GEOexplorer: a webserver for gene expression analysis and visualisation.

Gene Expression Omnibus (GEO) is a database repository hosting a substantial proportion of publicly available high throughput gene expression data. Gene expression analysis is a powerful tool to gain insight into the mechanisms and processes underlying the biological and phenotypic differences between sample groups. Despite the wide availability of gene expression datasets, their access, analysis, and integration are not trivial and require specific expertise and programming proficiency. We developed the GEOexplorer webserver to allow scientists to access, integrate and analyse gene expression datasets without requiring programming proficiency. Via its user-friendly graphic interface, users can easily apply GEOexplorer to perform interactive and reproducible gene expression analysis of microarray and RNA-seq datasets, while producing a wealth of interactive visualisations to facilitate data exploration and interpretation, and generating a range of publication ready figures. The webserver allows users to search and retrieve datasets from GEO as well as to upload user-generated data and combine and harmonise two datasets to perform joint analyses. GEOexplorer, available at https://geoexplorer.rosalind.kcl.ac.uk, provides a solution for performing interactive and reproducible analyses of microarray and RNA-seq gene expression data, empowering life scientists to perform exploratory data analysis and differential gene expression analysis on-the-fly without informatics proficiency.

Discharge summary hospital course summarisation of in patient Electronic Health Record text with clinical concept guided deep pre-trained Transformer models.

Brief Hospital Course (BHC) summaries are succinct summaries of an entire hospital encounter, embedded within discharge summaries, written by senior clinicians responsible for the overall care of a patient. Methods to automatically produce summaries from inpatient documentation would be invaluable in reducing clinician manual burden of summarising documents under high time-pressure to admit and discharge patients. Automatically producing these summaries from the inpatient course, is a complex, multi-document summarisation task, as source notes are written from various perspectives (e.g. nursing, doctor, radiology), during the course of the hospitalisation. We demonstrate a range of methods for BHC summarisation demonstrating the performance of deep learning summarisation models across extractive and abstractive summarisation scenarios. We also test a novel ensemble extractive and abstractive summarisation model that incorporates a medical concept ontology (SNOMED) as a clinical guidance signal and shows superior performance in 2 real-world clinical data sets.

Assessing the feasibility of a web-based outcome measurement system in child and adolescent mental health services - myHealthE a randomised controlled feasibility pilot study.

BACKGROUND: Interest in internet-based patient reported outcome measure (PROM) collection is increasing. The NHS myHealthE (MHE) web-based monitoring system was developed to address the limitations of paper-based PROM completion. MHE provides a simple and secure way for families accessing Child and Adolescent Mental Health Services to report clinical information and track their child's progress. This study aimed to assess whether MHE improves the completion of the Strengths and Difficulties Questionnaire (SDQ) compared with paper collection. Secondary objectives were to explore caregiver satisfaction and application acceptability. METHODS: A 12-week single-blinded randomised controlled feasibility pilot trial of MHE was conducted with 196 families accessing neurodevelopmental services in south London to examine whether electronic questionnaires are completed more readily than paper-based questionnaires over a 3-month period. Follow up process evaluation phone calls with a subset (n = 8) of caregivers explored system satisfaction and usability. RESULTS: MHE group assignment was significantly associated with an increased probability of completing an SDQ-P in the study period (adjusted hazard ratio (HR) 12.1, 95% CI 4.7-31.0; p = <.001). Of those caregivers' who received the MHE invitation (n = 68) 69.1% completed an SDQ using the platform compared to 8.8% in the control group (n = 68). The system was well received by caregivers, who cited numerous benefits of using MHE, for example, real-time feedback and ease of completion. CONCLUSIONS: MHE holds promise for improving PROM completion rates. Research is needed to refine MHE, evaluate large-scale MHE implementation, cost effectiveness and explore factors associated with differences in electronic questionnaire uptake.

Viral reduction of human blood by ultraviolet A-photosensitized vitamin K5.

Blood product transfusion can transmit viral pathogens. Pathogen reduction methods for blood products have been developed but, so far, are not available for whole blood. We evaluated if vitamin K5 (VK5) and ultraviolet A (UVA) irradiation could be used for virus inactivation in plasma and whole blood. Undiluted human plasma and whole blood diluted to 20% were spiked with high levels of vaccinia or Zika viruses. Infectious titers were measured by standard TCID50 assay before and after VK5/UVA treatments. Up to 3.6 log of vaccinia and 3.2 log of Zika were reduced in plasma by the combination of 500 µM VK5 and 3 J/cm2 UVA, and 3.1 log of vaccinia and 2.9 log of Zika were reduced in diluted human blood (20%) by the combination of 500 µM VK5 and 70 J/cm2 UVA. At end of whole blood treatment, hemolysis increased from 0.18% to 0.41% but remained below 1% hemolysis, which is acceptable to the Food and Drug Administration for red cell transfusion products. No significant alteration of biochemical parameters of red blood cells occurred with treatment. Our results provide proof of the concept that a viral pathogen reduction method based on VK5/UVA may be developed for whole blood.

Synergistic bactericidal effects of pairs of photosensitizer molecules activated by ultraviolet A light against bacteria in plasma.

BACKGROUND: The current approach to reducing bacterial contamination in blood transfusion products is through detection or pathogen reduction methods, some of which utilize ultraviolet (UV) light photosensitizers. A small number of photosensitizers are being used as single agents in combination with UV light, but their efficacy can be limited against some pathogens. Benzophenone (BP) and vitamins B1, B6, and K3 have been identified as effective UVA photosensitizers for inactivation of bacteria. We evaluated whether combining pairs of photosensitizers in this group would have synergistic bactericidal effects on Gram-negative and Gram-positive bacteria. STUDY DESIGN AND METHODS: Bacteria species of Escherichia coli, Bacillus cereus, Staphylococcus aureus, and Klebsiella pneumoniae were mixed with 0 to 100 mM concentrations of photosensitizers and exposed to UVA irradiation at 18 J/cm2 to assess their bactericidal effects. RESULTS: Single photosensitizers irradiated with UVA produced a range of bactericidal activity. When combined in pairs, all demonstrated some synergistic bactericidal effects with up to 4-log reduction above the sum of activities of individual molecules in the pair against bacteria in plasma. Photosensitizer pairs with BP had the highest synergism across all bacteria. With vitamin K3 in the pair, synergism was evident for Gram-positive but not for Gram-negative bacteria. Vitamin B1 and vitamin B6 had the least synergism. These results indicate that a combination approach with multiple photosensitizers may extend effectiveness of pathogen reduction in plasma. CONCLUSIONS: Combining photosensitizers in pathogen reduction methods could improve bactericidal efficacy and lead to use of lower concentrations of photosensitizers to reduce toxicities and unwanted side effects.

Estimating redundancy in clinical text.

The current mode of use of Electronic Health Records (EHR) elicits text redundancy. Clinicians often populate new documents by duplicating existing notes, then updating accordingly. Data duplication can lead to propagation of errors, inconsistencies and misreporting of care. Therefore, measures to quantify information redundancy play an essential role in evaluating innovations that operate on clinical narratives. This work is a quantitative examination of information redundancy in EHR notes. We present and evaluate two methods to measure redundancy: an information-theoretic approach and a lexicosyntactic and semantic model. Our first measure trains large Transformer-based language models using clinical text from a large openly available US-based ICU dataset and a large multi-site UK based Hospital. By comparing the information-theoretic efficient encoding of clinical text against open-domain corpora, we find that clinical text is ∼1.5× to ∼3× less efficient than open-domain corpora at conveying information. Our second measure, evaluates automated summarisation metrics Rouge and BERTScore to evaluate successive note pairs demonstrating lexicosyntactic and semantic redundancy, with averages from ∼43 to ∼65%.

The relative vertex clustering value--a new criterion for the fast discovery of functional modules in protein interaction networks.

BACKGROUND: Cellular processes are known to be modular and are realized by groups of proteins implicated in common biological functions. Such groups of proteins are called functional modules, and many community detection methods have been devised for their discovery from protein interaction networks (PINs) data. In current agglomerative clustering approaches, vertices with just a very few neighbors are often classified as separate clusters, which does not make sense biologically. Also, a major limitation of agglomerative techniques is that their computational efficiency do not scale well to large PINs. Finally, PIN data obtained from large scale experiments generally contain many false positives, and this makes it hard for agglomerative clustering methods to find the correct clusters, since they are known to be sensitive to noisy data. RESULTS: We propose a local similarity premetric, the relative vertex clustering value, as a new criterion allowing to decide when a node can be added to a given node's cluster and which addresses the above three issues. Based on this criterion, we introduce a novel and very fast agglomerative clustering technique, FAC-PIN, for discovering functional modules and protein complexes from a PIN data. CONCLUSIONS: Our proposed FAC-PIN algorithm is applied to nine PIN data from eight different species including the yeast PIN, and the identified functional modules are validated using Gene Ontology (GO) annotations from DAVID Bioinformatics Resources. Identified protein complexes are also validated using experimentally verified complexes. Computational results show that FAC-PIN can discover functional modules or protein complexes from PINs more accurately and more efficiently than HC-PIN and CNM, the current state-of-the-art approaches for clustering PINs in an agglomerative manner.

Moving from development to implementation of digital innovations within the NHS: myHealthE, a remote monitoring system for tracking patient outcomes in child and adolescent mental health services.

Objective: This paper aims to report our experience of developing, implementing, and evaluating myHealthE (MHE), a digital innovation for Child and Adolescents Mental Health Services (CAMHS), which automates the remote collection and reporting of Patient-Reported Outcome Measures (PROMs) into National Health Services (NHS) electronic healthcare records. Methods: We describe the logistical and governance issues encountered in developing the MHE interface with patient-identifiable information, and the steps taken to overcome these development barriers. We describe the application's architecture and hosting environment to enable its operability within the NHS, as well as the capabilities needed within the technical team to bridge the gap between academic development and NHS operational teams. Results: We present evidence on the feasibility and acceptability of this system within clinical services and the process of iterative development, highlighting additional functions that were incorporated to increase system utility. Conclusion: This article provides a framework with which to plan, develop, and implement automated PROM collection from remote devices back to NHS infrastructure. The challenges and solutions described in this paper will be pertinent to other digital health innovation researchers aspiring to deploy interoperable systems within NHS clinical systems.

Evaluating physical urban features in several mental illnesses using electronic health record data.

Objectives: Understanding the potential impact of physical characteristics of the urban environment on clinical outcomes on several mental illnesses. Materials and Methods: Physical features of the urban environment were examined as predictors for affective and non-affective several mental illnesses (SMI), the number and length of psychiatric hospital admissions, and the number of short and long-acting injectable antipsychotic prescriptions. In addition, the urban features with the greatest weight in the predicted model were determined. The data included 28 urban features and 6 clinical variables obtained from 30,210 people with SMI receiving care from the South London and Maudsley NHS Foundation Trust (SLaM) using the Clinical Record Interactive Search (CRIS) tool. Five machine learning regression models were evaluated for the highest prediction accuracy followed by the Self-Organising Map (SOM) to represent the results visually. Results: The prevalence of SMI, number and duration of psychiatric hospital admission, and antipsychotic prescribing were greater in urban areas. However, machine learning analysis was unable to accurately predict clinical outcomes using urban environmental data. Discussion: The urban environment is associated with an increased prevalence of SMI. However, urban features alone cannot explain the variation observed in psychotic disorder prevalence or clinical outcomes measured through psychiatric hospitalisation or exposure to antipsychotic treatments. Conclusion: Urban areas are associated with a greater prevalence of SMI but clinical outcomes are likely to depend on a combination of urban and individual patient-level factors. Future mental healthcare service planning should focus on providing appropriate resources to people with SMI in urban environments.

Multi-modal Latent-Space Self-alignment for Super-Resolution Cardiac MR Segmentation.

2D cardiac MR cine images provide data with a high signal-to-noise ratio for the segmentation and reconstruction of the heart. These images are frequently used in clinical practice and research. However, the segments have low resolution in the through-plane direction, and standard interpolation methods are unable to improve resolution and precision. We proposed an end-to-end pipeline for producing high-resolution segments from 2D MR images. This pipeline utilised a bilateral optical flow warping method to recover images in the through-plane direction, while a SegResNet automatically generated segments of the left and right ventricles. A multi-modal latent-space self-alignment network was implemented to guarantee that the segments maintain an anatomical prior derived from unpaired 3D high-resolution CT scans. On 3D MR angiograms, the trained pipeline produced high-resolution segments that preserve an anatomical prior derived from patients with various cardiovascular diseases.

A Knowledge Distillation Ensemble Framework for Predicting Short- and Long-Term Hospitalization Outcomes From Electronic Health Records Data.

The ability to perform accurate prognosis is crucial for proactive clinical decision making, informed resource management and personalised care. Existing outcome prediction models suffer from a low recall of infrequent positive outcomes. We present a highly-scalable and robust machine learning framework to automatically predict adversity represented by mortality and ICU admission and readmission from time-series of vital signs and laboratory results obtained within the first 24 hours of hospital admission. The stacked ensemble platform comprises two components: a) an unsupervised LSTM Autoencoder that learns an optimal representation of the time-series, using it to differentiate the less frequent patterns which conclude with an adverse event from the majority patterns that do not, and b) a gradient boosting model, which relies on the constructed representation to refine prediction by incorporating static features. The model is used to assess a patient's risk of adversity and provides visual justifications of its prediction. Results of three case studies show that the model outperforms existing platforms in ICU and general ward settings, achieving average Precision-Recall Areas Under the Curve (PR-AUCs) of 0.891 (95% CI: 0.878-0.939) for mortality and 0.908 (95% CI: 0.870-0.935) in predicting ICU admission and readmission.

The feasibility of a strategy for the remote recruitment, consenting and assessment of recent referrals: a protocol for phase 1 of the On-Line Parent Training for the Initial Management of ADHD referrals (OPTIMA).

BACKGROUND: In the UK, children with high levels of hyperactivity, impulsivity and inattention referred to clinical services with possible attention-deficit/hyperactivity disorder (ADHD) often wait a long time for specialist diagnostic assessment. Parent training (PT) has the potential to support parents during this difficult period, especially regarding the management of challenging and disruptive behaviours that often accompany ADHD. However, traditional face-to-face PT is costly and difficult to organise in a timely way. We have created a low-cost, easily accessible PT programme delivered via a phone app, Structured E-Parenting Support (STEPS), to address this problem. The overall OPTIMA programme will evaluate the efficacy and cost-effectiveness of STEPS as a way of helping parents manage their children behaviour while on the waitlist. To ensure the timely and efficient evaluation of STEPS in OPTIMA, we have worked with children's health services to implement a remote strategy for recruitment, screening and assessment of recently referred families. Part of this strategy is incorporated into routine clinical practice and part is OPTIMA specific. Here, we present the protocol for Phase 1 of OPTIMA-a study of the feasibility of this remote strategy, as a basis for a large-scale STEPS randomised controlled trial (RCT). METHODS: This is a single arm observational feasibility study. Participants will be parents of up to 100 children aged 5-11 years with high levels of hyperactivity/impulsivity, inattention and challenging behaviour who are waiting for assessment in one of five UK child and adolescent mental health or behavioural services. Recruitment, consenting and data collection will occur remotely. The primary outcome will be the rate at which the families, who meet inclusion criteria, agree in principle to take part in a full STEPS RCT. Secondary outcomes include acceptability of remote consenting and online data collection procedures; the feasibility of collecting teacher data remotely within the required timeframe, and technical difficulties with completing online questionnaires. All parents in the study will receive access to STEPS. DISCUSSION: Establishing the feasibility of our remote recruitment, consenting and assessment strategy is a pre-requisite for the full trial of OPTIMA. It can also provide a model for future trials conducted remotely.

Multi-domain clinical natural language processing with MedCAT: The Medical Concept Annotation Toolkit.

Electronic health records (EHR) contain large volumes of unstructured text, requiring the application of information extraction (IE) technologies to enable clinical analysis. We present the open source Medical Concept Annotation Toolkit (MedCAT) that provides: (a) a novel self-supervised machine learning algorithm for extracting concepts using any concept vocabulary including UMLS/SNOMED-CT; (b) a feature-rich annotation interface for customizing and training IE models; and (c) integrations to the broader CogStack ecosystem for vendor-agnostic health system deployment. We show improved performance in extracting UMLS concepts from open datasets (F1:0.448-0.738 vs 0.429-0.650). Further real-world validation demonstrates SNOMED-CT extraction at 3 large London hospitals with self-supervised training over ∼8.8B words from ∼17M clinical records and further fine-tuning with ∼6K clinician annotated examples. We show strong transferability (F1 > 0.94) between hospitals, datasets and concept types indicating cross-domain EHR-agnostic utility for accelerated clinical and research use cases.

Ensemble learning for poor prognosis predictions: A case study on SARS-CoV-2.

OBJECTIVE: Risk prediction models are widely used to inform evidence-based clinical decision making. However, few models developed from single cohorts can perform consistently well at population level where diverse prognoses exist (such as the SARS-CoV-2 [severe acute respiratory syndrome coronavirus 2] pandemic). This study aims at tackling this challenge by synergizing prediction models from the literature using ensemble learning. MATERIALS AND METHODS: In this study, we selected and reimplemented 7 prediction models for COVID-19 (coronavirus disease 2019) that were derived from diverse cohorts and used different implementation techniques. A novel ensemble learning framework was proposed to synergize them for realizing personalized predictions for individual patients. Four diverse international cohorts (2 from the United Kingdom and 2 from China; N = 5394) were used to validate all 8 models on discrimination, calibration, and clinical usefulness. RESULTS: Results showed that individual prediction models could perform well on some cohorts while poorly on others. Conversely, the ensemble model achieved the best performances consistently on all metrics quantifying discrimination, calibration, and clinical usefulness. Performance disparities were observed in cohorts from the 2 countries: all models achieved better performances on the China cohorts. DISCUSSION: When individual models were learned from complementary cohorts, the synergized model had the potential to achieve better performances than any individual model. Results indicate that blood parameters and physiological measurements might have better predictive powers when collected early, which remains to be confirmed by further studies. CONCLUSIONS: Combining a diverse set of individual prediction models, the ensemble method can synergize a robust and well-performing model by choosing the most competent ones for individual patients.

Inactivation of bacteria in plasma by photosensitizers benzophenone and vitamins K3, B1 and B6 with UV A light irradiation.

BACKGROUND: A photosensitizer is a light-activated molecule that can generate reactive oxygen species or directly interact with nucleic acids. Both consequences can be applied to reduction of pathogens in various media and to selectively attack tumor cells. Numerous natural and synthesized photosensitizers have been identified for pathogen reduction. METHODS: The photosensitizers of vitamins K3 (VK3), B1 (VB1), B6 (VB6) and benzophenone (BP) were prepared in 100-200 µM of PBS solution, irradiated with UVA at 0-48 J/cm2 for absorption spectrum alterations analysis. Bacteria species of E. coli, B. cereus, S. aureus and K. pneumoniae were mixed with 0-200 mM concentration of compounds and exposed to UVA irradiation of different dose at 6, 12 or 18 J/cm2 to assess the bactericidal effects. RESULTS AND CONCLUSIONS: Over six logs CFU/ml reduction of E. coli suspended in PBS occurred after treatment with either VB1, VB6, VK3 or BP combined with UVA irradiation. When bacteria were suspended in plasma, two to seven logs reduction occurred depending on the UVA dose, photosensitizer concentration, and bacteria species. Among these photosensitizers, BP had the most potent bactericidal effect and is a promising UVA photosensitizer for pathogen reduction. The level of absorption spectrum alteration after UVA irradiation was profound for VK3 and VB6 but minimal for BP and VB1. The UV-vis absorption spectrum changes did not correlate with the bactericidal effect indicating that molecule modification by UVA light is not required for the bactericidal activity.

On classifying sepsis heterogeneity in the ICU: insight using machine learning.

OBJECTIVES: Current machine learning models aiming to predict sepsis from electronic health records (EHR) do not account 20 for the heterogeneity of the condition despite its emerging importance in prognosis and treatment. This work demonstrates the added value of stratifying the types of organ dysfunction observed in patients who develop sepsis in the intensive care unit (ICU) in improving the ability to recognize patients at risk of sepsis from their EHR data. MATERIALS AND METHODS: Using an ICU dataset of 13 728 records, we identify clinically significant sepsis subpopulations with distinct organ dysfunction patterns. We perform classification experiments with random forest, gradient boost trees, and support vector machines, using the identified subpopulations to distinguish patients who develop sepsis in the ICU from those who do not. RESULTS: The classification results show that features selected using sepsis subpopulations as background knowledge yield a superior performance in distinguishing septic from non-septic patients regardless of the classification model used. The improved performance is especially pronounced in specificity, which is a current bottleneck in sepsis prediction machine learning models. CONCLUSION: Our findings can steer machine learning efforts toward more personalized models for complex conditions including sepsis.

The side effect profile of Clozapine in real world data of three large mental health hospitals.

OBJECTIVE: Mining the data contained within Electronic Health Records (EHRs) can potentially generate a greater understanding of medication effects in the real world, complementing what we know from Randomised control trials (RCTs). We Propose a text mining approach to detect adverse events and medication episodes from the clinical text to enhance our understanding of adverse effects related to Clozapine, the most effective antipsychotic drug for the management of treatment-resistant schizophrenia, but underutilised due to concerns over its side effects. MATERIAL AND METHODS: We used data from de-identified EHRs of three mental health trusts in the UK (>50 million documents, over 500,000 patients, 2835 of which were prescribed Clozapine). We explored the prevalence of 33 adverse effects by age, gender, ethnicity, smoking status and admission type three months before and after the patients started Clozapine treatment. Where possible, we compared the prevalence of adverse effects with those reported in the Side Effects Resource (SIDER). RESULTS: Sedation, fatigue, agitation, dizziness, hypersalivation, weight gain, tachycardia, headache, constipation and confusion were amongst the highest recorded Clozapine adverse effect in the three months following the start of treatment. Higher percentages of all adverse effects were found in the first month of Clozapine therapy. Using a significance level of (p< 0.05) our chi-square tests show a significant association between most of the ADRs and smoking status and hospital admission, and some in gender, ethnicity and age groups in all trusts hospitals. Later we combined the data from the three trusts hospitals to estimate the average effect of ADRs in each monthly interval. In gender and ethnicity, the results show significant association in 7 out of 33 ADRs, smoking status shows significant association in 21 out of 33 ADRs and hospital admission shows the significant association in 30 out of 33 ADRs. CONCLUSION: A better understanding of how drugs work in the real world can complement clinical trials.

An epigenome-wide association study of Alzheimer's disease blood highlights robust DNA hypermethylation in the HOXB6 gene.

A growing number of epigenome-wide association studies have demonstrated a role for DNA methylation in the brain in Alzheimer's disease. With the aim of exploring peripheral biomarker potential, we have examined DNA methylation patterns in whole blood collected from 284 individuals in the AddNeuroMed study, which included 89 nondemented controls, 86 patients with Alzheimer's disease, and 109 individuals with mild cognitive impairment, including 38 individuals who progressed to Alzheimer's disease within 1 year. We identified significant differentially methylated regions, including 12 adjacent hypermethylated probes in the HOXB6 gene in Alzheimer's disease, which we validated using pyrosequencing. Using weighted gene correlation network analysis, we identified comethylated modules of genes that were associated with key variables such as APOE genotype and diagnosis. In summary, this study represents the first large-scale epigenome-wide association study of Alzheimer's disease and mild cognitive impairment using blood. We highlight the differences in various loci and pathways in early disease, suggesting that these patterns relate to cognitive decline at an early stage.

How data science can advance mental health research.

Accessibility of powerful computers and availability of so-called big data from a variety of sources means that data science approaches are becoming pervasive. However, their application in mental health research is often considered to be at an earlier stage than in other areas despite the complexity of mental health and illness making such a sophisticated approach particularly suitable. In this Perspective, we discuss current and potential applications of data science in mental health research using the UK Clinical Research Collaboration classification: underpinning research; aetiology; detection and diagnosis; treatment development; treatment evaluation; disease management; and health services research. We demonstrate that data science is already being widely applied in mental health research, but there is much more to be done now and in the future. The possibilities for data science in mental health research are substantial.

Efficient Reuse of Natural Language Processing Models for Phenotype-Mention Identification in Free-text Electronic Medical Records: A Phenotype Embedding Approach.

BACKGROUND: Much effort has been put into the use of automated approaches, such as natural language processing (NLP), to mine or extract data from free-text medical records in order to construct comprehensive patient profiles for delivering better health care. Reusing NLP models in new settings, however, remains cumbersome, as it requires validation and retraining on new data iteratively to achieve convergent results. OBJECTIVE: The aim of this work is to minimize the effort involved in reusing NLP models on free-text medical records. METHODS: We formally define and analyze the model adaptation problem in phenotype-mention identification tasks. We identify "duplicate waste" and "imbalance waste," which collectively impede efficient model reuse. We propose a phenotype embedding-based approach to minimize these sources of waste without the need for labelled data from new settings. RESULTS: We conduct experiments on data from a large mental health registry to reuse NLP models in four phenotype-mention identification tasks. The proposed approach can choose the best model for a new task, identifying up to 76% waste (duplicate waste), that is, phenotype mentions without the need for validation and model retraining and with very good performance (93%-97% accuracy). It can also provide guidance for validating and retraining the selected model for novel language patterns in new tasks, saving around 80% waste (imbalance waste), that is, the effort required in "blind" model-adaptation approaches. CONCLUSIONS: Adapting pretrained NLP models for new tasks can be more efficient and effective if the language pattern landscapes of old settings and new settings can be made explicit and comparable. Our experiments show that the phenotype-mention embedding approach is an effective way to model language patterns for phenotype-mention identification tasks and that its use can guide efficient NLP model reuse.