The Co-Production and Evaluation of the Computerised Cognitive Assessment for Preclinical Alzheimer's Disease (CoCoA-PAD)

Alzheimer's disease (AD) healthcare is on the verge of significant change. Healthcare providers can now diagnose people with pre-dementia AD (i.e., preclinical and prodromal AD), and new disease modifying medications can slow cognitive decline. Health services are now faced with the challenge of implementing these developments into clinical practice. National Health Service (NHS) England and the National Dementia Team have developed an initial implementation strategy. It states that diagnostic assessments will be delivered in newly commissioned neurology and neuropsychology clinics, and that the completion of a 'comprehensive cognitive assessment' is required to diagnose the AD clinical syndrome, and stage the disease severity.

While AD biomarker assessments, such as plasma and cerebrospinal fluid assessments, provide information on the presence of AD pathology (Amyloid and Tau), they are not predictive of future dementia, i.e., only a proportion of 65-year-olds with preclinical AD biomarkers will develop Alzheimer's dementia in their lifetime. As such, the International Working Group (IWG) provide clear recommendations that the use of AD biomarkers in isolation in cognitive unimpaired older adults is not advisable and in fact may lead to harm. In contrast, meta-analytic research has demonstrated that the combination of biomarkers and cognitive assessments improves predictive accuracy for future AD dementia. The IWG have proposed that a clinical-biological definition of AD is adopted into clinical practice, and that the combination of plasma and sensitive cognitive markers of AD represent the most feasible strategy for a meaningful diagnosis of preclinical and prodromal AD.

The Implementation of Cognitive Assessment Pathways There are severe barriers to the implementation of comprehensive cognitive assessments for early AD into NHS practice. Most cognitive assessments are grossly insensitive to the early cognitive difficulties associated with AD and therefore not fit for purpose. Indeed, large scale research including over 5,000 participants, identified that traditional cognitive assessments do not add any predictive accuracy above demographic factors and genetics, to determine the risk of preclinical AD. There are some notable exceptions, but these rely on tests that can only be administered by practitioner psychologists, e.g., Logical Memory from Wechsler Memory Scales (WMS-IV). There is currently a national shortage of practitioner psychologists in the United Kingdom (UK). Two recent economic modelling studies demonstrate that the implementation of 'psychologist' delivered cognitive assessment onto early AD pathways would cost the NHS £4.2 billion to deliver, and without this investment, waiting lists would reach over 10 years by 2029. Therefore, the delivery of early AD cognitive assessment cannot be achieved by relying on existing cognitive tests or the existing workforce model. There is an urgent need for an implementation strategy for the delivery of large-scale cognitive assessment for early AD into the NHS, but there is no agreed upon strategy to achieve this.

Memory nurses are the largest staffing group in memory services and already have considerable experience administering and scoring brief cognitive screening tests, e.g., Montreal Cognitive Assessment (MoCA) and Addenbrooke's Cognitive Examination (ACE-III). Additionally, there is an impetus to increase the psychological workforce in the NHS through the delivery of the clinical associate psychologist (CAP) profession. In my opinion, the most realistic option of delivering comprehensive cognitive assessments at a large scale is to develop an assessment that can be reliably and robustly delivered by memory nurses and CAPs. Following consultation with memory nurses and neuropsychologists, there are two identified barriers to achieve this: 1) a lack of a standardised training protocol for non-practitioner psychologists on how to deliver cognitive assessments; and 2) a lack of available cognitive assessments that nurses and CAPs can use. This proposal seeks to co-design a nurse or non-practitioner psychologist administered cognitive assessment for preclinical AD, and co-produce a robust training protocol for nurses and allied health professionals to deliver cognitive assessments. If successful, this will contribute to the widespread delivery of early AD diagnosis and treatment through the NHS.

CoCoA-PAD Assessment The long-term aim of the Co-Production and Evaluation of the Computerised Cognitive Assessment for Preclinical Alzheimer's Disease (CoCoA-PAD) project is to develop a validated assessment app which can be used in clinical practice. The CoCoA-PAD project seeks to collate the best experimental neuropsychological tests in the academic literature, and work with a team of stakeholders to co-produce these tests into a viable clinical assessment. In its current format, the CoCoA-PAD assessment is experimental, and none of the subtests have been validated previously. Therefore, the CoCoA-PAD assessment in its current form is not a medical device and does not require Medicines and Healthcare products Regulatory Agency (MHRA) registration.

CoCoA-PAD is comprised of 16 standalone subtests, which are used to generate seven index scores (premorbid intelligence, memory, language and fluency, spatial processing, perceptual discriminability, cognitive control and performance validity). The CoCoA-PAD subtests are organised according to a theorised hierarchical structure. CoCoA-PAD is designed based on an up-to-date understanding of the cognitive neurology of preclinical and prodromal Alzheimer's disease and the temporal ordering of cognitive difficulties. The assessments have been designed to assess the cognitive sequelae of preclinical AD, and four preclinical syndromes.

CoCoA-PAD was designed to achieve two ambitions:

* Maximise Efficiency. CoCoA-PAD includes automatic and timed stimuli presentation, and automated scoring and psychometric calculations. This includes embedded machine learning approaches to maximise classification accuracy without the need for labour intensive scoring procedures. These steps ensure accuracy and reduce clinical time. * Maximise Clinical Value. CoCoA-PAD incorporates many features that are designed to maximise clinical value, which do not exist in conventional neuropsychological assessments. This includes embedded measures of premorbid ability and performance validity, structured observational assessments of validity, functional cognitive disorder, and other cognitive difficulties, e.g., language.

2.3. Co-Production of CoCoA-PAD The CoCoA-PAD assessment battery, and indeed all aspects of this research study, have been developed using a coproduction methodology, as outlined by the National Institute for Health and Care Research (NIHR). The co-production group consists of five older adults (including two with neurological conditions, one from an ethnic minority background, and one with educational deprivation), five assistant psychologists, two memory nurses, and seven clinical psychologists and neuropsychologists. The older adult coproduction group are involved in ensuring the assessment is well tolerated, the instructions are clear, that the test does not unduly discriminate against people with low computer literacy. The assistant psychologist group are involved in coproducing the usability of the assessment for the examiner. The memory nurses are coproducing the clinical observation checklists. The clinical psychologist and neuropsychology group are involved in coproducing the assessment governance and training requirements. All aspects of the assessment have been or are currently being co-produced.

CoCoA-PAD is delivered using two platforms. Some subtests are delivered using the CoCoA-PAD web-based application, including Colour and Object Binding and Location Test (COBALT), Matching Animals Name Exam (MANE), Binding of Allocentric Spatial and Location Transformations (BASALT), Connected Speech, Clock Drawing. The remaining subtests are delivered using Gorilla Experiment Builder, an online platform for designing and running behavioural experiments. Gorilla Experiment Builder provides precise stimulus delivery, secure data collection, and easy remote access, making it well-suited for this research.