NextDose: a web-based collaborative tool for target concentration intervention

Aims: The overarching aim of the NextDose project is to bring better dosing methods to the ward with target concentration intervention software. This includes two major steps of: providing a tool to allow pharmacometric model-based calculations useable by clinicians; and then actually getting them to use it. The goal is for doses to be determined by the best possible method, meaning real patients receiving doses that were determined (at least in part), by the most robust and accurate calculation method possible – not just the easiest to understand or most popular protocol.

Methods: The project started out with several iterations of a tool called FirstDose, beginning with a version written in Java, then an Adobe Flash version, followed by multiple rewrites in HTML and JavaScript until a useable and compatible calculator was fit for purpose for the wards. FirstDose was a fully client-side dose calculator used to recommend first and subsequent doses before concentration measurements became available for dose adjustment. It used published models for vancomycin, amikacin and gentamicin that took into account covariates such as height, weight, post-menstrual age, renal function and certain concurrent treatments. A small clinical trial in paediatric and neonatal intensive care units in Auckland was conducted that evaluated useability and improvements in serum concentrations compared to existing dosing protocols. Of significance, clinicians and lab scientists were more interested in learning how to improve dosing once concentration measurements became available, so work began on NextDose. NextDose would use Bayesian methodology to make the best use of information available for target concentration intervention. It would use a database to keep track of patients, and require a focus on security, stability and collaboration between team members from different locations. The initial release used published models of busulfan, methotrexate and tacrolimus and ran within a browser, communicating securely with a Windows server that performed Bayesian calculations using NONMEM. Output was parsed with Awk into text and csv files that were then processed by a PHP MySQL web server that interacted with a HTML5/JavaScript web app. As early release software, NextDose has been evaluated by audit, being used alongside existing departmental protocols to identify points of difference and opportunity for improvement.

Results: NextDose has been used in the care of over 90 patients receiving busulfan in Auckland. The majority were children under the age of 12. The median dose received has been close to that recommended by NextDose, being just 2% higher. Half of the dose changes made were within -6% and 10% of the NextDose recommended dose. However some doses were markedly different from the NextDose recommendation, in one case being 113% larger.

Conclusion: The NextDose project has met a number of practical challenges along the road to bring better dosing methods to the ward. Challenges include security of healthcare information, compatibility with ageing hospital computers, providing software support for an academic non-profit project, and trying to change routines in a clinical environment that is governed by multiple levels of management and policy. The limited data attained to date has been encouraging, supporting further research into software tools to improve target concentration intervention in hospitals.