Linkage Error Research Articles

Record linkage under suboptimal conditions for data-intensive evaluation of primary care in Rio de Janeiro, Brazil

BackgroundLinking Brazilian databases demands the development of algorithms and processes to deal with various challenges including the large size of the databases, the low number and poor quality of personal identifiers available to be compared (national security number not mandatory), and some characteristics of Brazilian names that make the linkage process prone to errors. This study aims to describe and evaluate the quality of the processes used to create an individual-linked database for data-intensive research on the impacts on health indicators of the expansion of primary care in Rio de Janeiro City, Brazil.MethodsWe created an individual-level dataset linking social benefits recipients, primary health care, hospital admission and mortality data. The databases were pre-processed, and we adopted a multiple approach strategy combining deterministic and probabilistic record linkage techniques, and an extensive clerical review of the potential matches. Relying on manual review as the gold standard, we estimated the false match (false-positive) proportion of each approach (deterministic, probabilistic, clerical review) and the missed match proportion (false-negative) of the clerical review approach. To assess the sensitivity (recall) to identifying social benefits recipients’ deaths, we used their vital status registered on the primary care database as the gold standard.ResultsIn all linkage processes, the deterministic approach identified most of the matches. However, the proportion of matches identified in each approach varied. The false match proportion was around 1% or less in almost all approaches. The missed match proportion in the clerical review approach of all linkage processes were under 3%. We estimated a recall of 93.6% (95% CI 92.8–94.3) for the linkage between social benefits recipients and mortality data.ConclusionThe adoption of a linkage strategy combining pre-processing routines, deterministic, and probabilistic strategies, as well as an extensive clerical review approach minimized linkage errors in the context of suboptimal data quality.

Variation in HIV care and treatment outcomes by facility in South Africa, 2011-2015: A cohort study.

Despite widespread availability of HIV treatment, patient outcomes differ across facilities. We propose and evaluate an approach to measure quality of HIV care at health facilities in South Africa's national HIV program using routine laboratory data. Data were extracted from South Africa's National Health Laboratory Service (NHLS) Corporate Data Warehouse. All CD4 counts, viral loads (VLs), and other laboratory tests used in HIV monitoring were linked, creating a validated patient identifier. We constructed longitudinal HIV care cascades for all patients in the national HIV program, excluding data from the Western Cape and very small facilities. We then estimated for each facility in each year (2011 to 2015) the following cascade measures identified a priori as reflecting quality of HIV care: median CD4 count among new patients; retention 12 months after presentation; 12-month retention among patients established in care; viral suppression; CD4 recovery; monitoring after an elevated VL. We used factor analysis to identify an underlying measure of quality of care, and we assessed the persistence of this quality measure over time. We then assessed spatiotemporal variation and facility and population predictors in a multivariable regression context. We analyzed data on 3,265 facilities with a median (IQR) annual size of 441 (189 to 988) lab-monitored HIV patients. Retention 12 months after presentation increased from 42% to 47% during the study period, and viral suppression increased from 66% to 79%, although there was substantial variability across facilities. We identified an underlying measure of quality of HIV care that correlated with all cascade measures except median CD4 count at presentation. Averaging across the 5 years of data, this quality score attained a reliability of 0.84. Quality was higher for clinics (versus hospitals), in rural (versus urban) areas, and for larger facilities. Quality was lower in high-poverty areas but was not independently associated with percent Black. Quality increased by 0.49 (95% CI 0.46 to 0.53) standard deviations from 2011 to 2015, and there was evidence of geospatial autocorrelation (p < 0.001). The study's limitations include an inability to fully adjust for underlying patient risk, reliance on laboratory data which do not capture all relevant domains of quality, potential for errors in record linkage, and the omission of Western Cape. We observed persistent differences in HIV care and treatment outcomes across South African facilities. Targeting low-performing facilities for additional support could reduce overall burden of disease.

Challenging racism in the use of health data

Challenging racism in the use of health data

Prelabeling Expansion Single-Molecule Localization Microscopy with Minimal Linkage Error.

Expansion microscopy combined with single-molecule localization microscopy (ExSMLM) has a potential for approaching molecular resolution. However, ExSMLM faces multiple challenges such as loss of fluorophores and proteins during polymerization, digestion or denaturation, and an increase in linkage error arising from the distance between the fluorophore and the target molecule. Here, we introduce a trifunctional streptavidin to link the target, fluorophore and gel matrix via a biotinylizable peptide tag. The resultant ExSMLM images of vimentin filaments demonstrated high labeling efficiency and a minimal linkage error of ∼5 nm. Our ExSMLM provides a simple and practical means for fluorescence imaging with molecular resolution.

Small Area Estimation with Linked Data

AbstractData linkage can be used to combine values of the variable of interest from a national survey with values of auxiliary variables obtained from another source, such as a population register, for use in small area estimation. However, linkage errors can induce bias when fitting regression models; moreover, they can create non-representative outliers in the linked data in addition to the presence of potential representative outliers. In this paper, we adopt a secondary analyst’s point of view, assuming that limited information is available on the linkage process, and develop small area estimators based on linear mixed models and M-quantile models to accommodate linked data containing a mix of both types of outliers. We illustrate the properties of these small area estimators, as well as estimators of their mean squared error, by means of model-based and design-based simulation experiments. We further illustrate the proposed methodology by applying it to linked data from the European Survey on Income and Living Conditions and the Italian integrated archive of economic and demographic micro data in order to obtain estimates of the average equivalised income for labour market areas in central Italy.

Linkage-Data Linear Regression

AbstractData linkage is increasingly being used to combine data from different sources with the aim of identifying and bringing together records from separate files, which correspond to the same entities. Usually, data linkage is not a trivial procedure and linkage errors, false and missed links, are unavoidable. In these cases, standard statistical techniques may produce misleading inference. In this paper, we propose a method for secondary linear regression analysis, where the linked data have to be prepared by someone else, and neither the match-key variables nor the unlinked records are available to the analyst. We develop also a diagnostic test for the assumption of non-informative linkage errors, which is required for all existing secondary analysis adjustment methods. Our approach provides important advantages: it relies on the realistic assumption that the probabilities of correct linkage vary across the records but it does not assume that one is able to estimate the probability of correct linkage for each individual record. Moreover, it accommodates in a simple manner the general situation where the files are of different sizes and none of them is a subset of another. The proposed methodology of adjustment and testing is studied by simulation and applied to real data.

National administrative record linkage between specialist community drug and alcohol treatment data (the National Drug Treatment Monitoring System (NDTMS)) and inpatient hospitalisation data (Hospital Episode Statistics (HES)) in England: design, method and evaluation

ObjectivesThe creation and evaluation of a national record linkage between substance misuse treatment, and inpatient hospitalisation data in England.DesignA deterministic record linkage using personal identifiers to link the National Drug...

Primary healthcare expansion and mortality in Brazil's urban poor: A cohort analysis of 1.2 million adults.

BackgroundExpanding delivery of primary healthcare to urban poor populations is a priority in many low- and middle-income countries. This remains a key challenge in Brazil despite expansion of the country’s internationally recognized Family Health Strategy (FHS) over the past two decades. This study evaluates the impact of an ambitious program to rapidly expand FHS coverage in the city of Rio de Janeiro, Brazil, since 2008.Methods and findingsA cohort of 1,241,351 low-income adults (observed January 2010–December 2016; total person-years 6,498,607) with linked FHS utilization and mortality records was analyzed using flexible parametric survival models. Time-to-death from all-causes and selected causes were estimated for FHS users and nonusers. Models employed inverse probability treatment weighting and regression adjustment (IPTW-RA).The cohort was 61% female (751,895) and had a mean age of 36 years (standard deviation 16.4). Only 18,721 individuals (1.5%) had higher education, whereas 102,899 (8%) had no formal education. Two thirds of individuals (827,250; 67%) were in receipt of conditional cash transfers (Bolsa Família). A total of 34,091 deaths were analyzed, of which 8,765 (26%) were due to cardiovascular disease; 5,777 (17%) were due to neoplasms; 5,683 (17%) were due to external causes; 3,152 (9%) were due to respiratory diseases; and 3,115 (9%) were due to infectious and parasitic diseases. One third of the cohort (467,155; 37.6%) used FHS services. In IPTW-RA survival analysis, an average FHS user had a 44% lower hazard of all-cause mortality (HR: 0.56, 95% CI 0.54–0.59, p < 0.001) and a 5-year risk reduction of 8.3 per 1,000 (95% CI 7.8–8.9, p < 0.001) compared with a non-FHS user. There were greater reductions in the risk of death for FHS users who were black (HR 0.50, 95% CI 0.46–0.54, p < 0.001) or pardo (HR 0.57, 95% CI 0.54–0.60, p < 0.001) compared with white (HR 0.59, 95% CI 0.56–0.63, p < 0.001); had lower educational attainment (HR 0.50, 95% CI 0.46–0.55, p < 0.001) for those with no education compared to no significant association for those with higher education (p = 0.758); or were in receipt of conditional cash transfers (Bolsa Família) (HR 0.51, 95% CI 0.49–0.54, p < 0.001) compared with nonrecipients (HR 0.63, 95% CI 0.60–0.67, p < 0.001).Key limitations in this study are potential unobserved confounding through selection into the program and linkage errors, although analytical approaches have minimized the potential for bias.ConclusionsFHS utilization in urban poor populations in Brazil was associated with a lower risk of death, with greater reductions among more deprived race/ethnic and socioeconomic groups. Increased investment in primary healthcare is likely to improve health and reduce health inequalities in urban poor populations globally.

A comparison of area level and unit level small area models in the presence of linkage errors

Abstract In Official Statistics, interest in data integration has grown enormously, but the effect of integration procedures on statistical analysis has not yet been sufficiently developed. Data integration is not an error-free procedure and linkage errors, as false links and missed links can invalidate standard estimates. Recently, increasing attention has been paid to the effect of linkage errors on the statistical analyses and on statistical predictions. Recently, methods to adjust the unit level small area estimators for linkage errors have been proposed when the domains are correctly specified. In this paper we compare the na¨ıve and the adjusted unit level estimators with the area level estimators that are not affected by the linkage errors. The comparison encourages the use of the adjusted unit level estimator.

Examining the quality of record linkage process using nationwide Brazilian administrative databases to build a large birth cohort

BackgroundResearch using linked routine population-based data collected for non-research purposes has increased in recent years because they are a rich and detailed source of information. The objective of this study is to present an approach to prepare and link data from administrative sources in a middle-income country, to estimate its quality and to identify potential sources of bias by comparing linked and non-linked individuals.MethodsWe linked two administrative datasets with data covering the period 2001 to 2015, using maternal attributes (name, age, date of birth, and municipally of residence) from Brazil: live birth information system and the 100 Million Brazilian Cohort (created using administrative records from over 114 million individuals whose families applied for social assistance via the Unified Register for Social Programmes) implementing an in house developed linkage tool CIDACS-RL. We then estimated the proportion of highly probably link and examined the characteristics of missed-matches to identify any potential source of bias.ResultsA total of 27,699,891 live births were submited to linkage with maternal information recorded in the baseline of the 100 Million Brazilian Cohort dataset of those, 16,447,414 (59.4%) children were found registered in the 100 Million Brazilian Cohort dataset. The proportion of highly probably link ranged from 39.3% in 2001 to 82.1% in 2014. A substantial improvement in the linkage after the introduction of maternal date of birth attribute, in 2011, was observed. Our analyses indicated a slightly higher proportion of missing data among missed matches and a higher proportion of people living in an urban area and self-declared as Caucasian among linked pairs when compared with non-linked sets.DiscussionWe demonstrated that CIDACS-RL is capable of performing high quality linkage even with a limited number of common attributes, using indexation as a blocking strategy in larg e routine databases from a middle-income country. However, residual records occurred more among people under worse living conditions. The results presented in this study reinforce the need of evaluating linkage quality and when necessary to take linkage error into account for the analyses of any generated dataset.

Impact of matching error on linked mortality outcome in a data linkage of secondary mental health data with Hospital Episode Statistics (HES) and mortality records in South East London: a cross-sectional study

ObjectivesLinkage of electronic health records (EHRs) to Hospital Episode Statistics (HES)-Office for National Statistics (ONS) mortality data has provided compelling evidence for lower life expectancy in people with severe mental...

Prevalence of Down's Syndrome in England, 1998-2013: Comparison of linked surveillance data and electronic health records.

IntroductionDisease registers and electronic health records are valuable resources for disease surveillance and research but can be limited by variation in data quality over time. Quality may be limited in terms of the accuracy of clinical information, of the internal linkage that supports person-based analysis of most administrative datasets, or by errors in linkage between multiple datasets.ObjectivesBy linking the National Down Syndrome Cytogenetic Register (NDSCR) to Hospital Episode Statistics for England (HES), we aimed to assess the quality of each and establish a consistent approach for analysis of trends in prevalence of Down’s syndrome among live births in England.MethodsProbabilistic record linkage of NDSCR to HES for the period 1998–2013 was supported by linkage of babies to mothers within HES. Comparison of prevalence estimates in England were made using NDSCR only, HES data only, and linked data. Capture-recapture analysis and quantitative bias analysis were used to account for potential errors, including false positive diagnostic codes, unrecorded diagnoses, and linkage error.ResultsAnalyses of single-source data indicated increasing live birth prevalence of Down’s Syndrome, particularly in the analysis of HES. Linked data indicated a contrastingly stable prevalence of 12.3 (plausible range: 11.6–12.7) cases per 10 000 live births.ConclusionCase ascertainment in NDSCR improved slightly over time, creating a picture of slowly increasing prevalence. The emerging epidemic suggested by HES primarily reflects improving linkage within HES (assignment of unique patient identifiers to hospital episodes). Administrative data are valuable but trends should be interpreted with caution, and with assessment of data quality over time. Data linkage with quantitative bias analysis can provide more robust estimation and, in this case, stronger evidence that prevalence is not increasing. Routine linkage of administrative and register data can enhance the value of each.KeywordsDown’s syndrome; data linkage; disease surveillance; linkage error; electronic health records; prevalence

Corrigendum to: Reflections on modern methods: linkage error bias.

Corrigendum to: Reflections on modern methods: linkage error bias.

Investigation of the Spatio-Temporal Dynamics of GLUT4 in Cardiomyocytes

Individuals with Type 2 diabetes mellitus (T2DM) are at significantly increased risk of cardiovascular disease and the development of diabetic cardiomyopathy. One of the major hallmarks of T2DM is a significant decrease in myocardial glucose uptake. Glucose transport in fat, skeletal and cardiac muscle is mediated by the glucose transporter GLUT4 that is mobilized to the cell surface from intracellular stores in response to insulin. Recently, total internal reflection microscopy was used to reveal that insulin not only regulates GLUT4 translocation to the plasma membrane (PM) but also its distribution within the PM. The emergence of super resolution microscopy techniques has allowed cell biologists to examine characteristics of molecules such as GLUT4 organization in the PM at the single-molecule level beyond the diffraction limit of conventional light microscopy. Stochastic reconstruction microscopy (STORM) investigated GLUT4 dispersal in basal and insulin-stimulated adipocytes and suggested that dispersal of GLUT4 within the plasma membrane was a key action of insulin. In my project, I have applied STORM with state of the art nanobody technology to investigate spatio-temporal dynamics of GLUT4 dispersal and significantly reduce fluorescent label linkage error using cardiomyocytes as a model system. Super resolution images showed that GLUT4 clusters are present in the basal state and retained at the plasma membrane fusion site; by contrast insulin stimulated a different kind of exocytosis that dispersed GLUT4 to the PM consistent with the findings of previous studies. GLUT4 trafficking is more complex in muscle tissues and therefore our ongoing work is focused on elucidating GLUT4 dispersal in cardiomyocytes. In particular key research questions include does exercise regulate GLUT4 dispersal, how specific is the dispersal process and it is impaired in conditions of insulin resistance.

Reconciliation of inconsistent data sources using hidden Markov models

This paper discusses how National Statistical Institutes (NSI’s) can use hidden Markov models (HMMs) to produce consistent official statistics for categorical, longitudinal variables using inconsistent sources. Two main challenges are addressed: first, the reconciliation of inconsistent sources with multi-indicator HMMs requires linking the sources on the micro level. Such linkage might lead to bias due to linkage error. Second, applying and estimating HMMs regularly is a complicated and expensive procedure. Therefore, it is preferable to use the error parameter estimates as a correction factor for a number of years. However, this might lead to biased structural estimates if measurement error changes over time or if the data collection process changes. Our results on these issues are highly encouraging and imply that the suggested method is appropriate for NSI’s. Specifically, linkage error only leads to (substantial) bias in very extreme scenarios. Moreover, measurement error parameters are largely stable over time if no major changes in the data collection process occur. However, when a substantial change in the data collection process occurs, such as a switch from dependent (DI) to independent (INDI) interviewing, re-using measurement error estimates is not advisable.

Record Linkage Approaches Using Prescription Drug Monitoring Program and Mortality Data for Public Health Analyses and Epidemiologic Studies.

The use of Prescription Drug Monitoring Program (PDMP) data has greatly increased in recent years as these data have accumulated as part of the response to the opioid epidemic in the United States. We evaluated the accuracy of record linkage approaches using the Controlled Substance Monitoring Database (Tennessee's [TN] PDMP, 2012-2016) and mortality data on all drug overdose decedents in Tennessee (2013-2016). We compared total, missed, and false positive (FP) matches (with manual verification of all FPs) across approaches that included a variety of data cleaning and matching methods (probabilistic/fuzzy vs. deterministic) for patient and death linkages, and prescription history. We evaluated the influence of linkage approaches on key prescription measures used in public health analyses. We evaluated characteristics (e.g., age, education, sex) of missed matches and incorrect matches to consider potential bias. The most accurate probabilistic/fuzzy matching approach identified 4,714 overdose deaths (vs. the deterministic approach, n = 4,572), with a low FP linkage error (<1%) and high correct match proportion (95% vs. 92% and ~90% for probabilistic approaches not using comprehensive data cleaning). Estimation of all prescription measures improved (vs. deterministic approach). For example, frequency (%) of decedents filling an oxycodone prescription in the last 60 days (n = 1,371 [32%] vs. n = 1,443 [33%]). Missed overdose decedents were more likely to be younger, male, nonwhite, and of higher education. Implications of study findings include underreporting, prescribing and outcome misclassification, and reduced generalizability to population risk groups, information of importance to epidemiologists and researchers using PDMP data.

Linking Errors between Two Populations and Tests: A Case Study in International Surveys in Education.

This simulation study was prompted by the current increased interest in linking national studies to international large-scale assessments (ILSAs) such as IEA’s TIMSS, IEA’s PIRLS, and OECD’s PISA. Linkage in this scenario is achieved by including items from the international assessments in the national assessments on the premise that the average achievement scores from the latter can be linked to the international metric. In addition to raising issues associated with different testing conditions, administrative procedures, and the like, this approach also poses psychometric challenges. This paper endeavors to shed some light on the effects that can be expected, the linkage errors in particular, by countries using this practice. The ILSA selected for this simulation study was IEA TIMSS 2011, and the three countries used as the national assessment cases were Botswana, Honduras, and Tunisia, all of which participated in TIMSS 2011. The items selected as items common to the simulated national tests and the international test came from the Grade 4 TIMSS 2011 mathematics items that IEA released into the public domain after completion of this assessment. The findings of the current study show that linkage errors seemed to achieve acceptable levels if 30 or more items were used for the linkage, although the errors were still significantly higher compared to the TIMSS’ cutoffs. Comparison of the estimated country averages based on the simulated national surveys and the averages based on the international TIMSS assessment revealed only one instance across the three countries of the estimates approaching parity. Also, the percentages of students in these countries who actually reached the defined benchmarks on the TIMSS achievement scale differed significantly from the results based on TIMSS and the results for the simulated national assessments. As a conclusion, we advise against using groups of released items from international assessments in national assessments in order to link the results of the former to the latter.

Allocating Unique Property Reference Numbers (UPRNs) to general practitioner-recorded patient addresses using a deterministic address-matching algorithm: evaluation of representativeness and bias in an ethnically-diverse inner city population

Background with rationalePseudonymised UPRNs based on patient addresses can be used to link environmental information to electronic health records (EHRs), however the representativeness and potential demographic or health-related biases in linkage using existing address-matching algorithms have not been evaluated using patient addresses.&#x0D; Main AimTo evaluate representativeness and bias in assigning UPRNs using an address-matching algorithm based on general practitioner (GP)-recorded patient addresses for a geographically-defined multi-ethnic inner city population.&#x0D; MethodsWe evaluated the Discovery Programme deterministic address-matching algorithm, comprising 213 rules applied, in rank order of minimising false positives, to the GP-recorded address of 879,286 (48% female) patients currently registered with all GP practices in four boroughs in inner east London.&#x0D; We used logistic regression to estimate the adjusted odds (aOR) of an address not being linked to a UPRN by: age band (reference group: &lt;1 year), sex (female), ethnic group (White British), Index of Multiple Deprivation (IMD) quintile (most deprived), number of long-term conditions (none); and timing of GP registration (most recent quartile). We evaluated the linkage and algorithm error rates in an independent validated NHS address dataset using best practice linkage reporting standards.&#x0D; Results99% of patients had a UPRN assigned. Men (aOR;95%CI:0.87;0.8,0.91), and patients aged 15-19 (0.51;0.39,0.68), 20-24 (0.67;0.51,0.89), or ≥90 years (0.35;0.83,0.91), of Chinese ethnic background (95% CI; 0.50; 0.45,0.56), or living in the least deprived IMD quintile (0.24; 0.20,0.30) were less likely, and those with a GP-registration preceding mid-2016 (p-value0.00) more likely, to have a UPRN assigned. The sensitivity, specificity, positive and negative predictive-values and F-measure of the algorithm were, respectively: 0.993, 0.019, 0.914, 0.204, and 0.9516.&#x0D; ConclusionWe have demonstrated, for the first time, a high GP-address UPRN match rate and quantified error rates and biases for users. Further work is needed to investigate addresses in patients with more complex address histories.

Prevalence three ways: Comparison of linked data from a patient register and electronic health records with allowance for linkage error

Background with rationalePatient registers and electronic health records are both valuable resources for disease surveillance but can be limited by variation in data quality over time. Variation may stem from changes in data collection methods, in the accuracy or completeness of clinical information, or in the quality of patient identifiers and the linkage that relies on these.&#x0D; Main AimBy linking the National Down Syndrome Cytogenetic Register (NDSCR) to Hospital Episode Statistics for England (HES), we aimed to assess the quality of each and establish a consistent approach for analysis of trends in prevalence of Down’s syndrome among live births in England.&#x0D; Methods/ApproachProbabilistic record linkage of NDSCR to HES for the period 1998–2013, supported by linkage of babies to mothers within HES. Comparison of prevalence estimates in England using NDSCR only, HES data only, and linked data. Capture-recapture analysis and quantitative bias analysis were used to account for potential errors, including false positive diagnostic codes, unrecorded diagnoses, and linkage error.&#x0D; ResultsAnalyses of single-source data indicated increasing live birth prevalence of Down’s syndrome, particularly steep in analysis of HES. Linked data indicated a contrastingly stable prevalence of 12.3 cases per 10,000 live births, with a plausible range of 11.6–12.7 cases per 10,000 live births allowing for potential errors.&#x0D; Conclusion Case ascertainment in NDSCR improved slightly over time, creating a picture of slowly increasing prevalence. The emerging epidemic suggested by HES primarily reflects improving linkage within HES (assignment of unique patient identifiers to hospital episodes). Administrative data are valuable but trends should be interpreted with caution, and with assessment of data quality over time. Linked data with quantitative bias analysis can provide more robust estimation and, in this case, reassurance that prevalence of Down’s syndrome is not increasing. Routine linkage of administrative and register data can enhance the value of each.

An instrumental variable approach to estimation of match probabilities or precision in linked data

Background with rationaleWhile probabilistic linkage methods are ostensibly based on the probabilities of record pairs being matches (i.e. their marginal precision or positive predictive value), in practice they are used principally for ranking candidate links and fall short of supporting estimation of absolute probabilities. A few variations on Fellegi and Sunter’s framework have been proposed to better accommodate the dependencies that limit transformation of match weights into match probabilities, but there are almost no alternative frameworks for match probability estimation.&#x0D; Main AimTo explore the feasibility, accuracy and limitations of a novel instrumental variable approach to estimation of match probabilities for use in either probabilistic record linkage or evaluation of linkage error.&#x0D; Methods/ApproachUsing both simulated data and a gold standard (labelled) dataset derived from real-world linked data, I assessed the accuracy of match probability estimation for a range of potential instruments and compared results to estimates produced using conventional probabilistic techniques.&#x0D; ResultsThe technique involves trading the potential value of one matching variable in discriminating between candidate links for improved estimation of match probabilities within groups of otherwise similar candidates. Analysis of simulated data confirmed the theoretical validity of the approach in supporting unbiased estimation of match probabilities despite dependencies between other matching variables. Analysis of real-world data demonstrated feasibility in terms of the availability of real-world instruments that provided sufficiently accurate estimation in groups of candidate links above a minimum size. Invalid instruments produced estimates that could be strongly biased.&#x0D; ConclusionThese early results are promising but the general availability of valid instruments, their ‘affordability’ in terms of sacrificed discrimination, and means for identifying valid instruments remain unclear. However, this approach represents a new variety of tool for the data linker’s toolkit, which may provide a useful angle on an otherwise difficult-to-estimate parameter and have applications yet to be envisaged.