Predicting pharmacotherapeutic outcomes for type 2 diabetes: An evaluation of three approaches to leveraging electronic health record data from multiple sources.

Electronic health record (EHR) data are increasingly used to develop prediction models to support clinical care, including the care of patients with common chronic conditions. A key challenge for individual healthcare systems in developing such models is that they may not be able to achieve the desired degree of robustness using only their own data. A potential solution-combining data from multiple sources-faces barriers such as the need for data normalization and concerns about sharing patient information across institutions. To address these challenges, we evaluated three alternative approaches to using EHR data from multiple healthcare systems in predicting the outcome of pharmacotherapy for type 2 diabetes mellitus(T2DM). Two of the three approaches, named Selecting Better (SB) and Weighted Average(WA), allowed the data to remain within institutional boundaries by using pre-built prediction models; the third, named Combining Data (CD), aggregated raw patient data into a single dataset. The prediction performance and prediction coverage of the resulting models were compared to single-institution models to help judge the relative value of adding external data and to determine the best method to generate optimal models for clinical decision support. The results showed that models using WA and CD achieved higher prediction performance than single-institution models for common treatment patterns. CD outperformed the other two approaches in prediction coverage, which we defined as the number of treatment patterns predicted with an Area Under Curve of 0.70 or more. We concluded that 1) WA is an effective option for improving prediction performance for common treatment patterns when data cannot be shared across institutional boundaries and 2) CD is the most effective approach when such sharing is possible, especially for increasing the range of treatment patterns that can be predicted to support clinical decision making.

The Quaternary evolutionary history of Bristol rock cress (Arabis scabra, Brassicaceae), a Mediterranean element with an outpost in the north-western Atlantic region.

BACKGROUND AND AIMS: Bristol rock cress is among the few plant species in the British Isles considered to have a Mediterranean-montane element. Spatiotemporal patterns of colonization of the British Isles since the last interglacial and after the Last Glacial Maximum (LGM) from mainland Europe are underexplored and have not yet included such floristic elements. Here we shed light on the evolutionary history of a relic and outpost metapopulation of Bristol rock cress in the south-western UK. METHODS: Amplified fragment length polymorphisms (AFLPs) were used to identify distinct gene pools. Plastome assembly and respective phylogenetic analysis revealed the temporal context. Herbarium material was largely used to exemplify the value of collections to obtain a representative sampling covering the entire distribution range. KEY RESULTS: The AFLPs recognized two distinct gene pools, with the Iberian Peninsula as the primary centre of genetic diversity and the origin of lineages expanding before and after the LGM towards mountain areas in France and Switzerland. No present-day lineages are older than 51 ky, which is in sharp contrast to the species stem group age of nearly 2 My, indicating severe extinction and bottlenecks throughout the Pleistocene. The British Isles were colonized after the LGM and feature high genetic diversity. CONCLUSIONS: The short-lived perennial herb Arabis scabra, which is restricted to limestone, has expanded its distribution range after the LGM, following corridors within an open landscape, and may have reached the British Isles via the desiccated Celtic Sea at about 16 kya. This study may shed light on the origin of other rare and peculiar species co-occurring in limestone regions in the south-western British Isles.

The Indiana Learning Health System Initiative: Early experience developing a collaborative, regional learning health system.

INTRODUCTION: Learning health systems (LHSs) are usually created and maintained by single institutions or healthcare systems. The Indiana Learning Health System Initiative (ILHSI) is a new multi-institutional, collaborative regional LHS initiative led by the Regenstrief Institute (RI) and developed in partnership with five additional organizations: two Indiana-based health systems, two schools at Indiana University, and our state-wide health information exchange. We report our experiences and lessons learned during the initial 2-year phase of developing and implementing the ILHSI. METHODS: The initial goals of the ILHSI were to instantiate the concept, establish partnerships, and perform LHS pilot projects to inform expansion. We established shared governance and technical capabilities, conducted a literature review-based and regional environmental scan, and convened key stakeholders to iteratively identify focus areas, and select and implement six initial joint projects. RESULTS: The ILHSI successfully collaborated with its partner organizations to establish a foundational governance structure, set goals and strategies, and prioritize projects and training activities. We developed and deployed strategies to effectively use health system and regional HIE infrastructure and minimize information silos, a frequent challenge for multi-organizational LHSs. Successful projects were diverse and included deploying a Fast Healthcare Interoperability Standards (FHIR)-based tool across emergency departments state-wide, analyzing free-text elements of cross-hospital surveys, and developing models to provide clinical decision support based on clinical and social determinants of health. We also experienced organizational challenges, including changes in key leadership personnel and varying levels of engagement with health system partners, which impacted initial ILHSI efforts and structures. Reflecting on these early experiences, we identified lessons learned and next steps. CONCLUSIONS: Multi-organizational LHSs can be challenging to develop but present the opportunity to leverage learning across multiple organizations and systems to benefit the general population. Attention to governance decisions, shared goal setting and monitoring, and careful selection of projects are important for early success.

Linkage of Indiana State Cancer Registry and Indiana Network for Patient Care Data.

BACKGROUND: Large automated electronic health records (EHRs), if brought together in a federated data model, have the potential to serve as valuable population-based tools in studying the patterns and effectiveness of treatment. The Indiana Network for Patient Care (INPC) is a unique federated EHR data repository that contains data collected from a large population across various health care settings throughout the state of Indiana. The INPC clinical data environment allows quick access and extraction of information from medical charts. The purpose of this project was to evaluate 2 different methods of record linkage between the Indiana State Cancer Registry (ISCR) and INPC, determine the match rate for linkage between the ISCR and INPC data for patients diagnosed with cancer, and to assess the completeness of the ISCR based on additional validated cancer cases identified in the INPC EHRs. METHODS: Deterministic and probabilistic algorithms were applied to link ISCR cases to the INPC. The linkage results were validated by manual review and the accuracy assessed with positive predictive value (PPV). Medical charts of melanoma and lung cancer cases identified in INPC but not linked to ISCR were manually reviewed to identify true incidence cancers missed by the ISCR, from which the completeness of the ISCR was estimated for each cancer. RESULTS: Both deterministic and probabilistic approaches to linking ISCR and INPC had extremely high PPV (>99%) for identifying true matches for the overall cohort and each subcohort. The combined match rate for melanoma and lung cancer cases identified in the ISCR that matched to any patient occurrence in INPC (not by disease) was 85.5% for the complete cohort, 94.4% for melanoma, and 84.4% for lung cancer. The estimated completeness of capture by the ISCR was 84% for melanoma and 98% for lung cancer. Conclusion: Cancer registries can be successfully linked to patients' EHR data from institutions participating in a regional health information organization (RHIO) with a high match rate. A pragmatic approach to data linkage may apply both deterministic and probabilistic approaches together for the diverse purposes of cancer control research. The RHIO has the potential to add value to the state cancer registry through the identification of additional true incident cases, but more advanced approaches, such as natural language processing, are needed.