As people are living longer, the number of people living with frailty is increasing dramatically. Frailty describes a loss of fitness that can occur as a result of natural ageing combined with the effects of multiple long-term medical conditions. This research will help to discover subtypes of frailty and to make more accurate predictions of an individual’s risk of worse health outcomes like falls, mortality or unanticipated increases in care needs.
This study will examine patterns of frailty, and has five elements.
1. We will compare the different ways for measuring frailty in electronic health records.
2. We will use a range of methods for finding patterns in both the number and combination of health issues that people with frailty experience. We will see if these methods give us similar answers, and answers that make sense to patients and healthcare professionals.
3. We will extend this analysis to examine changes in patterns over time.
4. We will use several different methods to examine how specific patterns found in (2) and (3) relate to worse outcomes (for example, deaths, hospital admissions, falls, fractures).
5. We will compare the patterns found in (2), (3) and (4) in terms of the specific characteristics of individuals displaying these found patterns (for example, different age groups; men vs. women).
We will share the results in a number of ways to ensure that as many diverse stakeholders as possible hear about them.
This study is part of a wider PhD project “What do data tell us about frailty?”, which is modelling observed health deficits and their relationship to adverse outcomes to discover subtypes of frailty with different prognostic profiles. Health deficits typically comprise a range of diseases, symptoms / signs, disabilities and abnormal lab values.
Objective 1: We aim to discover subtypes of frailty by inferring latent classes and clusters from the observed health deficits used in the electronic Frailty Index (eFI) and the eFI+. We will assess whether these subtypes are consistent (across methods), reproducible (across data sources) and explainable.
Objective 2: We will examine the demographic characteristics of individuals who exhibit each frailty subtype.
Objective 3: We will investigate the relationship between frailty subtypes and adverse health outcomes, taking into account relationships between different outcomes.
Objective 4: We will conduct a longitudinal analysis to explore how the sequential order in which health deficits occur can deepen our understanding of frailty subtypes.
The study design is retrospective and observational. It combines cross-sectional and longitudinal elements, allowing for frailty subtypes to be explored at a single time point and over time within subjects. The study population comprises individuals aged 50 years or older registered with a GP practice in England for at least a year in 2010 and 2019.
We will publish and disseminate our findings in a variety of ways to ensure we reach multiple audiences, and publish all our code to facilitate replication by others. Outside the scope of this application, we will examine whether the frailty subtypes derived by us in CPRD are replicable in other datasets.
Better understanding subtypes of frailty can help healthcare providers tailor interventions to better meet the specific needs of different individuals, resulting in more targeted and effective care for people living with frailty.
For the analysis on subtypes of frailty (objectives 1 and 4): the outcomes are the latent classes of frailty and clusters of deficits.
Rationale: Health deficit data is used to measure individual’s frailty in terms of the number and proportion of deficits recorded, which is quantified in a frailty index. It is the latent structure and clusters of these deficits which is the focus of analysis for discovering frailty subtypes. A critical issue is the choice of health deficits to examine, and how those deficits are defined in CPRD Aurum data using codelists in GP data. Our choice of deficits includes all of the deficits used in the electronic Frailty Index (eFI) [1] and the eFI+ [2].
For the analysis examining associations between individual characteristics and frailty subtypes (objective 2): the outcomes are frailty subtypes.
Rationale: Understanding who experiences different frailty subtypes is important to explore clinical implications, such as whether frailty manifests differently at different ages and in men vs. women (who are more frail but live longer).
For the analysis on adverse health outcomes (objective 3): the outcomes are subsequent adverse or other events, namely all-cause mortality; emergency department attendance; unplanned hospital admission (any admission; admission with falls or fractures; critical care admission), using ICD-10 codes in HES and ONS data.
Rationale: The selected outcomes have all been used in previous research, and are important to measure to provide comparability with previous research and to help evaluate whether the prognostic profiles for frailty subtypes have face validity / importance.
Sohan Seth - Chief Investigator - University of Edinburgh
Lara Johnson - Corresponding Applicant - University of Edinburgh
Alan Marshall - Collaborator - University of Edinburgh
Atul Anand - Collaborator - University of Edinburgh
Benjamin Bach - Collaborator - University of Edinburgh
Bruce Guthrie - Collaborator - University of Edinburgh
Sohan Seth - Collaborator - University of Edinburgh
HES Admitted Patient Care;ONS Death Registration Data