Final Report Summary - AGEING NETWORK (Frailty, disability, depression and cognitive impairment in ageing; untangling complex relationships in the older population)
Jorge Luis Borges once wrote “Everything touches everything”, and in doing so elegantly and simply described both the beauty of graph theory and the multi-factorial experience of ageing. Graph theory provides a framework for investigating the complex relationships. The ability to explore complex relationships in the context of ageing is of great importance. The intertwined relationships between frailty, disability, depression and cognitive impairment are poorly understood, and the methods currently used to understand and explore these relationships are limited by unrealistic assumptions. For example, frailty is often characterized by a total score that makes the assumption that it is the underlying construct of frailty that determines whether a person has a score of 3 or 4 and that the symptoms used to generate this score are locally independent of each other. Or in other words, the response given for a particular symptom is due to how frail the person is and not another symptom designed to measure frailty, however it is unlikely that weight loss and physical activity are not directly related to each other. Graph theory provides a framework to represent the complex relationships between symptoms and constructs that are important in ageing. Using graph theory, we built several networks to describe these processes and assess intra-individual differences.
The network model built using Wave 1 of the Irish Longitudinal Study on Ageing (TILDA) data set to specifically investigate the complex relationships between frailty, disability, depression and cognitive impairment lead to us identifying four clusters, though not quite as we had originally hypothesized. Initially, we believed that there would be a cluster representing frailty, another representing disability, a third representing depression and the final cluster representing cognition. What we found was that the frailty symptoms did not form their own unique cluster. Rather there was a disability cluster, a depression cluster and two cognitive clusters. One cognition cluster represented executive function like symptoms and the other the remaining symptoms from the mini-mental state exam (MMSE), a common measure used to screen for cognitive impairment.
Interestingly the 2 most central symptoms of frailty where self-report physical activity and weakness as measured by grip strength. The least central frailty symptom was weight loss; this may be because loss of lean body mass can occur without substantial weight loss to meet the criteria for this symptom of weight loss greater than 4.5kg without trying.
While these results do not directly support our hypothesis, the frailty phenotype is defined in the literature as a physiological cycle of decline that is multi-system, therefore, it appears that it may be more appropriate that the symptoms are embedded within the other clusters, which represents the multi-system nature of this syndrome. More importantly, where previous work has not made a distinction with respect to the specific symptoms present, in that a patient either meets the criteria for frailty (3 symptoms or more) or does not, we believe the type of symptom matters. For example, if a diagnosis of frailty is being driven by weight loss and exhaustion, ensuring the patient is not experiencing depression should be the first line of inquiry. From the research perspective, using longitudinal data to determine the trajectories of patients with different symptom clusters is going to be important to understand health care utilization, adverse outcomes (i.e. falls) and mortality.
Related to the Aims of “Frailty, disability, depression and cognitive impairment in ageing: Untangling complex relationships in the older population” Dr King-Kallimanis applied network analysis to obtain a broader understanding of the relationships between the indicators of cognition as measured via the comprehensive battery of cognitive tests in the TILDA data set. This work is a deviation from the original proposal but was deemed important for Dr. King-Kallimanis' integration as it allowed her to work closely with the A3 Frailty Action group. This work was also important to Aim 1 of the award as was used to provide an understanding of the cognition cluster of the network analysis. Finally, this work was deemed important as within the field of cognition there are differing opinions on the areas of cognition these tests measure we used network analysis to assess the relationships. For example, the term ‘executive function’ is frequently used to describe a set of abilities related to the processing and manipulation of information, and resultant decision making. However, the same basic construct has also been described as ‘cognitive control’ or ‘fluid reasoning’. Often tests require patients to use more than one domain of memory and the results from out analysis indicated where there was overlap.
Both network models provide insight into how symptoms of cognitive impairment relate to each other and also how the symptoms of frailty, disability, depression and cognitive impairment cluster together. Previous research has suggested that frailty, disability, depression and cognitive are distinct, but have overlap. By understanding common pathways between each cluster of symptoms and how symptoms overlap, we think that we can better understand this complex set of symptoms. The relationships between these disorders may actually be more important than the labels given to them when trying to understand the best action for treatment of older adults with complex health care needs. Further work, extending these models with longitudinal data will help provide insight into important clinical pathways and developing targeted interventions.
The network model built using Wave 1 of the Irish Longitudinal Study on Ageing (TILDA) data set to specifically investigate the complex relationships between frailty, disability, depression and cognitive impairment lead to us identifying four clusters, though not quite as we had originally hypothesized. Initially, we believed that there would be a cluster representing frailty, another representing disability, a third representing depression and the final cluster representing cognition. What we found was that the frailty symptoms did not form their own unique cluster. Rather there was a disability cluster, a depression cluster and two cognitive clusters. One cognition cluster represented executive function like symptoms and the other the remaining symptoms from the mini-mental state exam (MMSE), a common measure used to screen for cognitive impairment.
Interestingly the 2 most central symptoms of frailty where self-report physical activity and weakness as measured by grip strength. The least central frailty symptom was weight loss; this may be because loss of lean body mass can occur without substantial weight loss to meet the criteria for this symptom of weight loss greater than 4.5kg without trying.
While these results do not directly support our hypothesis, the frailty phenotype is defined in the literature as a physiological cycle of decline that is multi-system, therefore, it appears that it may be more appropriate that the symptoms are embedded within the other clusters, which represents the multi-system nature of this syndrome. More importantly, where previous work has not made a distinction with respect to the specific symptoms present, in that a patient either meets the criteria for frailty (3 symptoms or more) or does not, we believe the type of symptom matters. For example, if a diagnosis of frailty is being driven by weight loss and exhaustion, ensuring the patient is not experiencing depression should be the first line of inquiry. From the research perspective, using longitudinal data to determine the trajectories of patients with different symptom clusters is going to be important to understand health care utilization, adverse outcomes (i.e. falls) and mortality.
Related to the Aims of “Frailty, disability, depression and cognitive impairment in ageing: Untangling complex relationships in the older population” Dr King-Kallimanis applied network analysis to obtain a broader understanding of the relationships between the indicators of cognition as measured via the comprehensive battery of cognitive tests in the TILDA data set. This work is a deviation from the original proposal but was deemed important for Dr. King-Kallimanis' integration as it allowed her to work closely with the A3 Frailty Action group. This work was also important to Aim 1 of the award as was used to provide an understanding of the cognition cluster of the network analysis. Finally, this work was deemed important as within the field of cognition there are differing opinions on the areas of cognition these tests measure we used network analysis to assess the relationships. For example, the term ‘executive function’ is frequently used to describe a set of abilities related to the processing and manipulation of information, and resultant decision making. However, the same basic construct has also been described as ‘cognitive control’ or ‘fluid reasoning’. Often tests require patients to use more than one domain of memory and the results from out analysis indicated where there was overlap.
Both network models provide insight into how symptoms of cognitive impairment relate to each other and also how the symptoms of frailty, disability, depression and cognitive impairment cluster together. Previous research has suggested that frailty, disability, depression and cognitive are distinct, but have overlap. By understanding common pathways between each cluster of symptoms and how symptoms overlap, we think that we can better understand this complex set of symptoms. The relationships between these disorders may actually be more important than the labels given to them when trying to understand the best action for treatment of older adults with complex health care needs. Further work, extending these models with longitudinal data will help provide insight into important clinical pathways and developing targeted interventions.