Development and evaluation of an intelligent handheld insulin dose advisor for patients with Type 1 diabetes
Diabetes mellitus is an increasingly common, chronic, incurable disease requiring careful monitoring
and treatment so as to minimise the risk of serious long-term complications. It has been suggested
that computers used by healthcare professionals and/or patients themselves may playa useful role in
the diabetes care process.
Seven key systems (AIDA, ADICOL, DIABETES, DIAS, IIumaLink, T-IDDM, POIRO) in the area
of diabetes decision support, and their underlying techniques and approaches are summarised and
compared. The development of the Patient-Oriented Insulin Regimen Optimiser (POIRO) for insulindependent
(Type-I) diabetes, and its hybrid statistical and rule-based expert system is then taken
The re-implementation and updating of the system for the Palm OS family of modern Personal Digital
Assistants (PDAs) is described. The evaluation of this new version in a seven week, randomised, open,
cross-over clinical pilot study involving eight patients on short-acting plus long-acting insulin basalbolus
regimens showed it to be easy-to-operate, reliable, not time consuming and well liked by patients.
Following this, the characteristics and use of all currently available insulin formulations, and the corresponding
insulin regimens are summarised. Algorithms to provide dose advice and decision support
for patients taking the new rapid-acting, intermediate-acting and premixed insulin formulations are
then developed. The user interface is improved and extended, amongst others through the development
and use of a model describing individual user's meal time habits. Implementation-related issues
encountered are discussed, and further work and future directions are identified and outlined.
Motivated by the complex and safety-critical nature of systems such as POIRO, we also report on the
use of the B abstract machine notation for the formal specification of the original POIRO system, and
focusing on projects and published case studies review the use of formal methods in the development
of medical computer systems.