of birth, are required to minimize linkage bias. The amount of information available to identify a person for matching is more critical for linkage quality than the record linkage method applied. Differences in linkage quality between the best procedures (probabilistic linkage and multiple matchkeys) are minor.

Conclusions: Microsimulation is a valuable tool for designing record linkage procedures. By modelling the processes resulting in changes or errors in quasi-identifiers, predicting data quality to be expected after the implementation of a register seems possible.

Introduction: The Patient Master Index (PMI) plays an important role in management of patient information and epidemiological research, and the availability of unique patient identifiers improves the accuracy when linking patient records across disparate datasets. In our environment, however, a unique identifier is seldom present in all datasets containing patient information. Quasi identifiers are used to attempt to link patient records but sometimes present higher risk of over-linking. Data quality and completeness thus affect the ability to make correct linkages.

Aim: This paper describes the record linkage system that is currently implemented at the Provincial Health Data Centre (PHDC) in the Western Cape, South Africa, and assesses its output to date.

Methods: We apply a stepwise deterministic record linkage approach to link patient data that are routinely collected from health information systems in the Western Cape province of South Africa. Variables used in the linkage process include South African National Identity number (RSA ID), date of birth, year of birth, month of birth, day of birth, residential address and contact information. Descriptive analyses are used to estimate the level and extent of duplication in the provincial PMI.

Results: The percentage of duplicates in the provincial PMI lies between 10% and 20%. Duplicates mainly arise from spelling errors, and surname and first names carry most of the errors, with the first names and surname being different for the same individual in approximately 22% of duplicates. The RSA ID is the variable mostly affected by poor completeness with less than 30% of the records having an RSA ID.The current linkage algorithm requires refinement as it makes use of algorithms that have been developed and validated on anglicised names which might not work well for local names. Linkage is also affected by data quality-related issues that are associated with the routine nature of the data which often make it difficult to validate and enforce integrity at the point of data capture.

Data collection, analysis, and data driven action cycles have been viewed as vital components of healthcare for decades. Throughout the COVID-19 pandemic, case incidence and mortality data have consistently been used by various levels of governments and health institutions to inform pandemic strategies and service distribution. However, these responses are often inequitable, underscoring pre-existing healthcare disparities faced by marginalized populations. This has prompted governments to finally face these disparities and find ways to quickly deliver more equitable pandemic support. These rapid data informed supports proved that learning health systems (LHS) could be quickly mobilized and effectively used to develop healthcare actions that delivered healthcare interventions that matched diverse populations' needs in equitable and affordable ways. Within LHS, data are viewed as a starting point researchers can use to inform practice and subsequent research. Despite this innovative approach, the quality and depth of data collection and robust analyses varies throughout healthcare, with data lacking across the quadruple aims. Often, large data gaps pertaining to community socio-demographics, patient perceptions of healthcare quality and the social determinants of health exist. This prevents a robust understanding of the healthcare landscape, leaving marginalized populations uncounted and at the sidelines of improvement efforts. These gaps are often viewed by researchers as indication that more data is needed rather than an opportunity to critically analyze and iteratively learn from multiple sources of pre-existing data. This continued cycle of data collection and analysis leaves one to wonder if healthcare has a data problem or a learning problem. In this commentary, we discuss ways healthcare data are often used and how LHS disrupts this cycle, turning data into learning opportunities that inform healthcare practice and future research in real time. We conclude by proposing several ways to make learning from data just as important as the data itself.

Introduction: Longitudinal data that tracks student achievement over many years are crucial for understanding children's learning and for guiding effective policies and interventions. Despite being Canada's most populous province, Ontario lacks such large-scale and longitudinal data on student learning. Linking datasets across cohorts requires rigorous linkage protocols, flexible handling of complex cohort structures, methods to validate linked datasets, and viable organizational partnerships. We linked administrative data on early child development and educational achievement and merged two datasets on characteristics of students' neighborhoods and schools. We developed a linkage protocol and validated how the resulting database could be generalized to Ontario's student population.

Methods and analysis: Two main individual-level data sources were linked: 1) the Early Development Instrument (EDI), a school readiness assessment of all Ontario public school kindergartners that is administered in three-year cycles, and 2) Ontario's Educational Quality and Assessment Office's (EQAO) math and reading assessments in grades 3, 6, 9, and 10. To compensate for their lack of a common personal identification number, a deterministic linkage process was developed using several administrative variables. A school-level and a neighborhood-level dataset were also later linked. We examined differences between unlinked and linked cases across several variables.

Results and implications: We successfully linked 50% of the EDI's 374,239 cases, 86,778 of which contained all five datapoints, creating a database tracking achievement for multiple cohorts from kindergarten through grade 10, with covariates for their development, demographics, affect, neighborhoods, and schools. Analyses revealed only negligible differences between linked and unlinked cases across several demographic measures, while small differences were detected across a neighborhood socioeconomic index and some measures of child development. In conclusion, we recommend the filling of key voids in sustainable research capacity by creating representative data through linkage protocols and data verification.

This article describes the linkage methods that will be used in the Decennial Census Digitization and Linkage project (DCDL), which is completing the final four decades of a longitudinal census infrastructure covering the past 170 years of United States history. DCDL is digitizing and creating linkages between nearly a billion records across the 1960 through 1990 U.S. censuses, as well as to already-linked records from the censuses of 1940, 2000, 2010, and 2020. Our main goals in this article are to (1) describe the development of the DCDL and the protocol we will follow to build the linkages between the census files, (2) outline the techniques we will use to evaluate the quality of the links, and (3) show how the assignment and evaluation of these linkages leverages the joint use of routinely collected administrative data and non-routine survey data.

Databases covering all individuals of a population are increasingly used for research and decision-making. The massive size of such databases is often mistaken as a guarantee for valid inferences. However, population data have characteristics that make them challenging to use. Various assumptions on population coverage and data quality are commonly made, including how such data were captured and what types of processing have been applied to them. Furthermore, the full potential of population data can often only be unlocked when such data are linked to other databases. Record linkage often implies subtle technical problems, which are easily missed. We discuss a diverse range of myths and misconceptions relevant for anybody capturing, processing, linking, or analysing population data. Remarkably, many of these myths and misconceptions are due to the social nature of data collections and are therefore missed by purely technical accounts of data processing. Many are also not well documented in scientific publications. We conclude with a set of recommendations for using population data.