Record Linkage Research Articles

Polypharmacy as a risk factor for dementia: Scottish population‐based longitudinal record linkage study of 1 225 894 people

AbstractBackgroundPolypharmacy, the concurrent use of multiple medicines, is increasing. Population‐wide studies of its association with dementia are lacking. We examined this relationship longitudinally, at a national level.MethodWe used National Health Service community prescribing data from all adults in Scotland aged ≥50 years who received at least one drug in the first quarter of 2009. These data were linked to death records, including cause of death. We used Cox proportional hazards models to assess associations between the number of unique medicines dispensed in one quarter and mortality with any subtype of dementia over 8.5 years, in the whole sample and stratified by age.ResultThe sample contained 1,225,894 people aged ≥50 years (mean age 67.4 (SD = 10.8) years, 56.1% female, 3.8% care home residents). The mean number of drugs dispensed at baseline was 5.0 (SD = 3.7). Over 8.5 years, there were 336,244 deaths, of which 58,358 had any subtype of dementia on the death certificate.Among the whole sample, the hazard ratio (HR) for dementia mortality with each additional medicine was 1.027 (95% CI 1.024‐1.028). In people aged 50‐64 years, the HR was 1.075 (1.061‐1.089); for 65‐79 year‐olds, HR = 1.043 (1.040‐1.047) and for those aged ≥80 years, HR = 1.009 (1.006‐1.012). All models were adjusted for baseline age, gender, care home residence status and deprivation index based on postcode.ConclusionThere was higher mortality with dementia as the number of dispensed medicines increased. Age‐stratified analyses showed that the association was stronger in younger age groups, perhaps reflecting that younger people taking medication for more comorbidities had an increased risk of dying with dementia. These analyses did not allow adjustment for multimorbidity or the consideration of individual drug classes.

Data reconciliation and estimation in an agricultural survey

The Puerto Rico Census of Agriculture (Census) is the leading source of statistics about the island’s agricultural production. The Census is conducted every 5 years, in years ending in 2 and 7. However, the most recent Census was administered in 2018 due to delays caused by Hurricane Maria. Because the Census list frame is incomplete, a separate survey is conducted to inform measures of undercoverage, nonresponse, and misclassification, leading to adjusted weights. At the time of Hurricane Maria, the 2017 survey had already been conducted in preparation for the planned Census that year. In 2018, the survey was repeated using the same sample. Linking and analyzing the 2017 and 2018 survey data provide insights into the impacts of Hurricane Maria on the island’s agriculture. Furthermore, the 2018 survey gives an opportunity to evaluate the 2018 Census results. Although the same sample was used in 2017 and 2018, automated record linkage methods are not suitable to link records from the two surveys. This paper discusses record linkage and estimation approaches used to determine the number of farms and land in farms in Puerto Rico before and after Hurricane Maria based on the two surveys and the Census of Agriculture.

From Paper Files to Web-Based Application for Data-Driven Monitoring of HIV Programs: Nigeria's Journey to a National Data Repository for Decision-Making and Patient Care.

Timely and reliable data are crucial for clinical, epidemiologic, and program management decision making. Electronic health information systems provide platforms for managing large longitudinal patient records. Nigeria implemented the National Data Repository (NDR) to create a central data warehouse of all people living with human immunodeficiency virus (PLHIV) while providing useful functionalities to aid decision making at different levels of program implementation. We describe the Nigeria NDR and its development process, including its use for surveillance, research, and national HIV program monitoring toward achieving HIV epidemic control. Stakeholder engagement meetings were held in 2013 to gather information on data elements and vocabulary standards for reporting patient-level information, technical infrastructure, human capacity requirements, and information flow. Findings from these meetings guided the development of the NDR. An implementation guide provided common terminologies and data reporting structures for data exchange between the NDR and the electronic medical record (EMR) systems. Data from the EMR were encoded in extensible markup language and sent to the NDR over secure hypertext transfer protocol after going through a series of validation processes. By June 30, 2021, the NDR had up-to-date records of 1,477,064 (94.4%) patients receiving HIV treatment across 1,985 health facilities, of which 1,266,512 (85.7%) patient records had fingerprint template data to support unique patient identification and record linkage to prevent registration of the same patient under different identities. Data from the NDR was used to support HIV program monitoring, case-based surveillance and production of products like the monthly lists of patients who have treatment interruptions and dashboards for monitoring HIV test and start. The NDR enabled the availability of reliable and timely data for surveillance, research, and HIV program monitoring to guide program improvements to accelerate progress toward epidemic control.

Cluster analysis of angiotensin biomarkers to identify antihypertensive drug treatment in population studies

BackgroundThe recent progress in molecular biology generates an increasing interest in investigating molecular biomarkers as markers of response to treatments. The present work is motivated by a study, where the objective was to explore the potential of the molecular biomarkers of renin-angiotensin-aldosterone system (RAAS) to identify the undertaken antihypertensive treatments in the general population. Population-based studies offer an opportunity to assess the effectiveness of treatments in real-world scenarios. However, lack of quality documentation, especially when electronic health record linkage is unavailable, leads to inaccurate reporting and classification bias.MethodWe present a machine learning clustering technique to determine the potential of measured RAAS biomarkers for the identification of undertaken treatments in the general population. The biomarkers were simultaneously determined through a novel mass-spectrometry analysis in 800 participants of the Cooperative Health Research In South Tyrol (CHRIS) study with documented antihypertensive treatments. We assessed the agreement, sensitivity and specificity of the resulting clusters against known treatment types. Through the lasso penalized regression, we identified clinical characteristics associated with the biomarkers, accounting for the effects of cluster and treatment classifications.ResultsWe identified three well-separated clusters: cluster 1 (n = 444) preferentially including individuals not receiving RAAS-targeting drugs; cluster 2 (n = 235) identifying angiotensin type 1 receptor blockers (ARB) users (weighted kappa κw = 74%; sensitivity = 73%; specificity = 83%); and cluster 3 (n = 121) well discriminating angiotensin-converting enzyme inhibitors (ACEi) users (κw = 81%; sensitivity = 55%; specificity = 90%). Individuals in clusters 2 and 3 had higher frequency of diabetes as well as higher fasting glucose and BMI levels. Age, sex and kidney function were strong predictors of the RAAS biomarkers independently of the cluster structure.ConclusionsUnsupervised clustering of angiotensin-based biomarkers is a viable technique to identify individuals on specific antihypertensive treatments, pointing to a potential application of the biomarkers as useful clinical diagnostic tools even outside of a controlled clinical setting.

SLCO1B1*5 is protective against non-senile cataracts in cohort prescribed statins: analysis in a British-South Asian cohort.

Reported association between statin use and cataract risk is controversial. The SLCO1B1 gene encodes a transport protein responsible for statin clearance. The aim of this study was to investigate a possible association between the SLCO1B1*5 reduced function variant and cataract risk in statin users of South Asian ethnicity. The Genes & Health cohort consists of British-Bangladeshi and British-Pakistani participants from East London, Manchester and Bradford, UK. SLCO1B1*5 genotype was assessed with the Illumina GSAMD-24v3-0-EA chip. Medication data from primary care health record linkage was used to compare those who had regularly used statins compared to those who had not. Multivariable logistic regression was used to test for association between statin use and cataracts, adjusting for population characteristics and potential confounders in 36,513 participants. Multivariable logistic regression was used to test association between SLCO1B1*5 heterozygotes or homozygotes and cataracts, in subgroups having been regularly prescribed statins versus not. Statins were prescribed to 35% (12,704) of participants (average age 41 years old, 45% male). Non-senile cataract was diagnosed in 5% (1686) of participants. An apparent association between statins and non-senile cataract (12% in statin users and 0.8% in non-statin users) was negated by inclusion of confounders. In those prescribed a statin, presence of the SLCO1B1*5 genotype was independently associated with a decreased risk of non-senile cataract (OR 0.7 (CI 0.5-0.9, p 0.007)). Our findings suggest that there is no independent association between statin use and non-senile cataract risk after adjusting for confounders. Among statin users, the SLCO1B1*5 genotype is associated with a 30% risk reduction of non-senile cataracts. Stratification of on-drug cohorts by validated pharmacogenomic variants is a useful tool to support or repudiate adverse drug events in observational cohorts.

Comparison of record linkage methods

Record linkage is an important tool to enhance database integration. This is even more valuable in a scenario with more hefty budget cuts and a growing drop in response rate in traditional surveys. This strategy makes it possible to expand the crossing alternatives with variables not present in the original base. However, there are many different data pairing methods exposed in the literature. In this sense, the objective of this paper is to compare well-known methods of record linkage. The comparison was made in synthetic dataset. To compare the methods, it was adopted a quantitative approach based on the Precision, Recall, and F-Statistics metrics, using two comparison functions: Levenshtein and Jaro-Winkler. Among the six types of classifiers analyzed, the supervised methods had the best results.

The Development of Microhistorical Databases in Norway. A Historiography

Norwegian work on microdata started out with the full count 1801 census and census and vital records from around the capital. Today, most census and ministerial records from 1801 until the mid-20th century have been scanned, transcriptions are being completed, much is encoded and made available via the websites of the Digital National Archives and UiT The Arctic University of Norway. This article complements a previous publication on empirical results from historical microdata. It is primarily organized by technical issues: digitization of source materials, encoding and standardization, building of the Historical Population Register for the period since 1800, record linkage and source criticism as well as GIS. Presently, partner institutions are building the Historical Population Register with prolonged support from the Norwegian Research Council. This will contain longitudinal records of the nine million persons who lived in Norway since 1800. The register increasingly makes it possible to follow the entire population. Unique personal IDs with corresponding URLs to the person page providing links to many sources introduce a new level of historical documentation. Cross-sectional and vital records are being interlinked with automatic and manual record linkage software. Longitudinal data is available for searching as timelines and in Intermediate Data Structure format from UiT The Arctic University and for searching at Histreg.no, which also caters for manual editing. We are well on the way to creating a database that can fill the void in the two centuries before the Central Population Register starts in 1964.

Training Data Selection for Record Linkage Classification

This paper presents a new two-step approach for record linkage, focusing on the creation of high-quality training data in the first step. The approach employs the unsupervised random forest model as a similarity measure to produce a similarity score vector for record matching. Three constructions were proposed to select non-match pairs for the training data, with both balanced (symmetry) and imbalanced (asymmetry) distributions tested. The top and imbalanced construction was found to be the most effective in producing training data with 100% correct labels. Random forest and support vector machine classification algorithms were compared, and random forest with the top and imbalanced construction produced an F1-score comparable to probabilistic record linkage using the expectation maximisation algorithm and EpiLink. On average, the proposed approach using random forests and the top and imbalanced construction improved the F1-score by 1% and recall by 6.45% compared to existing record linkage methods. By emphasising the creation of high-quality training data, this new approach has the potential to improve the accuracy and efficiency of record linkage for a wide range of applications.

Frailty alone and interactively with obesity predicts heart failure: Kuopio Ischaemic Heart Disease Risk Factor Study.

We aim to evaluate the association of frailty and high body mass index with risk of incident heart failure. From the Kuopio Ischaemic Heart Disease Risk Factor Study, 408 women and 369 men, aged 61-74years were included in this study. Frailty was ascertained with the presence of 3-5 and prefrailty 1-2 of the following criteria: weight loss (highest 20% over 7years), self-reported tiredness, weakness (measured by handgrip strength), slow walking speed (walking pace), and low physical activity (lowest 20%). At the baseline, participants were allocated to frail (n=36), prefrail (n=340), and robust (n=441). HF incidents were obtained by record linkages from the national hospitalization registry in Finland up to 31 December 2019. Multivariate Cox proportional hazards regression estimated the hazard ratio (HR) of incident events, adjusted for potential confounders. Two hundred one HF events were recorded (111 in women and 90 in men) during the 14.2years follow-up. After adjustment for the age and sex, the risk of HF events was higher among prefrail (HR 1.42, 95% CI 1.08 to 1.79, P=0.02) and frail (HR 3.39, 95% CI 1.89 to 4.79, P≤0.001) compared with the robust group. After adjusting for multiple confounders result remained significant for HF indecent in prefrail [1.46 (HR 1.46, 95% CI 1.09 to 1.95, P=0.01] and frail (HR 3.33, 95% CI 1.86 to 5.70, P≤0.001). In the sensitivity analysis, significant interaction between high BMI (≥25kg/m2 ) and frailty was observed (P for interaction=0.02). The association of frailty [multivariate-adjusted HR: 2.88 (1.56 to 5.33), P≤0.001)] and prefrailty [multivariate-adjusted HR: 1.40 (1.08 to 1.91), P=0.03)] with risk of HF indecent was more pronounced in those with high BMI. Frailty is highly common in older age, and our results indicated the high risk of HF incident in frail and prefrail groups. While frailty is clinically recognized by weight loss phenotype, our finding showed that frailly and high BMI can coexist and worsen the risk of HF incidence. Further research is warranted to substantiate these results in large studies and clinical settings.

Inequity of antenatal influenza and pertussis vaccine coverage in Australia: the Links2HealthierBubs record linkage cohort study, 2012–2017

BackgroundPregnancy and early infancy are increased risk periods for severe adverse effects of respiratory infections. Aboriginal and/or Torres Strait Islander (respectfully referred to as First Nations) women and children in Australia bear a disproportionately higher burden of respiratory diseases compared to non-Indigenous women and infants. Influenza vaccines and whooping cough (pertussis) vaccines are recommended and free in every Australian pregnancy to combat these infections. We aimed to assess the equity of influenza and/or pertussis vaccination in pregnancy for three priority groups in Australia: First Nations women; women from culturally and linguistically diverse (CALD) backgrounds; and women living in remote areas or socio-economic disadvantage.MethodsWe conducted individual record linkage of Perinatal Data Collections with immunisation registers/databases between 2012 and 2017. Analysis included generalised linear mixed model, log-binomial regression with a random intercept for the unique maternal identifier to account for clustering, presented as prevalence ratios (PR) and 95% compatibility intervals (95%CI).ResultsThere were 445,590 individual women in the final cohort. Compared with other Australian women (n = 322,848), First Nations women (n = 29,181) were less likely to have received both recommended antenatal vaccines (PR 0.69, 95% CI 0.67–0.71) whereas women from CALD backgrounds (n = 93,561) were more likely to have (PR 1.16, 95% CI 1.10–1.13). Women living in remote areas were less likely to have received both vaccines (PR 0.75, 95% CI 0.72–0.78), and women living in the highest areas of advantage were more likely to have received both vaccines (PR 1.44, 95% CI 1.40–1.48).ConclusionsCompared to other groups, First Nations Australian families, those living in remote areas and/or families from lower socio-economic backgrounds did not receive recommended vaccinations during pregnancy that are the benchmark of equitable healthcare. Addressing these barriers must remain a core priority for Australian health care systems and vaccine providers. An extension of this cohort is necessary to reassess these study findings.

De-identified Bayesian personal identity matching for privacy-preserving record linkage despite errors: development and validation

BackgroundEpidemiological research may require linkage of information from multiple organizations. This can bring two problems: (1) the information governance desirability of linkage without sharing direct identifiers, and (2) a requirement to link databases without a common person-unique identifier.MethodsWe develop a Bayesian matching technique to solve both. We provide an open-source software implementation capable of de-identified probabilistic matching despite discrepancies, via fuzzy representations and complete mismatches, plus de-identified deterministic matching if required. We validate the technique by testing linkage between multiple medical records systems in a UK National Health Service Trust, examining the effects of decision thresholds on linkage accuracy. We report demographic factors associated with correct linkage.ResultsThe system supports dates of birth (DOBs), forenames, surnames, three-state gender, and UK postcodes. Fuzzy representations are supported for all except gender, and there is support for additional transformations, such as accent misrepresentation, variation for multi-part surnames, and name re-ordering. Calculated log odds predicted a proband’s presence in the sample database with an area under the receiver operating curve of 0.997–0.999 for non-self database comparisons. Log odds were converted to a decision via a consideration threshold θ and a leader advantage threshold δ. Defaults were chosen to penalize misidentification 20-fold versus linkage failure. By default, complete DOB mismatches were disallowed for computational efficiency. At these settings, for non-self database comparisons, the mean probability of a proband being correctly declared to be in the sample was 0.965 (range 0.931–0.994), and the misidentification rate was 0.00249 (range 0.00123–0.00429). Correct linkage was positively associated with male gender, Black or mixed ethnicity, and the presence of diagnostic codes for severe mental illnesses or other mental disorders, and negatively associated with birth year, unknown ethnicity, residential area deprivation, and presence of a pseudopostcode (e.g. indicating homelessness). Accuracy rates would be improved further if person-unique identifiers were also used, as supported by the software. Our two largest databases were linked in 44 min via an interpreted programming language.ConclusionsFully de-identified matching with high accuracy is feasible without a person-unique identifier and appropriate software is freely available.

Record linkage studies of drug-related deaths among adults who were released from prison to the community: a scoping review

BackgroundThere are public health concerns about an increased risk of mortality after release from prison. The objectives of this scoping review were to investigate, map and summarise evidence from record linkage studies about drug-related deaths among former adult prisoners.MethodsMEDLINE, EMBASE, PsychINFO and Web of Science were searched for studies (January 2011- September 2021) using keywords/index headings. Two authors independently screened all titles and abstracts using inclusion and exclusion criteria and subsequently screened full publications. Discrepancies were discussed with a third author. One author extracted data from all included publications using a data charting form. A second author independently extracted data from approximately one-third of the publications. Data were entered into Microsoft Excel sheets and cleaned for analysis. Standardised mortality ratios (SMRs) were pooled (where possible) using a random-effects DerSimonian-Laird model in STATA.ResultsA total of 3680 publications were screened by title and abstract, and 109 publications were fully screened; 45 publications were included. The pooled drug-related SMR was 27.07 (95%CI 13.32- 55.02; I 2 = 93.99%) for the first two weeks (4 studies), 10.17 (95%CI 3.74–27.66; I 2 = 83.83%) for the first 3–4 weeks (3 studies) and 15.58 (95%CI 7.05–34.40; I 2 = 97.99%) for the first 1 year after release (3 studies) and 6.99 (95%CI 4.13–11.83; I 2 = 99.14%) for any time after release (5 studies). However, the estimates varied markedly between studies. There was considerable heterogeneity in terms of study design, study size, location, methodology and findings. Only four studies reported the use of a quality assessment checklist/technique.ConclusionsThis scoping review found an increased risk of drug-related death after release from prison, particularly during the first two weeks after release, though drug-related mortality risk remained elevated for the first year among former prisoners. Evidence synthesis was limited as only a small number of studies were suitable for pooled analyses for SMRs due to inconsistencies in study design and methodology.

Severe Mental Illness and Type 2 Diabetes Outcomes and Complications: A Nationwide Cohort Study.

To compare cardiovascular and mortality outcomes in people with severe mental illness (SMI) versus no mental illness in a national cohort study of people with type 2 diabetes. We included adults diagnosed with type 2 diabetes between 2004 and 2018 from the national Scottish diabetes register, ascertaining history of mental illness from linked psychiatric and general hospital admission records. We identified major cardiovascular disease (CVD) events, all-cause mortality, and CVD-specific mortality through record linkage. Using Cox regression, we estimated hazard ratios (HRs) for associations between SMI and outcomes, adjusting for baseline sociodemographic and clinical characteristics, including history of CVD, comorbidity, hypertension, high cholesterol, HbA1c, BMI, alcohol use disorder, and smoking. Among 259,875 people with type 2 diabetes, 1.0%, 0.5%, and 3.0% had schizophrenia, bipolar disorder, and major depression, respectively. After adjusting for sociodemographic characteristics, the risk of major CVD events was higher in people with schizophrenia (HR 1.22, 95% CI 1.06-1.41), bipolar disorder (HR 1.58, 95% CI 1.33-1.87), and major depression (HR 1.59, 95% CI 1.49-1.70) compared with people without a history of mental illness. SMI was also associated with an approximately twofold increased risk of CVD-specific and all-cause mortality. All associations attenuated following further adjustment for clinical characteristics. Among people with diabetes, people with a history of SMI have poorer cardiovascular and mortality outcomes compared with those without mental illness. While the underlying mechanisms are further investigated, effective prevention and management of cardiovascular risk factors is needed in this high-risk group.

Completeness and Factors Affecting Community Workers' Reporting of Births and Deaths in the Countrywide Mortality Surveillance for Action in Mozambique.

Since March 2018, the Countrywide Mortality Surveillance for Action project, implemented as a national sample registration system by the Mozambique Instituto Nacional de Saude and the Instituto Nacional de Estatistica in 700 geographic clusters randomly distributed across the 11 provinces, has trained and deployed community surveillance agents (CSAs) to report births and deaths in each cluster prospectively. An independent, retrospective data collection was conducted to assess the completeness of surveillance data. Record linkage procedures were used to match households and vital events reported in the two data sources. We calculated birth and death reporting rates and used a regression model to determine factors associated with the likelihood of vital events being reported by the CSAs. Between March 2018 and December 2019, CSAs reported 54% of births (8,787/16,421) and 45% of deaths (1,726/3,867). Births of smaller cluster sizes (< 1,000 people) were more likely to be reported (adjusted odds ratio [aOR] = 1.45; 95% CI = 1.15-1.83) compared with those of larger cluster sizes (> 1,500 people). Deaths of rural clusters were more likely to be reported (aOR = 1.41; 95% CI = 1.07-1.85) than those of urban clusters. Adult deaths were more likely to be reported (aOR = 1.49; 95% CI = 1.10-2.02) than child deaths. Our findings suggest that a fully functioning sample vital registration system must adopt a dual system with high-quality surveys or other ways to estimate underregistration periodically, consider a smaller cluster size manageable by a community worker, and pay special attention to urban clusters as underreporting is larger.

ADAMS project: a genetic Association study in individuals from Diverse Ancestral backgrounds with Multiple Sclerosis based in the UK

PurposeGenetic studies of multiple sclerosis (MS) susceptibility and severity have focused on populations of European ancestry. Studying MS genetics in other ancestral groups is necessary to determine the generalisability of...

Evaluation of the reported data linkage process and associated quality issues for linked routinely collected healthcare data in multimorbidity research: a systematic methodology review

ObjectiveThe objective of this systematic review was to examine how the record linkage process is reported in multimorbidity research.MethodsA systematic search was conducted in Medline, Web of Science and Embase...

Assessing cervical precancer treatment rate among people utilising the Johannesburg public sector through record linkage of laboratory data

Assessing cervical precancer treatment rate among people utilising the Johannesburg public sector through record linkage of laboratory data

Impact of Hospitalizations due to Chronic Health Conditions on Early Child Development.

To assess the impact of hospitalization for chronic health conditions on early child development and wellbeing at school start. We conducted a longitudinal cohort study of children starting school using population-based record linkage of routinely collected admitted hospital data and standardized assessment of early childhood development (Australian Early Developmental Census: AEDC). Developmental vulnerability (DV) was defined as children scoring <10th centile in any one of five developmental domains. Children scoring <10th centile on two or more domains were considered developmentally high-risk (DHR). Children hospitalized with chronic health conditions were compared to children without hospitalizations prior to school start. Among 152,851 children with an AEDC record, 22,271 (14·6%) were hospitalized with a chronic condition. Children hospitalized with chronic health conditions were more likely to be DHR (adjusted odds ratio 1.25, 95% CI: 1.18-1.31) compared to children without hospitalizations. Children hospitalized more frequently (>7 times) or with longer duration (>2 weeks) had a 40% increased risk of being DHR (1.40, 95% CI: 1.05-1.88 and 1.40, 95% CI: 1.13-1.74, respectively). Children hospitalized with mental health/behavioral/developmental conditions had the highest risk of DHR (2.23, 95% CI: 1.72-2.90). Developmental vulnerability was increased for physical health (1.37, 95% CI: 1.30-1.45), language (1.28, 95% CI: 1.19-1.38), social competence (1.22, 95% CI: 1.16-1.29), communication (1.17, 95% CI: 1.10-1.23), and emotional maturity (1.16, 95% CI: 1.09-1.23). Frequent and longer duration hospitalizations for chronic health conditions can impact early childhood development. Research and interventions are required to support future development and well-being of children with chronic health conditions who are hospitalized.

Identifying risk factors for becoming obese: a record linkage longitudinal study in metropolitan Sydney using the 45 and Up Study

Identifying risk factors for becoming obese: a record linkage longitudinal study in metropolitan Sydney using the 45 and Up Study

Educational outcomes of children hospitalised for head injury: a population based cohort study utilising record linkage of 766,244 scottish schoolchildren

Educational outcomes of children hospitalised for head injury: a population based cohort study utilising record linkage of 766,244 scottish schoolchildren