Opening the black box of record linkage

Infant formula composition and educational performance: a protocol to extend follow-up for a set of randomised controlled trials using linked administrative education records

INTRODUCTION

The effect of infant nutrition on long-term cognition is important for parents and policy makers. However, most clinical trials typically have short follow-up periods, when measures of cognition are poorly predictive of later function. The few trials with longer-term follow-up have high levels of attrition, which can lead to selection bias, and in turn to erroneous interpretation of long-term harms and benefits of infant nutrition. We address the need for unbiased, long-term follow-up, by linking measures of educational performance from administrative education records. Educational performance is a meaningful marker of cognitive function in children and it is strongly correlated with IQ. We aim to evaluate educational performance for children who, as infants, were part of a series of trials that randomised participants to either nutritionally modified infant formula or standard formula. Most trialists anticipated positive effects of these interventions on later cognitive function.

METHODS AND ANALYSIS

Using data from 1923 participants of seven randomised infant formula trials linked to the English National Pupil Database (NPD), this study will provide new insights into the effect of nutrient intake in infancy on school achievement. Our primary outcome will be the mean differences in z-scores between intervention and control groups for a compulsory Mathematics exam sat at age 16. Secondary outcomes will be z-scores for a compulsory English exam at age 16 and z-scores for compulsory Mathematics and English exams at age 11. We will also evaluate intervention effects on the likelihood of receiving special educational needs (SEN) support. All analyses will be performed separately by trial.

ETHICS AND DISSEMINATION

Research ethics approval, and approval from the Health Research Authority Confidentiality Advisory Group, has been obtained for this study. The results of this study will be disseminated to scientific, practitioner, and lay audiences, submitted for publication in peer-reviewed journals, and will contribute towards a PhD dissertation.

Population Data Centre Profiles: Centre for Data Linkage

The Centre for Data Linkage (CDL) was established at Curtin University, Western Australia, to develop infrastructure to enable cross-jurisdictional record linkage in Australia. The CDL’s operating model makes use of the ‘separation principle’, with content data typically provided to researchers directly by the data custodian; jurisdictional linkage where available are used within the linkage process. Along with conducting record linkage, the team has also invested in establishing a research programme in record linkage methodology and in developing modern record linkage software which can handle the size and complexity of today’s workloads. The Centre has been instrumental in the development of practical methods for privacy-preserving record linkage, with this methodology now regularly used for real-world linkages. While the promise of a nation-wide linkage system in Australia has yet to be met, distributed models provide a potential solution.

Overcoming limitations of tuberculosis information systems: researcher and clinician perspectives

Setting

Tuberculosis (TB) diagnosis and treatment requires patients to have multiple encounters with health care systems and the different stakeholders who play a role in curing them to coordinate their efforts. To optimize this process, high-quality, readily available data are required. Data systems to facilitate these linkages are a neglected priority which, if weak, fundamentally undermine TB control interventions.

Objective

To describe lessons learnt from the use of programmatic data for TB patient care and research.

Design

We did a survey of researcher and clinical provider experiences with information systems and developed a tiered approach to addressing frequently reported barriers to high-quality care.

Results

Unreliable linkages, incomplete data, lack of a reliable unique patient identifier, and lack of data management expertise were the most important data-related barriers to high-quality patient care and research. We propose the creation of health service delivery environments that facilitate, prioritize, and evaluate high-quality data entry during patient or specimen registration.

Conclusion

An integrated approach, focused on high-quality data, and centered on unique patient identification will form the foundation for linkages across health systems that reduce patient management errors, bolster surveillance, and enhance the quality of research based on programmatic data.

Approach to record linkage of primary care data from Clinical Practice Research Datalink to other health-related patient data: overview and implications

Record linkage is increasingly used to expand the information available for public health research. An understanding of record linkage methods and the relevant strengths and limitations is important for robust analysis and interpretation of linked data. Here, we describe the approach used by Clinical Practice Research Datalink (CPRD) to link primary care data to other patient level datasets, and the potential implications of this approach for CPRD data analysis. General practice electronic health record software providers separately submit de-identified data to CPRD and patient identifiers to NHS Digital, excluding patients who have opted-out from contributing data. Data custodians for external datasets also send patient identifiers to NHS Digital. NHS Digital uses identifiers to link the datasets using an 8-stage deterministic methodology. CPRD subsequently receives a de-identified linked cohort file and provides researchers with anonymised linked data and metadata detailing the linkage process. This methodology has been used to generate routine primary care linked datasets, including data from Hospital Episode Statistics, Office for National Statistics and National Cancer Registration and Analysis Service. 10.6 million (M) patients from 411 English general practices were included in record linkage in June 2018. 9.1M (86%) patients were of research quality, of which 8.0M (88%) had a valid NHS number and were eligible for linkage in the CPRD standard linked dataset release. Linking CPRD data to other sources improves the range and validity of research studies. This manuscript, together with metadata generated on match strength and linkage eligibility, can be used to inform study design and explore potential linkage-related selection and misclassification biases.

Approach to record linkage of primary care data from Clinical Practice Research Datalink to other health-related patient data: overview and implications

Record linkage is increasingly used to expand the information available for public health research. An understanding of record linkage methods and the relevant strengths and limitations is important for robust analysis and interpretation of linked data. Here, we describe the approach used by Clinical Practice Research Datalink (CPRD) to link primary care data to other patient level datasets, and the potential implications of this approach for CPRD data analysis. General practice electronic health record software providers separately submit de-identified data to CPRD and patient identifiers to NHS Digital, excluding patients who have opted-out from contributing data. Data custodians for external datasets also send patient identifiers to NHS Digital. NHS Digital uses identifiers to link the datasets using an 8-stage deterministic methodology. CPRD subsequently receives a de-identified linked cohort file and provides researchers with anonymised linked data and metadata detailing the linkage process. This methodology has been used to generate routine primary care linked datasets, including data from Hospital Episode Statistics, Office for National Statistics and National Cancer Registration and Analysis Service. 10.6 million (M) patients from 411 English general practices were included in record linkage in June 2018. 9.1M (86%) patients were of research quality, of which 8.0M (88%) had a valid NHS number and were eligible for linkage in the CPRD standard linked dataset release. Linking CPRD data to other sources improves the range and validity of research studies. This manuscript, together with metadata generated on match strength and linkage eligibility, can be used to inform study design and explore potential linkage-related selection and misclassification biases.

Using the RECORD guidelines to improve transparent reporting of studies based on routinely collected data

Transparent reporting of routinely-collected data studies is key to producing valid and reliable research that can inform decisions about patient care and health systems. This article discusses some of the unique challenges in using these data sources, and explains how the REporting of studies Conducted using Observational Routinely-collected Data (RECORD) guidelines were developed to help researchers and journals to maintain a high level of quality in reporting of healthcare studies using routinely-collected data.

Individual Data Linkage of Survey Data with Claims Data in Germany-An Overview Based on a Cohort Study

Research based on health insurance data has a long tradition in Germany. By contrast, data linkage of survey data with such claims data is a relatively new field of research with high potential. Data linkage opens up new opportunities for analyses in the field of health services research and public health. Germany has comprehensive rules and regulations of data protection that have to be followed. Therefore, a written informed consent is needed for individual data linkage. Additionally, the health system is characterized by heterogeneity of health insurance. The lidA-living at work-study is a cohort study on work, age and health, which linked survey data with claims data of a large number of statutory health insurance data. All health insurance funds were contacted, of whom a written consent was given. This paper will give an overview of individual data linkage of survey data with German claims data on the example of the lidA-study results. The challenges and limitations of data linkage will be presented. Despite heterogeneity, such kind of studies is possible with a negligibly small influence of bias. The experience we gain in lidA will be shown and provide important insights for other studies focusing on data linkage.

Linking routinely collected social work, education and health data to enable monitoring of the health and health care of school-aged children in state care ('looked after children') in Scotland: a national demonstration project

BACKGROUND AND OBJECTIVES

Children in state care ('looked after children') have poorer health than children who are not looked after. Recent developments in Scotland and elsewhere have aimed to improve services and outcomes for looked after children. Routine monitoring of the health outcomes of looked after children compared to those of their non-looked after peers is currently lacking. Developing capacity for comparative monitoring of population-based outcomes based on linkage of routinely collected administrative data has been identified as a priority. To our knowledge there are no existing population-based data linkage studies providing data on the health of looked after and non-looked after children at national level. Smaller scale studies that are available generally provide very limited information on linkage methods and hence do not allow scrutiny of bias that may be introduced through the linkage process.

STUDY DESIGN AND METHODS

National demonstration project testing the feasibility of linking routinely collected looked after children, education and health data.

PARTICIPANTS

All children in publicly funded school in Scotland in 2011/12.

RESULTS

Linkage between looked after children data and the national pupil census classified 10,009 (1.5%) and 1757 (0.3%) of 670,952 children as, respectively, currently and previously looked after. Recording of the unique pupil identifier (Scottish Candidate Number, SCN) on looked after children returns is incomplete, with 66% of looked after records for 2011/12 for children of possible school age containing a valid SCN. This will have resulted in some under-ascertainment of currently and, particularly, previously looked after children within the general pupil population. Further linkage of the pupil census to the National Health Service Scotland master patient index demonstrated that a safe link to the child's unique health service (Community Health Index) number could be obtained for a very high proportion of children in each group (94%, 95% and 95% of children classified as currently, previously, and non-looked after, respectively). In general, linkage rates were higher for older children and those living in more affluent areas. Within the looked after group, linkage rates were highest for children with the fewest placements and for those in permanent fostering.

CONCLUSIONS

This novel data linkage demonstrates the feasibility of monitoring population-based health outcomes of school-aged looked after and non-looked after children using linked routine administrative data. Improved recording of the unique pupil identifier number on looked after data returns would be beneficial. Extending the range of personal identifiers on looked after children returns would enable linkage to health data for looked after children who are not in publicly funded schooling (i.e. those who are preschool or postschool, home schooled or in independent schooling).

Utilising identifier error variation in linkage of large administrative data sources

Background

Linkage of administrative data sources often relies on probabilistic methods using a set of common identifiers (e.g. sex, date of birth, postcode). Variation in data quality on an individual or organisational level (e.g. by hospital) can result in clustering of identifier errors, violating the assumption of independence between identifiers required for traditional probabilistic match weight estimation. This potentially introduces selection bias to the resulting linked dataset. We aimed to measure variation in identifier error rates in a large English administrative data source (Hospital Episode Statistics; HES) and to incorporate this information into match weight calculation.

Methods

We used 30,000 randomly selected HES hospital admissions records of patients aged 0–1, 5–6 and 18–19 years, for 2011/2012, linked via NHS number with data from the Personal Demographic Service (PDS; our gold-standard). We calculated identifier error rates for sex, date of birth and postcode and used multi-level logistic regression to investigate associations with individual-level attributes (age, ethnicity, and gender) and organisational variation. We then derived: i) weights incorporating dependence between identifiers; ii) attribute-specific weights (varying by age, ethnicity and gender); and iii) organisation-specific weights (by hospital). Results were compared with traditional match weights using a simulation study.

Results

Identifier errors (where values disagreed in linked HES-PDS records) or missing values were found in 0.11% of records for sex and date of birth and in 53% of records for postcode. Identifier error rates differed significantly by age, ethnicity and sex (p < 0.0005). Errors were less frequent in males, in 5–6 year olds and 18–19 year olds compared with infants, and were lowest for the Asian ethic group. A simulation study demonstrated that substantial bias was introduced into estimated readmission rates in the presence of identifier errors. Attribute- and organisational-specific weights reduced this bias compared with weights estimated using traditional probabilistic matching algorithms.

Conclusions

We provide empirical evidence on variation in rates of identifier error in a widely-used administrative data source and propose a new method for deriving match weights that incorporates additional data attributes. Our results demonstrate that incorporating information on variation by individual-level characteristics can help to reduce bias due to linkage error.

Linking Data for Mothers and Babies in De-Identified Electronic Health Data

OBJECTIVE

Linkage of longitudinal administrative data for mothers and babies supports research and service evaluation in several populations around the world. We established a linked mother-baby cohort using pseudonymised, population-level data for England.

DESIGN AND SETTING

Retrospective linkage study using electronic hospital records of mothers and babies admitted to NHS hospitals in England, captured in Hospital Episode Statistics between April 2001 and March 2013.

RESULTS

Of 672,955 baby records in 2012/13, 280,470 (42%) linked deterministically to a maternal record using hospital, GP practice, maternal age, birthweight, gestation, birth order and sex. A further 380,164 (56%) records linked using probabilistic methods incorporating additional variables that could differ between mother/baby records (admission dates, ethnicity, 3/4-character postcode district) or that include missing values (delivery variables). The false-match rate was estimated at 0.15% using synthetic data. Data quality improved over time: for 2001/02, 91% of baby records were linked (holding the estimated false-match rate at 0.15%). The linked cohort was representative of national distributions of gender, gestation, birth weight and maternal age, and captured approximately 97% of births in England.

CONCLUSION

Probabilistic linkage of maternal and baby healthcare characteristics offers an efficient way to enrich maternity data, improve data quality, and create longitudinal cohorts for research and service evaluation. This approach could be extended to linkage of other datasets that have non-disclosive characteristics in common.

Black-White Disparities in Adult Mortality: Implications of Differential Record Linkage for Understanding the Mortality Crossover

Mortality rates among black individuals exceed those of white individuals throughout much of the life course. The black-white disparity in mortality rates is widest in young adulthood, and then rates converge with increasing age until a crossover occurs at about age 85 years, after which black older adults exhibit a lower mortality rate relative to white older adults. Data quality issues in survey-linked mortality studies may hinder accurate estimation of this disparity and may even be responsible for the observed black-white mortality crossover, especially if the linkage of surveys to death records during mortality follow-up is less accurate for black older adults. This study assesses black-white differences in the linkage of the 1986-2009 National Health Interview Survey to the National Death Index through 2011 and the implications of racial/ethnic differences in record linkage for mortality disparity estimates. Match class and match score (i.e., indicators of linkage quality) differ by race/ethnicity, with black adults exhibiting less certain matches than white adults in all age groups. The magnitude of the black-white mortality disparity varies with alternative linkage scenarios, but convergence and crossover continue to be observed in each case. Beyond black-white differences in linkage quality, this study also identifies declines over time in linkage quality and even eligibility for linkage among all adults. Although linkage quality is lower among black adults than white adults, differential record linkage does not account for the black-white mortality crossover.

[The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement]

Zunehmend werden routinemäßig gesammelte Gesundheitsdaten, die zu administrativen und klinischen Zwecken und ohne spezifische, a priori festgelegte Forschungsziele erhoben wurden, auch für die Forschung eingesetzt. Die rasche Entwicklung und Verfügbarkeit dieser Daten machten Probleme deutlich, die in den bestehenden Berichts-Leitlinien, wie dem STROBE-Statement (Strengthening the Reporting of Observational Studies in Epidemiology) nicht behandelt werden. Das RECORD-Statement (REporting of studies Conducted using Observational Routinely-collected health Data) wurde entwickelt, um diese Lücken zu schließen. RECORD ist als Erweiterung des STROBE-Statements gedacht, um Punkte abzudecken, die spezifisch sind beim Berichten von Beobachtungsstudien, die routinemäßig gesammelte Gesundheitsdaten verwenden. RECORD besteht aus einer Checkliste von 13 Punkten mit Bezug zu Titel, Abstract, Einleitung, Methoden-, Ergebnis- und Diskussionsteil von Artikeln sowie zu anderen Informationen, die in Forschungsberichten dieser Art enthalten sein sollten. Dieses Dokument enthält die Checkliste sowie Erläuterungen und weitere Erklärungen, um die Verwendung der Checkliste zu verbessern. Beispiele für ein gutes Berichten der einzelnen Punkte der RECORD-Checkliste sind ebenfalls in diesem Dokument enthalten. Dieses Dokument sowie die zugehörige Website und ein Forum (http://www.record-statement.org) werden die Umsetzung und das Verständnis von RECORD verbessern. Autoren, Redakteure von Fachzeitschriften und Peer-Reviewer können die Transparenz beim Berichten von Forschungsergebnissen erhöhen, indem sie RECORD anwenden.

A Simple Sampling Method for Estimating the Accuracy of Large Scale Record Linkage Projects

BACKGROUND

Record linkage techniques allow different data collections to be brought together to provide a wider picture of the health status of individuals. Ensuring high linkage quality is important to guarantee the quality and integrity of research. Current methods for measuring linkage quality typically focus on precision (the proportion of incorrect links), given the difficulty of measuring the proportion of false negatives.

OBJECTIVES

The aim of this work is to introduce and evaluate a sampling based method to estimate both precision and recall following record linkage.

METHODS

In the sampling based method, record-pairs from each threshold (including those below the identified cut-off for acceptance) are sampled and clerically reviewed. These results are then applied to the entire set of record-pairs, providing estimates of false positives and false negatives. This method was evaluated on a synthetically generated dataset, where the true match status (which records belonged to the same person) was known.

RESULTS

The sampled estimates of linkage quality were relatively close to actual linkage quality metrics calculated for the whole synthetic dataset. The precision and recall measures for seven reviewers were very consistent with little variation in the clerical assessment results (overall agreement using the Fleiss Kappa statistics was 0.601).

CONCLUSIONS

This method presents as a possible means of accurately estimating matching quality and refining linkages in population level linkage studies. The sampling approach is especially important for large project linkages where the number of record pairs produced may be very large often running into millions.

Probabilistic record linkage

Studies involving the use of probabilistic record linkage are becoming increasingly common. However, the methods underpinning probabilistic record linkage are not widely taught or understood, and therefore these studies can appear to be a ‘black box’ research tool. In this article, we aim to describe the process of probabilistic record linkage through a simple exemplar. We first introduce the concept of deterministic linkage and contrast this with probabilistic linkage. We illustrate each step of the process using a simple exemplar and describe the data structure required to perform a probabilistic linkage. We describe the process of calculating and interpreting matched weights and how to convert matched weights into posterior probabilities of a match using Bayes theorem. We conclude this article with a brief discussion of some of the computational demands of record linkage, how you might assess the quality of your linkage algorithm, and how epidemiologists can maximize the value of their record-linked research using robust record linkage methods.

The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement

Routinely collected health data, obtained for administrative and clinical purposes without specific a priori research goals, are increasingly used for research. The rapid evolution and availability of these data have revealed issues not addressed by existing reporting guidelines, such as Strengthening the Reporting of Observational Studies in Epidemiology (STROBE). The REporting of studies Conducted using Observational Routinely collected health Data (RECORD) statement was created to fill these gaps. RECORD was created as an extension to the STROBE statement to address reporting items specific to observational studies using routinely collected health data. RECORD consists of a checklist of 13 items related to the title, abstract, introduction, methods, results, and discussion section of articles, and other information required for inclusion in such research reports. This document contains the checklist and explanatory and elaboration information to enhance the use of the checklist. Examples of good reporting for each RECORD checklist item are also included herein. This document, as well as the accompanying website and message board (http://www.record-statement.org), will enhance the implementation and understanding of RECORD. Through implementation of RECORD, authors, journals editors, and peer reviewers can encourage transparency of research reporting.

Optimization is required when using linked hospital and laboratory data to investigate respiratory infections

OBJECTIVE

Despite a recommendation for microbiological testing, only 45% of children hospitalized for respiratory infections in our previous data linkage study linked to a microbiological record. We conducted a chart review to validate linked microbiological data.

STUDY DESIGN AND SETTING

The chart review consisted of children aged <5 years admitted to seven selected hospitals for respiratory infections in Western Australia, 2000-2011. We calculated the proportion of admissions where testing was performed and any pathogens detected. We compared these proportions between the chart review and our previous data linkage study. Poisson regression was used to identify factors predicting the likelihood of microbiological tests in the chart review cohort.

RESULTS

From the chart review, 77% of 746 records had a microbiological test performed compared with 46% of 18,687 records from our previous data linkage study. Of those undergoing testing, 66% of the chart review and 64% of data linkage records had ≥1 respiratory pathogen(s) detected. In the chart review cohort, frequency of testing was highest in children admitted to metropolitan hospitals.

CONCLUSION

Validation studies are essential to ensure the quality of linked data. Our previous data linkage study failed to capture all relevant microbiological records. Findings will be used to optimize extraction protocols for future linkage studies.

Accuracy and completeness of patient pathways--the benefits of national data linkage in Australia

BACKGROUND

The technical challenges associated with national data linkage, and the extent of cross-border population movements, are explored as part of a pioneering research project. The project involved linking state-based hospital admission records and death registrations across Australia for a national study of hospital related deaths.

METHODS

The project linked over 44 million morbidity and mortality records from four Australian states between 1st July 1999 and 31st December 2009 using probabilistic methods. The accuracy of the linkage was measured through a comparison with jurisdictional keys sourced from individual states. The extent of cross-border population movement between these states was also assessed.

RESULTS

Data matching identified almost twelve million individuals across the four Australian states. The percentage of individuals from one state with records found in another ranged from 3-5%. Using jurisdictional keys to measure linkage quality, results indicate a high matching efficiency (F measure 97 to 99%), with linkage processing taking only a matter of days.

CONCLUSIONS

The results demonstrate the feasibility and accuracy of undertaking cross jurisdictional linkage for national research. The benefits are substantial, particularly in relation to capturing the full complement of records in patient pathways as a result of cross-border population movements. The project identified a sizeable 'mobile' population with hospital records in more than one state. Research studies that focus on a single jurisdiction will under-enumerate the extent of hospital usage by individuals in the population. It is important that researchers understand and are aware of the impact of this missing hospital activity on their studies. The project highlights the need for an efficient and accurate data linkage system to support national research across Australia.

The REporting of Studies Conducted Using Observational Routinely-Collected Health Data (RECORD) Statement: Methods for Arriving at Consensus and Developing Reporting Guidelines

Objective

Routinely collected health data, collected for administrative and clinical purposes, without specific a priori research questions, are increasingly used for observational, comparative effectiveness, health services research, and clinical trials. The rapid evolution and availability of routinely collected data for research has brought to light specific issues not addressed by existing reporting guidelines. The aim of the present project was to determine the priorities of stakeholders in order to guide the development of the REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement.

Methods

Two modified electronic Delphi surveys were sent to stakeholders. The first determined themes deemed important to include in the RECORD statement, and was analyzed using qualitative methods. The second determined quantitative prioritization of the themes based on categorization of manuscript headings. The surveys were followed by a meeting of RECORD working committee, and re-engagement with stakeholders via an online commentary period.

Results

The qualitative survey (76 responses of 123 surveys sent) generated 10 overarching themes and 13 themes derived from existing STROBE categories. Highest-rated overall items for inclusion were: Disease/exposure identification algorithms; Characteristics of the population included in databases; and Characteristics of the data. In the quantitative survey (71 responses of 135 sent), the importance assigned to each of the compiled themes varied depending on the manuscript section to which they were assigned. Following the working committee meeting, online ranking by stakeholders provided feedback and resulted in revision of the final checklist.

Conclusions

The RECORD statement incorporated the suggestions provided by a large, diverse group of stakeholders to create a reporting checklist specific to observational research using routinely collected health data. Our findings point to unique aspects of studies conducted with routinely collected health data and the perceived need for better reporting of methodological issues.

Evaluating bias due to data linkage error in electronic healthcare records

Background

Linkage of electronic healthcare records is becoming increasingly important for research purposes. However, linkage error due to mis-recorded or missing identifiers can lead to biased results. We evaluated the impact of linkage error on estimated infection rates using two different methods for classifying links: highest-weight (HW) classification using probabilistic match weights and prior-informed imputation (PII) using match probabilities.

Methods

A gold-standard dataset was created through deterministic linkage of unique identifiers in admission data from two hospitals and infection data recorded at the hospital laboratories (original data). Unique identifiers were then removed and data were re-linked by date of birth, sex and Soundex using two classification methods: i) HW classification - accepting the candidate record with the highest weight exceeding a threshold and ii) PII–imputing values from a match probability distribution. To evaluate methods for linking data with different error rates, non-random error and different match rates, we generated simulation data. Each set of simulated files was linked using both classification methods. Infection rates in the linked data were compared with those in the gold-standard data.

Results

In the original gold-standard data, 1496/20924 admissions linked to an infection. In the linked original data, PII provided least biased results: 1481 and 1457 infections (upper/lower thresholds) compared with 1316 and 1287 (HW upper/lower thresholds). In the simulated data, substantial bias (up to 112%) was introduced when linkage error varied by hospital. Bias was also greater when the match rate was low or the identifier error rate was high and in these cases, PII performed better than HW classification at reducing bias due to false-matches.

Conclusions

This study highlights the importance of evaluating the potential impact of linkage error on results. PII can help incorporate linkage uncertainty into analysis and reduce bias due to linkage error, without requiring identifiers.

Linkage, evaluation and analysis of national electronic healthcare data: application to providing enhanced blood-stream infection surveillance in paediatric intensive care

Background

Linkage of risk-factor data for blood-stream infection (BSI) in paediatric intensive care (PICU) with bacteraemia surveillance data to monitor risk-adjusted infection rates in PICU is complicated by a lack of unique identifiers and under-ascertainment in the national surveillance system. We linked, evaluated and performed preliminary analyses on these data to provide a practical guide on the steps required to handle linkage of such complex data sources.

Methods

Data on PICU admissions in England and Wales for 2003-2010 were extracted from the Paediatric Intensive Care Audit Network. Records of all positive isolates from blood cultures taken for children <16 years and captured by the national voluntary laboratory surveillance system for 2003-2010 were extracted from the Public Health England database, LabBase2. “Gold-standard” datasets with unique identifiers were obtained directly from three laboratories, containing microbiology reports that were eligible for submission to LabBase2 (defined as “clinically significant” by laboratory microbiologists). Reports in the gold-standard datasets were compared to those in LabBase2 to estimate ascertainment in LabBase2. Linkage evaluated by comparing results from two classification methods (highest-weight classification of match weights and prior-informed imputation using match probabilities) with linked records in the gold-standard data. BSI rate was estimated as the proportion of admissions associated with at least one BSI.

Results

Reporting gaps were identified in 548/2596 lab-months of LabBase2. Ascertainment of clinically significant BSI in the remaining months was approximately 80-95%. Prior-informed imputation provided the least biased estimate of BSI rate (5.8% of admissions). Adjusting for ascertainment, the estimated BSI rate was 6.1-7.3%.

Conclusion

Linkage of PICU admission data with national BSI surveillance provides the opportunity for enhanced surveillance but analyses based on these data need to take account of biases due to ascertainment and linkage error. This study provides a generalisable guide for linkage, evaluation and analysis of complex electronic healthcare data.

Can colorectal cancer survivors recall their medications and doctor visits reliably?

Background

The evidence on the agreement between self-reported health resource use and administrative records is mixed and no gold standard exists. The objective of this study was to assess self-reported general practitioner (GP) and specialist doctor visits, as well as medication use via telephone interview against national insurance administrative data for colorectal cancer survivors.

Methods

In a sample of 76 adults recently diagnosed with colorectal cancer, data was abstracted from telephone survey items on GP visits, specialist visits and medication use over the previous six months and compared with data on the same individuals from administrative data. Intraclass correlation coefficients (ICC) were used to assess the reliability of frequency of visits and kappa statistics were derived for four broad categories of medicines used for gastrointestinal conditions, cardiovascular disease, psychological conditions and chronic obstructive pulmonary disease. Logistic regression was undertaken to assess factors associated with agreement (yes/no) between the two data sources for doctors’ visits.

Results

Good agreement was found for GP visits (ICC 0.62, 95%CI: 0.38, 0.86) and specialist visits (ICC 0.73, 95%CI: 0.56, 0.91) across the two data sources. When costs were assigned to frequencies, mean costs for the two methods were not significantly different over six months. Over-reporting was more common among men and participants with frequent doctor encounters. Large discrepancies between self-reports and administration records were found for broad types of medications used (44% agreement, kappa 0.13).

Conclusion

Self-reported frequency of doctor’s visits using telephone interviews may be a reasonable substitute for administratively recorded data however, medication use by self-report appears to be unreliable. Administrative records are preferable to self-report for health service use in colorectal cancer survivors with high and complex service needs.