Suchergebnisse

IntroductionData linkage has considerable potential to improve health and society. Linking vast and detailed information across multiple administrative population-scaled data sources enhances the quality of existing data, empowers population health research, and produces objective evidence to inform policy decisions. In this context, data cleaning is crucial to minimise linkage errors.
Objectives and ApproachAs dealing with the heterogeneity of administrative datasets is an acknowledged time-consuming task, the objective is creating a public and open-source R package to automate and report steps of data cleaning in a reproducible fashion.
The package automatically assesses variables and reports relevant information and issues for linkage purposes, then cleans the dataset based on problems found, reassesses the variables and reports results again.
It has a default cleaning procedure based on years of accumulated linkage knowledge and an interactive exploratory session to check variables individually. The report also includes all settings from both default and interactive session.
ResultsThe package accurately detected, cleaned and reported potential linkage problems in variables such as names, addresses and dates in so far 15 actual populational datasets from multiple sources, with a diverse range of format, content, and inconsistencies. The entire process took minutes rather than hours.
The reports correctly gathered, organised and presented all relevant information for linkage, in all distinct sections of the hyperlinked document, such as those related to the dataset, individual variables or settings used for cleaning. The different types of information included text, figures, data dictionaries, and frequency tables of detected issues, such as non-alphanumeric characters, annotation terms or suffixes and prefixes.
The output datasets had all evaluated variables with cleaned data plus extra columns containing only issues themselves or problematic records.
Conclusion/ImplicationsThe package accelerates the data cleaning of linkage variables, automating time-consuming steps, providing pertinent information for linkage as well as cleaned datasets. The complete process is time-efficient and reproducible.
As the output dataset contains variables with cleaned data and detected issues, it allows assessment of the level of cleaning performed.

IntroductionMany well established string comparators are currently used in data linkage. Jaro-Winkler distance is SA NT DataLink's metric of choice for comparing personal names. However, due to Jaro-Winkler's lower specificity we investigated if output scores could be transformed to produce scores more closely matching those assigned manually.
Objectives and ApproachOur objective was to reduce the need for clerical review by modifying the Jaro-Winkler distance metric output scores. Clerical reviewers assigned similarity scores to pairs of first or last names from a database of approximately 2,000 random cases. By plotting the Jaro-Winkler scores against those assigned by the reviewers, a distinct radical function shape was observed. We then transformed the Jaro-Winkler scores by applying a power function where we gradually changed the exponent until we obtained the best fit with our clerically assigned scores. From the next linkage, two separate outputs were created (original and modified) and the results compared.
ResultsTo assess the best fit we calculated the sum of squared errors for each of tested exponent values ranging from 1.1 to 6.0 (with 0.1 steps). The minimum sum of squared errors was achieved with exponent value of 4.6. We performed a probabilistic linkage for one decade of the Birth Registry records looking for familial links. Two separate linkage runs were conducted and clerically reviewed. In the second run, names were compared using the modified Jaro-Winkler comparator. This resulted in a reduced number of false positives. Though the lower-end threshold of the clerically reviewed "grey area" had to be lowered, the overall range was narrower resulting in less record pairs for clerical review.
Conclusion/ImplicationsBy transforming the Jaro-Winkler scores, we reduced the number of records requiring clerical review. While only three linkage variables were affected, the resultant outcome was encouraging enough to consider exploring other possibilities for replicating clerical review knowledge in other comparators and metrics to reduce the demands for clerical review.

IntroductionRecord linkage has existed for decades, regarded as an activity where an entity is ascertained across different data sources. Initially medically driven to determine further outcomes. Developments in technology led to cheaper computing and storage, which ushered capture of information beyond the primary entities such as carer's details.
Objectives and ApproachAdministrative datasets that capture an event involving a child would more often than not, capture information about their parents or carers including relationship to the primary entity within the record.
The main objective was to exploit the connections across the entities captured in a record and have that as an additional layer of information that may contribute to context. All entities extracted from one record were clearly defined as related. Information was considered true at the point of capture. Then record linkage for all the individual entities extracted were performed.
ResultsFamily structures evolve and a family structure during birth may change by the time the child attends school where carer information was recorded during enrolment.
Each data source inevitably becomes a "source of truth" and understanding the different aspects of a dataset affects its credibility. Authority for collection dictates if the recording of the event is mandatory or optional and directly influences the coverage. The process of data capture and storage is also important; manual vs digital input and if any validation was present.
Application of family connection to specific types of research will dictate the relevant "source of truth". If one is to use family connections in the context of hereditary diseases, pregnancy outcomes would be more suitable than a birth registry.
Conclusion/ImplicationsFamily unit as defined in the traditional sense have changed and all variations are reflected in the data.
Assessing the impact of social connections to ones health may now be possible. Extracting these connections from administrative datasets can serve as additional intelligence that may inform if not shape longitudinal studies.

ApproachUniquely governed by a broad range of consortium partners, SA NT DataLink's business model operates with flexibility to adapt to researcher priorities and government requirements. Its Data Linkage Unit routinely links data from over 50 sources with more than 57 million records on approximately 2.9 million individuals. It arguably provides the broadest range of linked data sources in Australia, focusing on administrative datasets and clinical registries from various health and human services domains. Operating in strict adherence with the separation principal, SA NT DataLink's Data Integration Unit separately manages anonymised clinical and service use data in collaboration with the respective data custodians through the Custodian Controlled Data Repository, allowing approved analysts to efficiently access high quality linked anonymised data. To protect individual privacy throughout the process of data linkage and data provision, SA NT DataLink's processes align with state, territory and federal privacy legislations. Operating consistently with National Health and Medical Research Council guidelines, linkage projects are subject to approvals by the relevant data custodians and approved Human Research Ethics Committees.
Noteworthy OutputsSA NT DataLink has provided linkage services to over 160 data linkage projects, informing nationally significant research and policy initiatives, including initiatives to improve indigenous children's hearing and child development.
ConclusionTo respond to a changing data linkage landscape, SA NT DataLink is continuously reviewing and improving its systems, linkage processes and governance, addressing administrative, funding, data access, social licence and data linkage challenges and opportunities to meet increasing demand and new business developments.