Towards Drug Safety Surveillance and Pharmacovigilance: Current Progress in Detecting Medication and Adverse Drug Events from Electronic Health Records

Monitoring Adverse Drug Events in Web Forums: Evaluation of a Pipeline and Use Case Study

BACKGROUND

To mitigate safety concerns, regulatory agencies must make informed decisions regarding drug usage and adverse drug events (ADEs). The primary pharmacovigilance data stem from spontaneous reports by health care professionals. However, underreporting poses a notable challenge within the current system. Explorations into alternative sources, including electronic patient records and social media, have been undertaken. Nevertheless, social media's potential remains largely untapped in real-world scenarios.

OBJECTIVE

The challenge faced by regulatory agencies in using social media is primarily attributed to the absence of suitable tools to support decision makers. An effective tool should enable access to information via a graphical user interface, presenting data in a user-friendly manner rather than in their raw form. This interface should offer various visualization options, empowering users to choose representations that best convey the data and facilitate informed decision-making. Thus, this study aims to assess the potential of integrating social media into pharmacovigilance and enhancing decision-making with this novel data source. To achieve this, our objective was to develop and assess a pipeline that processes data from the extraction of web forum posts to the generation of indicators and alerts within a visual and interactive environment. The goal was to create a user-friendly tool that enables regulatory authorities to make better-informed decisions effectively.

METHODS

To enhance pharmacovigilance efforts, we have devised a pipeline comprising 4 distinct modules, each independently editable, aimed at efficiently analyzing health-related French web forums. These modules were (1) web forums' posts extraction, (2) web forums' posts annotation, (3) statistics and signal detection algorithm, and (4) a graphical user interface (GUI). We showcase the efficacy of the GUI through an illustrative case study involving the introduction of the new formula of Levothyrox in France. This event led to a surge in reports to the French regulatory authority.

RESULTS

Between January 1, 2017, and February 28, 2021, a total of 2,081,296 posts were extracted from 23 French web forums. These posts contained 437,192 normalized drug-ADE couples, annotated with the Anatomical Therapeutic Chemical (ATC) Classification and Medical Dictionary for Regulatory Activities (MedDRA). The analysis of the Levothyrox new formula revealed a notable pattern. In August 2017, there was a sharp increase in posts related to this medication on social media platforms, which coincided with a substantial uptick in reports submitted by patients to the national regulatory authority during the same period.

CONCLUSIONS

We demonstrated that conducting quantitative analysis using the GUI is straightforward and requires no coding. The results aligned with prior research and also offered potential insights into drug-related matters. Our hypothesis received partial confirmation because the final users were not involved in the evaluation process. Further studies, concentrating on ergonomics and the impact on professionals within regulatory agencies, are imperative for future research endeavors. We emphasized the versatility of our approach and the seamless interoperability between different modules over the performance of individual modules. Specifically, the annotation module was integrated early in the development process and could undergo substantial enhancement by leveraging contemporary techniques rooted in the Transformers architecture. Our pipeline holds potential applications in health surveillance by regulatory agencies or pharmaceutical companies, aiding in the identification of safety concerns. Moreover, it could be used by research teams for retrospective analysis of events.

Clinical pharmacology facing the real-world setting: Pharmacovigilance, pharmacoepidemiology and the economic evaluation of drugs

Traditionally, clinical pharmacology has focused its activities on drug-organism interaction, from an individual or collective perspective. Drug efficacy assessment by performing randomized clinical trials and analysis of drug use in clinical practice by carrying out drug utilization studies have also been other areas of interest. From now on, Clinical pharmacology should move from the analysis of the drug-individual interaction to the analysis of the drug-individual-society interaction. It should also analyze the clinical and economic consequences of the use of drugs in the conditions of normal clinical practice, beyond clinical trials. The current exponential technological development that facilitates the analysis of real-life data offers us a golden opportunity to move to all these other areas of interest. This review describes the role that clinical pharmacology has played at the beginning and during the evolution of pharmacovigilance, pharmacoepidemiology and economic drug evaluations in Spain. In addition, the challenges that clinical pharmacology is going to face in the following years in these three areas are going to be outlined too.

Cardiovascular Care Innovation through Data-Driven Discoveries in the Electronic Health Record

The electronic health record (EHR) represents a rich source of patient information, increasingly being leveraged for cardiovascular research. Although its primary use remains the seamless delivery of health care, the various longitudinally aggregated structured and unstructured data elements for each patient within the EHR can define the computational phenotypes of disease and care signatures and their association with outcomes. Although structured data elements, such as demographic characteristics, laboratory measurements, problem lists, and medications, are easily extracted, unstructured data are underused. The latter include free text in clinical narratives, documentation of procedures, and reports of imaging and pathology. Rapid scaling up of data storage and rapid innovation in natural language processing and computer vision can power insights from unstructured data streams. However, despite an array of opportunities for research using the EHR, specific expertise is necessary to adequately address confidentiality, accuracy, completeness, and heterogeneity challenges in EHR-based research. These often require methodological innovation and best practices to design and conduct successful research studies. Our review discusses these challenges and their proposed solutions. In addition, we highlight the ongoing innovations in federated learning in the EHR through a greater focus on common data models and discuss ongoing work that defines such an approach to large-scale, multicenter, federated studies. Such parallel improvements in technology and research methods enable innovative care and optimization of patient outcomes.

Unknown adverse drug reactions from spontaneous reports in a hospital setting: characterization, follow-up, and contribution to the pharmacovigilance system

Introduction: Post-marketing identification and report of unknown adverse drug reactions (ADRs) are crucial for patient safety. However, complete information on unknown ADRs seldom is available at the time of spontaneous ADR reports and this can hamper their contribution to the pharmacovigilance system. Methods: In order to characterize the seriousness and outcome of unknown ADRs at the time of report and at follow-up, and analyze their contribution to generate pharmacovigilance regulatory actions, a retrospective observational study of those identified in the spontaneous ADR reports of patients assisted at a hospital (January, 2016-December, 2021) was carried out. Information on demographic, clinical and complementary tests was retrieved from patients' hospital medical records. To evaluate the contribution to pharmacovigilance system we reviewed the European Union SmPCs, the list of the pharmacovigilance signals discussed by the Pharmacovigilance Risk Assessment Committee, and its recommendations reports on safety signals. Results: A total of 15.2% of the spontaneous reported cases during the study contained at least one unknown drug-ADR pair. After exclusions, 295 unknown drug-ADR pairs were included, within them the most frequently affected organs or systems were: skin and subcutaneous tissue (34, 11.5%), hepatobiliary disorders (28, 9.5%), cardiac disorders (28, 9.5%) and central nervous system disorders (27, 9.2%). The most frequent ADRs were pemphigus (7, 2.4%), and cytolytic hepatitis, sudden death, cutaneous vasculitis and fetal growth restriction with 6 (2%) each. Vaccines such as covid-19 and pneumococcus (68, 21.3%), antineoplastics such as paclitaxel, trastuzumab and vincristine (39, 12.2%) and immunosuppressants such as methotrexate and tocilizumab (35, 11%) were the most frequent drug subgroups involved. Sudden death due to hydroxychloroquine alone or in combination (4, 1.4%) and hypertransaminasemia by vincristine (n = 3, 1%) were the most frequent unknown drug-ADR pairs. A total of 269 (91.2%) of them were serious. Complementary tests were performed in 82.7% of unknown-ADR pairs and helped to reinforce their association in 18.3% of them. A total of 18 (6.1%) unknown drug-ADR pairs were evaluated by the EMA, in 8 (2.7%) the information was added to the drug's SmPC and in 1 case the risk prevention material was updated. Conclusion: Identification and follow-up of unknown ADRs can be of great relevance for patient safety and for the enrichment of the pharmacovigilance system.

Clinical insights into drug-associated pancreatic injury

PURPOSE OF REVIEW

Drug-induced pancreatitis is one of the top three causes of acute pancreatitis. A drug exposure is traditionally determined to be the cause of pancreatitis only after other possible and common causes of pancreatitis have been excluded.

RECENT FINDINGS

In this review, we challenge this traditional notion of drug-induced pancreatitis as a diagnosis of exclusion. Instead, we propose to shift the paradigm of conceptualizing what we term drug-associated pancreatic injury (DAPI); as a continuum of pancreatic injury that can be concomitant with other risk factors. The aims of this targeted review are to harness recent literature to build a foundation for conceptualizing DAPI, to highlight specific drugs associated with DAPI, and to describe a framework for future studies of DAPI.

SUMMARY

Our hope is that probing and characterizing the mechanisms underlying the various types of DAPI will lead to safer use of the DAPI-inducing drugs by minimizing the adverse event of pancreatitis.

Mobile app to perform anonymized longitudinal studies in the context of COVID-19 adverse drug reaction monitoring, leveraging the citizenship engagement

BACKGROUND

Over the last few years, increasingly studies focused on the development of mobile apps as complementary tools to existing pharmacovigilance traditional surveillance systems for improving and facilitating adverse drug reactions reporting.

OBJECTIVE

In this study, we evaluated the potentiality of a new mobile app (vaxEffect@UniMiB) to perform longitudinal studies while preserving the anonymity of the respondents. We applied it to monitor the adverse drug reactions during COVID-19 vaccination campaign in a sample of Italian population.

METHODS

We administered vaxEffect@UniMiB to a convenience sample of academic subjects vaccinated at Milano-Bicocca University hub for COVID-19 during the Italian national vaccination campaign. vaxEffect@UniMiB was developed for both Android and iOS devices. The mobile app asks users to send their medical history and, upon every vaccine administration, their vaccination data and the adverse reactions that occurred within seven days after the vaccination, allowing the follow of reactions dynamic for each respondent. The app sends data over the web to an application server. The web server, along with receiving all user data, saves them in a SQL database server, and reminds patients to submit vaccine and adverse reactions' data by push notifications sent to the mobile app through Firebase Cloud Messaging. On initial startup of the app, a unique user identifier was generated for each respondent, so that its anonymity is completely ensured, while enabling longitudinal studies.

RESULTS

A total of 3712 people have been vaccinated during the first vaccination wave. A total of 2733 respondents between the ages of 19 and 80, coming from the University of Milano-Bicocca and the Politecnico of Milan, participated in the survey. Overall, we collected the information about vaccination and adverse reactions to the first vaccine dose for 2226 subjects (60.0% of vaccinated), to the second dose for 1610 subjects (43.3%), and, in a non-sponsored fashion, to the third dose for 169 individuals.

CONCLUSIONS

vaxEffect@UniMiB revealed to be the first attempt in performing longitudinal studies to monitor the same subject over time in terms of the reported ADRs after each vaccine administration, while guaranteeing at the same time complete anonymity of the subjects. A series of aspects contributed to a positive involvement from people in using this application to report their ADRs to vaccination: ease of use, availability from multiple platforms, anonymity of all the survey participants and protection of the submitted data and the healthcare workers' support.

CLINICALTRIAL

Over the counter low-dose cannabidiol: A viewpoint from the ACRE Capacity Building Group

Amidst growing global acceptance of medicinal cannabinoids as a potential therapeutic interest in cannabidiol (CBD) is increasing. In Australia in 2020, a government inquiry examined the barriers that the public are experiencing in accessing medicinal cannabis. A number of recommendations to improve access were made. In response to these recommendations, the Australian therapeutics regulatory authority down-scheduled CBD from Prescription Only (Schedule 4) to Pharmacist Only (Schedule 3). As a group of early to mid-career researchers of the Australian Centre for Cannabinoid Clinical and Research Excellence (ACRE), we propose some considerations in relation to over-the-counter availability of CBD and opportunities to improve knowledge about its potential therapeutic benefits alongside its increased uptake.

Strategies prescribers and pharmacists use to identify and mitigate adverse drug reactions in inpatient and outpatient care: a cognitive task analysis at a US Veterans Affairs Medical Center

OBJECTIVE

To develop a descriptive model of the cognitive processes used to identify and resolve adverse drug reactions (ADRs) from the perspective of healthcare providers in order to inform future informatics efforts SETTING: Inpatient and outpatient care at a tertiary care US Veterans Affairs Medical Center.

PARTICIPANTS

Physicians, nurse practitioners and pharmacists who report ADRs.

OUTCOMES

Descriptive model and emerging themes from interviews.

RESULTS

We conducted critical decision method interviews with 10 physicians and 10 pharmacists. No nurse practitioners submitted ADR incidents. We generated a descriptive model of an ADR decision-making process and analysed emerging themes, categorised into four stages: detection of potential ADR, investigation of the problem's cause, risk/benefit consideration, and plan, action and follow-up. Healthcare professionals (HCPs) relied on several confirmatory or disconfirmatory cues to detect and investigate potential ADRs. Evaluating risks and benefits of related medications played an essential role in HCPs' pursuits of solutions CONCLUSIONS: This study provides an illustrative model of how HCPs detect problems and make decisions regarding ADRs. The design of supporting technology for potential ADR problems should align with HCPs' real-world cognitive strategies, to assist fully in detecting and preventing ADRs for patients.

Real-world evaluation of supportive care using an electronic health record text-mining tool: G-CSF use in breast cancer patients

PURPOSE

Chemotherapy-induced febrile neutropenia (FN) is a life-threatening and chemotherapy dose-limiting adverse event. FN can be prevented with granulocyte-colony stimulating factors (G-CSFs). Guidelines recommend primary G-CSF use for patients receiving either high (> 20%) FN risk (HR) chemotherapy, or intermediate (10-20%) FN risk (IR) chemotherapy if the overall risk with additional patient-related risk factors exceeds 20%. In this study, we applied an EHR text-mining tool for real-world G-CSF treatment evaluation in breast cancer patients.

METHODS

Breast cancer patients receiving IR or HR chemotherapy treatments between January 2015 and February 2021 at LUMC, the Netherlands, were included. We retrospectively collected data from EHR with a text-mining tool and assessed G-CSF use, risk factors, and the FN and neutropenia (grades 3-4) and incidence.

RESULTS

A total of 190 female patients were included, who received 77 HR and 113 IR treatments. In 88.3% of the HR regimens, G-CSF was administered; 7.3% of these patients developed FN vs. 33.3% without G-CSF. Although most IR regimen patients had ≥ 2 risk factors, only 4% received G-CSF, of which none developed neutropenia. However, without G-CSF, 11.9% developed FN and 31.2% severe neutropenia.

CONCLUSIONS

Our text-mining study shows high G-CSF use among HR regimen patients, and low use among IR regimen patients, although most had ≥ 2 risk factors. Therefore, current practice is not completely in accordance with the guidelines. This shows the need for increased awareness and clarity regarding risk factors. Also, text-mining can effectively be implemented for the evaluation of patient care.

Applying Machine Learning in Distributed Data Networks for Pharmacoepidemiologic and Pharmacovigilance Studies: Opportunities, Challenges, and Considerations

Increasing availability of electronic health databases capturing real-world experiences with medical products has garnered much interest in their use for pharmacoepidemiologic and pharmacovigilance studies. The traditional practice of having numerous groups use single databases to accomplish similar tasks and address common questions about medical products can be made more efficient through well-coordinated multi-database studies, greatly facilitated through distributed data network (DDN) architectures. Access to larger amounts of electronic health data within DDNs has created a growing interest in using data-adaptive machine learning (ML) techniques that can automatically model complex associations in high-dimensional data with minimal human guidance. However, the siloed storage and diverse nature of the databases in DDNs create unique challenges for using ML. In this paper, we discuss opportunities, challenges, and considerations for applying ML in DDNs for pharmacoepidemiologic and pharmacovigilance studies. We first discuss major types of activities performed by DDNs and how ML may be used. Next, we discuss practical data-related factors influencing how DDNs work in practice. We then combine these discussions and jointly consider how opportunities for ML are affected by practical data-related factors for DDNs, leading to several challenges. We present different approaches for addressing these challenges and highlight efforts that real-world DDNs have taken or are currently taking to help mitigate them. Despite these challenges, the time is ripe for the emerging interest to use ML in DDNs, and the utility of these data-adaptive modeling techniques in pharmacoepidemiologic and pharmacovigilance studies will likely continue to increase in the coming years.

Intelligent Telehealth in Pharmacovigilance: A Future Perspective

Pharmacovigilance improves patient safety by detecting and preventing adverse drug events. However, challenges exist that limit adverse drug event detection, resulting in many adverse drug events being underreported or inaccurately reported. One challenge includes having access to large data sets from various sources including electronic health records and wearable medical devices. Artificial intelligence, including machine learning methods, such as natural language processing and deep learning, can detect and extract information about adverse drug events, thus automating the pharmacovigilance process and improving the surveillance of known and documented adverse drug events. In addition, with the increased demand for telehealth services, for managing both acute and chronic diseases, artificial intelligence methods can play a role in detecting and preventing adverse drug events. In this review, we discuss two use cases of how artificial intelligence methods may be useful to improve the quality of pharmacovigilance and the role of artificial intelligence in telehealth practices.

Development and Implementation of an e-Trigger Tool for Adverse Drug Events in a Swiss University Hospital

PURPOSE

The purpose of the study was to develop and implement an institution-specific trigger tool based on the Institute for Healthcare Improvement medication module trigger tool (IHI MMTT) in order to detect and monitor ADEs.

METHODS

We performed an investigator-driven, single-center study using retrospective and prospective patient data to develop ("development phase") and implement ("implementation phase") an efficient, institution-specific trigger tool based on the IHI MMTT. Complete medical data from 1008 patients hospitalized in 2018 were used in the development phase. ADEs were identified by chart review. The performance of two versions of the tool was assessed by comparing their sensitivities and specificities. Tool A employed only digitally extracted triggers ("e-trigger-tool") while Tool B employed an additional manually extracted trigger. The superior tool - taking efficiency into account - was applied prospectively to 19-22 randomly chosen charts per month for 26 months during the implementation phase.

RESULTS

In the development phase, 189 (19%) patients had ≥1 ADE (total 277 ADEs). The time needed to identify these ADEs was 15 minutes/chart. A total of 203 patients had ≥1 trigger (total 273 triggers - Tool B). The sensitivities and specificities of Tools A and B were 0.41 and 0.86, and 0.43 and 0.86, respectively. Tool A was more time-efficient than Tool B (4 vs 9 minutes/chart) and was therefore used in the implementation phase. During the 26-month implementation phase, 22 patients experienced trigger-identified ADEs and 529 did not. The median number of ADEs per 1000 patient days was 6 (range 0-13). Patients with at least one ADE had a mean hospital stay of 22.3 ± 19.7 days, compared to 8.0 ± 7.6 days for those without an ADE (p = 2.7×10-14).

CONCLUSION

We developed and implemented an e-trigger tool that was specific and moderately sensitive, gave consistent results and required minimal resources.

Broadening the reach of the FDA Sentinel system: A roadmap for integrating electronic health record data in a causal analysis framework

The Sentinel System is a major component of the United States Food and Drug Administration's (FDA) approach to active medical product safety surveillance. While Sentinel has historically relied on large quantities of health insurance claims data, leveraging longitudinal electronic health records (EHRs) that contain more detailed clinical information, as structured and unstructured features, may address some of the current gaps in capabilities. We identify key challenges when using EHR data to investigate medical product safety in a scalable and accelerated way, outline potential solutions, and describe the Sentinel Innovation Center's initiatives to put solutions into practice by expanding and strengthening the existing system with a query-ready, large-scale data infrastructure of linked EHR and claims data. We describe our initiatives in four strategic priority areas: (1) data infrastructure, (2) feature engineering, (3) causal inference, and (4) detection analytics, with the goal of incorporating emerging data science innovations to maximize the utility of EHR data for medical product safety surveillance.

Effect of CYP3A5 and CYP3A4 Genetic Variants on Fentanyl Pharmacokinetics in a Pediatric Population

Fentanyl is an anesthetic/analgesic commonly used in surgical and recovery settings. CYP3A4 and CYP3A5 encode enzymes, which metabolize fentanyl; genetic variants in these genes impact fentanyl pharmacokinetics in adults. Pharmacokinetic (PK) studies are difficult to replicate in children due to the burden of additional blood taken solely for research purposes. The aim of this study is to test the effect of CYP3A5 and CYP3A4 genetic variants on fentanyl PKs in children using opportunistically collected samples. Fentanyl concentrations were measured from remnant blood specimens and dosing data were extracted from electronic health records. Variant data defining CYP3A4*1G and CYP3A5*3 and *6 alleles were available from prior genotyping; alleles with no variant were defined as *1. The study cohort included 434 individuals (median age 9 months, 52% male subjects) and 1,937 fentanyl concentrations were available. A two-compartment model was selected as the base model, and the final covariate model included age, weight, and surgical severity score. Clearance was significantly associated with either CYP3A5*3 or CYP3A5*6 alleles, but not the CYP3A4*1G allele. A genotype of CYP3A5*1/*3 or CYP3A5*1/*6 (i.e., intermediate metabolizer status) was associated with a 0.84-fold (95% confidence interval (CI): 0.71-1.00) reduction in clearance vs. CYP3A5*1/*1 (i.e., normal metabolizer status). CYP3A5*3/*3, CYP3A5*3/*6, or CYP3A5*6/*6 (i.e., poor metabolizer status) was associated with a 0.76-fold (95% CI: 0.58-0.99) reduction in clearance. In the final model, expected clearance was 8.9 and 6.8 L/hour for a normal and poor metabolizer, respectively, with median population covariates (9 months old, 7.7 kg, low surgical severity).

A weakly supervised model for the automated detection of adverse events using clinical notes

With clinical trials unable to detect all potential adverse reactions to drugs and medical devices prior to their release into the market, accurate post-market surveillance is critical to ensure their safety and efficacy. Electronic health records (EHR) contain rich observational patient data, making them a valuable source to actively monitor the safety of drugs and devices. While structured EHR data and spontaneous reporting systems often underreport the complexities of patient encounters and outcomes, free-text clinical notes offer greater detail about a patient's status. Previous studies have proposed machine learning methods to detect adverse events from clinical notes, but suffer from manually extracted features, reliance on costly hand-labeled data, and lack of validation on external datasets. To address these challenges, we develop a weakly-supervised machine learning framework for adverse event detection from unstructured clinical notes and evaluate it on insulin pump failure as a test case. Our model accurately detected cases of pump failure with 0.842 PR AUC on the holdout test set and 0.815 PR AUC when validated on an external dataset. Our approach allowed us to leverage a large dataset with far less hand-labeled data and can be easily transferred to additional adverse events for scalable post-market surveillance.

Pharmacosurveillance in Juvenile Idiopathic Arthritis

The advent of biologic disease-modifying antirheumatic drugs targeting specific cytokines or cell-cell interactions has dramatically changed the outlook of patients with juvenile idiopathic arthritis. However, safety concerns remain around the use of therapeutic agents for children with juvenile idiopathic arthritis. Foremost among these are the risks of serious infections and malignancy. This article provides an overview of methodologies for pharmacosurveillance in juvenile idiopathic arthritis, including spontaneous reporting systems and the use of diverse data sources, such as electronic health records, administrative claims, and clinical registries. The risks of infections and malignancies are then briefly reviewed.

A Survey on Recent Named Entity Recognition and Relationship Extraction Techniques on Clinical Texts

Significant growth in Electronic Health Records (EHR) over the last decade has provided an abundance of clinical text that is mostly unstructured and untapped. This huge amount of clinical text data has motivated the development of new information extraction and text mining techniques. Named Entity Recognition (NER) and Relationship Extraction (RE) are key components of information extraction tasks in the clinical domain. In this paper, we highlight the present status of clinical NER and RE techniques in detail by discussing the existing proposed NLP models for the two tasks and their performances and discuss the current challenges. Our comprehensive survey on clinical NER and RE encompass current challenges, state-of-the-art practices, and future directions in information extraction from clinical text. This is the first attempt to discuss both of these interrelated topics together in the clinical context. We identified many research articles published based on different approaches and looked at applications of these tasks. We also discuss the evaluation metrics that are used in the literature to measure the effectiveness of the two these NLP methods and future research directions.

Drug-Related Deaths in a Tertiary Hospital: Characteristics of Spontaneously Reported Cases and Comparison to Cases Detected from a Retrospective Study

Drug-related deaths (DRDs) are a common cause of hospital death. Pharmacovigilance, either as spontaneous reporting or active surveillance, plays a key role in the detection and reporting of suspected adverse drug reactions (ADRs). We conducted a retrospective analysis of all DRDs spontaneously reported to a pharmacovigilance program of a tertiary hospital, by health care professionals. We compared these results to those of a previous retrospective study conducted in the same hospital from the hospital's mortality registry. From 1460 spontaneous reported ADRs in a 10-year period, 73 (5%) were DRDs. The median age of DRD was 75 years (range 1 month-94) and 60.3% were men. The most frequent DRDs were hemorrhages (41.1%), followed by infections (17.8%). The most frequently involved drugs were anticoagulants and/or antithrombotic (30%), and antineoplastics (26.3%). When comparing both studies, spontaneous reporting detected more type B reactions (p < 0.001) and hospital-acquired DRD (p < 0.001); the number of concomitant drugs was higher (p = 0.0035); and the kind of ADR were different. The combination of several methods is mandatory to detect, assess, understand, and design strategies to prevent ADRs in a hospital setting, to ensure patient safety.

Spontaneously Reported Adverse Drug Reactions and Their Description in Hospital Discharge Reports: A Retrospective Study

The inclusion of spontaneously reported adverse drug reactions (ADRs) in hospital discharge reports was examined, in addition to the factors associated with their inclusion, the resulting therapeutic decisions, and any recommendations made upon patient discharge regarding the suspected offending drugs. ADRs that were spontaneously reported during 2017 and 2018 to the pharmacovigilance program were retrospectively analyzed. Information regarding patient characteristics, drug treatments, and ADRs was collected from the ADR notifications and from patient electronic medical records. The dependent variable was the mentioning of ADRs in the discharge reports, while characteristics of the ADRs, pharmacovigilance causality algorithms, and some of the suspected drugs themselves were the independent variables during bivariant analysis. A total of 286 reports of suspected ADRs from 271 patients (50.2% female; 77% adults) were included. Information regarding the ADRs was present in the discharge reports for 238 reports (83.2%); the ADR seriousness and the lack of potential alternative causes were the only associated factors. Withdrawal or withdrawal and substitution by an alternative drug were the most common therapeutic decisions, although often no recommendation was made. Overall, there is still room for improvement in terms of including information related to ADRs in hospital discharge reports.

Comprehensive evaluation of post-approval regulatory actions during the drug lifecycle - a focus on benefits and risks

Background: Prior studies investigated regulatory actions that reflected a negative impact on drug risks. We aimed to evaluate occurrence of regulatory actions that reflected a negative or positive impact on benefits or risks, as well as relations between them.Research design and methods: We followed EMA-approved innovative drugs from approval (2009-2010) until July 2020 or withdrawal to identify regulatory actions. We assessed these for impact on benefits or risks and relations between actions. Additionally, we scrutinized drug lifecycles for time-variant characteristics that may contribute to specific patterns of regulatory actions.Results: We identified 14 letters and 361 label updates for 40 drugs. Of the label updates, 85 (24%) reflected a positive impact, mostly concerning indications, and 276 (76%) a negative impact, mostly adverse drug reactions. Many updates (54%) occurred simultaneously with other updates, also if these reflected a different impact. Furthermore, levels of patient exposure, innovativeness, needs for regulatory learning and unexpected risks may contribute to patterns of regulatory actions.Conclusions: Almost a quarter of regulatory actions reflected a positive impact on benefits and risks. Also, simultaneous learning about benefits and risks suggests an important role for drug development in risk characterization. These findings may impact regulatory analyses and decision-making.

Identifying Drugs Inducing Prematurity by Mining Claims Data with High-Dimensional Confounder Score Strategies

BACKGROUND

Pregnant women are largely exposed to medications. However, knowledge is lacking about their effects on pregnancy and the fetus.

OBJECTIVE

This study sought to evaluate the potential of high-dimensional propensity scores and high-dimensional disease risk scores for automated signal detection in pregnant women from medico-administrative databases in the context of drug-induced prematurity.

METHODS

We used healthcare claims and hospitalization discharges of a 1/97th representative sample of the French population. We tested the association between prematurity and drug exposure during the trimester before delivery, for all drugs prescribed to at least five pregnancies. We compared different strategies (1) for building the two scores, including two machine-learning methods and (2) to account for these scores in the final logistic regression models: adjustment, weighting, and matching. We also proposed a new signal detection criterion derived from these scores: the p value relative decrease. Evaluation was performed by assessing the relevance of the signals using a literature review and clinical expertise.

RESULTS

Screening 400 drugs from a cohort of 57,407 pregnancies, we observed that choosing between the two machine-learning methods had little impact on the generated signals. Score adjustment performed better than weighting and matching. Using the p value relative decrease efficiently filtered out spurious signals while maintaining a number of relevant signals similar to score adjustment. Most of the relevant signals belonged to the psychotropic class with benzodiazepines, antidepressants, and antipsychotics.

CONCLUSIONS

Mining complex healthcare databases with statistical methods from the high-dimensional inference field may improve signal detection in pregnant women.

Characteristics of voluntary reporting of adverse drug events related to antipsychotics in Australia: 14-year analysis

BACKGROUND

Retrospective analyses of large databases of treated patients can provide useful links to the presence of drug misuse or rare and infrequent adverse effects, such as agranulocytosis, diabetic ketoacidosis or neuroleptic malignant syndrome. The aim of this study is to describe the adverse effects to antipsychotics reported in the Australian Database of Adverse Event Notifications (DAEN).

METHODS

Data were collected from the DAEN - a spontaneous reporting database. The database, which covered the period from January 2004 to December 2017, was obtained from the Therapeutic Goods Administration (TGA) website (www.TGA.gov). The drugs selected for this investigation are the following: aripiprazole, clozapine, olanzapine, paliperidone, risperidone, ziprasidone, quetiapine, haloperidol and pimozide. All data were analysed descriptively. Comparison of reporting and management of adverse events between adults (older than 20 years) and children (5-19 years) was undertaken using chi squared test, where p < 0.05 is significant.

RESULTS

A total of 7122 adverse events associated with the antipsychotics aripiprazole, clozapine, haloperidol, olanzapine, paliperidone, pimozide, quetiapine and risperidone were reported to the TGA between January 2004 and December 2017. On average, there were 2.6 adverse events reported for each case. The most common adverse event reported for antipsychotics was neuroleptic malignant syndrome. There were no significant differences in the number of co-medications, formulations, indications, therapeutic dose, hospital admission and overdose among the antipsychotics between paediatric and adult populations. However, there were significant differences between causality, death and the management of adverse events between adult and paediatric populations (5-19 years) (p < 0.05, chi squared test).

CONCLUSION

The antipsychotic drug associated with the highest adverse events in adults was clozapine, followed by olanzapine. The most common adverse event in adults, and reported with a number of antipsychotic drugs, was neuroleptic malignant syndrome. In children, the highest numbers of adverse events reported in the database were associated with risperidone, clozapine and olanzapine.

PLAIN LANGUAGE SUMMARY

Adverse events reported of antipsychoticsBackground: Retrospective analyses of large databases of treated patients can provide useful clues to the presence of drug misuse or rare and infrequent adverse effects associated with antipsychotics. The drugs selected for this investigation are the following: aripiprazole, clozapine, olanzapine, paliperidone, risperidone, ziprasidone, quetiapine, haloperidol and pimozide.Methods: All data were analysed descriptively and investigated for any associations between the variables collected. Comparison of reporting and management of adverse events between adults (older than 20 years) and children (5-19 years) was undertaken using chi squared test, where p < 0.05 is significant.Results: The antipsychotic drug associated with the highest adverse events was clozapine, followed by olanzapine. In children, the highest numbers of adverse events reported in the database were associated with risperidone, clozapine and olanzapine. The most common adverse event in adults, and reported with a number of antipsychotic drugs, was neuroleptic malignant syndrome.Conclusion: There were significant differences between causality, death and the management of adverse events between adult and paediatric populations (5-19 years).Keywords: Antipsychotics, adverse effects, adverse events, safety.

A combined recall and rank framework with online negative sampling for chinese procedure terminology normalization

MOTIVATION

Medical terminology normalization aims to map the clinical mention to terminologies coming from a knowledge base, which plays an important role in analyzing Electronic Health Record (EHR) and many downstream tasks. In this paper, we focus on Chinese procedure terminology normalization. The expressions of terminology are various and one medical mention may be linked to multiple terminologies. Existing studies based on Learning To Rank (LTR) does not fully consider the quality of negative samples during model training and the importance of keywords in this domain-specific task.

RESULTS

We propose a combined recall and rank framework to solve these problems. A pair-wise Bert model with deep metric learning is used to recall candidates. Previous methods either train Bert in a point-wise way or based on a multi-class classification problem, which may lead serious efficiency problems or not be effective enough. During model training, we design a novel online negative sampling algorithm to activate the pair-wise method. To deal with multi-implication scenarios, we train the task of implication number prediction together with the recall task in a multi-task learning setting, since these two tasks are highly complementary. In rank step, we propose a keywords attentive mechanism to focus on domain-specific information such as procedure sites and procedure types. Finally, a fusion block merges the results of the recall and the rank model. Detailed experimental analysis shows our proposed framework has a remarkable improvement on both performance and efficiency.

AVAILABILITY

The source code will be available at https://github.com/sxthunder/CMTN upon publication.

Identifying relations of medications with adverse drug events using recurrent convolutional neural networks and gradient boosting

OBJECTIVE

To develop a natural language processing system that identifies relations of medications with adverse drug events from clinical narratives. This project is part of the 2018 n2c2 challenge.

MATERIALS AND METHODS

We developed a novel clinical named entity recognition method based on an recurrent convolutional neural network and compared it to a recurrent neural network implemented using the long-short term memory architecture, explored methods to integrate medical knowledge as embedding layers in neural networks, and investigated 3 machine learning models, including support vector machines, random forests and gradient boosting for relation classification. The performance of our system was evaluated using annotated data and scripts provided by the 2018 n2c2 organizers.

RESULTS

Our system was among the top ranked. Our best model submitted during this challenge (based on recurrent neural networks and support vector machines) achieved lenient F1 scores of 0.9287 for concept extraction (ranked third), 0.9459 for relation classification (ranked fourth), and 0.8778 for the end-to-end relation extraction (ranked second). We developed a novel named entity recognition model based on a recurrent convolutional neural network and further investigated gradient boosting for relation classification. The new methods improved the lenient F1 scores of the 3 subtasks to 0.9292, 0.9633, and 0.8880, respectively, which are comparable to the best performance reported in this challenge.

CONCLUSION

This study demonstrated the feasibility of using machine learning methods to extract the relations of medications with adverse drug events from clinical narratives.

Withdrawal of hospital outpatient treatments in severe diseases due to unacceptable toxicity: A retrospective study from the register of patients and treatments

AIM

To retrospectively analyse hospital outpatient treatment (HOT) withdrawal due to unacceptable toxicity at our hospital. Information regarding unacceptable toxicity leading to treatment withdrawal was recorded.

METHODS

HOT interruptions because of unacceptable toxicity were identified from the Register of Patients and Treatments (RPT) (January 2014 to December 2017). Information regarding the demographic and clinical characteristics of patients, adverse drug reactions (ADRs) and drug treatments was retrieved from electronic health records. Causality and previous knowledge of ADRs were assessed according to the Spanish Pharmacovigilance System algorithm. Information regarding HOT risk management plans (RMPs) and their classification as inverted black triangle medicines was obtained from the European Medicines Agency (EMA).

RESULTS

HOTs were withdrawn due to unacceptable toxicity in 136 (1.5%) registries corresponding to 135 (1.7%) patients. Fifty-one different HOTs (38.6% of those registered) were involved in 240 ADR/HOT pairs: 24 (47%) were additional monitoring medicines and 37 (72.5%) were EMA RMPs. The most frequent medicines involved in ADRs were lenalidomide (30, 12.5%) (mainly neutropenia, thrombocytopenia and bicytopenia), bevacizumab (19, 7.9%) (mainly venous and pulmonary thromboembolism) and sunitinib (13, 5.4%) (mainly thromboembolic events, diarrhoea and worsening of chronic renal failure). Cytopenia (40, 17.3%), diarrhoea (15, 6.5%), asthenia (9, 3.9%) and neuropathy (6, 2.6%) were the most frequent ADRs. All ADRs were severe, 10 (6 patients) had been poorly described or were unknown and only 9 (5 patients) had been reported by spontaneous notification.

CONCLUSIONS

Valuable information regarding severe and unknown ADRs was obtained from the RPT. Such registers are useful tools to complement spontaneous ADR notifications.

Unified Medical Language System resources improve sieve-based generation and Bidirectional Encoder Representations from Transformers (BERT)-based ranking for concept normalization

OBJECTIVE

Concept normalization, the task of linking phrases in text to concepts in an ontology, is useful for many downstream tasks including relation extraction, information retrieval, etc. We present a generate-and-rank concept normalization system based on our participation in the 2019 National NLP Clinical Challenges Shared Task Track 3 Concept Normalization.

MATERIALS AND METHODS

The shared task provided 13 609 concept mentions drawn from 100 discharge summaries. We first design a sieve-based system that uses Lucene indices over the training data, Unified Medical Language System (UMLS) preferred terms, and UMLS synonyms to generate a list of possible concepts for each mention. We then design a listwise classifier based on the BERT (Bidirectional Encoder Representations from Transformers) neural network to rank the candidate concepts, integrating UMLS semantic types through a regularizer.

RESULTS

Our generate-and-rank system was third of 33 in the competition, outperforming the candidate generator alone (81.66% vs 79.44%) and the previous state of the art (76.35%). During postevaluation, the model's accuracy was increased to 83.56% via improvements to how training data are generated from UMLS and incorporation of our UMLS semantic type regularizer.

DISCUSSION

Analysis of the model shows that prioritizing UMLS preferred terms yields better performance, that the UMLS semantic type regularizer results in qualitatively better concept predictions, and that the model performs well even on concepts not seen during training.

CONCLUSIONS

Our generate-and-rank framework for UMLS concept normalization integrates key UMLS features like preferred terms and semantic types with a neural network-based ranking model to accurately link phrases in text to UMLS concepts.

Adverse Drug Reaction Case Safety Practices in Large Biopharmaceutical Organizations from 2007 to 2017: An Industry Survey

INTRODUCTION

Drug safety remains a top global public health concern. An increase in the number of data sources available has increased the complexity of pharmacovigilance operations, so the US FDA has created draft guidance focusing on optimizing drug safety data for well-characterized medicines. However, to date, no data demonstrating changes in reports have been presented.

OBJECTIVES

This study provided data assessing changes in individual case safety reports (ICSRs) and aggregate reports (ARs) for large biopharmaceutical companies from 2007 to 2017. This study also evaluated current trends on the use of advanced machine and deep learning in order to process all data captured for ICSRs as well as opinions from industry thought leaders on creating a sustainable case-processing operation.

METHODOLOGY

Using data captured from Navitas Life Science's annual pvnet^® benchmark, we calculated workload indicators characterizing pharmacovigilance operations for large biopharmaceutical organizations. Workload indicators included the number of ICSRs by organization, the number of ARs, and the number and types of data sources used. We also conducted structured in-depth interviews with seven biopharmaceutical executives to discover the reasons for changes in workload indicators across time as well as current strategies for increasing efficiencies in drug safety reporting.

RESULTS

The median number of ICSRs increased from 84,960 cases in 2007 to over 200,000 cases in 2017; this increase was largely attributable to an increase in both nonserious cases and follow-up cases. Member companies reported using 12 ± 3 data sources for case identification. The number of ARs also increased from a median of 70 reports in 2007 to 258 reports in 2017. To address these increases, 61% of the biopharmaceutical organizations we surveyed planned to adopt machine learning for full ICSR processing; however, as of 2018, none of the organizations surveyed had mechanisms in place.

CONCLUSION

This study demonstrated that pharmacovigilance departments are currently burdened by ever-increasing case volumes. With increased guidance from regulatory agencies, as well as improvements in artificial intelligence and natural language processing, biopharmaceutical organizations must determine the most resource-efficient and sustainable methods to process the growing volume of cases.

Detecting adverse drug reactions in discharge summaries of electronic medical records using Readpeer

BACKGROUND

Hospital discharge summaries offer a potentially rich resource to enhance pharmacovigilance efforts to evaluate drug safety in real-world clinical practice. However, it is infeasible for experts to read through all discharge summaries to find cases of drug-adverse event (AE) relations.

PURPOSE

The objective of this paper is to develop a natural language processing (NLP) framework to detect drug-AE relations from unstructured hospital discharge summaries.

BASIC PROCEDURES

An NLP algorithm was designed using customized dictionaries of drugs, adverse event (AE) terms, and rules based on trigger phrases, negations, fuzzy logic and word distances to recognize drug, AE terms and to detect drug-AE relations. Furthermore, a customized annotation tool was developed to facilitate expert review of discharge summaries from a tertiary hospital in Singapore in 2011.

MAIN FINDINGS

A total of 33 trial sets with 50 to 100 records per set were evaluated (1620 discharge summaries) by our algorithm and reviewed by pharmacovigilance experts. After every 6 trial sets, drug and AE dictionaries were updated, and rules were modified to improve the system. Excellent performance was achieved for drug and AE entity recognition with over 92% precision and recall. On the final 6 sets of discharge summaries (600 records), our algorithm achieved 75% precision and 59% recall for identification of valid drug-AE relations.

PRINCIPAL CONCLUSIONS

Adverse drug reactions are a significant contributor to health care costs and utilization. Our algorithm is not restricted to particular drugs, drug classes or specific medical specialties, which is an important attribute for a national regulatory authority to carry out comprehensive safety monitoring of drug products. Drug and AE dictionaries may be updated periodically to ensure that the tool remains relevant for performing surveillance activities. The development of the algorithm, and the ease of reviewing and correcting the results of the algorithm as part of an iterative machine learning process, is an important step towards use of hospital discharge summaries for an active pharmacovigilance program.

Creation of a Multicenter Pediatric Inpatient Data Repository Derived from Electronic Health Records

BACKGROUND

Integration of electronic health records (EHRs) data across sites and access to that data remain limited.

OBJECTIVE

We developed an EHR-based pediatric inpatient repository using nine U.S. centers from the National Institute of Child Health and Human Development Pediatric Trials Network.

METHODS

A data model encompassing 147 mandatory and 99 optional elements was developed to provide an EHR data extract of all inpatient encounters from patients <17 years of age discharged between January 6, 2013 and June 30, 2017. Sites received instructions on extractions, transformation, testing, and transmission to the coordinating center.

RESULTS

We generated 177 staging reports to process all nine sites' 147 mandatory and 99 optional data elements to the repository. Based on 520 prespecified criteria, all sites achieved 0% errors and <2% warnings. The repository includes 386,159 inpatient encounters from 264,709 children to support study design and conduct of future trials in children.

CONCLUSION

Our EHR-based data repository of pediatric inpatient encounters utilized a customized data model heavily influenced by the PCORnet format, site-based data mapping, a comprehensive set of data testing rules, and an iterative process of data submission. The common data model, site-based extraction, and technical expertise were key to our success. Data from this repository will be used in support of Pediatric Trials Network studies and the labeling of drugs and devices for children.