Analyzing the therapy of tight glycemic controlled patients for better treatment

OData support
Supervisor:
Dr. Benyó Balázs István
Department of Control Engineering and Information Technology

Hyperglycemia caused by stress, high level of insulin resistance, and severe blood glucose level variability are the most common problems with patients in intensive care units. A strong correlation can be shown between these symptoms and mortality level.

The beneficial effect of tight glycemic control methods is proved in numerous studies. A proper insulin and nutrition dosage can significantly decrease the rate of the mortality, in front of poor level insulin and nutrition therapy. STAR (Stochastic TARgeted) is a flexible, model-based TGC approach, which directly accounts for intra- and interpatient variability with a stochastically derived maximum 5% risk of blood glucose below 72 mg/dl.

During the semester I evaluated two groups of patients. The first measurements were obtained from the Hungarian Békés Megyei Pándy Kálmán Hospital, and the other set of data was from the Christchurch Hospital from New-Zeeland. Both groups contain non-diabetic adult patients, who were treated with STAR protocol during intensive care. In Hungary, STAR protocol was combined with an advanced feeding regime. The Hungarian group got more and continual enteral and parenteral nutrition during the treatment, while the patient group from Christchurch Hospital got a lower nutrition level. The differences in the feeding are due to the different hospital protocols, which have different clinical considerations.

During my work I processed the raw data and extracted the useful parameters for statistic tests. I compared the results of the groups with experimental and observational statistical studies and confirmed that the feeding protocol, used in Hungary helped resulted a faster recovery and a decreased the number of the spent days in hospital. I proved that the Hungarian protocol is efficient and worth for a deeper level of analysis in my future work.

Downloads

Please sign in to download the files of this thesis.