Routinely-collected Healthcare Data Research Articles

Accuracy of heart failure ascertainment using routinely collected healthcare data: a systematic review and meta-analysis

BackgroundAscertainment of heart failure (HF) hospitalizations in cardiovascular trials is costly and complex, involving processes that could be streamlined by using routinely collected healthcare data (RCD). The utility of coded RCD for HF outcome ascertainment in randomized trials requires assessment. We systematically reviewed studies assessing RCD-based HF outcome ascertainment against “gold standard” (GS) methods to study the feasibility of using such methods in clinical trials.MethodsStudies assessing International Classification of Disease (ICD) coded RCD-based HF outcome ascertainment against GS methods and reporting at least one agreement statistic were identified by searching MEDLINE and Embase from inception to May 2021. Data on study characteristics, details of RCD and GS data sources and definitions, and test statistics were reviewed. Summary sensitivities and specificities for studies ascertaining acute and prevalent HF were estimated using a bivariate random effects meta-analysis. Heterogeneity was evaluated using I2 statistics and hierarchical summary receiver operating characteristic (HSROC) curves.ResultsA total of 58 studies of 48,643 GS-adjudicated HF events were included in this review. Strategies used to improve case identification included the use of broader coding definitions, combining multiple data sources, and using machine learning algorithms to search free text data, but these methods were not always successful and at times reduced specificity in individual studies. Meta-analysis of 17 acute HF studies showed that RCD algorithms have high specificity (96.2%, 95% confidence interval [CI] 91.5–98.3), but lacked sensitivity (63.5%, 95% CI 51.3–74.1) with similar results for 21 prevalent HF studies. There was considerable heterogeneity between studies.ConclusionsRCD can correctly identify HF outcomes but may miss approximately one-third of events. Methods used to improve case identification should also focus on minimizing false positives.

Validation and impact of algorithms for identifying variables in observational studies of routinely collected data

Background and ObjectivesAmong observational studies of routinely collected health data (RCD) for exploring treatment effects, algorithms are used to identify study variables. However, the extent to which algorithms are reliable and impact the credibility of effect estimates is far from clear. This study aimed to investigate the validation of algorithms for identifying study variables from RCD, and examine the impact of alternative algorithms on treatment effects. MethodsWe searched PubMed for observational studies published in 2018 that used RCD to explore drug treatment effects. Information regarding the reporting, validation, and interpretation of algorithms was extracted. We summarized the reporting and methodological characteristics of algorithms and validation. We also assessed the divergence in effect estimates given alternative algorithms by calculating the ratio of estimates of the primary vs. alternative analyses. ResultsA total of 222 studies were included, of which 93 (41.9%) provided a complete list of algorithms for identifying participants, 36 (16.2%) for exposure, and 132 (59.5%) for outcomes, and 15 (6.8%) for all study variables including population, exposure, and outcomes. Fifty-nine (26.6%) studies stated that the algorithms were validated, and 54 (24.3%) studies reported methodological characteristics of 66 validations, among which 61 validations in 49 studies were from the cross-referenced validation studies. Of those 66 validations, 22 (33.3%) reported sensitivity and 16 (24.2%) reported specificity. A total of 63.6% of studies reporting sensitivity and 56.3% reporting specificity used test-result-based sampling, an approach that potentially biases effect estimates. Twenty-eight (12.6%) studies used alternative algorithms to identify study variables, and 24 reported the effects estimated by primary analyses and sensitivity analyses. Of these, 20% had differential effect estimates when using alternative algorithms for identifying population, 18.2% for identifying exposure, and 45.5% for classifying outcomes. Only 32 (14.4%) studies discussed how the algorithms may affect treatment estimates. ConclusionIn observational studies of RCD, the algorithms for variable identification were not regularly validated, and–even if validated–the methodological approach and performance of the validation were often poor. More seriously, different algorithms may yield differential treatment effects, but their impact is often ignored by researchers. Strong efforts, including recommendations, are warranted to improve good practice.

Can routinely collected administrative data effectively be used to evaluate and validate endpoints used in breast cancer clinical trials? Protocol for a scoping review of the literature

BackgroundRandomized controlled trials (RCTs) are a critical component of evidence-based medicine and the evolution of patient care. However, the costs of conducting a RCT can be prohibitive. A promising approach toward reduction of costs and lessening of the burden of intensive and lengthy patient follow-up is the use of routinely collected healthcare data (RCHD), commonly called real-world data. We propose a scoping review to identify existing RCHD case definitions of breast cancer progression and survival and their diagnostic performance.MethodsWe will search MEDLINE, EMBASE, and CINAHL to identify primary studies of women with either early-stage or metastatic breast cancer, managed with established therapies, that evaluated the diagnostic accuracy of one or more RCHD-based case definitions or algorithms of disease progression (i.e., recurrence, progression-free survival, disease-free survival, or invasive disease-free survival) or survival (i.e., breast-cancer-free survival or overall survival) compared with a reference standard measure (e.g., chart review or a clinical trial dataset). Study characteristics and descriptions of algorithms will be extracted along with measures of the diagnostic accuracy of each algorithm (e.g., sensitivity, specificity, positive predictive value, negative predictive value), which will be summarized both descriptively and in structured figures/tables.DiscussionFindings from this scoping review will be clinically meaningful for breast cancer researchers globally. Identification of feasible and accurate strategies to measure patient-important outcomes will potentially reduce RCT budgets as well as lessen the burden of intensive trial follow-up on patients.Systematic review registrationOpen Science Framework (https://doi.org/10.17605/OSF.IO/6D9RS)

MP11-13 CLINICAL FRACTURE INCIDENCE IN METASTATIC HORMONE-SENSITIVE PROSTATE CANCER (mHSPC) AND RISK-REDUCTION FOLLOWING ADDITION OF ZOLEDRONIC ACID TO ANDROGEN DEPRIVATION THERAPY (ADT) WITH OR WITHOUT DOCETAXEL (DOC): LONG-TERM RESULTS FROM 2 PHASE 3 TRIALS FROM THE STAMPEDE PLATFORM PROTOCOL

You have accessJournal of UrologyCME1 Apr 2023MP11-13 CLINICAL FRACTURE INCIDENCE IN METASTATIC HORMONE-SENSITIVE PROSTATE CANCER (mHSPC) AND RISK-REDUCTION FOLLOWING ADDITION OF ZOLEDRONIC ACID TO ANDROGEN DEPRIVATION THERAPY (ADT) WITH OR WITHOUT DOCETAXEL (DOC): LONG-TERM RESULTS FROM 2 PHASE 3 TRIALS FROM THE STAMPEDE PLATFORM PROTOCOL Craig Jones, Ashwin Sachdeva, Laura Murphy, Macey Murray, Louise Brown, Eugene McCloskey, Janet Brown, Gerhardt Attard, Mahesh Parmar, Nicholas James, Matthew Sydes, and Noel Clarke Craig JonesCraig Jones More articles by this author , Ashwin SachdevaAshwin Sachdeva More articles by this author , Laura MurphyLaura Murphy More articles by this author , Macey MurrayMacey Murray More articles by this author , Louise BrownLouise Brown More articles by this author , Eugene McCloskeyEugene McCloskey More articles by this author , Janet BrownJanet Brown More articles by this author , Gerhardt AttardGerhardt Attard More articles by this author , Mahesh ParmarMahesh Parmar More articles by this author , Nicholas JamesNicholas James More articles by this author , Matthew SydesMatthew Sydes More articles by this author , and Noel ClarkeNoel Clarke More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000003226.13AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Long-term ADT is recommended for men with mHSPC: bone loss and increased fracture risk are known complications. The STAMPEDE trial compared ADT±docetaxel (Doc)±zoledronic acid (ZA): adding ZA did not improve overall survival (OS) but long-term effects of bisphosphonate treatment on skeletal health was not formally collected as part of the trial. Accessible, high quality, routinely-collected healthcare data (RCHD) enables analysis of outcome data beyond standard trial follow-up to evaluate long-term treatment toxicity and clinical efficacy of ZA as a bone protective agent in mHSPC. The Hospital Episode Statistics (HES) database provides RCHD for patients in England and Wales. METHODS: HES data up to 2018 were obtained for patients (pts) randomised 2:1 between ADT (Arm A) and ADT+ZA (Arm B), or ADT+Doc (Arm C) or ADT+Doc+ZA (Arm E). ZA (4 mg) was six 3-weekly cycles, then 4-weekly for 2 years. Clinical fracture events were identified using a bespoke pre-specified coding framework of International Classification of Diseases (ICD 10) diagnosis and Classification of Interventions and Procedures (OPCS 4) procedure codes. Multivariate Cox regression models were developed to adjust for age, N stage, WHO performance status, Gleason score and NSAID use to determine impact of ZA on fracture risk. RESULTS: Connected datasets were available for 2,145 eligible pts (796 M0, 1,349 M1) of 2,962 (72%). Overall, 5-year fracture incidence was 6.4%; higher in M1 pts (9.6% in M1 vs 2.1% in M0, p<0.0001), and lower amongst M1 pts allocated ZA (4.55% with ZA vs 12.9% without ZA, p<0.0001) (Figure 1). ZA significantly reduced fracture risk in M1 pts (HR 0.36, 95% CI 0.22-0.57, p<0.0005) but not in M0 pts (HR 0.67, 95% CI 0.32-1.39, p=0.28). CONCLUSIONS: By accessing RCHD, we evaluated long-term fracture events amongst patients in STAMPEDE. Fracture rates were clinically notable in M1 and reduced significantly in pts allocated ZA with ADT±Doc. These data support using bone protective agents to reduce clinically significant fractures in mHSPC. Source of Funding: None © 2023 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 209Issue Supplement 4April 2023Page: e129 Advertisement Copyright & Permissions© 2023 by American Urological Association Education and Research, Inc.MetricsAuthor Information Craig Jones More articles by this author Ashwin Sachdeva More articles by this author Laura Murphy More articles by this author Macey Murray More articles by this author Louise Brown More articles by this author Eugene McCloskey More articles by this author Janet Brown More articles by this author Gerhardt Attard More articles by this author Mahesh Parmar More articles by this author Nicholas James More articles by this author Matthew Sydes More articles by this author Noel Clarke More articles by this author Expand All Advertisement PDF downloadLoading ...

Epidemiology of multiple sclerosis in the Campania Region (Italy): Derivation and validation of an algorithm to calculate the 2015-2020 incidence.

We aim to validate an algorithm based on routinely-collected healthcare data to detect incidence of multiple sclerosis (MS) in the Campania Region (South Italy) and to explore its spatial and temporal variations. We included individuals resident in the Campania Region who had at least one MS record in administrative datasets (drug prescriptions, hospital discharge, outpatients), from 2015 to 2020. We merged administrative to the clinical datasets to ascertain the actual date of diagnosis, and validated the minimum interval from our study baseline (Jan 1, 2015) to first MS records in administrative datasets to detect incident cases. We used Bayesian approach to explore geographical distribution, also including deprivation index as a covariate in the estimation model. We used the capture-recapture method to estimate the proportion of undetected cases. The best performance was achieved by the 12-month interval algorithm, detecting 2,150 incident MS cases, with 74.4% sensitivity (95%CI=64.1%, 85.9%) and 95.3% specificity (95%CI=90.7%, 99.8%). The cumulative incidence was 36.68 (95%CI=35.15, 38.26) per 100,000 from 2016 to 2020. The mean annual incidence was 7.34 (95%CI=7.03, 7.65) per 100,000 people-year. The geographical distribution of MS relative risk shows a decreasing east-west incidence gradient. The number of expected MS cases was 11% higher than the detected cases. We validated a case-finding algorithm based on administrative data to estimate MS incidence, and its spatial/temporal variations. This algorithm provides up-to-date estimates of MS incidence, and will be used in future studies to evaluate changes in MS incidence in relation to different risk factors.

Linkage of routinely-collected healthcare data and bespoke research questionnaire data to best serve NHS patient study participants.

ObjectivesPre-surgical ‘health optimisation’ programmes aim to improve patients’ health and to reduce their need for surgery. Routinely-collected data does not provide sufficient insight into whether or how these programmes meet their intended outcomes. The aim was to link routinely-collected NHS data with bespoke research questionnaire data to address this insufficiency.&#x0D; ApproachA study was designed to identify and collect data from a patient cohort eligible for a health optimisation programme (offering weight management and smoking cessation support) to determine programme uptake and patient outcomes. Patient and public involvement (PPI) group engagement revealed support for the use of data linkage between routinely-collected NHS data and study questionnaires, and shaped the study proposal. The HDRUK-supported regional ‘Systemwide Dataset’ of NHS routinely-collected data would provide demographics, healthcare usage and clinical outcomes. Questionnaires would provide data on patient decision-making and engagement with weight management or smoking cessation and detailed pain, function, wellbeing and activation measures.&#x0D; ResultsThe study received Health Research Authority approval and recruited six Participant Identification Centres. PPI co-production of the study’s participant-facing documents addressed the need for appropriate description of the novel data linkage process for potential participants. The data linkage design will offer an important advantage of a reduction in the participants’ time and recall accuracy needed in completing questionnaires due to the ability to omit duplicate demographic and healthcare usage fields. Further potential advantages include the ability to compare participants with patients who decline participation in the questionnaires, and long-term follow up of patient outcomes due to the continuation of routinely-collected data. Challenges have been faced in operationalizing the data collection and linkage processes due to decisions over data controllership of the combined dataset.&#x0D; ConclusionLinkage between routinely-collected NHS data and bespoke research questionnaires was deemed ethically acceptable, and could offer significant advantages in combining demographics, healthcare usage, outcomes and experience data of a cohort. Systemic decision-making to specify and streamline acceptable processes for this data linkage are important next steps.

RSV testing practice and positivity by patient demographics in the United States: integrated analyses of MarketScan and NREVSS databases

BackgroundRSV-incidence estimates obtained from routinely-collected healthcare data (e.g., MarketScan) are commonly adjusted for under-reporting using test positivity reported in national Surveillance Systems (NREVSS). However, NREVSS lacks detail on patient-level characteristics and the validity of applying a single positivity estimate across diverse patient groups is uncertain. We aimed to describe testing practices and test positivity across subgroups of private health insurance enrollees in the US and illustrate the possible magnitude of misclassification when using NREVSS to correct for RSV under ascertainment.MethodsUsing billing records, we determined distributions of RSV-test claims and test positivity among a national sample of private insurance enrollees. Tests were considered positive if they coincided with an RSV-diagnosis. We illustrated the influence of positivity variation across sub-populations when accounting for untested acute respiratory infections.ResultsMost tests were for children (age 0–4: 65.8%) and outpatient encounters (78.3%). Test positivity varied across age (0–4: 19.8%, 5–17: 1.8%, adults: 0.7%), regions (7.6–16.1%), settings (inpatient 4.7%, outpatient 14.2%), and test indication (5.0–35.9%). When compared to age, setting or indication-specific positivity, bias due to using NREVSS positivity to correct for untested ARIs ranged from − 76% to 3556%.ConclusionsRSV-test positivity depends on the characteristics of patients for whom those tests were ordered. NREVSS-based correction for RSV-under-ascertainment underestimates the true incidence among children and overestimate rates among adults. Demographic-specific detail on testing practice and positivity can improve the accuracy of RSV-incidence estimates.

Data provenance and integrity of health-care systems data for clinical trials

The need to run clinical trials quickly and efficiently is well recognised, none more so than during the push to find life-saving treatments and preventative measures for COVID-19. Late phase clinical trials can take many years and are expensive, due to the immense efforts necessary to deliver them. There are various ways in which the conduct of clinical trials could be improved, one is through judicious use of data already collected in health-care interactions. These data might be known as health-care systems data, routinely collected health-care data (RCHD), or real-world data.

Huntington's Disease in Israel: A Population-Based Study Using 20 Years of Routinely-Collected Healthcare Data.

Huntington's disease (HD) is a rare, genetic, neurodegenerative disease. Obtaining population-level data on epidemiology and disease management is challenging. To investigate the epidemiology, clinical manifestations, treatment, and healthcare utilization of patients with HD in Israel. Retrospective population-based cohort study, including 20 years of routinely collected data from Maccabi Healthcare Services, an insurer and healthcare provider for one-quarter of the Israeli population. The study cohort included 109 adult patients (aged ≥18 years) diagnosed with HD, with mean age of 49.9 years and 56%females. The most common HD-related conditions were anxiety (40%), behavioral problems (34%), sleep disorders (21%), and falls (13%). Annual incidence rates for HD ranged from 0.17 to 1.34 per 100,000 from 2000 to 2018; the 2018 crude prevalence in adults was 4.36 per 100,000. Median survival from diagnosis was approximately 12 years (95%CI: 10.4-15.3). The most frequent symptomatic treatments were antidepressants (69%), antipsychotics (63%), and tetrabenazine (63%), the only drug approved for the treatment of HD chorea in Israel during the examined period. Patterns of healthcare utilization changed as disease duration increased, reflected by increased frequency of emergency department visits and home visits. This retrospective population-based study provides insights into the prevalence, incidence, clinical profile, survival, and resource utilization of patients with HD in ethnically diverse Israel. The findings in this study are generally consistent with the international literature and demonstrate the value of routinely collected healthcare data as a complementary resource in HD research.

Interferon beta for the treatment of multiple sclerosis in the Campania Region of Italy: Merging the real-life to routinely collected healthcare data.

We aim to overcome limitations of previous clinical and population-based studies by merging a clinical registry to routinely-collected healthcare data, and to specifically describe differences in clinical outcomes, healthcare resource utilization and costs between interferon beta formulations for multiple sclerosis (MS). We included 850 patients with MS treated with interferon beta formulations, from 2015 to 2019, seen at the MS Clinical Care and Research Centre (Federico II University of Naples, Italy) and with linkage to routinely-collected healthcare data (prescription data, hospital admissions, outpatient services). We extracted and computed clinical outcomes (relapses, 6-month EDSS progression using a roving EDSS as reference), persistence (time spent on a specific interferon beta formulation), adherence (medication possession ratio (MPR)), healthcare resource utilization and costs (annualized hospitalization rate (AHR), costs for hospital admissions and DMTs). To evaluate differences between interferon beta formulations, we used linear regression (adherence), Poisson regression (AHR), mixed-effect regression (costs), and Cox-regression models (time varying variables); covariates were age, sex, treatment duration, baseline EDSS and adherence. Looking at clinical outcomes, rates of relapses and EDSS progression were lower than studies run on previous cohorts; there was no differences in relapse risk between interferon beta formulations. Risk of discontinuation was higher for Betaferon®/Extavia® (HR = 3.28; 95%CI = 2.11, 5.12; p<0.01). Adherence was lower for Betaferon®/Extavia® (Coeff = -0.05; 95%CI = -0.10, -0.01; p = 0.02), and Avonex® (Coeff = -0.06; 95%CI = -0.11, -0.02; p<0.01), when compared with Rebif® and Plegridy® (Coeff = 0.08; 95%CI = 0.01, 0.16; p = 0.02). AHR and costs for MS hospital admissions were higher for Betaferon®/Extavia® (IRR = 2.38; 95%CI = 1.01, 5.55; p = 0.04; Coeff = 14.95; 95%CI = 1.39, 28.51; p = 0.03). We have showed the feasibility of merging routinely-collected healthcare data to a clinical registry for future MS research, and have confirmed interferon beta formulations play an important role in the management of MS, with positive clinical outcomes. Differences between interferon beta formulations are mostly driven by adherence and healthcare resource utilization.

Do men avoid seeking medical advice? A register-based analysis of gender-specific changes in primary healthcare use after first hospitalisation at ages 60+ in Denmark

BackgroundIt remains unclear whether women’s greater primary healthcare use reflects a lower treatment-seeking threshold or a health disadvantage. We address this question by studying primary healthcare use surrounding a major...

The Secure Anonymised Information Linkage databank Dementia e-cohort (SAIL-DeC).

IntroductionThe rising burden of dementia is a global concern, and there is a need to study its causes, natural history and outcomes. The Secure Anonymised Information Linkage (SAIL) Databank contains anonymised, routinely-collected healthcare data for the population of Wales, UK. It has potential to be a valuable resource for dementia research owing to its size, long follow-up time and prospective collection of data during clinical care.ObjectivesWe aimed to apply reproducible methods to create the SAIL dementia e-cohort (SAIL-DeC). We created SAIL-DeC with a view to maximising its utility for a broad range of research questions whilst minimising duplication of effort for researchers.MethodsSAIL contains individual-level, linked primary care, hospital admission, mortality and demographic data. Data are currently available until 2018 and future updates will extend participant follow-up time. We included participants who were born between 1st January 1900 and 1st January 1958 and for whom primary care data were available. We applied algorithms consisting of International Classification of Diseases (versions 9 and 10) and Read (version 2) codes to identify participants with and without all-cause dementia and dementia subtypes. We also created derived variables for comorbidities and risk factors.ResultsFrom 4.4 million unique participants in SAIL, 1.2 million met the cohort inclusion criteria, resulting in 18.8 million person-years of follow-up. Of these, 129,650 (10%) developed all-cause dementia, with 77,978 (60%) having dementia subtype codes. Alzheimer’s disease was the most common subtype diagnosis (62%). Among the dementia cases, the median duration of observation time was 14 years.ConclusionWe have created a generalisable, national dementia e-cohort, aimed at facilitating epidemiological dementia research.

Identifying dementia outcomes in UK Biobank: a validation study of primary care, hospital admissions and mortality data

Prospective, population-based studies that recruit participants in mid-life are valuable resources for dementia research. Follow-up in these studies is often through linkage to routinely-collected healthcare datasets. We investigated the accuracy of these datasets for dementia case ascertainment in a validation study using data from UK Biobank—an open access, population-based study of > 500,000 adults aged 40–69 years at recruitment in 2006–2010. From 17,198 UK Biobank participants recruited in Edinburgh, we identified those with ≥ 1 dementia code in their linked primary care, hospital admissions or mortality data and compared their coded diagnoses to clinical expert adjudication of their full-text medical record. We calculated the positive predictive value (PPV, the proportion of cases identified that were true positives) for all-cause dementia, Alzheimer’s disease and vascular dementia for each dataset alone and in combination, and explored algorithmic code combinations to improve PPV. Among 120 participants, PPVs for all-cause dementia were 86.8%, 87.3% and 80.0% for primary care, hospital admissions and mortality data respectively and 82.5% across all datasets. We identified three algorithms that balanced a high PPV with reasonable case ascertainment. For Alzheimer’s disease, PPVs were 74.1% for primary care, 68.2% for hospital admissions, 50.0% for mortality data and 71.4% in combination. PPV for vascular dementia was 43.8% across all sources. UK routinely-collected healthcare data can be used to identify all-cause dementia in prospective studies. PPVs for Alzheimer’s disease and vascular dementia are lower. Further research is required to explore the geographic generalisability of these findings.

Investigating maternal risk factors for stillbirth in a population-based cohort in the South of England

Introduction In the UK there has been little improvement on the rates of stillbirth; the stillbirth rate was 2.9 per 1000 in 2015 (≥ 28 weeks gestation) and the annual rate reduction (ARR) of stillbirths was 1.4% from 2000–2015, which is roughly in the bottom one-third of high-income countries. Until now, most studies have only identified individual maternal risk factors associated with stillbirth and, to our knowledge, only one study has produced a stillbirth prediction model in the UK. Our study aimed to investigate risk factors for stillbirth (defined as a baby delivered with no signs of life after 24 weeks of pregnancy) using routinely-collected healthcare data at a regional level in the South of England. Methods Analysis of a population-based cohort using routine antenatal care and delivery data collected between 2003–2017 in Southampton, UK. Stillbirth was defined as born with no signs of life after 24 weeks gestation. Univariable comparisons between livebirth and stillbirth pregnancies were made using chi-squared test for categorical variables and independent t-test for continuous variables. Logistic regression modelling produced unadjusted and adjusted odd ratios (aOR) for stillbirth risk, adjusting for within-participant clustering of multiple pregnancies. A stillbirth prediction model was developed using backward stepwise regression. The model's derivation area under receiver operator curve (AUROC) was calculated. For the purpose of generating and testing a stillbirth risk score, the cohort was divided chronologically into derivation (n = 44,108) and internal validation cohorts (n = 24,684). An unweighted risk score variable consisting of the number of risk factors was generated for each pregnancy. A weighted risk score variable was generated taking into account the magnitude of the association of each risk factor with stillbirth (OR). Results The total sample size was 81,798 (81,366 livebirth pregnancies and 432 stillbirths), giving a stillbirth incidence rate (≥ 24 weeks gestation) of 5.28/1000 (95% CI 4.80–5.80). A total of15,813 births were excluded in the multivariable model due to missing data, hence giving a total number of 65,985 births (65,663 livebirths and 322 stillbirths). Significant risk factors for stillbirth in the multivariable model of the whole cohort (n = 65,985) included: maternal age ≥ 35 years (aOR 1.39, 95% CI 1.03–1.87), Asian ethnicity (aOR 1.80, 95% CI 1.22–2.66), mothers with unemployed partners (aOR 1.53 95% CI 1.07–2.18]), no previous livebirths (aOR 1.49, 95% CI 1.13–1.97), maternal educational qualification of secondary school or below (aOR 1.46, 95% CI 1.07–2.00) and college (aOR 1.49, 95% CI 1.05–2.12) (compared to university qualification or above), maternal obesity (aOR 1.58, 95% CI 1.18, 2.12), diastolic hypertension (aOR 4.3, 95% CI 2.04–9.06) and pre-existing diabetes (aOR 2.35, 95% CI 1.35, 4.11). Upon investigating the proportion of women in the study population (n = 65 985) with significant risk factors identified in the multivariable analysis; 5126 (7.42%) had none, 25,430 (36.80%) had one, 27,442 (39.71%) had two, 9518 (13.77%) had three and 1586 (2.30%) had ≥ 4 significant risk factors. When conducting the multivariable model in the derivation cohort (n = 44 108) and testing the risk score in the validation cohort (n = 24 684), for each additional significant risk factor, there was a 16% increased risk of stillbirth (95% CI 6–27%, P = 0.001). The stillbirth prediction model's AUROC was 0.64 (95% CI 0.61–0.67). Conclusion This study identifies the characteristics associated with stillbirth in this English cohort and that the risk of stillbirth is amplified in women with co-existing risk factors. Derived using routinely-collected healthcare data, this prediction model, if externally validated, can help early identification of pregnant mothers at risk of stillbirth from their first antenatal appointment. Interventions attempting to prevent stillbirth could then be targeted at the high-risk mothers.

Identifying diabetes cases from administrative data: a population-based validation study

BackgroundHealth care data allow for the study and surveillance of chronic diseases such as diabetes. The objective of this study was to identify and validate optimal algorithms for diabetes cases within health care administrative databases for different research purposes, populations, and data sources.MethodsWe linked health care administrative databases from Ontario, Canada to a reference standard of primary care electronic medical records (EMRs). We then identified and calculated the performance characteristics of multiple adult diabetes case definitions, using combinations of data sources and time windows.ResultsThe best algorithm to identify diabetes cases was the presence at any time of one hospitalization or physician claim for diabetes AND either one prescription for an anti-diabetic medication or one physician claim with a diabetes-specific fee code [sensitivity 84.2%, specificity 99.2%, positive predictive value (PPV) 92.5%]. Use of physician claims alone performed almost as well: three physician claims for diabetes within one year was highly specific (sensitivity 79.9%, specificity 99.1%, PPV 91.4%) and one physician claim at any time was highly sensitive (sensitivity 93.6%, specificity 91.9%, PPV 58.5%).ConclusionsThis study identifies validated algorithms to capture diabetes cases within health care administrative databases for a range of purposes, populations and data availability. These findings are useful to study trends and outcomes of diabetes using routinely-collected health care data.

Validating epilepsy diagnoses in routinely collected data

PurposeAnonymised, routinely-collected healthcare data is increasingly being used for epilepsy research. We validated algorithms using general practitioner (GP) primary healthcare records to identify people with epilepsy from anonymised healthcare data within the Secure Anonymised Information Linkage (SAIL) databank in Wales, UK. MethodA reference population of 150 people with definite epilepsy and 150 people without epilepsy was ascertained from hospital records and linked to records contained within SAIL (containing GP records for 2.4 million people). We used three different algorithms, using combinations of GP epilepsy diagnosis and anti-epileptic drug (AED) prescription codes, to identify the reference population. ResultsCombining diagnosis and AED prescription codes had a sensitivity of 84% (95% ci 77–90) and specificity of 98% (95–100) in identifying people with epilepsy; diagnosis codes alone had a sensitivity of 86% (80–91) and a specificity of 97% (92–99); and AED prescription codes alone achieved a sensitivity of 92% (70–83) and a specificity of 73% (65–80). Using AED codes only was more accurate in children achieving a sensitivity of 88% (75–95) and specificity of 98% (88–100). ConclusionGP epilepsy diagnosis and AED prescription codes can be confidently used to identify people with epilepsy using anonymised healthcare records in Wales, UK.

Accuracy of routinely-collected healthcare data for identifying motor neurone disease cases: A systematic review.

BackgroundMotor neurone disease (MND) is a rare neurodegenerative condition, with poorly understood aetiology. Large, population-based, prospective cohorts will enable powerful studies of the determinants of MND, provided identification of disease cases is sufficiently accurate. Follow-up in many such studies relies on linkage to routinely-collected health datasets. We systematically evaluated the accuracy of such datasets in identifying MND cases.MethodsWe performed an electronic search of MEDLINE, EMBASE, Cochrane Library and Web of Science for studies published between 01/01/1990-16/11/2015 that compared MND cases identified in routinely-collected, coded datasets to a reference standard. We recorded study characteristics and two key measures of diagnostic accuracy—positive predictive value (PPV) and sensitivity. We conducted descriptive analyses and quality assessments of included studies.ResultsThirteen eligible studies provided 13 estimates of PPV and five estimates of sensitivity. Twelve studies assessed hospital and/or death certificate-derived datasets; one evaluated a primary care dataset. All studies were from high income countries (UK, Europe, USA, Hong Kong). Study methods varied widely, but quality was generally good. PPV estimates ranged from 55–92% and sensitivities from 75–93%. The single (UK-based) study of primary care data reported a PPV of 85%.ConclusionsDiagnostic accuracy of routinely-collected health datasets is likely to be sufficient for identifying cases of MND in large-scale prospective epidemiological studies in high income country settings. Primary care datasets, particularly from countries with a widely-accessible national healthcare system, are potentially valuable data sources warranting further investigation.