Page 147 - Petelin, Ana, ur. 2021. Zdravje starostnikov / Health of the Elderly. Zbornik povzetkov z recenzijo / Book of Abstracts. Koper: Založba Univerze na Primorskem/University of Primorska Press
P. 147
ited lecture tehnološke in ostale ikt rešitve | technological an ict solutions
Using aging population big data to develop prediction models
Leona Cilar, Gregor Štiglic
University of Maribor, Faculty of Health Sciences, Maribor, Slovenia
Introduction. With the increase of the population age worldwide, there are
more and more people living with the type 2 diabetes mellitus. Type 2 diabetes
mellitus is leading disease, causing mortality and morbidity worldwide and pre-
sents a major economic burden for healthcare systems. The aim of this study
was to develop and validate a type 2 diabetes mellitus risk prediction model.
Methods. A secondary data analysis was performed. Data from the Survey of
Health, Ageing and Retirement in Europe waves 1 to 7 collected between 2004
and 2017 were used to develop and validate prognostic models for 10-year
type 2 diabetes mellitus prediction in European countries.
Results. Sample of 16,363 participants was used to develop and validate a global
regularized logistic regression model. Model achieved an area under the receiv-
er operating characteristic curve of 0.702 (95 % CI: 0.698 – 0.706). Measured
performance of local country-specific models where an area under the receiv-
er operating characteristic curve ranged from 0.578 (0.565 - 0.592) to 0.768
(0.749 0.787). The Danish model was assessed as model with the best predic-
tion performance where body mass index was the only variable that was se-
lected in all cross-validation runs, followed by alcohol consumption in the last
six months and smoking status which were selected in 48 % and 47% of the
models.
Discussion and conclusion. Study findings demonstrate the importance of re-cali-
bration of the models as well as strengths of pooling the data from multiple
countries to reduce the variance and consequently increase the precision of
the results.
Keywords: type 2 diabetes; prediction model; elderly person; healthcare
145
Using aging population big data to develop prediction models
Leona Cilar, Gregor Štiglic
University of Maribor, Faculty of Health Sciences, Maribor, Slovenia
Introduction. With the increase of the population age worldwide, there are
more and more people living with the type 2 diabetes mellitus. Type 2 diabetes
mellitus is leading disease, causing mortality and morbidity worldwide and pre-
sents a major economic burden for healthcare systems. The aim of this study
was to develop and validate a type 2 diabetes mellitus risk prediction model.
Methods. A secondary data analysis was performed. Data from the Survey of
Health, Ageing and Retirement in Europe waves 1 to 7 collected between 2004
and 2017 were used to develop and validate prognostic models for 10-year
type 2 diabetes mellitus prediction in European countries.
Results. Sample of 16,363 participants was used to develop and validate a global
regularized logistic regression model. Model achieved an area under the receiv-
er operating characteristic curve of 0.702 (95 % CI: 0.698 – 0.706). Measured
performance of local country-specific models where an area under the receiv-
er operating characteristic curve ranged from 0.578 (0.565 - 0.592) to 0.768
(0.749 0.787). The Danish model was assessed as model with the best predic-
tion performance where body mass index was the only variable that was se-
lected in all cross-validation runs, followed by alcohol consumption in the last
six months and smoking status which were selected in 48 % and 47% of the
models.
Discussion and conclusion. Study findings demonstrate the importance of re-cali-
bration of the models as well as strengths of pooling the data from multiple
countries to reduce the variance and consequently increase the precision of
the results.
Keywords: type 2 diabetes; prediction model; elderly person; healthcare
145
