Common Misconceptions about Population Data

Häntschel

et al. 2022

Int J Data Sci Anal

Privacy-preserving record linkage (PPRL) is the process aimed at identifying records that represent the same real-world entity across different data sources while guaranteeing the privacy of sensitive information about these entities. A popular PPRL method is to encode sensitive plain-text data into Bloom filters (BFs), bit vectors that enable the efficient calculation of similarities between records that is required for PPRL. However, BF encoding cannot completely prevent the re-identification of plain-text values because sets of BFs can contain bit patterns that can be mapped to plain-text values using cryptanalysis attacks. Various hardening techniques have therefore been proposed that modify the bit patterns in BFs with the aim to prevent such attacks. However, it has been shown that even hardened BFs can still be vulnerable to attacks. To avoid any such attacks, we propose a novel encoding technique for PPRL based on autoencoders that transforms BFs into vectors of real numbers. To achieve a high comparison quality of the generated numerical vectors, we propose a method that guarantees the comparability of encodings generated by the different data owners. Experiments on real-world data sets show that our technique achieves high linkage quality and prevents known cryptanalysis attacks on BF encoding.

Section: Introductionmentioning

confidence: 99%

Privacy-preserving record linkage using autoencoders

Häntschel

et al. 2022

Int J Data Sci Anal

“…4 O segundo desafio diz respeito à falta de familiaridade de pesquisadores e gestores com a utilização de bases de dados secundárias volumosas e complexas para fins de vigilância, avaliação e pesquisa, o que pode levar a interpretações incorretas das evidências geradas. 2,6 Ao contrário do que ocorre com dados primários coletados para resposta a uma pergunta específica de pesquisa, pesquisadores e gestores, em geral, não têm controle sobre os processos de geração e processamento dos conjuntos de dados secundários. Ao se utilizar um banco de dados secundários…”

unclassified

“…These data, especially when linked, have been increasingly used in health, research, surveillance and evaluation activities, as well as in decision-making. 2 …”

mentioning

confidence: 99%

“…The second challenge is related to the lack of familiarity of researchers and managers with the use of large and complex secondary databases for surveillance, evaluation and research purposes, which may lead to misinterpretation of the evidence generated. 2 , 6 Unlike primary data collected to answer a specific research question, researchers and managers generally have no control over the generation and processing of secondary datasets. When using a linked secondary database, it is necessary to know several aspects, including population coverage, field completeness, presence of duplicate records, proportion of missing data, reliability and validity of data, the attribute coding systems, the algorithms used for data transformation, and the data linkage process, including linking errors.…”

mentioning

confidence: 99%

“…When using a linked secondary database, it is necessary to know several aspects, including population coverage, field completeness, presence of duplicate records, proportion of missing data, reliability and validity of data, the attribute coding systems, the algorithms used for data transformation, and the data linkage process, including linking errors. 2 Additionally, taking into consideration that these data are collected in different locations and over long periods of time, it is important to evaluate whether these aspects are stable over time, and whether there are regional inequalities.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Ciência de dados populacionais

Coeli

2022

Epidemiol. Serv. Saúde

A rápida evolução das tecnologias de informação proporciona um crescimento sem precedentes da capacidade de produção, armazenamento, troca, processamento e análise de dados. 1 Além dos dados estruturados, vários formatos de dados não estruturados são hoje disponibilizados. Dados estruturados são usados em bases administrativas produzidas e utilizadas nas atividades típicas da Saúde Coletiva. Nesse tipo de base, cada linha corresponde a uma entidade, geralmente um indivíduo, e cada coluna, a um atributo dessa entidade (por exemplo, a data de nascimento). Por seu turno, os dados não estruturados apresentam diferentes formatos, nos quais se incluem, por exemplo, textos em documentos ou em redes sociais, imagens, e saídas de sensores. Esses dados, especialmente quando vinculados, vêm, cada vez mais, sendo empregados na saúde, em atividades de pesquisa, vigilância e avaliação, bem como na tomada de decisão. 2

Establishment of a birth-to-education cohort of 1 million Palestinian refugees using electronic medical records and electronic education records

Jamaluddine,

Seita,

Ballout

et al. 2023

IJPDS

IntroductionBy linking datasets, electronic records can be used to build large birth-cohorts, enabling researchers to cost-effectively answer questions relevant to populations over the life-course. Currently, around 5.8 million Palestinian refugees live in five settings: Jordan, Lebanon, Syria, West Bank, and Gaza Strip. The United Nations Relief and Works Agency for Palestine Refugees in the Near East (UNRWA) provides them with free primary health and elementary-school services. It maintains electronic records to do so. We aimed to establish a birth cohort of Palestinian refugees born between 1st January 2010 and 31st December 2020 living in five settings by linking mother obstetric records with child health and education records and to describe some of the cohort characteristics. In future, we plan to assess effects of size-at-birth on growth, health and educational attainment, among other questions. MethodsWe extracted all available data from 140 health centres and 702 schools across five settings, i.e. all UNRWA service users. Creating the cohort involved examining IDs and other data, preparing data, de-duplicating records, and identifying live-births, linking the mothers' and children's data using different deterministic linking algorithms, and understanding reasons for non-linkage. ResultsWe established a birth cohort of Palestinian refugees using electronic records of 972,743 live births. We found high levels of linkage to health records overall (83%), which improved over time (from 73% to 86%), and variations in linkage rates by setting: these averaged 93% in Gaza, 89% in Lebanon, 75% in Jordan, 73% in West Bank and 68% in Syria. Of the 423,580 children age-eligible to go to school, 47% went to UNRWA schools and comprised of 197,479 children with both health and education records, and 2,447 children with only education records. In addition to year and setting, other factors associated with non-linkage included mortality and having a non-refugee mother. Misclassification errors were minimal. ConclusionThis linked open birth-cohort is unique for refugees and the Arab region and forms the basis for many future studies, including to elucidate pathways for improved health and education in this vulnerable, understudied population. Our characterization of the cohort leads us to recommend using different sub-sets of the cohort depending on the research question and analytic purposes.