A Method for Constructing Geographical Knowledge Graph from Multisource Data

Global problems all occur at a particular location on or near the Earth’s surface. Sitting at the junction of artificial intelligence (AI) and big data, knowledge graphs (KGs) organize, interlink, and create semantic knowledge, thus attracting much attention worldwide. Although the existing KGs are constructed from internet encyclopedias and contain abundant knowledge, they lack exact coordinates and geographical relationships. In light of this, a geographical knowledge graph (GeoKG) construction method based on multisource data is proposed, consisting of a modeling schema layer and a filling data layer. This method has two advantages: (1) the knowledge can be extracted from geographic datasets; (2) the knowledge on multisource data can be represented and integrated. Firstly, the schema layer is designed to represent geographical knowledge. Then, the methods of extraction and integration from multisource data are designed to fill the data layer, and a storage method is developed to associate semantics with geospatial knowledge. Finally, the GeoKG is verified through linkage rate, semantic relationship rate, and application cases. The experiments indicate that the method could automatically extract and integrate knowledge from multisource data. Additionally, our GeoKG has a higher success rate of linking web pages with geographic datasets, and its exact coordinates have increased to 100%. This paper could bridge the distance between a Geographic Information System and a KG, thus facilitating more geospatial applications.

Evaluating PPRL Vs Clear Text Linkage with Real-World Data

IntroductionPrivacy-preserving Record Linkage (PPRL) is a record linkage technique that can increase the security of personal information. PPRL uses techniques of either hashing identifiers (where exact matches are required) or Blooming identifiers (where partial matches are of interest before they are provided for linkage. Objectives and ApproachWe use LinXmart software to evaluate performance of PPRL linkage compared to linkage using clear text identifiers. The test linkage dataset is one that is routinely linked (N=2,672,257) at our linkage centre. The population spine (N=8,440,442) includes a record for every person who has resided in British Columbia, Canada over the past 30 years. Weights were determined using LinXmart’s implementation of the Expectation Maximization (EM) algorithm. For both linkages, accepted links were the highest-weighted candidate link with a weight above the threshold suggested by EM estimation. We compare linkage rates and quality and differences in weight and threshold estimations between clear-text and PPRL linkages ResultsClear-text and PPRL methods resulted in 97% and 90% linkage rates, respectively. Approximately 67% of records in the linked datasets contained a nominally unique ID. Records with a unique ID linked at higher rates (>99% for both clear-text and PPRL) while the linkage rate for records missing the ID differed substantially (92% /70% for clear-text/PPRL). Comparing PPRL linkage to the clear-text linkage, we obtain F-measures of 0.99 and 0.80 for records with and without the unique ID, respectively. Conclusion / ImplicationsLinkage performance may be attributable to differences in comparison operators between the two methods. Bloomed fields compared with Dice coefficient allow for partial matching but may not be as sensitive as clear-text string comparisons. Numerical comparisons in PPRL are exact matches while clear-text comparisons allow for more sophisticated matching. Further refinements in PPRL are being explored to improve these results.

Improving Inter-Agency Data Sharing Through Linkage Spine Interoperability

IntroductionLinkage of Federal Government data in Australia is conducted primarily through Accredited Integrating Authorities (AIAs). These agencies hold different dataset from Commonwealth and state/territory government agencies. Historically, linkage projects involving data held by different AIAs has been inefficient, requiring the transfer of identifiable data between agencies, and relinking data that have already been linked by another agency. Objectives and ApproachTwo AIAs (the AIHW and ABS) have developed a system of interoperable linkage spines to address this issue. By using common datasets as a base, the agencies have improved the efficiency and security of linkage projects. This process was developed through an analysis of spine datasets, and two test projects to share data between the agencies. ResultsThe two test projects were successfully able to link cross-portfolio and cross-jurisdictional data without the need to share additional identifying information between the AIAs. Preliminary results suggest a high linkage rate from this process, and work is underway to quantify the linkage quality compared to traditional linkage methodologies. The ABS and AIHW are also investigating the implications for linkage quality as more datasets are included in the agencies’ linkage spines. Conclusion / ImplicationsThe success of this project will increase the efficiency of cross-jurisdictional and cross-portfolio linkage in Australia. It will also allow specialised AIAs to work on datasets where they have specific expertise, and feed these into broader projects. This is expected to have an additional impact on public trust in the linkage system, by minimising the sharing of personally identifiable information while still maintaining high quality linkage.

A continuous energy-efficiency optimization controller for field-orientation induction motor drives

The current paper discusses the adaptive choice of a filter time constant for filtering the steady-state flux reference in the energy-efficient control problem of field-orientation induction machines in transient behavior when load and speed conditions are changing taking into account the effect of the main induction saturation. It is shown that by appropriately managing the flux linkage rate of change the energy losses per full operation cycle under torque changes can be significantly reduced compared to the conventional cases. The analysis for the appropriate choice of the filter time constant as a fraction of the rotor time constant is based on a numerical study and simulation results for three different induction machines with different rated powers.

Abstract 20: Use of Administrative Claims to Ascertain Outcomes in Randomized Clinical Trials for Transcatheter Aortic Valve Replacement: Findings From the Extending Trial-based Evaluations of Medical Therapies Using Novel Sources of Data (EXTEND) Study

Introduction: Administrative claims may be a low cost alternative to traditional clinical trial event adjudication, but whether claims data can validly substitute for adjudicated outcomes to reproduce trial-derived treatment effects is uncertain. Methods: We linked adults aged ≥65 in the US CoreValve Pivotal High Risk (HiR) and Surgical or Transcatheter Aortic Valve Replacement in Intermediate-Risk Patients (SURTAVI) Trials to 100% Medicare inpatient claims (2003-2016). Primary (i.e. death and stroke) and secondary trial endpoints (i.e. bleeding, acute kidney injury, cardiogenic shock, pacemaker implantation, aortic valve reintervention, myocardial infarction, and major adverse cerebrovascular and cardiovascular events [MACCE]) were compared across treatment arms (i.e. TAVR vs. SAVR) using outcomes derived from claims at 1 year (HiR) or 2 years (SURTAVI), and compared with trial-adjudicated outcomes. Results: Among 600 linked HiR participants (linkage rate 80.0%), the rate of the trial’s primary endpoint of all-cause mortality at 1-year was 13.7% for TAVR and 16.4% for SAVR using both trial data (HR 0.84, 95% CI 0.65-1.09; p = 0.33) and claims data (HR 0.86, 95% CI 0.66-1.11; p = 0.34; interaction p-value = 0.80) ( Figure ). Among 1004 linked SURTAVI trial participants (linkage rate 60.5%), the trial’s primary endpoint of combined death and stroke at 2-years was 12.9% for TAVR and 13.1% for SAVR using trial data (HR 1.08, 95% CI 0.79-1.48, p = 0.90), and 11.3% for TAVR and 12.5% for SAVR patients using claims data (HR 1.02, 95% CI 0.73-1.41, p = 0.58; interaction p-value = 0.89) ( Figure ). Procedural secondary outcomes (e.g., aortic valve reintervention, pacemaker implantation) were more concordant between trials and claims than non-procedural outcomes (e.g. bleeding, cardiogenic shock, stroke). Acute kidney injury, myocardial infarction, and MACCE outcomes were also concordant between data types. Conclusions: In both HiR and SURTAVI trials, ascertainment of trial primary endpoints using claims reproduced both the magnitude and direction of treatment effect observed in adjudicated event data. Non-fatal secondary outcomes displayed variable concordance. Use of claims to substitute for adjudicated outcomes in traditional trial treatment comparisons may be a valid alternative for some endpoints.

Meeting the challenge of data linkage for special populations

IntroductionA file’s overall linkage rate may hide significant bias in unlinked records. Newborns, for example, may represent a small percentage of a file but a large proportion of unlinked records. This creates challenges for researchers focused on the affected populations. Decreasing bias is a clear goal for data linkage science. Objectives and ApproachA standard spine-based linkage process with a typical set of identifiers was used to link perinatal records in British Columbia. This included multiple iterations and some manual linkage, stopping when there were only minor improvements. Despite high overall linkage rates, poor rates existed for babies who were stillborn or died near birth. Given research interests in stillbirth, there was a desire to use a creative approach to improve linkage for that sub-population. The solution was to expand the scope of the spine linkage to include birth and stillbirth records from other files, e.g. hospital discharge and Vital Statistics Registrations. ResultsOriginal linkage methods produce overall baby linkage rates of 98.1% but only 3.9% of records discharged to death/stillbirth were linked. Revised methods provide a small change in the overall rate to 99.4% but increase the discharge to death/stillbirth linkage rate to 90.3%. All numbers exclude terminations. Additions to our base representation of a linked entity enabled the association of records for babies either stillborn or never registered for health insurance coverage with the Ministry of Health. Additional variables addressed high rates of missing identifiers, while new linkage processing to Vital Statistics data addressed gaps within multiple newborn related datasets. Conclusion/ImplicationsThe complexity of adding BC Perinatal Data as a linked dataset to Population Data BC holdings was underappreciated. There are always adjustments in linkage approach across different data sets, but the linkage of babies with adverse outcomes required considerably more change, including moving beyond our usual spine linkage approach.

Intermediate-linkage steps used to obtain longitudinal data (containing heath service use, morbidity and mortality) for a large cohort of patients who are homeless that visited, and were discharged from hospital in England.

IntroductionIn England, details on hospital admissions and mortality are recorded nationally, but housing status and patients’ hospital discharge arrangements are only recorded locally within discharge services. These data are required to evaluate specialist homeless hospital discharge (HHD) services in England, and can be obtained through linkage within and across sectors. Objectives and ApproachWe aimed to improve linkage to enable the evaluation of HHD schemes. 16 sites with a scheme were recruited along with a specialist facility that deliver screening and treatment services to homeless hostels (Find\&Treat). Linkage fields including National Health Service number (NHS number), name, gender and birthdate for clinical contacts between November 2013 and November 2016 were collected and linked to national hospital data, Hospital Episodes Statistics (HES). To improve linkage with HES, intermediate-linkage to a gender-names dictionary and a national demographic database (NDD) was performed. Ethics, access permissions were obtained through HRA-REC (REC16/EE/0018) and NHS CAG (16/CAG/0021). Results47,569 clinical contacts among people who were homeless were collected from Find&Treat and 12,931 from sites. The median age at mid-study period (15th May 2015) among contacts with sites compared to Find&Treat were similar at 44 (IQR 34-53, n=12,905) and 45 (IQR 35-54, n=47,569), respectively. Among Find&Treat, 82% (n=38,905) were contacts with Males and 18% (n=8,650) with Females. Gender was not collected at all HHD sites or for all admissions. 70% of contacts had missing gender and among these contacts, gender was assigned using the gender-names dictionary. After imputing gender, 52% of contacts all linkage fields and 47% had all but NHS number. These data were linked to the NDD, an approximate 60% linkage rate was achieved retrieving complete linkage fields for these contacts. Conclusion/ImplicationsIntermediate linkage steps described here provides the largest dataset of it’s kind, enabling investigations into effectiveness of hospital discharge schemes in England. The study provides generally a proof of concept that large cohorts of hard-to-reach population groups can be obtained through data linkage.

Calgary Thrives: Data sharing and linkage in the not-for-profit sector.

IntroductionCompared to the public sector, not-for-profits are less focused on cross-agency data linkage. Technical capacity is often secondary to addressing caseloads and protecting clients. Clients’ privacy is paramount and can be perceived as a barrier to collaboration between agencies. However, collaboration could streamline referrals and better assist vulnerable populations. Objectives and ApproachSix not-for-profit agencies in Calgary, Canada participated in a data sharing project to measure various aspects of poverty and link data to determine cross-agency service usage. With this goal in mind, agencies examined their consent and data sharing practices to assess barriers to data sharing. There was a thorough exploration of client consent and how a client’s context can enable or limit data sharing. Cross-agency program usage was assessed among participating agencies using a privacy-preserving record linkage (PPRL) methodology. ResultsAmongst the six participating agencies, four were deemed to have adequate technical capacity to share data. A contributing factor to the willingness of agencies to share data was the development of LinkWise: a PPRL software created and developed by PolicyWise. Linkage rates amongst three agencies were compared. Rates ranged from 47.8% to 0.23%. A higher linkage rate between two agencies indicated a small community based agency which provided many referrals to a larger agency, such as a food bank. Lower linkage rates on client intake may indicate an agency with many clients. It may also indicate differing socio-economic brackets for their clients’ catchment area. Conclusion/ImplicationsWhile capacity, caseloads, and privacy protection restrict data sharing, not-for-profit agencies would benefit from a data sharing strategy. Linking data represent opportunities for collaboration within a significantly resource constrained sector. Moreover, it could more effectively address issues of vulnerable populations, streamline referrals for services, and facilitate quality improvement.

Unlocking First Nations health information through data linkage

IntroductionThe importance of Indigenous data sovereignty and Indigenous-led research processes is increasingly being recognized in Canada and internationally. For First Nations in Ontario, Canada, access to routinely-collected demographic and health systems data is critical to planning and measuring health status and outcomes in their populations. Linkage of this data with the Indian Register (IR), under First Nations data governance, has unlocked data for use by First Nations organizations and communities. ObjectivesTo describe the linkage of the IR database to the Ontario Registered Persons Database (RPDB) within the context of Indigenous data sovereignty principles. MethodsDeterministic and probabilistic record linkage methods were used to link the IR to the RPDB. There is no established population of First Nations people living in Ontario with which we could establish a linkage rate. Accordingly, several approaches were taken to determine a denominator that would represent the total population of First Nations we would hope to link to the RPDB. ResultsOverall, 201,678 individuals in the national IR database matched to Ontario health records by way of the RPDB, of which 98,562 were female and 103,116 were male. Of those First Nations individuals linked to the RPDB, 90.2% (n=181,915) lived in Ontario when they first registered with IR, or were affiliated with an Ontario First Nation Community. The proportion of registered First Nations people linking to the RPDB improved across time, from 62.8% in the 1960s to 94.5% in 2012. Conclusions This linkage of the IR and RPDB has resulted in the creation of the largest First Nations health research study cohort in Canada. The linked data are being used by First Nations communities to answer questions that ultimately promote wellbeing, effective policy, and healing.

Linkage of Maternity Hospital Episode Statistics data to birth registration and notification records for births in England 2005-2014: methods. A population-based birth cohort study

IntroductionMaternity Hospital Episode Statistics (HES) data for 2005–2014 were linked to birth registration and birth notification data (previously known as NHS Numbers for Babies or NN4B) to bring together some key demographic and clinical data items not otherwise available at a national level. The linkage algorithm that was previously used to link 2005–2007 data was revised to improve the linkage rate and reduce the number of duplicate HES records.MethodsBirth registration and notification linked records from the Office for National Statistics (‘ONS birth records’) were further linked to Maternity HES delivery and birth records using the NHS Number and other direct identifiers if the NHS Number was missing.ResultsFor the period 2005–2014, over 94% of birth registration and notification records were correctly linked to HES delivery records. Two per cent of the ONS birth records were incorrectly linked to the HES delivery record and 5% of ONS birth records were linked to more than one HES delivery record. Therefore, a considerable amount of time was spent in quality assuring these files.ConclusionThe linkage rate for birth registration and notification records to HES delivery records steadily improved from 2005 to 2014 due to improvement in the quality and completeness of patient identifiers in both HES and birth notification data.