Search

Objective Health data standardized to a common data model (CDM) simplifies and facilitates research. This study examines the factors that make standardizing observational health data to the Observational Medical Outcomes Partnership (OMOP) CDM successful. Materials and methods Twenty-five data partners (DPs) from 11 countries received funding from the European Health Data Evidence Network (EHDEN) to standardize their data. Three surveys, DataQualityDashboard results, and statistics from the conversion process were analyzed qualitatively and quantitatively. Our measures of success were the total number of days to transform source data into the OMOP CDM and participation in network research. Results The health data converted to CDM represented more than 133 million patients. 100%, 88%, and 84% of DPs took Surveys 1, 2, and 3. The median duration of the 6 key extract, transform, and load (ETL) processes ranged from 4 to 115 days. Of the 25 DPs, 21 DPs were considered applicable for analysis of which 52% standardized their data on time, and 48% participated in an international collaborative study. Discussion This study shows that the consistent workflow used by EHDEN proves appropriate to support the successful standardization of observational data across Europe. Over the 25 successful transformations, we confirmed that getting the right people for the ETL is critical and vocabulary mapping requires specific expertise and support of tools. Additionally, we learned that teams that proactively prepared for data governance issues were able to avoid considerable delays improving their ability to finish on time. Conclusion This study provides guidance for future DPs to standardize to the OMOP CDM and participate in distributed networks. We demonstrate that the Observational Health Data Sciences and Informatics community must continue to evaluate and provide guidance and support for what ultimately develops the backbone

Background: Researchers developing prediction models are faced with numerous design choices that may impact model performance. One key decision is how to include patients who are lost to follow-up. In this paper we perform a large-scale empirical evaluation investigating the impact of this decision. In addition, we aim to provide guidelines for how to deal with loss to follow-up. Methods: We generate a partially synthetic dataset with complete follow-up and simulate loss to follow-up based either on random selection or on selection based on comorbidity. In addition to our synthetic data study we investigate 21 real-world data prediction problems. We compare four simple strategies for developing models when using a cohort design that encounters loss to follow-up. Three strategies employ a binary classifier with data that: (1) include all patients (including those lost to follow-up), (2) exclude all patients lost to follow-up or (3) only exclude patients lost to follow-up who do not have the outcome before being lost to follow-up. The fourth strategy uses a survival model with data that include all patients. We empirically evaluate the discrimination and calibration performance. Results: The partially synthetic data study results show that excluding patients who are lost to follow-up can introduce bias when loss to follow-up is common and does not occur at random. However, when loss to follow-up was completely at random, the choice of addressing it had negligible impact on model discrimination performance. Our empirical real-world data results showed that the four design choices investigated to deal with loss to follow-up resulted in comparable performance when the time-at-risk was 1-year but demonstrated differential bias when we looked into 3-year time-at-risk. Removing patients who are lost to follow-up before experiencing the outcome but keeping patients who are lost to follow-up after the outcome can bias a model and should be avoided. Conclusion: Based on thi

INTRODUCTION: Angiotensin converting enzyme inhibitors (ACEs) and angiotensin receptor blockers (ARBs) could influence infection risk of coronavirus disease (COVID-19). Observational studies to date lack pre-specification, transparency, rigorous ascertainment adjustment and international generalizability, with contradictory results. METHODS: Using electronic health records from Spain (SIDIAP) and the United States (Columbia University Irving Medical Center and Department of Veterans Affairs), we conducted a systematic cohort study with prevalent ACE, ARB, calcium channel blocker (CCB) and thiazide diuretic (THZ) use to determine relative risk of COVID-19 diagnosis and related hospitalization outcomes. The study addressed confounding through large-scale propensity score adjustment and negative control experiments. RESULTS: Following over 1.1 million antihypertensive users identified between November 2019 and January 2020, we observed no significant difference in relative COVID-19 diagnosis risk comparing ACE/ARB vs CCB/THZ monotherapy (hazard ratio: 0.98; 95% CI 0.84 - 1.14), nor any difference for mono/combination use (1.01; 0.90 - 1.15). ACE alone and ARB alone similarly showed no relative risk difference when compared to CCB/THZ monotherapy or mono/combination use. Directly comparing ACE vs. ARB demonstrated a moderately lower risk with ACE, non-significant for monotherapy (0.85; 0.69 - 1.05) and marginally significant for mono/combination users (0.88; 0.79 - 0.99). We observed, however, no significant difference between drug- classes for COVID-19 hospitalization or pneumonia risk across all comparisons. CONCLUSION: There is no clinically significant increased risk of COVID-19 diagnosis or hospitalization with ACE or ARB use. Users should not discontinue or change their treatment to avoid COVID-19.

Objectives: The Italian project MATRICE aimed to assess how well cases of type 2 diabetes (T2DM), hypertension, ischaemic heart disease (IHD) and heart failure (HF) and their levels of severity can be automatically extracted from the Health Search/CSD Longitudinal Patient Database (HSD). From the medical records of the general practitioners (GP) who volunteered to participate, cases were extracted by algorithms based on diagnosis codes, keywords, drug prescriptions and results of diagnostic tests. A random sample of identified cases was validated by interviewing their GPs. Setting: HSD is a database of primary care medical records. A panel of 12 GPs participated in this validation study. Participants: 300 patients were sampled for each disease, except for HF, where 243 patients were assessed. Outcome measures: The positive predictive value (PPV) was assessed for the presence/absence of each condition against the GP's response to the questionnaire, and Cohen's ? was calculated for agreement on the severity level. Results: The PPV was 100% (99% to 100%) for T2DM and hypertension, 98% (96% to 100%) for IHD and 55% (49% to 61%) for HF. Cohen's kappa for agreement on the severity level was 0.70 for T2DM and 0.69 for hypertension and IHD. Conclusions: This study shows that individuals with T2DM, hypertension or IHD can be validly identified in HSD by automated identification algorithms. Automatic queries for levels of severity of the same diseases compare well with the corresponding clinical definitions, but some misclassification occurs. For HF, further research is needed to refine the current algorithm.

Aim Electronic healthcare record (EHR)-based surveillance systems are increasingly being developed to support early detection of safety signals. It is unknown what the power of such a system is for surveillance among children and adolescents. In this paper we provide estimates of the number and classes of drugs, and incidence rates (IRs) of events, that can be monitored in children and adolescents (0-18 years). Methods Data were obtained from seven population-based EHR databases in Denmark, Italy, and the Netherlands during the period 1996-2010. We estimated the number of drugs for which specific adverse events can be monitored as a function of actual drug use, minimally detectable relative risk (RR) and IRs for 10 events. Results The population comprised 4 838 146 individuals (25 575 132 person years (PYs)), who were prescribed 2170 drugs (1 610 631 PYs drug-exposure). Half of the total drug-exposure in PYs was covered by only 18 drugs (0.8%). For a relatively frequent event like upper gastrointestinal bleeding there were 39 drugs for which an association with a RR ≥4, if present, could be investigated. The corresponding number of drugs was eight for a rare event like anaphylactic shock. Conclusion Drug use in children is rare and shows little variation. The number of drugs with enough exposure to detect rare adverse events in children and adolescents within an EHR-based surveillance system such as EU-ADR is limited. Use of additional sources of paediatric drug exposure information and global collaboration are imperative in order to optimize EHR data for paediatric safety surveillance.

Background Data mining in spontaneous reporting databases has shown that drug-induced liver injury is infrequently reported in children. Objectives Our objectives were to (i) identify drugs potentially associated with acute liver injury (ALI) in children and adolescents using electronic healthcare record (EHR) data; and (ii) to evaluate the significance and novelty of these associations. Methods We identified potential cases of ALI during exposure to any prescribed/dispensed drug for individuals <18 years old from the EU-ADR network, which includes seven databases from three countries, covering the years 1996-2010. Several new methods for signal detection were applied to identify all statistically significant associations between drugs and ALI. A drug was considered statistically significantly associated with ALI, using all other time as a reference category, if the 95 % CI lower band of the relative risk was >1 and in the presence of at least three exposed cases of ALI. Potentially new signals were distinguished from already known associations concerning ALI (whether in adults and/or in the paediatric population) through manual review of published literature and drug product labels. Results The study population comprised 4,838,146 individuals aged <18 years, who contributed an overall 25,575,132 person-years of follow-up. Within this population, we identified 1,015 potential cases of ALI. Overall, 20 positive drug-ALI associations were detected. The associations between ALI and domperidone, flunisolide and human insulin were considered as potentially new signals. Citalopram and cetirizine have been previously described as hepatotoxic in adults but not in children, while all remaining associations were already known in both adults and children. Conclusions Data mining of multiple EHR databases for signal detection confirmed known associations between ALI and several drugs, and identified some potentially new signals in children that require further investig

Background: Administrative databases are widely available and have been extensively used to provide estimates of chronic disease prevalence for the purpose of surveillance of both geographical and temporal trends. There are, however, other sources of data available, such as medical records from primary care and national surveys. In this paper we compare disease prevalence estimates obtained from these three different data sources. Methods: Data from general practitioners (GP) and administrative transactions for health services were collected from five Italian regions (Veneto, Emilia Romagna, Tuscany, Marche and Sicily) belonging to all the three macroareas of the country (North, Center, South). Crude prevalence estimates were calculated by data source and region for diabetes, ischaemic heart disease, heart failure and chronic obstructive pulmonary disease (COPD). For diabetes and COPD, prevalence estimates were also obtained from a national health survey. When necessary, estimates were adjusted for completeness of data ascertainment. Results: Crude prevalence estimates of diabetes in administrative databases (range: from 4.8% to 7.1%) were lower than corresponding GP (6.2%-8.5%) and survey-based estimates (5.1%-7.5%). Geographical trends were similar in the three sources and estimates based on treatment were the same, while estimates adjusted for completeness of ascertainment (6.1%-8.8%) were slightly higher. For ischaemic heart disease administrative and GP data sources were fairly consistent, with prevalence ranging from 3.7% to 4.7% and from 3.3% to 4.9%, respectively. In the case of heart failure administrative estimates were consistently higher than GPs' estimates in all five regions, the highest difference being 1.4% vs 1.1%. For COPD the estimates from administrative data, ranging from 3.1% to 5.2%, fell into the confidence interval of the Survey estimates in four regions, but failed to detect the higher prevalence in the most Southern region (4.0% in adminis

Background The Observational Medical Outcomes Partnership (OMOP) has just completed a large scale empirical evaluation of statistical methods and analysis choices for risks identification in longitudinal observational healthcare data. This experiment drew data from four large US health insurance claims databases and one US electronic health record (EHR) database, but it is unclear to what extend the findings of this study apply to other data sources. Objective To replicate the OMOP experiment in six European ERR databases. Research Design Six databases of the EU-ADR (Exploring and Understanding Adverse Drug Reactions) database network participated in this study: Aarhus (Denmark), ARS (Italy), HealthSearch (Italy), IPCI (the Netherlands), Pedianet (Italy), and Pharmo (the Netherlands). All methods in the OMOP experiment were applied to a collection of 165 positive and 234 negative control drug outcome pairs across four outcomes: acute liver injury, acute myocardial infarction, acute kidney injury, and upper gastrointestinal bleeding. Area under the receiver operator characteristics curve (AUC) was computed per database and for a combination of all six databases using metaanalysis for random effects. We provide expected values of estimation error as well, based on negative controls. Results Similarly to the US experiment, high predictive accuracy was found (AUC >0.8) for some analyses. Selfcontrolled designs, such as self-controlled case series, IC temporal pattern discovery and self-controlled cohort achieved higher performance than other methods, both in terms of predictive accuracy and observed bias. Conclusions The major findings of the recent OMOP experiment were also observed in the European databases.

Objective: To evaluate positive predictive value (PPV) of different disease codes and free text in identifying acute myocardial infarction (AMI) from electronic healthcare records (EHRs). Design: Validation study of cases of AMI identified from general practitioner records and hospital discharge diagnoses using free text and codes from the International Classification of Primary Care (ICPC), International Classification of Diseases 9th revision-clinical modification (ICD9-CM) and ICD-10th revision (ICD-10). Setting: Population-based databases comprising routinely collected data from primary care in Italy and the Netherlands and from secondary care in Denmark from 1996 to 2009. Participants: A total of 4 034 232 individuals with 22 428 883 person-years of follow-up contributed to the data, from which 42 774 potential AMI cases were identified. A random sample of 800 cases was subsequently obtained for validation. Main outcome measures: PPVs were calculated overall and for each code/free text. `Best-case scenario' and `worst-case scenario' PPVs were calculated, the latter taking into account non-retrievable/non-assessable cases. We further assessed the effects of AMI misclassification on estimates of risk during drug exposure. Results: Records of 748 cases (93.5% of sample) were retrieved. ICD-10 codes had a `best-case scenario' PPV of 100% while ICD9-CM codes had a PPV of 96.6% (95% CI 93.2% to 99.9%). ICPC codes had a `best-case scenario' PPV of 75% (95% CI 67.4% to 82.6%) and free text had PPV ranging from 20% to 60%. Corresponding PPVs in the `worst-case scenario' all decreased. Use of codes with lower PPV generally resulted in small changes in AMI risk during drug exposure, but codes with higher PPV resulted in attenuation of risk for positive associations. Conclusions: ICD9-CM and ICD-10 codes have good PPV in identifying AMI from EHRs; strategies are necessary to further optimise utility of ICPC codes and free-text search. Use of specific AMI disease codes in e

Acute liver failure is idiopathic and drug-related in, respectively, around 50 and 15 % of children. Population-based, epidemiologic data about the pattern of disease manifestation and incidence of less severe acute liver injury, either idiopathic or potentially drug-attributed are limited in children and adolescents. (i) To assess the incidence of idiopathic acute liver injury (ALI) and its clinical features in children and adolescent outpatients; and (ii) to investigate the role of the drug as a potential cause of ALI which is considered idiopathic. A retrospective cohort study was performed during the years 2000-2008. Data were retrieved from three longitudinal electronic healthcare databases in two European countries: Pedianet and Health Search/CSD Longitudinal Patient Database from Italy and the Integrated Primary Care Information database from The Netherlands. Cases of idiopathic acute liver injury in population aged < 18 years were identified by exclusion of all competing causes of liver injury (e.g. viral, autoimmune hepatitis), according to CIOMS criteria. The potential role of drug exposure as actual underlying cause of idiopathic ALI was detected through signal detection mining techniques. Both pooled and country-specific incidence rates [IR/100,000 person-years (PYs)] of idiopathic ALI and pooled adjusted rate ratios (RR) of drugs identified as a potential cause of idiopathic ALI, plus 95 % confidence intervals (CI) were estimated using the custom-built software Jerboa. Among 785 definite cases of idiopathic ALI, the pooled IR was 62.4/100,000 PYs (95 % CI 58.1-66.8). The country-specific IR was higher in Italy (73.0/100,000 PYs, 95 % CI 67.8-78.4) than in The Netherlands (21.0/100,000 PYs, 95 % CI 16.0-27.2) and increased with age in both countries. Isolated elevations of liver enzymes were reported in around two-thirds of cases in Italy, while in The Netherlands the cases were more often identified by a combination of signs/symptoms. Among drugs detect

Background: There is huge potential for mining electronic healthcare records (EHR) databases to augment current systems in pharmacovigilance.[1-3] Like any signal detection system, there is a need to establish ‘rules’ how to trigger an alert, when to consider a signal likely enough to be true to warrant follow-up or even to require immediate health policy intervention.[4,5] Objectives: To describe the process of prioritisation of drug-adverse event associations derived from signal detection using EHR databases in the EU-ADR Project. Methods: Association measures between drug use and acute myocardial infarction (AMI) were generated by first applying various statistical methods on healthcare data from seven databases of the EUADR network.[6] Association estimates were ranked based on the best performing method (Longitudinal Gamma Poisson Shrinker). Matched case-control and self-controlled case series methods were additionallyconducted to deal with temporality and confounding effect, while the LEOPARD method was applied to specifically detect protopathic bias. Consistency of the association among drugs of the same class and the number of excess cases attributable to the drug exposure were further assessed to prioritize the list of potential signals. Finally, signal filtering and signal substantiation were done using different bioinformatics workflows to determine the novelty and plausibility of the identified signals. Results: Demographic, clinical and prescription/dispensing data in three European Countries were obtained from 21 171 291 individuals with 154 474 063 person-years of follow-up within the period 1995–2011. Overall, 163 potential signals forAMI were identified based on statistical association. Of these, 72 signals were flagged by LEOPARDas likely due to protopathic bias. Further signal refinement to reduce possible confounding decreased the number of signals to 39. Nine signals remained after applying the criteria for novelty and plausibility. Conclusion

Purpose To provide estimates of the number and types of drugs that can be monitored for safety surveillance using electronic healthcare databases. Methods Using data from eight European databases (administrative claims, medical records) and in the context of a cohort study, we determined the amount of drug exposure required for signal detection across varying magnitudes of relative risk (RR). We provide estimates of the number and types of drugs that can be monitored as a function of actual use, minimal detectable RR, and empirically derived incidence rates for the following adverse events: (i) acute myocardial infarction; (ii) acute renal failure; (iii) anaphylactic shock; (iv) bullous eruptions; (v) rhabdomyolysis; and (vi) upper gastrointestinal bleeding. We performed data simulation to see how expansion of database size would influence the capabilities of such system. Results Data from 19 647 452 individuals (59 594 132 person-years follow-up) who used 2289 drugs in the EU-ADR network show that for a frequent event such as acute myocardial infarction, there are 531 drugs (23% of total) for which an association with RR = 2, if present, can be investigated. For a rare event such as rhabdomyolysis, there are 19 drugs (1%) for which an association of same magnitude can be investigated. Conclusion Active surveillance using healthcare data-based networks for signal detection is feasible, although the leverage to do so may be low for infrequently used drugs and for rare outcomes. Extending database network size to include data from heterogeneous populations and increasing follow-up time are warranted to maximize leverage of these surveillance systems. Copyright (C) 2012 John Wiley & Sons, Ltd.

BACKGROUND: Drug safety monitoring relies primarily on spontaneous reporting, but electronic health care record databases offer a possible alternative for the detection of adverse drug reactions (ADRs). OBJECTIVES: To evaluate the relative performance of different statistical methods for detecting drug-adverse event associations in electronic health care record data representing potential ADRs. RESEARCH DESIGN: Data from 7 databases across 3 countries in Europe comprising over 20 million subjects were used to compute the relative risk estimates for drug-event pairs using 10 different methods, including those developed for spontaneous reporting systems, cohort methods such as the longitudinal gamma poisson shrinker, and case-based methods such as case-control. The newly developed method "longitudinal evaluation of observational profiles of adverse events related to drugs" (LEOPARD) was used to remove associations likely caused by protopathic bias. Data from the different databases were combined by pooling of data, and by meta-analysis for random effects. A reference standard of known ADRs and negative controls was created to evaluate the performance of the method. MEASURES: The area under the curve of the receiver operator characteristic curve was calculated for each method, both with and without LEOPARD filtering. RESULTS: The highest area under the curve (0.83) was achieved by the combination of either longitudinal gamma poisson shrinker or case-control with LEOPARD filtering, but the performance between methods differed little. LEOPARD increased the overall performance, but flagged several known ADRs as caused by protopathic bias. CONCLUSIONS: Combinations of methods demonstrate good performance in distinguishing known ADRs from negative controls, and we assume that these could also be used to detect new drug safety signals.

Background: Traditional pharmacovigilance activities do not focus specifically on children and medicines in children are frequently prescribed off-label based on extrapolating experience from adults.[1] The EU-ADR Project aims to use information from various electronic healthcare record (EHR) databases to produce a computerised integrated system for early detection of drug safety signals.[2] Objectives: To provide estimates of the number of drugs and incidence rates (IRs) of adverse events that can be monitored in children and adolescents in the EU-ADR network. Methods: Demographic, clinical events and outpatient drug prescription/dispensing data were obtained for individuals 0 to 18 years of age from seven databases from Denmark, Italy, and the Netherlands. Data were analysed from 1996–2008. We estimated the number and types of drugs for which specific adverse events can be monitored as function of actual drug use, minimally detectable RR, and empiricallyderived incidence rates for 10 events deemed to be important in pharmacovigilance.[3,4] The same was done for adverse events frequently reported in children, using IRs described in literature. Results: The paediatric population comprised 4 838 146 individuals contributing 25 575 132 person-years (PYs) of follow-up during the study period. Within this population a total of 2170 drugs were prescribed, with a total drug exposure of 1 610 631 PYs. Eighteen of the 2170 drugs (0.8%) comprised half of total drug exposure while 90% of the total drug exposure in PYs was represented by 158 drugs (7.3%). For a relatively frequent event such as upper gastrointestinal bleeding (IR = 14.4/100 000 PYs), there were 39 drugs (comprising 66% of total exposure in PY) for which an association with a RR ‡4, if present, can be investigated. For a rare event such as anaphylactic shock, there were 8 drugs (comprising 35% of total exposure) for which an association of same magnitude can be investigated. Based on literature-derived IR, ther

Introduction: Data mining in spontaneous reporting databases has shown that drug-induced acute liver injury (ALI) is infrequently reported in children,[1] but such system still has limitations in terms of signal detection. The EU-ADR project aimed to develop and validate a computerised system combining data from multiple European electronic healthcare records (EHR) for early detection of drug safety signals.[2] Aims: To identify drugs potentially associated with ALI in children and adolescents using EHR data. Methods: We extracted data on ALI cases and prescribed/dispensed drugs for individuals 0–18 years old registered with seven European population-based medical records/claims databases of the EU-ADR network during the years 1996–2008. Based on the estimated background incidence rate of ALI, power = 80%, alpha = 5% and the number of person-years (PYs) of drug exposure by ATC classification, we calculated the minimal drug exposure required to detect a signal concerning paediatric ALI.[3] Using the Longitudinal Gamma Poisson Shrinker (LGPS) method,[4] we defined as threshold for potential signals in EU-ADR a value of relative risk (RRLGPS) ‡2 and a lower 95% CI of RRLGPS >1 for each drug associated with at least three events of ALI. We discriminated among potentially relevant new signals and already known signals either in adults or in children by data mining of published literature as well as product labels. Results: Overall 4 838 146 children 0–18 years old contributed 25 575 132 PYs of follow-up time to the EU-ADR database network. Within this population, 1015 events of ALI were identified. The IR of paediatric ALI was estimated to be 3.96 (3.73–4.21)/100 000 PYs. The total amount of drug exposure that is required to detect a weak, moderate and strong association with paediatric ALI (RR‡ 2.0, 4.0 and 6.0) was 202 733 PYs, and 31 041 PYs, and 13 860 PYs, respectively. Among drugs associated with at least three cases, 20 drugs were found to be potential signals for

Background: There is growing interest in the use of electronic healthcare records (EHR) databases for drug safety signal detection, but an important concern is whether these databases have the data and adequate power to monitor all types of drugs and all types of adverse events. Objectives: To put into perspective expectations about use of EHR data-based systems as new approach to signal detection and active drug safety surveillance. Methods: Within the setting of EU-ADR network comprising eight European EHR databases (administrative claims and general practice medical records) and in the context of a cohort study, we determined the amount of drug exposure required to enable detection of safety signals of varying magnitudes of risk. We provide estimates of the number and types of drugs (WHO ATC Classification, 5th level) that can be monitored as a function of actual use, minimal detectable relative risk (RR), and empirically-derived Background incidence rates of the following adverse events deemed to be important in pharmacovigilance: (1) acute myocardial infarction; (2) acute renal failure; (3) anaphylactic shock; (4) bullous eruptions; (5) rhabdomyolysis; and (6) upper gastrointestinal bleeding.[1] We further perform data simulation in an attempt to see if, and how, expansion of database size would make a difference in the capabilities of the system. Results: Data from 19 647 452 individuals (59 594 132 person-years follow-up) who used 2289 drugs show that for a frequent event such as acute myocardial infarction there are 531 drugs (23% of total) for which an association with RR= 2, if present, can be investigated and 939 drugs (41%) for which an association with RR= 4 can be investigated. For a rare event such as rhabdomyolysis there are 19 drugs (1%) for which an association with RR= 2, if present, can be investigated and 170 drugs (7%) for which an association with RR= 4 can be investigated. With expansion of the database network to 10 times its current size (»

Background: The EU-ADR project aims to exploit different European electronic healthcare records (EHR) databases for drug safety signal detection. Currently, it is unknown what could be the additional value of using EHR databases for signal detection, with respect to traditional pharmacovigilance system. Objectives: To describe the preliminary results of the comparison of signal detection, as conducted in EU-ADR database network versus US Food and Drug Administration and World Health Organization spontaneous reporting databases. Methods: EU-ADR data sources consist of eight databases in four countries (Denmark, Italy, Netherlands, and United Kingdom) that are combined through distributed data network.[1] A custom-built software (Jerboaa) elaborates harmonized input data that are produced locally and generates aggregated data which are then stored in a central repository. These data are subsequently analyzed using a variety of signal detection methods adapted to longitudinal data (e.g., Longitudinal Gamma Poisson Shrinker). As potential signals, all drugs that are associated with ten events of interest (bullous eruptions - BE, acute renal failure - ARF, acute myocardial infarction - AMI, anaphylactic shock - AS, rhabdomyolysis - RHABD, upper gastrointestinal bleeding – UGIB, neutropenia – NEUTROP, pancytopenia – PANCYT, acute liver injury – ALI, and cardiac valve fibrosis) have been detected via different data mining techniques in the two systems. The number of drugs that could be investigated and the potential signals detected for each event of interest were then compared between spontaneous reporting systems (SRS) and EU-ADR network. Results: SRSs could explore, as potential signals, a larger number of drugs for the ten events, in comparison to EU-ADR, particularly for rare events generally thought to be highly drug attributable (i.e. BE: 3393 vs 228). The highest proportion of signals detected in SRSs was found for BE, ARF and AS, while for AMI, ARF, and UGIB in EU

Purpose: In this proof-of-concept paper we describe the framework, process, and preliminary results of combining data from European electronic healthcare record (EHR) databases for large-scale monitoring of drug safety. Methods: Aggregated demographic, clinical, and prescription data from eight databases in four countries (Denmark, Italy, Netherlands, the UK) were pooled using a distributed network approach by generation of common input data followed by local aggregation through custom-built software, Jerboa©. Comparison of incidence rates of upper gastrointestinal bleeding (UGIB) and nonsteroidal anti-inflammatory drug (NSAID) utilization patterns were used to evaluate data harmonization and quality across databases. The known association of NSAIDs and UGIB was employed to demonstrate sensitivity of the system by comparing incidence rate ratios (IRRs) of UGIB during NSAID use to UGIB during all other person-time. Results: The study population for this analysis comprised 19 647 445 individuals corresponding to 59 929 690 person-years of follow-up. 39 967 incident cases of UGIB were identified during the study period. Crude incidence rates varied between 38.8 and 109.5/100 000 person-years, depending on country and type of database, while age-standardized rates ranged from 25.1 to 65.4/100 000 person-years. NSAID use patterns were similar for databases within the same country but heterogeneous among different countries. A statistically significant age- and gender-adjusted association between use of any NSAID and increased risk for UGIB was confirmed in all databases, IRR from 2.0 (95%CI:1.7-2.2) to 4.3 (95%CI: 4.1-4.5). Conclusions: Combining data from EHR databases of different countries to identify drug-adverse event associations is feasible and can set the stage for changing and enlarging the scale for drug safety monitoring.

Background: Drug-induced acute liver injury (ALI) is one of the leading causes of drug withdrawal from the market. There is, however, lack of data on ALI in the pediatric population. Combining multi-country electronic healthcare databases offers the opportunity to explore the risk of ALI in children and adolescents and facilitate detection of potential drug safety signals. Objectives: To identify drugs potentially associated with ALI in the pediatric population using the EU-ADR database network and to estimate the power of such a network for signal detection concerning ALI as a function of actual drug use, minimal detectable relative risk (RR) and Background incidence rate (IR) of ALI in children and adolescents. Methods: We extracted data on all potential cases of ALI and data concerning prescribed/dispensed drugs among individuals 0–15 years registered within eight European, population-based, electronicmedical record and claims databases of the EU-ADR network during the period 1995–2010. We estimated the Background IR of ALI in the pediatric population and assessed drug use according to number of person-years (PYs) of exposure by Anatomical Therapeutic and Chemical (ATC) classification, 5th level. Based on IR of ALI derived within EU-ADR, power = 80% and alpha = 5%, we estimated how much drug exposure would be necessary to allow for detection of a signal concerning ALI. Finally, among drugs with enough exposure, all potential signals were identified by measuring age- and sex-adjusted relative risks of ALI, with all other drugs as comparator. Results: Children 0–15 years of age contributed 20 088 726 PYs of follow-up to the EU-ADR database network. The incidence rate for ALI was estimated to be 4.3 per 100 000 PYs in this pediatric population. The total amount of drug exposure that would be required to detect a ‘weak’ association (RR= 2) with ALI, if present, was 186 490 PYs and 28 554 PYs to detect a ‘moderate’ association (RR = 4). The following drugs were identi