On the evaluation of synthetic longitudinal electronic health records.

BACKGROUND: Synthetic Electronic Health Records (EHRs) are becoming increasingly popular as a privacy enhancing technology. However, for longitudinal EHRs specifically, little research has been done into how to properly evaluate synthetically generated samples. In this article, we provide a discussion on existing methods and recommendations when evaluating the quality of synthetic longitudinal EHRs. METHODS: We recommend to assess synthetic EHR quality through similarity to real EHRs in low-dimensional projections, accuracy of a classifier discriminating synthetic from real samples, performance of synthetic versus real trained algorithms in clinical tasks, and privacy risk through risk of attribute inference. For each metric we discuss strengths and weaknesses, next to showing how it can be applied on a longitudinal dataset. RESULTS: To support the discussion on evaluation metrics, we apply discussed metrics on a dataset of synthetic EHRs generated from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) repository. CONCLUSIONS: The discussion on evaluation metrics provide guidance for researchers on how to use and interpret different metrics when evaluating the quality of synthetic longitudinal EHRs.

Federated learning for privacy-preserving depression detection with multilingual language models in social media posts.

The incidences of mental health illnesses, such as suicidal ideation and depression, are increasing, which highlights the urgent need for early detection methods. There is a growing interest in using natural language processing (NLP) models to analyze textual data from patients, but accessing patients' data for research purposes can be challenging due to privacy concerns. Federated learning (FL) is a promising approach that can balance the need for centralized learning with data ownership sensitivity. In this study, we examine the effectiveness of FL models in detecting depression by using a simulated multilingual dataset. We analyzed social media posts in five different languages with varying sample sizes. Our findings indicate that FL achieves strong performance in most cases while maintaining clients' privacy for both independent and non-independent client partitioning.

Extracting patient lifestyle characteristics from Dutch clinical text with BERT models.

BACKGROUND: BERT models have seen widespread use on unstructured text within the clinical domain. However, little to no research has been conducted into classifying unstructured clinical notes on the basis of patient lifestyle indicators, especially in Dutch. This article aims to test the feasibility of deep BERT models on the task of patient lifestyle classification, as well as introducing an experimental framework that is easily reproducible in future research. METHODS: This study makes use of unstructured general patient text data from HagaZiekenhuis, a large hospital in The Netherlands. Over 148 000 notes were provided to us, which were each automatically labelled on the basis of the respective patients' smoking, alcohol usage and drug usage statuses. In this paper we test feasibility of automatically assigning labels, and justify it using hand-labelled input. Ultimately, we compare macro F1-scores of string matching, SGD and several BERT models on the task of classifying smoking, alcohol and drug usage. We test Dutch BERT models and English models with translated input. RESULTS: We find that our further pre-trained MedRoBERTa.nl-HAGA model outperformed every other model on smoking (0.93) and drug usage (0.77). Interestingly, our ClinicalBERT model that was merely fine-tuned on translated text performed best on the alcohol task (0.80). In t-SNE visualisations, we show our MedRoBERTa.nl-HAGA model is the best model to differentiate between classes in the embedding space, explaining its superior classification performance. CONCLUSIONS: We suggest MedRoBERTa.nl-HAGA to be used as a baseline in future research on Dutch free text patient lifestyle classification. We furthermore strongly suggest further exploring the application of translation to input text in non-English clinical BERT research, as we only translated a subset of the full set and yet achieved very promising results.

A mixed methods analysis of the medication review intervention centered around the use of the 'Systematic Tool to Reduce Inappropriate Prescribing' Assistant (STRIPA) in Swiss primary care practices.

BACKGROUND: Electronic clinical decision support systems (eCDSS), such as the 'Systematic Tool to Reduce Inappropriate Prescribing' Assistant (STRIPA), have become promising tools for assisting general practitioners (GPs) with conducting medication reviews in older adults. Little is known about how GPs perceive eCDSS-assisted recommendations for pharmacotherapy optimization. The aim of this study was to explore the implementation of a medication review intervention centered around STRIPA in the 'Optimising PharmacoTherapy In the multimorbid elderly in primary CAre' (OPTICA) trial. METHODS: We used an explanatory mixed methods design combining quantitative and qualitative data. First, quantitative data about the acceptance and implementation of eCDSS-generated recommendations from GPs (n = 21) and their patients (n = 160) in the OPTICA intervention group were collected. Then, semi-structured qualitative interviews were conducted with GPs from the OPTICA intervention group (n = 8), and interview data were analyzed through thematic analysis. RESULTS: In quantitative findings, GPs reported averages of 13 min spent per patient preparing the eCDSS, 10 min performing medication reviews, and 5 min discussing prescribing recommendations with patients. On average, out of the mean generated 3.7 recommendations (SD=1.8). One recommendation to stop or start a medication was reported to be implemented per patient in the intervention group (SD=1.2). Overall, GPs found the STRIPA useful and acceptable. They particularly appreciated its ability to generate recommendations based on large amounts of patient information. During qualitative interviews, GPs reported the main reasons for limited implementation of STRIPA were related to problems with data sourcing (e.g., incomplete data imports), preparation of the eCDSS (e.g., time expenditure for updating and adapting information), its functionality (e.g., technical problems downloading PDF recommendation reports), and appropriateness of recommendations. CONCLUSIONS: Qualitative findings help explain the relatively low implementation of recommendations demonstrated by quantitative findings, but also show GPs' overall acceptance of STRIPA. Our results provide crucial insights for adapting STRIPA to make it more suitable for regular use in future primary care settings (e.g., necessity to improve data imports). TRIAL REGISTRATION: Clinicaltrials.gov NCT03724539, date of first registration: 29/10/2018.

Optimising prescribing in older adults with multimorbidity and polypharmacy in primary care (OPTICA): cluster randomised clinical trial.

OBJECTIVE: To study the effects of a primary care medication review intervention centred around an electronic clinical decision support system (eCDSS) on appropriateness of medication and the number of prescribing omissions in older adults with multimorbidity and polypharmacy compared with a discussion about medication in line with usual care. DESIGN: Cluster randomised clinical trial. SETTING: Swiss primary care, between December 2018 and February 2021. PARTICIPANTS: Eligible patients were ≥65 years of age with three or more chronic conditions and five or more long term medications. INTERVENTION: The intervention to optimise pharmacotherapy centred around an eCDSS was conducted by general practitioners, followed by shared decision making between general practitioners and patients, and was compared with a discussion about medication in line with usual care between patients and general practitioners. MAIN OUTCOME MEASURES: Primary outcomes were improvement in the Medication Appropriateness Index (MAI) and the Assessment of Underutilisation (AOU) at 12 months. Secondary outcomes included number of medications, falls, fractures, and quality of life. RESULTS: In 43 general practitioner clusters, 323 patients were recruited (median age 77 (interquartile range 73-83) years; 45% (n=146) women). Twenty one general practitioners with 160 patients were assigned to the intervention group and 22 general practitioners with 163 patients to the control group. On average, one recommendation to stop or start a medication was reported to be implemented per patient. At 12 months, the results of the intention-to-treat analysis of the improvement in appropriateness of medication (odds ratio 1.05, 95% confidence interval 0.59 to 1.87) and the number of prescribing omissions (0.90, 0.41 to 1.96) were inconclusive. The same was the case for the per protocol analysis. No clear evidence was found for a difference in safety outcomes at the 12 month follow-up, but fewer safety events were reported in the intervention group than in the control group at six and 12 months. CONCLUSIONS: In this randomised trial of general practitioners and older adults, the results were inconclusive as to whether the medication review intervention centred around the use of an eCDSS led to an improvement in appropriateness of medication or a reduction in prescribing omissions at 12 months compared with a discussion about medication in line with usual care. Nevertheless, the intervention could be safely delivered without causing any harm to patients. TRIAL REGISTRATION: NCT03724539Clinicaltrials.gov NCT03724539.

Topic Modeling for Interpretable Text Classification From EHRs.

The clinical notes in electronic health records have many possibilities for predictive tasks in text classification. The interpretability of these classification models for the clinical domain is critical for decision making. Using topic models for text classification of electronic health records for a predictive task allows for the use of topics as features, thus making the text classification more interpretable. However, selecting the most effective topic model is not trivial. In this work, we propose considerations for selecting a suitable topic model based on the predictive performance and interpretability measure for text classification. We compare 17 different topic models in terms of both interpretability and predictive performance in an inpatient violence prediction task using clinical notes. We find no correlation between interpretability and predictive performance. In addition, our results show that although no model outperforms the other models on both variables, our proposed fuzzy topic modeling algorithm (FLSA-W) performs best in most settings for interpretability, whereas two state-of-the-art methods (ProdLDA and LSI) achieve the best predictive performance.

Development of a Pipeline for Adverse Drug Reaction Identification in Clinical Notes: Word Embedding Models and String Matching.

BACKGROUND: Knowledge about adverse drug reactions (ADRs) in the population is limited because of underreporting, which hampers surveillance and assessment of drug safety. Therefore, gathering accurate information that can be retrieved from clinical notes about the incidence of ADRs is of great relevance. However, manual labeling of these notes is time-consuming, and automatization can improve the use of free-text clinical notes for the identification of ADRs. Furthermore, tools for language processing in languages other than English are not widely available. OBJECTIVE: The aim of this study is to design and evaluate a method for automatic extraction of medication and Adverse Drug Reaction Identification in Clinical Notes (ADRIN). METHODS: Dutch free-text clinical notes (N=277,398) and medication registrations (N=499,435) from the Cardiology Centers of the Netherlands database were used. All clinical notes were used to develop word embedding models. Vector representations of word embedding models and string matching with a medical dictionary (Medical Dictionary for Regulatory Activities [MedDRA]) were used for identification of ADRs and medication in a test set of clinical notes that were manually labeled. Several settings, including search area and punctuation, could be adjusted in the prototype to evaluate the optimal version of the prototype. RESULTS: The ADRIN method was evaluated using a test set of 988 clinical notes written on the stop date of a drug. Multiple versions of the prototype were evaluated for a variety of tasks. Binary classification of ADR presence achieved the highest accuracy of 0.84. Reduced search area and inclusion of punctuation improved performance, whereas incorporation of the MedDRA did not improve the performance of the pipeline. CONCLUSIONS: The ADRIN method and prototype are effective in recognizing ADRs in Dutch clinical notes from cardiac diagnostic screening centers. Surprisingly, incorporation of the MedDRA did not result in improved identification on top of word embedding models. The implementation of the ADRIN tool may help increase the identification of ADRs, resulting in better care and saving substantial health care costs.

Frequency and Acceptance of Clinical Decision Support System-Generated STOPP/START Signals for Hospitalised Older Patients with Polypharmacy and Multimorbidity.

BACKGROUND: The Screening Tool of Older Persons' Prescriptions (STOPP)/Screening Tool to Alert to Right Treatment (START) instrument is used to evaluate the appropriateness of medication in older people. STOPP/START criteria have been converted into software algorithms and implemented in a clinical decision support system (CDSS) to facilitate their use in clinical practice. OBJECTIVE: Our objective was to determine the frequency of CDSS-generated STOPP/START signals and their subsequent acceptance by a pharmacotherapy team in a hospital setting. DESIGN AND METHODS: Hospitalised older patients with polypharmacy and multimorbidity allocated to the intervention arm of the OPERAM (OPtimising thERapy to prevent Avoidable hospital admissions in the Multimorbid elderly) trial underwent a CDSS-assisted structured medication review in four European hospitals. We evaluated the frequency of CDSS-generated STOPP/START signals and the subsequent acceptance of these signals by a trained pharmacotherapy team consisting of a physician and pharmacist after evaluation of clinical applicability to the individual patient, prior to discussing pharmacotherapy optimisation recommendations with the patient and attending physicians. Multivariate linear regression analysis was used to investigate potential patient-related (e.g. age, number of co-morbidities and medications) and setting-related (e.g. ward type, country of inclusion) determinants for acceptance of STOPP and START signals. RESULTS: In 819/826 (99%) of the patients, at least one STOPP/START signal was generated using a set of 110 algorithms based on STOPP/START v2 criteria. Overall, 39% of the 5080 signals were accepted by the pharmacotherapy team. There was a high variability in the frequency and the subsequent acceptance of the individual STOPP/START criteria. The acceptance ranged from 2.5 to 75.8% for the top ten most frequently generated STOPP and START signals. The signal to stop a drug without a clinical indication was most frequently generated (28%), with more than half of the signals accepted (54%). No difference in mean acceptance of STOPP versus START signals was found. In multivariate analysis, most patient-related determinants did not predict acceptance, although the acceptance of START signals increased in patients with one or more hospital admissions (+ 7.9; 95% confidence interval [CI] 1.6-14.1) or one or more falls in the previous year (+ 7.1; 95% CI 0.7-13.4). A higher number of co-morbidities was associated with lower acceptance of STOPP (- 11.8%; 95% CI - 19.2 to - 4.5) and START (- 11.0%; 95% CI - 19.4 to - 2.6) signals for patients with more than nine and between seven and nine co-morbidities, respectively. For setting-related determinants, the acceptance differed significantly between the participating trial sites. Compared with Switzerland, the acceptance was higher in Ireland (STOPP: + 26.8%; 95% CI 16.8-36.7; START: + 31.1%; 95% CI 18.2-44.0) and in the Netherlands (STOPP: + 14.7%; 95% CI 7.8-21.7). Admission to a surgical ward was positively associated with acceptance of STOPP signals (+ 10.3%; 95% CI 3.8-16.8). CONCLUSION: The involvement of an expert team in translating population-based CDSS signals to individual patients is essential, as more than half of the signals for potential overuse, underuse, and misuse were not deemed clinically appropriate in a hospital setting. Patient-related potential determinants were poor predictors of acceptance. Future research investigating factors that affect patients' and physicians' agreement with medication changes recommended by expert teams may provide further insight for implementation in clinical practice. REGISTRATION: ClinicalTrials.gov Identifier: NCT02986425.

Optimizing Therapy to Prevent Avoidable Hospital Admissions in Multimorbid Older Adults (OPERAM): cluster randomised controlled trial.

OBJECTIVE: To examine the effect of optimising drug treatment on drug related hospital admissions in older adults with multimorbidity and polypharmacy admitted to hospital. DESIGN: Cluster randomised controlled trial. SETTING: 110 clusters of inpatient wards within university based hospitals in four European countries (Switzerland, Netherlands, Belgium, and Republic of Ireland) defined by attending hospital doctors. PARTICIPANTS: 2008 older adults (≥70 years) with multimorbidity (≥3 chronic conditions) and polypharmacy (≥5 drugs used long term). INTERVENTION: Clinical staff clusters were randomised to usual care or a structured pharmacotherapy optimisation intervention performed at the individual level jointly by a doctor and a pharmacist, with the support of a clinical decision software system deploying the screening tool of older person's prescriptions and screening tool to alert to the right treatment (STOPP/START) criteria to identify potentially inappropriate prescribing. MAIN OUTCOME MEASURE: Primary outcome was first drug related hospital admission within 12 months. RESULTS: 2008 older adults (median nine drugs) were randomised and enrolled in 54 intervention clusters (963 participants) and 56 control clusters (1045 participants) receiving usual care. In the intervention arm, 86.1% of participants (n=789) had inappropriate prescribing, with a mean of 2.75 (SD 2.24) STOPP/START recommendations for each participant. 62.2% (n=491) had ≥1 recommendation successfully implemented at two months, predominantly discontinuation of potentially inappropriate drugs. In the intervention group, 211 participants (21.9%) experienced a first drug related hospital admission compared with 234 (22.4%) in the control group. In the intention-to-treat analysis censored for death as competing event (n=375, 18.7%), the hazard ratio for first drug related hospital admission was 0.95 (95% confidence interval 0.77 to 1.17). In the per protocol analysis, the hazard ratio for a drug related hospital admission was 0.91 (0.69 to 1.19). The hazard ratio for first fall was 0.96 (0.79 to 1.15; 237 v 263 first falls) and for death was 0.90 (0.71 to 1.13; 172 v 203 deaths). CONCLUSIONS: Inappropriate prescribing was common in older adults with multimorbidity and polypharmacy admitted to hospital and was reduced through an intervention to optimise pharmacotherapy, but without effect on drug related hospital admissions. Additional efforts are needed to identify pharmacotherapy optimisation interventions that reduce inappropriate prescribing and improve patient outcomes. TRIAL REGISTRATION: ClinicalTrials.gov NCT02986425.

Baseline characteristics and comparability of older multimorbid patients with polypharmacy and general practitioners participating in a randomized controlled primary care trial.

OBJECTIVES: Recruiting general practitioners (GPs) and their multimorbid older patients for trials is challenging for multiple reasons (e.g., high workload, limited mobility). The comparability of study participants is important for interpreting study findings. This manuscript describes the baseline characteristics of GPs and patients participating in the 'Optimizing PharmacoTherapy in older multimorbid adults In primary CAre' (OPTICA) trial, a study of optimization of pharmacotherapy for multimorbid older adults. The overall aim of this study was to determine if the GPs and patients participating in the OPTICA trial are comparable to the real-world population in Swiss primary care. DESIGN: Analysis of baseline data from GPs and patients in the OPTICA trial and a reference cohort from the FIRE ('Family medicine ICPC Research using Electronic medical records') project. SETTING: Primary care, Switzerland. PARTICIPANTS: Three hundred twenty-three multimorbid (≥ 3 chronic conditions) patients with polypharmacy (≥ 5 regular medications) aged ≥ 65 years and 43 GPs recruited for the OPTICA trial were compared to 22,907 older multimorbid patients with polypharmacy and 227 GPs from the FIRE database. METHODS: We compared the characteristics of GPs and patients participating in the OPTICA trial with other GPs and other older multimorbid adults with polypharmacy in the FIRE database. We described the baseline willingness to have medications deprescribed of the patients participating in the OPTICA trial using the revised Patients' Attitudes Towards Deprescribing (rPATD) questionnaire. RESULTS: The GPs in the FIRE project and OPTICA were similar in terms of sociodemographic characteristics and their work as a GP (e.g. aged in their fifties, ≥ 10 years of experience, ≥ 60% are self-employed, ≥ 80% work in a group practice). The median age of patients in the OPTICA trial was 77 years and 45% of trial participants were women. Patients participating in the OPTICA trial and patients in the FIRE database were comparable in terms of age, certain clinical characteristics (e.g. systolic blood pressure, body mass index) and health services use (e.g. selected lab and vital data measurements). More than 80% of older multimorbid patients reported to be willing to stop ≥ 1 of their medications if their doctor said that this would be possible. CONCLUSION: The characteristics of patients and GPs recruited into the OPTICA trial are relatively comparable to characteristics of a real-world Swiss population, which indicates that recruiting a generalizable patient sample is possible in the primary care setting. Multimorbid patients in the OPTICA trial reported a high willingness to have medications deprescribed. TRIAL REGISTRATION: Clinicaltrials.gov ( NCT03724539 ), KOFAM (Swiss national portal) ( SNCTP000003060 ), Universal Trial Number (U1111-1226-8013).

SYMBALS: A Systematic Review Methodology Blending Active Learning and Snowballing.

Research output has grown significantly in recent years, often making it difficult to see the forest for the trees. Systematic reviews are the natural scientific tool to provide clarity in these situations. However, they are protracted processes that require expertise to execute. These are problematic characteristics in a constantly changing environment. To solve these challenges, we introduce an innovative systematic review methodology: SYMBALS. SYMBALS blends the traditional method of backward snowballing with the machine learning method of active learning. We applied our methodology in a case study, demonstrating its ability to swiftly yield broad research coverage. We proved the validity of our method using a replication study, where SYMBALS was shown to accelerate title and abstract screening by a factor of 6. Additionally, four benchmarking experiments demonstrated the ability of our methodology to outperform the state-of-the-art systematic review methodology FAST2.

A data mining-based cross-industry process for predicting major bleeding in mechanical circulatory support.

Aims: Over a third of patients, treated with mechanical circulatory support (MCS) for end-stage heart failure, experience major bleeding. Currently, the prediction of a major bleeding in the near future is difficult because of many contributing factors. Predictive analytics using data mining could help calculating the risk of bleeding; however, its application is generally reserved for experienced researchers on this subject. We propose an easily applicable data mining tool to predict major bleeding in MCS patients. Methods and results: All data of electronic health records of MCS patients in the University Medical Centre Utrecht were included. Based on the cross-industry standard process for data mining (CRISP-DM) methodology, an application named Auto-Crisp was developed. Auto-Crisp was used to evaluate the predictive models for a major bleeding in the next 3, 7, and 30 days after the first 30 days post-operatively following MCS implantation. The performance of the predictive models is investigated by the area under the curve (AUC) evaluation measure. In 25.6% of 273 patients, a total of 142 major bleedings occurred during a median follow-up period of 542 [interquartile range (IQR) 205-1044] days. The best predictive models assessed by Auto-Crisp had AUC values of 0.792, 0.788, and 0.776 for bleedings in the next 3, 7, and 30 days, respectively. Conclusion: The Auto-Crisp-based predictive model created in this study had an acceptable performance to predict major bleeding in MCS patients in the near future. However, further validation of the application is needed to evaluate Auto-Crisp in other research projects.

Combining Supervised and Unsupervised Machine Learning Methods for Phenotypic Functional Genomics Screening.

There has been an increase in the use of machine learning and artificial intelligence (AI) for the analysis of image-based cellular screens. The accuracy of these analyses, however, is greatly dependent on the quality of the training sets used for building the machine learning models. We propose that unsupervised exploratory methods should first be applied to the data set to gain a better insight into the quality of the data. This improves the selection and labeling of data for creating training sets before the application of machine learning. We demonstrate this using a high-content genome-wide small interfering RNA screen. We perform an unsupervised exploratory data analysis to facilitate the identification of four robust phenotypes, which we subsequently use as a training set for building a high-quality random forest machine learning model to differentiate four phenotypes with an accuracy of 91.1% and a kappa of 0.85. Our approach enhanced our ability to extract new knowledge from the screen when compared with the use of unsupervised methods alone.

Intervention protocol: OPtimising thERapy to prevent avoidable hospital Admission in the Multi-morbid elderly (OPERAM): a structured medication review with support of a computerised decision support system.

BACKGROUND: Several approaches to medication optimisation by identifying drug-related problems in older people have been described. Although some interventions have shown reductions in drug-related problems (DRPs), evidence supporting the effectiveness of medication reviews on clinical and economic outcomes is lacking. Application of the STOPP/START (version 2) explicit screening tool for inappropriate prescribing has decreased inappropriate prescribing and significantly reduced adverse drug reactions (ADRs) and associated healthcare costs in older patients with multi-morbidity and polypharmacy. Therefore, application of STOPP/START criteria during a medication review is likely to be beneficial. Incorporation of explicit screening tools into clinical decision support systems (CDSS) has gained traction as a means to improve both quality and efficiency in the rather time-consuming medication review process. Although CDSS can generate more potential inappropriate medication recommendations, some of these have been shown to be less clinically relevant, resulting in alert fatigue. Moreover, explicit tools such as STOPP/START do not cover all relevant DRPs on an individual patient level. The OPERAM study aims to assess the impact of a structured drug review on the quality of pharmacotherapy in older people with multi-morbidity and polypharmacy. The aim of this paper is to describe the structured, multi-component intervention of the OPERAM trial and compare it with the approach in the comparator arm. METHOD: This paper describes a multi-component intervention, integrating interventions that have demonstrated effectiveness in defining DRPs. The intervention involves a structured history-taking of medication (SHiM), a medication review according to the systemic tool to reduce inappropriate prescribing (STRIP) method, assisted by a clinical decision support system (STRIP Assistant, STRIPA) with integrated STOPP/START criteria (version 2), followed by shared decision-making with both patient and attending physician. The developed method integrates patient input, patient data, involvement from other healthcare professionals and CDSS-assistance into one structured intervention. DISCUSSION: The clinical and economical effectiveness of this experimental intervention will be evaluated in a cohort of hospitalised, older patients with multi-morbidity and polypharmacy in the multicentre, randomized controlled OPERAM trial (OPtimising thERapy to prevent Avoidable hospital admissions in the Multi-morbid elderly), which will be completed in the last quarter of 2019. TRIAL REGISTRATION: Universal Trial Number: U1111-1181-9400 Clinicaltrials.gov: NCT02986425, Registered 08 December 2016. FOPH (Swiss national portal): SNCTP000002183. Netherlands Trial Register: NTR6012 (07-10-2016).

Evaluating sentence representations for biomedical text: Methods and experimental results.

Text representations ar one of the main inputs to various Natural Language Processing (NLP) methods. Given the fast developmental pace of new sentence embedding methods, we argue that there is a need for a unified methodology to assess these different techniques in the biomedical domain. This work introduces a comprehensive evaluation of novel methods across ten medical classification tasks. The tasks cover a variety of BioNLP problems such as semantic similarity, question answering, citation sentiment analysis and others with binary and multi-class datasets. Our goal is to assess the transferability of different sentence representation schemes to the medical and clinical domain. Our analysis shows that embeddings based on Language Models which account for the context-dependent nature of words, usually outperform others in terms of performance. Nonetheless, there is no single embedding model that perfectly represents biomedical and clinical texts with consistent performance across all tasks. This illustrates the need for a more suitable bio-encoder. Our MedSentEval source code, pre-trained embeddings and examples have been made available on GitHub.

A Lightweight API-Based Approach for Building Flexible Clinical NLP Systems.

Natural language processing (NLP) has become essential for secondary use of clinical data. Over the last two decades, many clinical NLP systems were developed in both academia and industry. However, nearly all existing systems are restricted to specific clinical settings mainly because they were developed for and tested with specific datasets, and they often fail to scale up. Therefore, using existing NLP systems for one's own clinical purposes requires substantial resources and long-term time commitments for customization and testing. Moreover, the maintenance is also troublesome and time-consuming. This research presents a lightweight approach for building clinical NLP systems with limited resources. Following the design science research approach, we propose a lightweight architecture which is designed to be composable, extensible, and configurable. It takes NLP as an external component which can be accessed independently and orchestrated in a pipeline via web APIs. To validate its feasibility, we developed a web-based prototype for clinical concept extraction with six well-known NLP APIs and evaluated it on three clinical datasets. In comparison with available benchmarks for the datasets, three high F1 scores (0.861, 0.724, and 0.805) were obtained from the evaluation. It also gained a low F1 score (0.373) on one of the tests, which probably is due to the small size of the test dataset. The development and evaluation of the prototype demonstrates that our approach has a great potential for building effective clinical NLP systems with limited resources.

'Optimising PharmacoTherapy In the multimorbid elderly in primary CAre' (OPTICA) to improve medication appropriateness: study protocol of a cluster randomised controlled trial.

INTRODUCTION: Multimorbidity and polypharmacy are major risk factors for potentially inappropriate prescribing (eg, overprescribing and underprescribing), and systematic medication reviews are complex and time consuming. In this trial, the investigators aim to determine if a systematic software-based medication review improves medication appropriateness more than standard care in older, multimorbid patients with polypharmacy. METHODS AND ANALYSIS: Optimising PharmacoTherapy In the multimorbid elderly in primary CAre is a cluster randomised controlled trial that will include outpatients from the Swiss primary care setting, aged ≥65 years with ≥three chronic medical conditions and concurrent use of ≥five chronic medications. Patients treated by the same general practitioner (GP) constitute a cluster, and clusters are randomised 1:1 to either a standard care sham intervention, in which the GP discusses with the patient if the medication list is complete, or a systematic medication review intervention based on the use of the 'Systematic Tool to Reduce Inappropriate Prescribing'-Assistant (STRIPA). STRIPA is a web-based clinical decision support system that helps customise medication reviews. It is based on the validated 'Screening Tool of Older Person's Prescriptions' (STOPP) and 'Screening Tool to Alert doctors to Right Treatment' (START) criteria to detect potentially inappropriate prescribing. The trial's follow-up period is 12 months. Outcomes will be assessed at baseline, 6 and 12 months. The primary endpoint is medication appropriateness, as measured jointly by the change in the Medication Appropriateness Index (MAI) and Assessment of Underutilisation (AOU). Secondary endpoints include the degree of polypharmacy, overprescribing and underprescribing, the number of falls and fractures, quality of life, the amount of formal and informal care received by patients, survival, patients' quality adjusted life years, patients' medical costs, cost-effectiveness of the intervention, percentage of recommendations accepted by GPs, percentage of recommendation rejected by GPs and patients' willingness to have medications deprescribed. ETHICS AND DISSEMINATION: The ethics committee of the canton of Bern in Switzerland approved the trial protocol. The results of this trial will be published in a peer-reviewed journal. MAIN FUNDING: Swiss National Science Foundation, National Research Programme (NRP 74) 'Smarter Healthcare'. TRIAL REGISTRATION NUMBERS: Clinicaltrials.gov (NCT03724539), KOFAM (Swiss national portal) (SNCTP000003060), Universal Trial Number (U1111-1226-8013).