Background
Population ageing affects most of the industrialized countries. In Italy, life expectancy at birth is the second highest in the world, being 80.6 years for men and 85.3 years for women in 2018 [
1].
The increase in life expectancy is related to an increase in age-related conditions, including cognitive impairment and dementia [
2‐
4]. In older subjects, those conditions may depend on demographic characteristics, such as age, education, and sex. A close association between age and dementia has been reported, with a higher speed of cognitive decline generally found with increasing age [
5]. In addition, previous studies have supported the key finding that a higher education level is a strong predictor of sustained cognitive function in old ages [
6‐
8] and may protect against age-related decline [
9‐
11]. Although no significant differences were found between men and women on verbal, spatial, or other cognitive abilities in one paper [
12], sex-related differences have been identified in episodic memory [
12], verbal memory [
8], cognitive speed and memory tasks [
13]: women outperformed men on these tasks, despite their generally lower level of formal education [
8].
Due to the higher life expectancy, an increased number of older drivers has been recently registered [
14]. In Italy, about 19% of all driving license holders are aged over 60 years and only 63% of the women (versus 85% of men) has a driving license, with the lowest percentages being in the south of Italy; in the north of Italy, the Lombardy region shows about 17% of all registered driving licenses, as to the more recent official report on the topic [
15].
As compared to those < 65 years, older subjects have a higher risk of car accidents [
16,
17]. In a meta-analysis [
18] based on 62 studies on impairment and risk of injury accidents, male and female drivers aged 75 years or more had a relative risk of being involved in car accidents of about 3.1 and 3.25, respectively, as compared to the corresponding groups with the lowest accident risks (male drivers aged 45–54 and female drivers aged 35–54 years, respectively). In addition, as also confirmed in [
16] on a per person-mile of travel basis, accident involvement rate was found to be a U-shaped function of driver age, both for men and for women: the relative risk decreased from the 16–19 years cohort until it reached its bottom for drivers aged 45–54 and consistently increased for all the older age cohorts [
18].
However, chronological age may be a crude criterion to prohibit driving. Driving allows independence, prevents social isolation and, on this way, provides a higher quality of life [
19,
20]. In addition, when measures are taken per capita or per number of drivers, older drivers cause fewer crashes compared to younger age groups [
16,
21]. Similarly, there has been a long-recognized association between extent of driving and crash involvement at any age group: the lower the annual mileage driven, the higher the per-distance crash rate. Because older drivers generally drive less distance per year than others, this association has been used to explain much of their apparent over-involvement in crashes [
22]. For instance, in a Dutch travel survey, when crash rates were compared after being matched for yearly driving distance, most drivers aged 75 years and above were indicatively safer than all other drivers, with only those travelling less than 3000 km/year (~ 10% of all older drivers in the survey) being at elevated crash rates [
22].
It is, therefore, essential to develop suitable criteria for identifying older subjects still able to drive. One conceivable possibility is that older drivers will be tested based on their physical and neuropsychological abilities [
23,
24]. The British Psychological Society [
25] has suggested to assess the following cognitive and psychological functions: perception, executive (or frontal) functions (e.g. plan, anticipate and make decisions, practical skills), language, memory, and personality characteristics (e.g. self-control, emotional stability, and social responsibility). With additional details, Wagner and colleagues (2011) [
21] highlighted that driving requires the integration of vision, motor, and high-level cognitive functions, including visual attention, visual perception, executive function, as well as the following three dimensions of long-term memory: episodic, semantic, and procedural memory (for detailed definitions of the single cognitive functions, see
eTable 1). A similar indication has come from two papers [
26,
27] which posed self-regulated driving abilities within the context of the “Michon model” (1985) [
28] and distinguished the three hierarchical levels of cognitive control of driving into operational, tactical, and strategic levels. In detail, the operational level implies automatic operations typically related to driving, the tactical level deals with modifications of the driving behaviour for adapting to changes in the environment, and strategic behaviors refer to trip planning, including decisions to avoid dangerous situations (for detailed definitions of hierarchical levels, see
eTable 2).
Among the available tests targeting the previous domains, the Mini Mental State Examination (MMSE) test [
29] has been suggested for the assessment of cognitive abilities related to a safe driving [
30]. Although it is widely used as a brief screening instrument for dementia and as a proxy for Alzheimer’s Disease staging [
31,
32], it is not sensitive enough to distinguish mild cognitive impairment (MCI) subjects from healthy ones [
33,
34]. The Montreal Cognitive Assessment (MoCA) [
35] can predict dementia in people with MCI, and, because it tests for executive function, it is useful for subjects with normal MMSE scores. Nonetheless, both tests are able to verify the presence of behavioral and neuropsychological signs of cortical dementia (impaired memory, orientation, and language functions) and of subcortical dementia, where slow movement and loss of attention are observed [
36]. To our knowledge, only 2 papers have compared the 2 tests in terms of driving performance. The former one [
37] concluded that they similarly predicted driving test outcomes. The latter one [
38] indicated that MoCA reliably identified at-risk individuals who had a pre-existing diagnosis of cognitive impairment, but not those without diagnosis; in addition, no significant results were found for MMSE in either sample.
Upon request of the Agency for Health Protection of Milan (Agenzia di Tutela della Salute della Città Metropolitana di Milano - ATS Milano), Milan, Italy, the University of Milan is currently carrying out a neuropsychological screening assessment of oldest old subjects [i.e., subjects > = 80 years [
39,
40]] who would like to renew their driving license and are sent there for evaluation by the Local Medical Commission in charge of license renewals. Since 2011, cognitive and psychomotor functions have been assessed through the MMSE test (odd visits) and the MoCA test (even visits), whenever subjects ask for additional renewals. Although collected with a different aim, this information provides an overall picture of the cognitive status, and its determinants, of a very interesting subpopulation of oldest old subjects.
The main objective of this paper is to assess if total and single domain scores of either test depend on socio-demographic characteristics, including age, sex, and education. Within this highly selected and functioning subpopulation of Italian oldest old subjects, we hypothesize that:
1. Regardless of the test used, age, sex, and education will still influence total cognitive performance. In detail, we expect that cognitive performance will: 1.a decrease with increasing age; 1.b increase with increasing levels of education; 1.c not depend on sex;
2. Regardless of the test used, age, sex, and education may affect the single cognitive domains under investigation. In detail, we expect that visuospatial and time orientation, memory, attention and calculation, language, and executing function will: 2.a worsen with increasing age; 2.b be higher with higher levels of education; in addition, we expect that: 2.c women will perform better than men in language and long-term memory tasks;
3. Results will be different when MoCA, instead of MMSE, test is considered. In detail, we expect MMSE scores to be higher than MOCA ones, at the overall and single-domain levels.
Discussion
The present study considers a convenience sample of oldest old persons (i.e. those with 80 years of age or more) from the metropolitan area of Milan, Lombardy, Italy, and assesses the role of age, sex, and education in influencing their cognitive performance, as measured cross-sectionally by either the MMSE or the MoCA tests. Our findings suggest that sex and education, as well as age to a lesser extent, are strong predictors of preserved cognitive functions, even in this highly functioning oldest old population, where, in median, domains targeting operational and tactical driving abilities were reached. Domains targeting strategic abilities were not fully met, due to sub-optimal performances in long-term memory domains for both tests.
Determinants of total test scores (hypothesis 1)
Concerning hypothesis 1.a, we confirm that age is a significant predictor of the mean total performance for the MMSE, but not for the MoCA test [
5,
43‐
47]. Four factors dealing with the presence of preserved driving abilities in older subjects are of great relevance [
18]. First, the health of the drivers: as they become older, reaction time worsens and the ability to drive safely is likely reduced. Second, morbidity and use of medicines in general increase with age. Third, at older ages the probability that a driver has dementia increases [
48,
49] and cognitive impairment also means reduced ability to process information and make decisions. Forth, older drivers have likely increased frailty, making them more susceptible to serious injury. With its negative significant relationship with age for subjects aged 87 years or more, the less sensitive MMSE test [
21] could have captured one or more of these aspects. Notably, although the relationship with age was stronger for increasing age categories, the effect was still moderate and of modest statistical significance; this may allow to speculate that, although oldest old subjects become older over time, the impact of ageing on cognitive performance was globally modest in our sample of highly functioning subjects.
A failure in finding a significant effect of age on MCI-targeted MoCA test may be similarly interpreted. Despite likely reduced levels of physical functioning - which are the main reasons to be evaluated by the Local Medical Commission - our oldest old subjects were mainly in good mental health at visit occasions. When compared with a subset of 48 80+ years old Italian healthy subjects from the first normative study on the MoCA test as translated into Italian (including 415 subjects from different Italian districts, representing the Italian healthy population, and with a mean raw MMSE score of 28) [
50], the mean raw MoCA score in our sample was 24.54 (see
Additional file1– eTable 4) versus 17.8 from the 80+ years old subjects from the normative study; however, only 10% of the 80+ years old subjects interviewed in the normative study had 13 or more years of education versus 32% in our sample, so we cannot exclude that the higher education level in our sample is responsible in part for our higher mean score. These latter observations supports the hypothesis of the high resilience and ability to adapt to age-associated challenges of our subjects, independently from being 80- or 90-years old or more [
51].
In conclusion, although age itself is a proxy of several other aspects related to driving abilities, it should not necessarily be used to assume driving capability or inability; however, it’s still possible that age acts more strongly on some specific domain scores, even though not on all of them.
Moreover, we confirm the key role of education in increasing mean test scores (hypothesis 1.b). No matter of the test under investigation, education showed a consistent trend, where increasing levels of education provided increasing total scores in both tests. More years of education and life-time participation in intellectual activities contribute to the “cognitive reserve”, here intended as the ability to cope with higher levels of brain damage without presenting clinical symptoms of dementia [
52]. In addition, lifestyle is a key component in the pathway including education and cognitive reserve. An active involvement in hobbies, outdoor and social activities prevents loneliness and social isolation; at the extreme consequences, it also contributes to an increase of the brain functioning against mental decline [
52,
53]. On the other hand, driving license itself is a key component of an independent lifestyle and contributes to maintain good cognitive functions as age increases. In conclusion, the highest the education level of the subject the highest should be the probability that he/she still has preserved driving abilities.
Contrary to hypothesis 1.c, in our sample we found that sex is a significant predictor of the total test score, with females having a higher mean total score in both tests [
54]. As compared to males, females in general may have lower levels of age-associated cognitive decline, possibly because of biological mechanisms such as the effects of estrogen or sex-specific cognitive reserve, but also due to sociocultural factors, as emerged for Whites – including a cohort in the neiborhood of Milan - in an updated pooled analysis of individual-level longitudinal data from 20 population-based cohorts (2–15 years of follow-up) from 15 countries over 5 continents, including 48,522 individuals (58.4% women) aged 54–105 (mean = 72.7) years and without dementia at baseline [
55].
Based on the neuropsychological evaluation, women should be more likely than men to receive license renewals at the same general health conditions. Evidence from a meta-analysis on 62 studies suggested that, when compared to the lowest risk categories of the same sex, male and female drivers aged 75+ had a similar relative risk of injury of ~ 3 [
18]; however, when the analysis was restricted to the 9 studies assessing the relationship between age and gender of car drivers and their involvement in injury accidents, 75+ years-old females showed the second highest relative risk of being involved in injury accidents, immediately following the 16–19 years old males, with a relative risk over 5 versus 3 for men aged 75+ [
18]. Although women are generally considered to be more careful drivers than men, and are charged for traffic offences much less often than men, the following aspects should be taken into account. Firstly, females usually drive less than men (e.g. [
56,
57]) and the accident involvement rate per kilometre of driving decreases as driving distances increase. In detail, Italian female drivers (63% of all Italian women) of any age drove 10% less than men in 2016 (women: 10600 km/year; men: 11500 km/year, from the Italian website (in Italian)
https://www.scuolaguida.it/it/Statistiche/art/2232-guidano-meglio-gli-uomini-o-le-donne/, date last access: June 17, 2020) and we expect this difference to be definitely higher for older women, although we were not able to find out Italian data on this aspect. Secondly, women tend to drive smaller cars than men (e.g. [
18,
56]), with likely not as good protection against injury in an accident as larger cars. Thirdly, women tend to drive more in cities (e.g. [
57]), where the risk of accidents is higher than in rural areas. All the mentioned aspects fit perfectly with our sample of oldest old drivers in the urban area of Milan, with its 1352 millions of citizens in 2017.
Determinants of single domain scores (hypothesis 2)
Concerning hypothesis 2.a, age was found to be a significant predictor of the long-term memory domain in the MMSE test, but not in the MoCA. The former finding has been previously supported in the literature [
5], whereas the latter one may be attributed to the different scoring system of the long-term memory domain in the two tests (see the Results section, paragraph 3.1). Indeed, although both tests ask to recall words within approximately 5 min, the MoCA considered 5 words, as compared to the 3 of MMSE, in accordance with its main objective of targeting subjects with MCI.
We also confirm hypothesis 2.b on the key role of education in increasing single domain scores of both tests, with the only exception of visuospatial orientation in MoCA. This is in line with previous literature [
8] on the role of age, sex, and education on executive functioning, verbal fluency, verbal memory, and cognitive speed tasks in healthy 64–81 years old subjects from the Maastricht Aging Study.
A similar indication has indirectly come from the literature examining risk factors for Alzheimer’s disease and other dementias (e.g. [
58,
59]), where low education was found to be a risk factor. In detail, based on a meta-analysis of ~ 20 studies up to 2005, the relative risks for low versus high education level were significantly higher for Alzheimer’s disease and for all dementias. In addition, the estimated population-attributable risk of Alzheimer’s disease worldwide was the highest for low educational attainment, as compared to diabetes, midlife hypertension, midlife obesity, physical inactivity, depression, and smoking [
59]. Within the “cognitive reserve” hypothesis, a higher level of education can delay the clinical expression of the disease in subjects with brain damage. Indeed, education provides strategies to solve the cognitive requirements and modifies neural connectivity and plasticity [
58].
Finally, models for successful ageing (e.g. Rowe and Kahn’s (1987, 1997) model [
60,
61], where avoidance of disease and disability, maintenance of high physical and cognitive function, and sustained engagement in social and productive activities were acknowledged) further suggest that higher levels of education support the good functioning of oldest old subjects. Indeed, the percentage of adults aging successfully increases markedly with increasing level of education (e.g. [
62‐
64]). In detail, models for estimating the prevalence of successful agers within and across countries (e.g. [
62,
64]) and over time (e.g. [
62]) provide control for age, sex, and education, as well as racial-ethnic background and socio-economic status, possibly at the individual level. In one of the mentioned studies on retired older adults from the US [
63], after adjusting for the previous covariates, the odds of aging successfully increased substantially for those with higher levels of education, income, and wealth. Notably, significant differences remained for each indicator of socio-economic status after simultaneously controlling for the others [
63]. This suggests parallel roles of each socio-economic component on successful aging. For instance, whereas higher income allows greater access to health promoting resources, the cognitive resources (e.g. capacity for problem solving, or knowledge) garnered through higher education may foster a sense of control that results in better health practices [
65].
Similarly, each socio-economic variable may show a positive relationship with purpose in life. High levels of income and education, as well as high professional status, may be a direct source of purpose in life; they provide more knowledge about how to reach desired goals more effectively, reflect success in life, and contribute to the perception of one’s past life as successful and meaningful. There are also indirect associations between socioeconomic status and purpose in life, mediated through activities and attainments. For example, education gives access to a broad range of interests and meaningful activities; ample financial resources enable individuals to engage in a greater variety of activities that may contribute to purpose in life [
66].
Concerning the effect of sex on single domains, we have speculated that females would have performed better in language and long-term memory tasks than males. In our sample, sex was found to be a significant predictor of long-term memory, but not of language. In addition, we can argue that previous findings [
8,
54] on better performance of women on verbal memory tasks (but not on other cognitive domains, such as speed of information processing and attention) are related to ours, as most of the memory tasks rely on verbal abilities [
12].
Finally, when relating results on single domain scores to driving abilities, we showed that a median subject asking for driving license renewal reached the maximum MMSE or MoCA scores in the (easiest) operational and tactical levels indicated in the Michon model, but failed in doing that at the more complex strategic level; this failure was due to a sub-optimal performance in the long-term memory domain. All the previous observations taken together provided support to the following findings:
1.
As a proxy for more complex abilities required for a safe driving, long-term memory should receive a higher weight in the evaluation of cognitive performance in driving license renewals;
2.
A preserved long-term memory is more likely to be present when relatively younger females with the highest level of education are interviewed;
3.
After confirmation of education being the more consistent and strong determinant of cognitive function at the single domain level, interviews on driving license renewals should always ask about the highest level of education reached by the subject in his/her life.
Comparison of results between mini mental state examination and Montreal cognitive assessment tests (hypothesis 3)
Finally, we have hypothesized that results would have been different in the 2 tests, with higher scores in the MMSE one (hypothesis 3). Indeed, in the regression models, we observed a higher mean total score in the reference category (i.e., [80–86) years old females with up to the fifth grade of primary school) with MMSE test (intercept: 27.6) as compared to MoCA (intercept: 22.2). This aspect likely reflects differences in test administration
: MoCA is longer than MMSE to fill in and our subjects generally reported more difficulties in carrying out the single tests of the MoCA battery. However, both education and sex had a stronger effect when MoCA test was considered: the highest level of education (i.e., “>High school graduate”) added to the mean total score 4.2 points (versus 1.52 of the MMSE) and males showed a mean total score decreasing by 0.71 with MoCA (versus 0.31 with MMSE). This may be due to the fact that the MoCA total score has a reduced ceiling effect in comparison with the MMSE (3rd quartile of total score distribution: 30 for the MMSE versus 27 for the MoCA test), as also emerged for MCI and healthy controls in Trezpack et al. [
67].
Both knowledge of the theory and the practical administration of the two tests to our sample would suggest that MoCA should perform better in capturing driving abilities (see for instance comments in [
21]). However, literature comparing the two tests in terms of driving performance [
37,
38] is still scanty and inconclusive. Indeed, the ability of MoCA to reliably identify at-risk individuals who had a pre-existing diagnosis of cognitive impairment was highly expected (e.g., [
67]) and no significant results were found for the MMSE test in either sample [
37,
38]. Within a public health perspective, where both traffic safety of the population and independence of oldest old subjects are major priorities, a fair evaluation of driving ability should integrate the neuropsychological assessment from both tests with practical sessions at driving simulators in a sufficiently large sample of subjects.
Strengths and limitations of the study
Among study limitations, the major one is that our results have a limited generalizability. Indeed, we interviewed oldest old subjects who live in a metropolitan area of the industrialized northern Italy, have a very high level of education, and do not present severe cognitive impairment (minimum MMSE total score: 17; 1st quartile for MoCA score: 23), although a few MCI cases are likely to be present (2.5% percentile of MMSE total score: 23; minimum MoCA total score: 7). In addition, our sample included only ~ 10% of women. This sex ratio is, however, in line with existing data on Italian car drivers > 65 years of age, with > 80% of male drivers registered in 2008 [
14]. We expect the sex ratio to change drastically in the next birth cohorts, with possibly different relationships with measures of cognitive impairment and memory too. We also lack of important individual-level information, including subject’s major diseases and treatments; in addition, we were not able to relate test performance to driving test outcomes or accidents [
38]. Similarly, longitudinal analyses including either MMSE results from visits 1, 3, and 5 or MoCA results from visits 2, 4, and 6 could not be carried out so far, due to the limited number of subjects available at later visit occasions. Finally, our design did not allow for a formal comparison of MMSE and MoCA tests, as they were administered at different time points. However, we were able to cross-sectionally confirm some previous evidence on the higher sensitivity of MoCA test in identifying MCI cases [
67].
Among study strengths, to our knowledge, this is the first Italian study that describes socio-demographic characteristics of subjects asking for driving license renewal and relates them to cognitive impairment. Our sample is large and it is expected to grow in future years, as far as there is a legal requirement for collecting information on cognitive functions of oldest old subjects asking for driving license renewal. This provided us enough power to analyze each test domain separately and to conclude that long-term memory is a critical domain to target as a proxy for more complex driving abilities, indicated as “strategic abilities” within the Michon model. Although there is no previous literature on such an Italian population, our results on total and single domain scores are generally sound.
Conclusions
In conclusion, our work has shown that sex and education, as well as age to a lesser extent, are strong predictors of cognitive function, even in this highly functioning oldest old subpopulation.
Future longitudinal studies should extend the current evaluation in several directions. The administration of the MMSE and MoCA test at the same visit occasion should be integrated within a psycho-social questionnaire-based evaluation of subject’s needs, support network, medical condition, and daily life abilities to assess subject’s attitude towards successful ageing. If in doubt, the Medical Commission for the renewals should request results of practical sessions with driving simulators; although collected on a case-by-case basis, future studies on fragile subpopulations of major interest could start from these additional examinations. Alternatively, a record linkage procedure with data from the Department of Motor Vehicles of Milan would provide access to car injuries for this large population of oldest old subjects. This would allow to test our hypothesis of long-term memory being the most important cognitive domain to be targeted for assessing the presence of strategic abilities in safe driving.
Within a unified approach where cognitive reserve and successful ageing represent the key theoretical criteria, our analysis and future extensions highlight that it is indeed possible to use information on socio-demographic characteristics and cognitive performance to support the identification of older subjects still able to drive and to overcome socio-cultural barriers preventing a healthy aging. Well-being and a high quality of life in oldest old subjects would require them to maintain complex abilities, including driving, that we have targeted with disentangling overall cognitive performance into single domains with the necessary statistical power.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.