Using Electronic Health Records to Identify Adverse Drug Events in Ambulatory Care: A Systematic Review

Accelerating adverse pregnancy outcomes research amidst rising medication use: parallel retrospective cohort analyses for signal prioritization

BACKGROUND

Pregnant women are significantly underrepresented in clinical trials, yet most of them take medication during pregnancy despite the limited safety data. The objective of this study was to characterize medication use during pregnancy and apply propensity score matching method at scale on patient records to accelerate and prioritize the drug effect signal detection associated with the risk of preterm birth and other adverse pregnancy outcomes.

METHODS

This was a retrospective study on continuously enrolled women who delivered live births between 2013/01/01 and 2022/12/31 (n = 365,075) at Providence St. Joseph Health. Our exposures of interest were all outpatient medications prescribed during pregnancy. We limited our analyses to medication that met the minimal sample size (n = 600). The primary outcome of interest was preterm birth. Secondary outcomes of interest were small for gestational age and low birth weight. We used propensity score matching at scale to evaluate the risk of these adverse pregnancy outcomes associated with drug exposure after adjusting for demographics, pregnancy characteristics, and comorbidities.

RESULTS

The total medication prescription rate increased from 58.5 to 75.3% (P < 0.0001) from 2013 to 2022. The prevalence rate of preterm birth was 7.7%. One hundred seventy-five out of 1329 prenatally prescribed outpatient medications met the minimum sample size. We identified 58 medications statistically significantly associated with the risk of preterm birth (P ≤ 0.1; decreased: 12, increased: 46).

CONCLUSIONS

Most pregnant women are prescribed medication during pregnancy. This highlights the need to utilize existing real-world data to enhance our knowledge of the safety of medications in pregnancy. We narrowed down from 1329 to 58 medications that showed statistically significant association with the risk of preterm birth even after addressing numerous covariates through propensity score matching. This data-driven approach demonstrated that multiple testable hypotheses in pregnancy pharmacology can be prioritized at scale and lays the foundation for application in other pregnancy outcomes.

Scalable incident detection via natural language processing and probabilistic language models

Post marketing safety surveillance depends in part on the ability to detect concerning clinical events at scale. Spontaneous reporting might be an effective component of safety surveillance, but it requires awareness and understanding among healthcare professionals to achieve its potential. Reliance on readily available structured data such as diagnostic codes risks under-coding and imprecision. Clinical textual data might bridge these gaps, and natural language processing (NLP) has been shown to aid in scalable phenotyping across healthcare records in multiple clinical domains. In this study, we developed and validated a novel incident phenotyping approach using unstructured clinical textual data agnostic to Electronic Health Record (EHR) and note type. It's based on a published, validated approach (PheRe) used to ascertain social determinants of health and suicidality across entire healthcare records. To demonstrate generalizability, we validated this approach on two separate phenotypes that share common challenges with respect to accurate ascertainment: (1) suicide attempt; (2) sleep-related behaviors. With samples of 89,428 records and 35,863 records for suicide attempt and sleep-related behaviors, respectively, we conducted silver standard (diagnostic coding) and gold standard (manual chart review) validation. We showed Area Under the Precision-Recall Curve of ~ 0.77 (95% CI 0.75-0.78) for suicide attempt and AUPR ~ 0.31 (95% CI 0.28-0.34) for sleep-related behaviors. We also evaluated performance by coded race and demonstrated differences in performance by race differed across phenotypes. Scalable phenotyping models, like most healthcare AI, require algorithmovigilance and debiasing prior to implementation.

Transforming Hospital Quality Improvement Through Harnessing the Power of Artificial Intelligence

This policy analysis focuses on harnessing the power of artificial intelligence (AI) in hospital quality improvement to transform quality and patient safety. It examines the application of AI at the two following fundamental levels: (1) diagnostic and treatment and (2) clinical operations. AI applications in diagnostics directly impact patient care and safety. At the same time, AI indirectly influences patient safety at the clinical operations level by streamlining (1) operational efficiency, (2) risk assessment, (3) predictive analytics, (4) quality indicators reporting, and (5) staff training and education. The challenges and future perspectives of AI application in healthcare, encompassing technological, ethical, and other considerations, are also critically analyzed.

Extracting adverse drug events from clinical Notes: A systematic review of approaches used

BACKGROUND

An adverse drug event (ADE) is any unfavorable effect that occurs due to the use of a drug. Extracting ADEs from unstructured clinical notes is essential to biomedical text extraction research because it helps with pharmacovigilance and patient medication studies.

OBJECTIVE

From the considerable amount of clinical narrative text, natural language processing (NLP) researchers have developed methods for extracting ADEs and their related attributes. This work presents a systematic review of current methods.

METHODOLOGY

Two biomedical databases have been searched from June 2022 until December 2023 for relevant publications regarding this review, namely the databases PubMed and Medline. Similarly, we searched the multi-disciplinary databases IEEE Xplore, Scopus, ScienceDirect, and the ACL Anthology. We adopted the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement guidelines and recommendations for reporting systematic reviews in conducting this review. Initially, we obtained 5,537 articles from the search results from the various databases between 2015 and 2023. Based on predefined inclusion and exclusion criteria for article selection, 100 publications have undergone full-text review, of which we consider 82 for our analysis.

RESULTS

We determined the general pattern for extracting ADEs from clinical notes, with named entity recognition (NER) and relation extraction (RE) being the dual tasks considered. Researchers that tackled both NER and RE simultaneously have approached ADE extraction as a "pipeline extraction" problem (n = 22), as a "joint task extraction" problem (n = 7), and as a "multi-task learning" problem (n = 6), while others have tackled only NER (n = 27) or RE (n = 20). We further grouped the reviews based on the approaches for data extraction, namely rule-based (n = 8), machine learning (n = 11), deep learning (n = 32), comparison of two or more approaches (n = 11), hybrid (n = 12) and large language models (n = 8). The most used datasets are MADE 1.0, TAC 2017 and n2c2 2018.

CONCLUSION

Extracting ADEs is crucial, especially for pharmacovigilance studies and patient medications. This survey showcases advances in ADE extraction research, approaches, datasets, and state-of-the-art performance in them. Challenges and future research directions are highlighted. We hope this review will guide researchers in gaining background knowledge and developing more innovative ways to address the challenges.

Scalable Incident Detection via Natural Language Processing and Probabilistic Language Models

Post marketing safety surveillance depends in part on the ability to detect concerning clinical events at scale. Spontaneous reporting might be an effective component of safety surveillance, but it requires awareness and understanding among healthcare professionals to achieve its potential. Reliance on readily available structured data such as diagnostic codes risk under-coding and imprecision. Clinical textual data might bridge these gaps, and natural language processing (NLP) has been shown to aid in scalable phenotyping across healthcare records in multiple clinical domains. In this study, we developed and validated a novel incident phenotyping approach using unstructured clinical textual data agnostic to Electronic Health Record (EHR) and note type. It's based on a published, validated approach (PheRe) used to ascertain social determinants of health and suicidality across entire healthcare records. To demonstrate generalizability, we validated this approach on two separate phenotypes that share common challenges with respect to accurate ascertainment: 1) suicide attempt; 2) sleep-related behaviors. With samples of 89,428 records and 35,863 records for suicide attempt and sleep-related behaviors, respectively, we conducted silver standard (diagnostic coding) and gold standard (manual chart review) validation. We showed Area Under the Precision-Recall Curve of ∼ 0.77 (95% CI 0.75-0.78) for suicide attempt and AUPR ∼ 0.31 (95% CI 0.28-0.34) for sleep-related behaviors. We also evaluated performance by coded race and demonstrated differences in performance by race were dissimilar across phenotypes and require algorithmovigilance and debiasing prior to implementation.

The Effects of CYP2C19 Genotype on Proxies of SSRI Antidepressant Response in the UK Biobank

Selective serotonin reuptake inhibitors (SSRIs) are the most commonly used psychopharmaceutical treatment for major depressive disorder (MDD), but individual responses to SSRIs vary greatly. CYP2C19 is a key enzyme involved in the metabolism of several drugs, including SSRIs. Variations in the CYP2C19 gene are associated with differential metabolic activity, and thus differential SSRI exposure; accordingly, the CYP2C19 genotype may affect the therapeutic response and clinical outcomes, though existing evidence of this link is not entirely consistent. Therefore, we analysed data from the UK Biobank, a large, deeply phenotyped prospective study, to investigate the effects of CYP2C19 metaboliser phenotypes on several clinical outcomes derived from primary care records, including multiple measures of antidepressant switching, discontinuation, duration, and side effects. In this dataset, 24,729 individuals were prescribed citalopram, 3012 individuals were prescribed escitalopram, and 12,544 individuals were prescribed sertraline. Consistent with pharmacological expectations, CYP2C19 poor metabolisers on escitalopram were more likely to switch antidepressants, have side effects following first prescription, and be on escitalopram for a shorter duration compared to normal metabolisers. CYP2C19 poor and intermediate metabolisers on citalopram also exhibited increased odds of discontinuation and shorter durations relative to normal metabolisers. Generally, no associations were found between metabolic phenotypes and proxies of response to sertraline. Sensitivity analyses in a depression subgroup and metabolic activity scores corroborated results from the primary analysis. In summary, our findings suggest that CYP2C19 genotypes, and thus metabolic phenotypes, may have utility in determining clinical responses to SSRIs, particularly escitalopram and citalopram, though further investigation of such a relationship is warranted.

Causality, Severity, and Avoidability of Adverse Drug Reactions in Hospitalized Children: A Prospective Cohort Study

BACKGROUND

 Adverse drug reactions are an important cause of morbidity and mortality in all patients. Information regarding adverse drug reactions in the pediatric age group, especially with regard to the drugs involved and the clinical presentations is scanty. The aim of our study is to determine the incidence of adverse drug reactions and to study their features in terms of causality, type, severity, avoidability, drugs implicated and their clinical presentations.

METHODS

 The study was carried out on patients admitted to the pediatric ward and the pediatric intensive care unit over a one-year period (January 1, 2013 to December 31, 2013). Patients either presenting with or developing an adverse drug reaction in the hospital were included in the study.

RESULTS

 The incidence rate for adverse drug reaction causing hospital admission was 1.79% (95% CI 1.48, 2.16) whereas it was 1.23% (95% CI 0.97, 1.53) for children exposed to a drug during their hospital stay. Type B (bizarre or idiosyncratic type) was seen in 114 (62.6%) of the ADRs whereas 53 (29.1%) were of type A (augmented pharmacologic effect). Severe ADRs were seen in 25 (13.7%) of the total ADRs. ADR was responsible for the death of two patients. 15.4% were rated as avoidable. Anti-microbials were the most common group responsible for ADRs (43.4%), followed by drugs acting on the immune system (15.9%) and drugs acting on the nervous system (14.3%). The most common ADRs were metabolic (29.3%) followed by neurological (17.6%).

CONCLUSIONS

 Adverse drug reactions can occur in a substantial proportion of hospitalized patients with some of them being severe and potentially avoidable. Awareness among physicians should be encouraged regarding monitoring, documentation and notification of adverse drug reactions.

Malpractice Claims and Incident Reporting: Two Faces of the Same Coin?

Incident reporting is an important method to identify risks because learning from the reports is crucial in developing and implementing effective improvements. A medical malpractice claims analysis is an important tool in any case. Both incident reports and claims show cases of damage caused to patients, despite incident reporting comprising near misses, cases where no event occurred and no-harm events. We therefore compare the two worlds to assess whether they are similar or definitively different. From 1 January 2014 to 31 December 2021, the claims database of Policlinico Universitario A. Gemelli IRCCS collected 843 claims. From 1 January 2020 to 31 December 2021, the incident-reporting database collected 1919 events. In order to compare the two, we used IBNR calculation, usually adopted by the insurance industry to determine loss to a company and to evaluate the real number of adverse events that occurred. Indeed, the number of reported adverse events almost overlapped with the total number of events, which is indicative that incurred-but-not-reported events are practically irrelevant. The distribution of damage events reported as claims in the period from 1 January 2020 to 31 December 2021 and related to incidents that occurred in the months of the same period, grouped by quarter, was then compared with the distribution of damage events reported as adverse events and sentinel events in the same period, grouped by quarter. The analysis of the claims database showed that the claims trend is slightly decreasing. However, the analysis of the reports database showed that, in the period 2020-2021, the reports trend was increasing. In our study, the comparison of the two, malpractice claims and incident reporting, documented many differences and weak areas of overlap. Nevertheless, this contribution represents the first attempt to compare the two and new studies focusing on single types of adverse events are, therefore, desirable.

Ambulatory Medication Safety in Primary Care: A Systematic Review

PURPOSE

To review the literature on medication safety in primary care in the electronic health record era.

METHODS

Included studies measured rates and outcomes of medication safety in patients whose prescriptions were written in primary care clinics with electronic prescribing. Four investigators independently reviewed titles and analyzed abstracts with dual-reviewer review for eligibility, characteristics, and risk of bias.

RESULTS

Of 1464 articles identified, 56 met the inclusion criteria. Forty-three studies were noninterventional and 13 included an intervention. The majority of the studies (30) used their own definition of error. The most common outcomes were potentially inappropriate prescribing/medications (PIPs), adverse drug events (ADEs), and potential prescribing omissions (PPOs). Most of the studies only included high-risk subpopulations (39), usually older adults taking > 4 medications. The rate of PIPs varied widely (0.19% to 98.2%). The rate of ADEs was lower (0.47% to 14.7%). There was poor correlation of PIP and PPO with documented ADEs leading to physical harm.

CONCLUSIONS

This literature is limited by its inconsistent and highly variable outcomes. The majority of medication safety studies in primary care were in high-risk populations and measured potential harms rather than actual harms. Applying algorithms to primary care medication lists significantly overestimates rate of actual harms.

A retrospective study of sugammadex for reversal of neuromuscular blockade induced by rocuronium in critically ill patients in the ICU

Sugammadex has been approved for reversal of neuromuscular blockade by vecuronium and rocuronium in adults undergoing surgery. Although widely used in the operating room, sugammadex has not been investigated in the intensive care unit setting. This study aimed to evaluate the use of sugammadex in critically ill patients with a focus on known drug-related adverse reactions. In this single-center, retrospective, observational study, 91 critically ill patients who were administered sugammadex while in the ICU were evaluated. Electronic health records were reviewed, and baseline data, as well as indication and incidence of complications possibly related to sugammadex, were retrospectively collected. The most common procedures requiring neuromuscular blockade followed by reversal with sugammadex were bronchoscopy, percutaneous dilatative tracheostomy, and percutaneous endoscopic gastrostomy. Within 2 h following administration of sugammadex, skin rash and use of antihistamines were reported in 4 patients (4.4%) in total; bradycardia was observed in 9 patients (9.9%), and respiratory adverse events were described in 3 patients (3.3%). New-onset bleeding up to 24 h after sugammadex was reported in 7 patients (7.7%), 3of whom received transfusions of packed red blood cells. Sugammadex was well tolerated in critically ill patients and could be considered for reversal of neuromuscular blockade in this population. Larger prospective studies are required to determine the safety profile and evaluate the potential benefit and indications of sugammadex in the critical care setting.

Comparing the dangers of a stay in English and German hospitals for high-need patients

OBJECTIVE

To estimate the risk of an avoidable adverse event for high-need patients in England and Germany and the causal impact that has on outcomes.

DATA SOURCES

We use administrative, secondary data for all hospital inpatients in 2018. Patient records for the English National Health Service are provided by the Hospital Episode Statistics database and for the German health care system accessed through the Research Data Center of the Federal Statistical Office.

STUDY DESIGN

We calculated rates of three hospital-acquired adverse events and their causal impact on mortality and length of stay through propensity score matching and estimation of average treatment effects.

DATA COLLECTION/EXTRACTION METHODS

Patients were identified based on diagnoses codes and translated Patient Safety Indicators developed by the Agency for Healthcare Research and Quality.

PRINCIPAL FINDINGS

For the average hospital stay, the risk of an adverse event was 5.37% in the English National Health Service and 3.26% in the German health care system. High-need patients are more likely to experience an adverse event, driven by hospital-acquired infections (2.06%-4.45%), adverse drug reactions (2.37%-2.49%), and pressure ulcers (2.25%-0.45%). Adverse event risk is particularly high for patients with advancing illnesses (10.50%-27.11%) and the frail elderly (17.75%-28.19%). Compared to the counterfactual, high-need patients with an adverse event are more likely to die during their hospital stay and experience a longer length of stay.

CONCLUSIONS

High-need patients are particularly vulnerable with an adverse event risking further deterioration of health status and adding resource use. Our results indicate the need to assess the costs and benefits of a hospital stay, particularly when care could be provided in settings considered less hazardous.

Optimizing Clinical Monitoring Tools to Enhance Patient Review by Pharmacists

BACKGROUND

The Clinical Monitoring List (CML) is a real-time scoring system and intervention tool used by Mayo Clinic pharmacists caring for hospitalized patients.

OBJECTIVE

The study aimed to describe the iterative development and implementation of pharmacist clinical monitoring tools within the electronic health record at a multicampus health system enterprise.

METHODS

Between October 2018 and January 2019, pharmacists across the enterprise were surveyed to determine opportunities and gaps in CML functionality. Responses were received from 39% (n = 162) of actively staffing inpatient pharmacists. Survey responses identified three main gaps in CML functionality: (1) the desire for automated checklists of tasks, (2) additional rule logic closely aligning with clinical practice guidelines, and (3) the ability to dismiss and defer rules. The failure mode and effect analysis were used to assess risk areas within the CML. To address identified gaps, two A/B testing pilots were undertaken. The first pilot analyzed the effect of updated CML rule logic on pharmacist satisfaction in the domains of automated checklists and guideline alignment. The second pilot assessed the utility of a Clinical Monitoring Navigator (CMN) functioning in conjunction with the CML to display rules with selections to dismiss or defer rules until a user-specified date. The CMN is a workspace to guide clinical end user workflows; permitting the review and actions to be completed within one screen using EHR functionality.

RESULTS

A total of 27 pharmacists across a broad range of practice specialties were selected for two separate two-week pilot tests. Upon pilot completion, participants were surveyed to assess the effect of updates on performance gaps.

CONCLUSION

Findings from the enterprise-wide survey and A/B pilot tests were used to inform final build decisions and planned enterprise-wide updated CML and CMN launch. This project serves as an example of the utility of end-user feedback and pilot testing to inform project decisions, optimize usability, and streamline build activities.

Collecting specialty-related medical terms: Development and evaluation of a resource for Spanish

BACKGROUND

Controlled vocabularies are fundamental resources for information extraction from clinical texts using natural language processing (NLP). Standard language resources available in the healthcare domain such as the UMLS metathesaurus or SNOMED CT are widely used for this purpose, but with limitations such as lexical ambiguity of clinical terms. However, most of them are unambiguous within text limited to a given clinical specialty. This is one rationale besides others to classify clinical text by the clinical specialty to which they belong.

RESULTS

This paper addresses this limitation by proposing and applying a method that automatically extracts Spanish medical terms classified and weighted per sub-domain, using Spanish MEDLINE titles and abstracts as input. The hypothesis is biomedical NLP tasks benefit from collections of domain terms that are specific to clinical subdomains. We use PubMed queries that generate sub-domain specific corpora from Spanish titles and abstracts, from which token n-grams are collected and metrics of relevance, discriminatory power, and broadness per sub-domain are computed. The generated term set, called Spanish core vocabulary about clinical specialties (SCOVACLIS), was made available to the scientific community and used in a text classification problem obtaining improvements of 6 percentage points in the F-measure compared to the baseline using Multilayer Perceptron, thus demonstrating the hypothesis that a specialized term set improves NLP tasks.

CONCLUSION

The creation and validation of SCOVACLIS support the hypothesis that specific term sets reduce the level of ambiguity when compared to a specialty-independent and broad-scope vocabulary.

Development of a Taxonomy for Medication-Related Patient Safety Events Related to Health Information Technology in Pediatrics

BACKGROUND

Although electronic health records (EHRs) are designed to improve patient safety, they have been associated with serious patient harm. An agreed-upon and standard taxonomy for classifying health information technology (HIT) related patient safety events does not exist.

OBJECTIVES

We aimed to develop and evaluate a taxonomy for medication-related patient safety events associated with HIT and validate it using a set of events involving pediatric patients.

METHODS

We performed a literature search to identify existing classifications for HIT-related safety events, which were assessed using real-world pediatric medication-related patient safety events extracted from two sources: patient safety event reporting system (ERS) reports and information technology help desk (HD) tickets. A team of clinical and patient safety experts used iterative tests of change and consensus building to converge on a single taxonomy. The final devised taxonomy was applied to pediatric medication-related events assess its characteristics, including interrater reliability and agreement.

RESULTS

Literature review identified four existing classifications for HIT-related patient safety events, and one was iteratively adapted to converge on a singular taxonomy. Safety events relating to usability accounted for a greater proportion of ERS reports, compared with HD tickets (37 vs. 20%, p = 0.022). Conversely, events pertaining to incorrect configuration accounted for a greater proportion of HD tickets, compared with ERS reports (63 vs. 8%, p < 0.01). Interrater agreement (%) and reliability (kappa) were 87.8% and 0.688 for ERS reports and 73.6% and 0.556 for HD tickets, respectively.

DISCUSSION

A standardized taxonomy for medication-related patient safety events related to HIT is presented. The taxonomy was validated using pediatric events. Further evaluation can assess whether the taxonomy is suitable for nonmedication-related events and those occurring in other patient populations.

CONCLUSION

Wider application of standardized taxonomies will allow for peer benchmarking and facilitate collaborative interinstitutional patient safety improvement efforts.

Multiple Regression Analysis and Frequent Itemset Mining of Electronic Medical Records: A Visual Analytics Approach Using VISA_M3R3

Medication-induced acute kidney injury (AKI) is a well-known problem in clinical medicine. This paper reports the first development of a visual analytics (VA) system that examines how different medications associate with AKI. In this paper, we introduce and describe VISA_M3R3, a VA system designed to assist healthcare researchers in identifying medications and medication combinations that associate with a higher risk of AKI using electronic medical records (EMRs). By integrating multiple regression models, frequent itemset mining, data visualization, and human-data interaction mechanisms, VISA_M3R3 allows users to explore complex relationships between medications and AKI in such a way that would be difficult or sometimes even impossible without the help of a VA system. Through an analysis of 595 medications using VISA_M3R3, we have identified 55 AKI-inducing medications, 24,212 frequent medication groups, and 78 medication groups that are associated with AKI. The purpose of this paper is to demonstrate the usefulness of VISA_M3R3 in the investigation of medication-induced AKI in particular and other clinical problems in general. Furthermore, this research highlights what needs to be considered in the future when designing VA systems that are intended to support gaining novel and deep insights into massive existing EMRs.