Evaluation Considerations for Secondary Uses of Clinical Data: Principles for an Evidence-based Approach to Policy and Implementation of Secondary Analysis.

Objectives: To set the scientific context and then suggest principles for an evidence-based approach to secondary uses of clinical data, covering both evaluation of the secondary uses of data and evaluation of health systems and services based upon secondary uses of data. Method: Working Group review of selected literature and policy approaches. Results: We present important considerations in the evaluation of secondary uses of clinical data from the angles of governance and trust, theory, semantics, and policy. We make the case for a multi-level and multi-factorial approach to the evaluation of secondary uses of clinical data and describe a methodological framework for best practice. We emphasise the importance of evaluating the governance of secondary uses of health data in maintaining trust, which is essential for such uses. We also offer examples of the re-use of routine health data to demonstrate how it can support evaluation of clinical performance and optimize health IT system design. Conclusions: Great expectations are resting upon "Big Data" and innovative analytics. However, to build and maintain public trust, improve data reliability, and assure the validity of analytic inferences, there must be independent and transparent evaluation. A mature and evidence-based approach needs not merely data science, but must be guided by the broader concerns of applied health informatics.

Safety of a web-based insulin titration system for patients with type 2 diabetes mellitus - pilot study.

The treatment of type 2 diabetes patients with insulin requires active dose titration to obtain optimal glycemic control. We developed a web-based decision support system to guide patients in performing the titration task autonomously, at their homes. The system is based on a clinically validated algorithm. The aim of this study was to test the safety of the system in a pilot implementation in clinical practice. Patients were blinded from the advice given by the system and instead received insulin dosing advice given by caregivers. At the end of the pilot, advice of the system were evaluated on safety by an expert panel. In this pilot study six patients used the web-based system at their home. In total, 48 advice were logged in the system resulting in eighteen deviating systems dosing advice as compared to the advice of the caregiver. Evaluation of the eighteen deviating systems advice lead to the detection of one unsafe advice indicating a need to extend the algorithm with an additional safety decision rule.