A concentric circles view of health data relations facilitates understanding of sociotechnical challenges for learning health systems and the role of federated data networks.

The ability to use clinical and research data at scale is central to hopes for data-driven medicine. However, in using such data researchers often encounter hurdles-both technical, such as differing data security requirements, and social, such as the terms of informed consent, legal requirements and patient and public trust. Federated or distributed data networks have been proposed and adopted in response to these hurdles. However, to date there has been little consideration of how FDNs respond to both technical and social constraints on data use. In this Perspective we propose an approach to thinking about data in terms that make it easier to navigate the health data space and understand the value of differing approaches to data collection, storage and sharing. We set out a socio-technical model of data systems that we call the "Concentric Circles View" (CCV) of data-relationships. The aim is to enable a consistent understanding of the fit between the local relationships within which data are produced and the extended socio-technical systems that enable their use. The paper suggests this model can help understand and tackle challenges associated with the use of real-world data in the health setting. We use the model to understand not only how but why federated networks may be well placed to address emerging issues and adapt to the evolving needs of health research for patient benefit. We conclude that the CCV provides a useful model with broader application in mapping, understanding, and tackling the major challenges associated with using real world data in the health setting.

Robust Classification of Histology Images Exploiting Adversarial Auto Encoders.

Deep learning (DL) thrives on the availability of a large number of high quality images with reliable labels. Due to the large size of whole slide images in digital pathology, patches of manageable size are often mined for use in DL models. These patches are variable in quality, weakly supervised, individually less informative, and noisily labelled. To improve classification accuracy even with these noisy inputs and labels in histopathology, we propose a novel method for robust feature generation using an adversarial autoencoder (AAE). We utilize the likelihood of the features in the latent space of AAE as a criterion to weigh the training samples. We propose different weighting schemes for our framework and evaluate the effectiveness of our methods on the publically available BreakHis and BACH histopathology datasets. We observe consistent improvement in AUC scores using our methods, and conclude that robust supervision strategies should be further explored for computational pathology.

Federated electronic health records research technology to support clinical trial protocol optimization: Evidence from EHR4CR and the InSite platform.

OBJECTIVE: To determine if inclusion/exclusion (I/E) criteria of clinical trial protocols can be represented as structured queries and executed using a secure federated research platform (InSite) on hospital electronic health records (EHR) systems, to estimate the number of potentially eligible patients. METHODS: Twenty-three clinical trial protocols completed during 2011-2017 across diverse disease areas were analyzed to construct queries that were executed with InSite using EHR records from 24 European hospitals containing records of >14 million patients. The number of patients matching I/E criteria of each protocol was estimated. RESULTS: All protocols could be formalized to some extent into a medical coding system (e.g. ICD-10CM, ATC, LOINC, SNOMED) and mapped to local hospital coding systems. The median number of I/E criteria of protocols tested was 29 (range: 14-47). A median of 55% (range 38-89%) of I/E criteria in each protocol could be transformed into a computable format. The median number of eligible patients identified was 26 per hospital site (range: 1-134). CONCLUSION: Clinical trial I/E eligibility criteria can be structured computationally and executed as queries on EHR systems to estimate the patient recruitment pool at each site. The results further suggest that an increase in structured coded information in EHRs would increase the number of I/E criteria that could be evaluated. Additional work is needed on broader deployment of federated platforms such as InSite.

Factors influencing longitudinal changes of circulating liver enzyme concentrations in subjects randomized to placebo in four clinical trials.

Liver enzyme concentrations are measured as safety end points in clinical trials to detect drug-related hepatotoxicity, but little is known about the epidemiology of these biomarkers in subjects without hepatic dysfunction who are enrolled in drug trials. We studied alanine and aspartate aminotransferase (ALT and AST) in subjects randomized to placebo who completed assessments over 36 mo in a cardiovascular outcome trial [the Stabilisation of Atherosclerotic Plaque by Initiation of Darapladib Therapy ("STABILITY") trial; n = 4,264; mean age: 64.2 yr] or over 12 mo in three trials that enrolled only subjects with type 2 diabetes (T2D) [the DIA trials; n = 308; mean age: 62.4 yr] to investigate time-dependent relationships and the factors that might affect ALT and AST, including body mass index (BMI), T2D, and renal function. Multivariate linear mixed models examined time-dependent relationships between liver enzyme concentrations as response variables and BMI, baseline T2D status, hemoglobin A1c levels, and renal function, as explanatory variables. At baseline, ALT was higher in individuals who were men, <65 yr old, and obese and who had glomerular filtration rate (GFR) >60 ml·min-1·1.73 m-2. ALT was not significantly associated with T2D at baseline, although it was positively associated with HbA1c. GFR had a greater impact on ALT than T2D. ALT concentrations decreased over time in subjects who lost weight but remained stable in individuals with increasing BMI. Weight change did not alter AST concentrations. We provide new insights on the influence of time, GFR, and HbA1c on ALT and AST concentrations and confirm the effect of sex, age, T2D, BMI, and BMI change in subjects receiving placebo in clinical trials. NEW & NOTEWORTHY Clinical trials provide high-quality data on liver enzyme concentrations from subjects randomized to placebo that can be used to investigate the epidemiology of these biomarkers. The adjusted models show the influence of sex, age, time, renal function, type 2 diabetes, HbA1c, and body mass index on alanine aminotransferase and aspartate aminotransferase concentrations and their relative importance. These factors need to be considered when assessing potential signals of hepatotoxicity in trials of new drugs and in clinical trials investigating subjects with nonalcoholic fatty liver disease.

Real world big data for clinical research and drug development.

The objective of this paper is to identify the extent to which real world data (RWD) is being utilized, or could be utilized, at scale in drug development. Through screening peer-reviewed literature, we have cited specific examples where RWD can be used for biomarker discovery or validation, gaining a new understanding of a disease or disease associations, discovering new markers for patient stratification and targeted therapies, new markers for identifying persons with a disease, and pharmacovigilance. None of the papers meeting our criteria was specifically geared toward novel targets or indications in the biopharmaceutical sector; the majority were focused on the area of public health, often sponsored by universities, insurance providers or in combination with public health bodies such as national insurers. The field is still in an early phase of practical application, and is being harnessed broadly where it serves the most direct need in public health applications in early, rare and novel disease incidents. However, these exemplars provide a valuable contribution to insights on the use of RWD to create novel, faster and less invasive approaches to advance disease understanding and biomarker discovery. We believe that pharma needs to invest in making better use of Electronic Health Records and the need for more precompetitive collaboration to grow the scale of this 'big denominator' capability, especially given the needs of precision medicine research.