Bond, Ashley, Harrison, Simon and Gerber, Luciano ORCID: https://orcid.org/0000-0002-8423-4642
(2025)
Development and utility of a computer-based selection algorithm for the prescribing of home parenteral nutrition.
[Conference or Workshop Item]
![]() |
Published Version
File not available for download. Available under License In Copyright. Download (49kB) |
Abstract
Introduction: There is a drive to increase the use of MCBs, as an alternative to compounded parenteral support, for CIF patients. Currently in the UK, there are, in excess of 70 MCBs available on the NHS England framework, together with >30 IV fluid products. If a required regime is that of an MCB plus one IV fluid product, then the potential combinations exceed 3000. Method: We have developed a computer-based selection algorithm known as PNMatch. It utilises a ‘kNearest Neighbour’ strategy for matching. Following completion of the computational design for the algorithm, a staged testing and development process was undertaken. We then performed a matching assessment against existing prescriptions. Cohen’s correlation testing was used to assess this matching. Results: Matching correlation was assessed in two main phases. Phase 1: Of the 20 prescriptions tested 18/20 (90%) were an exact product match, 1/20 (5%) were not match but improved and 1/20 (5%) not matching and inferior. Cohen’s kappa value for matching was 0.8. Also recorded by the operator was the time taken to complete each individual selection process, with the mean time being 116 seconds (range 79-168 seconds). Phase 2: Existing formulation request for 31 patients receiving MCB+/- IV fluids regimes were collated, in total 44 MCB containing requests were assessed. Off the 44 different formulation requests performed 43 had a product selected that was deemed suitable for prescribing. With 42 appearing in the top 5 ranking selection from the computer algorithm. Of the 2 that did not appear in the top 5 ranking, one regime was unmatched (MCB + IV fluid), with the remaining being unmatched but deemed superior to the existing home PN prescription by the physician. A Cohen’s matching coefficient for the computer algorithm top 5 position and the existing home PN prescription was found to be 0.95, with a figure of 0.85 for top 1 rank and home prescription. There was no significant difference in the aggregate difference between the formulation request and the algorithm selected product across a whole week. Conclusion: We have designed, developed and tested a computer-based selection algorithm for the prescription of MCB based PS for patients with CIF. This has the potential to improve efficiency and reduced variability across the intestinal failure network, supporting the initiative to expand the use of MCBs nationally.
Impact and Reach
Statistics
Additional statistics for this dataset are available via IRStats2.