| Abstract |
Introduction
With the gradual increase in the population longevity, chronic conditions have become more prevalent
particularly in the elderly. Cognitive disorders like Alzheimer’s disease and mild cognitive impairment
(MCI) being amongst the most common comorbidities in that age group. Cognitive disorders have a
significant impact in the quality of life of patients, their families and their caregivers. In the absence of a
disease-modifying drug treatment, early detection and management of key risk-factors remains a key
strategy for health services. Primary Health Care (PHC) is the first point of entry of patients with the health
care services plays a significant role in the early detection as well as in the management of comorbidities
parallel to cognitive disorders. The aim of this doctoral thesis is to apply multilevel statistical models as
well as machine learning methods towards the detection of cognitive disorders, modifiable risk-factors
and comorbidities.
Methods
The current doctoral thesis made use of the data from the research project named “Thalis University of
Crete: A multi-disciplinary network for the study of Alzheimer’s disease and related disorders”. The study
took place in the district of Heraklion, Crete, Greece between March 2013 and December 2014 and took
place in two phases. During the 1st phase of the study 3,140 participant were recruited from 14 selected
PHC units located within the district. All participants completed a structured and pre-tested questionnaire
which elicited information regarding basic socio-demographic characteristics, health-related habits,
chronic illnesses and prescribed medication and finally the Mini Mental State Examination (MMSE)
cognitive test. Participants who scored below 24 units in the MMSE as well as a selected matched sample
of those with MMSE score > 24 were invited to participate in the 2nd phase of the study where a complete
neuropsychologic and neuropsychiatric evaluation by a team of experts took place. For the diagnosis of
dementia (all types) and mild cognitive impairment the Diagnostic and Statistical Manual of Mental
Disorders (DSM IV) was used. A two-level multiple logistic regression model was used in order to
investigate the impact of selected modifiable risk-factors in the presence of probable cognitive
impairment (according to MMSE score). The impact of selected comorbidities on MMSE scores was
investigated with the use of multiple regression analyses. Generalized Linear Model Lasso Regularization
was used for feature selection in the MMSE items. Finally, two-layered artificial neural networks were
used in order to classify patients as cognitively impaired (dementia or MCI) versus non-impaired.
Results
A total of 3140 participants were recruited in the first phase of the study (43.2% were males; with a mean
age of 73.7±7.8 years). The average MMSE score for the total population was 26.0±3.8; 26.7±3.5 in male
and 25.4±3.9 in female participants (p<0.0001). Low MMSE scores were detected in 20.2% of participants;
25.9% for females vs 12.8% for males; p<0.0001. Female gender (Odds ratio -OR- 2.72; 95% CI 2.31 to
3.47), age (OR=1.11; 95% CI 1.10 to 1.13), having received only primary or no formal education (OR=2.87;
12
95% CI 2.26 to 3.65), alcohol intake (OR=1.19; 95% CI 1.03 to 1.37), reporting one or more sleep complaints
(OR 1.63; 95% CI 1.14 to 2.32), dyslipidemia (OR=0.80; 95% CI 0.65 to 0.98) and history of depression
(OR=1.90; 95% CI 1.43 to 2.52) were associated with the presence of low MMSE scores. Among
participants with low MMSE scores 344 (54.1%) underwent comprehensive neuropsychiatric evaluation
and 185 (53.8%) were diagnosed with MCI 118 (34.3%) with dementia. Mental and behavioral disorders
(F00-F99) and diseases of the nervous system (G00-G99) increased the odds of low MMSE scores in both
genders. Generalized linear model lasso regularization indicated that 7/30 MMSE questions contributed
the most to the classification of patients as impaired (dementia/MCI) vs. non-impaired with a combined
accuracy of 82.0%. These MMSE items were questions 5, 13, 19, 20, 22, 23, and 26 of the Greek version
of MMSE assessing orientation in time, repetition, calculation, registration, and visuo-constructive ability.
Conclusions
Results of the present doctoral thesis indicated a relatively high prevalence of low MMSE scores amongst
elderly PHC visitors and validated the associations with selected modifiable risk factors. Findings of this
thesis have also identified the associations of certain chronic illness complexes (according to ICD-10) with
low MMSE scores. Finally, machine learning algorithms have provided evidence that seven of the MMSE
items could provide sufficient power to classify participants as cognitively impaired (dementia or MCI)
|
Search
