Integrated Data System Person Identification: Accuracy Requirements and Methods

Repeated and Combined Participation in Workforce, Education and Social Program Services: Preliminary Evidence from Maryland Administrative Data Sources

Measuring Hospital Performance in Congenital Heart Surgery: Administrative Versus Clinical Registry Data

While linking records across large administrative datasets ["big data"] has the potential to revolutionize empirical social science research, many administrative data files do not have common identifiers and are thus not designed to be linked to others. To address this problem, researchers have developed probabilistic record linkage algorithms which use statistical patterns in identifying characteristics to perform linking tasks. Naturally, the accuracy of a candidate linking algorithm can be substantially improved when an algorithm has access to "ground-truth" examples-matches which can be validated using institutional knowledge or auxiliary data. Unfortunately, the cost of obtaining these examples is typically high, often requiring a researcher to manually review pairs of records in order to make an informed judgement about whether they are a match. When a pool of ground-truth information is unavailable, researchers can use "active learning" algorithms for linking, which ask the user to provide ground-truth information for select candidate pairs. In this paper, we investigate the value of providing ground-truth examples via active learning for linking performance. We confirm popular intuition that data linking can be dramatically improved with the availability of ground truth examples. But critically, in many real-world applications, only a relatively small number of tactically-selected ground-truth examples are needed to obtain most of the achievable gains. With a modest investment in ground truth, researchers can approximate the performance of a supervised learning algorithm that has access to a large database of ground truth examples using a readily available off-the-shelf tool.

This article describes two administrative data sources - UCAS applicant data and Higher Education Statistical Agency (HESA) data - and demonstrates how they can be utilised to monitor the progression of students from vocational and educational training (VET) programmes in to higher education (HE) in the UK. First the article provides a general discussion of the administrative data sets, their coverage and resulting limitations. Second, it illustrates how a combined HESA-UCAS data set can be utilised to provide a descriptive statistical analysis that demonstrates that including VET students within HE programmes widens participation for under-represented groups in the UK but shows that VET students are at a disadvantage compared with those progressing through academic pathways in terms of their probability of gaining access, which types of higher education institution (HEI) they gain access to, and an increased risk of dropping out during their first year of study. Third, multilevel modelling indicates that drop-out risks differ across HEIs and suggests that HEIs can take actions that will reduce the risk of VET students dropping out. Finally some conclusions about the utility of administrative data to answer substantive research questions are presented, with suggestions for improving the quality and access to such data for researchers.

Measures of the gap in living standards, life expectancy, education, health and employment between Indigenous and non‐Indigenous Australians are primarily derived from administrative data sources. However, Indigenous identification in these data sources is affected by administrative practices, missing data, inconsistency, and error. As these factors have changed over time, assessing whether the gap between Indigenous and non‐Indigenous Australians has changed over time, based on data unadjusted for these sources of error can potentially lead to misguided conclusions. Combining administrative data on the same individuals collected from different sources provides a method by which a more consistent derived Indigenous status can be applied across all records for an individual within a linked data environment. We used the Western Australian Data Linkage system to produce derived Indigenous statuses for individuals using a range of algorithms. We found that these algorithms reduced the amount of missing data and improved within‐individual consistency. Based on these findings, we recommend our Multi‐Stage Median algorithm be used as the standard indicator of Indigenous status for any reporting based on administrative datasets when multiple datasets are available for linkage, and that algorithmic approaches also be considered for improving the quality of other demographic variables from administrative data sources.

The quantity of administrative data created, stored and processed in the world has grown exponentially over recent years, but the statistical theory to support its use in official statistics has not. That’s not to say that within the statistical community we are not using or investigating a wide variety of statistical sources, but the methods used are not based on well-established theory like we have for surveys. As a National Statistical Institute (NSI), we are transforming to put administrative and alternative data sources at the core of our statistics. Combining new sources with surveys will allow us to meet the ever-increasing user demand for improved and more detailed statistics. However, using administrative data involves addressing a range of statistical challenges as identified in “Statistical challenges of administrative and transaction data” (Hand, 2018).
 The Administrative Data Methods Research Programme was set up to address these challenges by developing a statistical framework for using and integrating administrative and transactional data to produce official statistics and analysis. The key themes of the programme include quality, linkage, privacy, estimation, statistical design, analysis and innovative techniques.
 The programme will ensure the rigor and assurance of methods developed, allowing where possible wide-reaching application, and to create and share best practice and guidance. A key part of this programme will be engaging and working collaboratively with the statistical community in academia, government and private sector to help prioritise the challenges, shape the work programme and drive the work forward.

The use of "clinical" versus "administrative" data sets for health care quality assessment continues to be debated. This study directly compares the University HealthSystem Consortium Clinical Database (UHC CDB) and the National Surgical Quality Improvement Program (NSQIP) in terms of their assessment of complications and death for 26 322 surgery patients using analyses of variance, correlation, and multivariable logistic regression. The NSQIP had more variables with significant correlation with outcomes. The NSQIP was better at predicting death (c-index 0.94 vs 0.90, P < .05) and complications (c-index 0.78 vs 0.76, P = .07), especially for higher risk patients. The UHC CDB missed and misclassified several major complications. The data sets are similar in their explanatory power relative to outcomes, but the clinical data set is better, particularly at identifying higher risk patients and specific complications. It should prove more useful for initiating and monitoring clinical process improvements because of more clinically relevant variables.

Linkage between health databases typically requires patient identifiers such as names and personal identification numbers. We developed and validated a record linkage strategy to combine administrative health databases without identifiers for South Africa’s public sector HIV program. We linked CD4 counts and HIV viral loads from South Africa’s TIER.Net with the National Health Laboratory Service (NHLS) database for patients receiving care between 2015–2019 in Ekurhuleni District (Gauteng Province). Linkage variables were result value, specimen collection date, facility of collection, year and month of birth, and sex. We used three matching strategies: exact matching on exact values of all variables, caliper matching allowing a ± 5 day window on result date, and specimen barcode matching using unique specimen identifiers. A sequential linkage approach applied specimen barcode, followed by exact, and then caliper matching. Exact and caliper matching were validated using barcodes (available for 34% of records in TIER.Net) as a “gold standard”. Performance measures were sensitivity, positive predictive value (PPV), share of patients linked, and percent increase in data points. We attempted to link 2,017,290 laboratory test results from TIER.Net (523,558 unique patients) with 2,414,059 NHLS test results. Exact matching achieved 69.0% sensitivity and 95.1% PPV. Caliper matching achieved 75% sensitivity and 94.5% PPV. Sequential linkage matched 41.9% using specimen barcodes, 51.3% through exact matching, and 6.8% through caliper matching, for 71.9% (95% CI: 71.9, 72.0) of test results matched overall, with 96.8% (95% CI: 96.7, 97.1) PPV and 85.9% (95% CI: 85.7, 85.9) sensitivity. This linked 86.0% (95% CI: 85.9, 86.1) of TIER.Net patients to the NHLS (N = 1,450,087), increasing laboratory results in TIER.Net by 62.6%. Linkage of TIER.Net and NHLS without patient identifiers attained high accuracy and yield without compromising privacy. The integrated cohort provides a more complete laboratory test history and supports more accurate HIV program indicator estimates.

BACKGROUND: Quantitative research on the social, demographic, and economic outcomes of sexual minorities has long been hampered by data shortfalls, with most surveys and censuses limited by sample sizes and/or a lack of direct questions on sexual identity. The growing availability of administrative data presents an opportunity to fill some of these gaps. OBJECTIVE: This article highlights the challenges and opportunities involved with using a novel administrative dataset – the Longitudinal Administrative Databank, which includes 20% of Canadian tax filers – to study sexual minority populations in Canada. We identify three sources of bias, propose strategies to adjust for this bias, and introduce a measure of “inferred sexual minority status” to improve the identification of sexual minorities in tax data. RESULTS: Administrative tax data offers significant advantages, including a large sample size, high-quality income data for individuals and linked family members, a longitudinal design, and the ability to trace individuals’ same-/different-sex partnership histories. Our adjustment strategies mitigate some biases in identifying same-sex couples, including underreporting, misclassification, and measurement errors. The estimated proportion of individuals in same-sex marriages closely aligns with Canadian census estimates from 2006–2021, while the proportion in same-sex common-law partnerships is underestimated. Finally, our earnings gaps analyses highlight the utility of the inferred sexual minority status measure. CONTRIBUTION: This article contributes to research on sexual minority data landscapes, offering new insights into the identification and measures of sexual minority populations using longitudinal administrative tax data. Our approach points to new opportunities for studying the long-term longitudinal income and family dynamics of sexual minority populations on the national level.

Linking fathers and children in administrative data for public health research: a systematic scoping review

IntroductionResearch data combined with administrative data provides a robust resource capable of answering unique research questions. However, in cases where personal health data are encrypted, due to ethics requirements or institutional restrictions, traditional methods of deterministic and probabilistic record linkages are not feasible. Instead, privacy-preserving record linkages must be used to protect patients’ personal data during data linkage.ObjectivesTo determine the feasibility and validity of a deterministic privacy preserving data linkage protocol using homomorphically encrypted data.MethodsFeasibility was measured by the number of records that successfully matched via direct identifiers. Validity was measured by the number of records that matched with multiple indirect identifiers. The threshold for feasibility and validity were both set at 95%. The datasets shared a single, direct identifier (health card number) and multiple indirect identifiers (sex and date of birth). Direct identifiers were encrypted in both datasets and then transferred to a third-party server capable of linking the encrypted identifiers without decrypting individual records. Once linked, the study team used indirect identifiers to verify the accuracy of the linkage in the final dataset.ResultsWith a combination of manual and automated data transfer in a sample of 8,128 individuals, the privacy-preserving data linkage took 36 days to match to a population sample of over 3.2 million records. 99.9% of the records were successfully matched with direct identifiers, and 99.8% successfully matched with multiple indirect identifiers. We deemed the linkage both feasible and valid.ConclusionsAs combining administrative and research data becomes increasingly common, it is imperative to understand options for linking data when direct linkage is not feasible. The current linkage process ensured the privacy and security of patient data and improved data quality. While the initial implementations required significant computational and human resources, increased automation keeps the requirements within feasible bounds.

SummaryWe compared measures of hospital performance by using both administrative and clinical data sources. Hospital-specific mortality outcomes on 10086 patients who had been admitted to 102 hospitals with a diagnosis of acute myocardial infarction in Ontario, Canada, were used as a test-case. Four and six hospitals were identified as having mortality that was statistically significantly higher than expected by using administrative and clinical data respectively, when model-based indirect standardization was used. When using random-effects models, zero and two hospitals were identified as having significantly higher mortality by using administrative and clinical data respectively. Approximately one in four hospitals changed at least two decile rankings when clinical data were used compared with when administrative data were used.

Expert groups have recommended ongoing monitoring of the public health workforce to determine its ability to execute designated objectives. Resource- and time-intensive surveys have been a primary data source to monitor the workforce. We evaluated an administrative data source containing US Department of Health and Human Services (HHS) aggregate federal civil service workforce-related data to determine its potential as a workforce surveillance system for this component of the workforce. We accessed FedScope, a publicly available online database containing federal administrative civilian HHS personnel data. Using established guidelines for evaluating surveillance systems and identified workforce characteristics, we evaluated FedScope attributes for workforce surveillance purposes. We determined FedScope to be a simple, highly accepted, flexible, stable, and timely system to support analyses of federal civil service workforce-related data. Data can be easily accessed, analyzed, and monitored for changes across years and draw conclusions about the workforce. FedScope data can be used to calculate demographics (eg, sex, race or ethnicity, age group, and education level), employment characteristics (ie, supervisory status, work schedule, and appointment type), retirement projections, and characterize the federal workforce into standard occupational categories. This study indicates that an administrative data source containing HHS personnel data can function as a workforce surveillance system valuable to researchers, public health leaders, and decision makers interested in the federal civil service public health workforce. Using administrative data for workforce development is a model that can be applicable to federal and nonfederal public health agencies and ultimately support improvements in public health.

Given the growing popularity of administrative data for health research, information on the differences and similarities between administrative data and customary data sources (e.g. surveillance) will help to inform the use of administrative data in the field of sexually transmissible infections (STIs). The objective of this study was to compare the incidence rates of three nonviral STIs from a large health insurance administrative database (MarketScan) with surveillance data. We computed and compared STI rates for 2005-10 from MarketScan and national surveillance data for three major nonviral STIs (i.e. chlamydia (Chlamydia trachomatis), gonorrhoea (Neisseria gonorrhoeae) and syphilis (Treponema pallidum)). For administrative data, we assessed the sensitivity of the rates to enrollee inclusion criteria: continuous (≥320 member-days) versus all enrollees. Relative rates were computed for 5-year age groups and by gender. The administrative database rates were significantly lower (P<0.01) than those in the national surveillance data, except for syphilis in females. Gonorrhoea and syphilis rates based on administrative data were significantly lower (P<0.01) for all enrollees versus continuous enrollees only. The relative STI rates by age group from the administrative data were similar to those in the surveillance data. Although absolute STI rates in administrative data were lower than in the surveillance data, relative STI rates from administrative data were consistent with national surveillance data. For gonorrhoea and syphilis, the estimated rates from administrative data were sensitive to the enrollee inclusion criteria. Future studies should examine the potential for administrative data to complement surveillance data.

Background with rationaleChildren born to opioid-dependent mothers are at a developmental disadvantage from pre-birth. They are additionally affected by the mother’s compromised ability to recognise and respond to the infant’s cues. Development is often compounded by environmental factors. Research to date has primarily focused on early infancy and small, clinical samples. This group is difficult to follow-up using traditional methods due to chaotic home environments, housing instability and parent-child separation. The use of administrative data circumnavigates such difficulties, allowing follow-up of children over longer periods, even when removed from the birth parent.&#x0D; Main AimThis paper will describe the complex creation of a cohort of children born to opioid-dependent women, using administrative data. It will also describe early results about pregnancy and neonatal outcomes.&#x0D; Methods/ApproachData were pooled from women who gave birth between 2007 and 2017 using five datasets (c.5,000 women): women who were recorded as using heroin, street methadone or opioid substitution therapy (OST) on the Drugs Misuse Database, or on OST prescription records; women admitted to hospital, or psychiatric care, for an opioid related reason; and/or women whose children were recorded as having Neonatal Abstinence Syndrome (NAS). Data on children’s neonatal outcomes will be described, including birth weight and gestation, congenital abnormalities, neonatal death and NAS treatment. Models will be fitted to investigate the associations between possible teratogenicity of prenatal opioids and developmental outcomes.&#x0D; ResultsThe development of this cohort using administrative data sources has been complex, requiring five different datasets to ensure all women of interest are captured. Descriptive results on outcomes will be available in the Autumn.&#x0D; ConclusionThis administrative data study demonstrates the value of using linked data sources to enhance our knowledge of the trajectories of this vulnerable group of children, and the additional support that they, and their carers, may require.