Traduction en anglais : Crossover validation depicts general similarity between clients for ambulatory, or non-ambulatory nursing care insurance according to German law.
In nursing insurance according to German legislation, the act of classifying the client into degrees of severity of disablement (0=none, 1 to 3=considerable, severe, most severe disablement) by expert assessment is based on legally defined distinct criteria (need for help items).
Over and above these criteria, information with regard to the activities of daily life (ADL) is also documented routinely.
Is there a fundamental difference between patients suitable for ambulatory or non-ambulatory nursing care, or are their positive pattern scores for needing help fundamentally different ?
Based on 7,000 electronic records of the assessments for ambulatory, and another 7,000 records for non-ambulatory nursing care, we fed artificial neural networks with all, or subsets, of the available items.
Thereafter, in a crossover design, we evaluated the nets'classification competencies on independent validation data, using the ambulatory net for non-ambulatory classification, and vice versa.
Weighted kappa (kw) was calculated as an index of performance.
The nets trained on ambulatory data uniformly performed slightly better on ambulatory than on non-ambulatory data (kw=0.78 versus 0.65, all items ; 0.68 versus 0.67, only ADL items), whereas the non-ambulatory nets showed no consistent difference in respect of the nature of the input data (kw=0.76/0.77, and 0.71/0.66).
Overall, the nets'classification competence was convincing. (...)
Mots-clés Pascal : Soin, Infirmier, Ambulatoire, Classification, Législation, Evaluation, Homme, Allemagne, Europe, Personnel sanitaire
Mots-clés Pascal anglais : Care, Nurse, Ambulatory, Classification, Legislation, Evaluation, Human, Germany, Europe, Health staff
Notice produite par :
Inist-CNRS - Institut de l'Information Scientifique et Technique
Cote : 99-0207833
Code Inist : 002B30A01C. Création : 16/11/1999.