Premium
Are existing dementia risk prediction models reliable?
Author(s) -
Goerdten Jantje,
Čukić Iva,
Danso Samuel O,
Carriere Isabelle,
Terrera Graciela Muniz
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.040814
Subject(s) - logistic regression , dementia , reliability (semiconductor) , identification (biology) , predictive modelling , regression analysis , risk assessment , proportional hazards model , computer science , coronary heart disease , regression , statistics , psychology , medicine , artificial intelligence , disease , machine learning , mathematics , power (physics) , physics , botany , computer security , quantum mechanics , biology
Abstract Background Numerous dementia risk prediction models have been developed in the past, however only a few are recommended for clinical use (Stephan et al., 2015; Tang et al., 2015). This might be due to methodological limitations that hamper their reliability and accuracy. Here we summarize and critically discuss the employed methodological approaches to develop dementia risk prediction models. Method We systematically reviewed the literature from March 2014 to September 2018 for publications presenting a dementia risk prediction model. Result In total 137 publications were included in the qualitative synthesis. Three analytical techniques are commonly used: machine learning, logistic regression and Cox regression. Additionally, we identified three major methodological weaknesses: (1) overreliance on one data source, (2) poor verification of statistical assumptions of Cox and logistic regression and (3) lack of validation. Conclusion Inaccurate application of analytical techniques can cause false identification of high‐risk groups and biased prognoses (Abrahamowicz et al.,1997; Exalto et al., 2014). Hence, we recommend the use and collection of larger and more diverse samples. Furthermore, the assumptions should be tested thoroughly, and actions taken if derivations are detected. Improved practice in data analysis and innovative data designs may improve dementia risk scores. References: (1) Abrahamowicz, M., du Berger, R., & Grover, S. A. (1997). Flexible modeling of the effects of serum cholesterol on coronary heart disease mortality. Am J Epidemiol, 145 (8), 714‐729. (2) Exalto, L. G., Quesenberry, C. P., Barnes, D., Kivipelto, M., Biessels, G. J., & Whitmer, R. A. (2014). Midlife risk score for the prediction of dementia four decades later. Alzheimers Dement, 10 (5), 562‐570. doi:10.1016/j.jalz.2013.05.1772. (3) Stephan, B. C., Tzourio, C., Auriacombe, S., Amieva, H., Dufouil, C., Alperovitch, A., & Kurth, T. (2015). Usefulness of data from magnetic resonance imaging to improve prediction of dementia: population based cohort study. BMJ, 350 , h2863. doi:10.1136/bmj.h2863. (4) Tang, E. Y., Harrison, S. L., Errington, L., Gordon, M. F., Visser, P. J., Novak, G., Stephan, B. C. (2015). Current Developments in Dementia Risk Prediction Modelling: An Updated Systematic Review. PLoS One, 10 (9), e0136181. doi:10.1371/journal.pone.0136181.