Ms. Bucci is a research intern at BIDMC and completing a degree in in Biological Aspects of Public Health. Dr. Torous is Clinical Fellow in Psychiatry, Beth Israel Deaconess Medical Center (BIDMC), Harvard Medical School (HMS) in Boston. Dr. Press is Associate Professor of Neurology, BIDMC and HMS. Dr. Lim is Assistant Professor of neurology, BIDMC and HMS.
Alzheimer disease (AD) is a devastating illness that impacts both patients and their families. Symptoms vary from person to person and fluctuate with time and across environments, making diagnosis a challenge.
In-clinic neuropsychological testing is the gold standard but cannot account for fluctuations in performance and the anxiety of being tested in a clinic. Despite research efforts, an affordable or straightforward AD diagnostic test does not exist.
Fortunately, several screening tools have been established to help clinicians identify cognitive impairment. The Montreal Cognitive Assessment\(MoCA), theMini Mental Status Exam (MMSE), and the Addenbrooke Cognitive Examination (ACE) (among others) are used to detect for evidence of cognitive impairment. But because these clinical scales are in often administered in "paper" form and in the clinic, their sensitivity and specificity are limited.
In an ongoing study, researchers at Beth Israel Deaconess Medical Center (BIDMC) in Boston are using computer adaptive testing to explore and identify aspects of cognitive performance in a series of computerized neuropsychological tests that best correlate with neurocognitive disease states, ie, mild cognitive impairment (MCI) and early stage AD. Computerized applications will check for spatial attention span, facial recognition, visual search, executive functioning, associative memory, naming, and attention.
Although these computerized tests are analogous to simple pen-and-paper tasks that help spot early warning signs of AD, they have a number of potential advantages. Apps are designed to fine-tune and analyze a patient’s responses in order to provide additional and personalized information. For example, apps can present tests in varying order with varying degrees of difficulty and measure variables such as response time, number of response changes, and other relevant measures that have previously been difficult to quantify.
With more data collection and analysis, machine learning can be utilized to narrow down and eliminate redundancy by identifying the optimal parameters for better decision-making regarding diagnosis. Computer analysis will identify the and customize a battery of tests to maximize sensitivity and specificity of differentiating patient populations.
Beyond personalizing assessments, smartphones can capture information about one’s immediate surroundings and habits (eg, how social a person has been recently; how often a patient has left the house). Such data may give clinicians the ability to discern changes in a patient's cognition or symptoms.
Studying neuropsychiatric illnesses with these technologies will enable BIDMC researchers to work toward developing a scalable smartphone tool to accurately and efficiently measure cognitive deficits and other factors associated with dementia. This approach is affordable and provides a multifaceted perspective of a patient’s cognitive abilities, capturing not only traditional testing metrics but also practical information.
Implications for neurologists
New digital technologies will enable personalized neuropsychiatric testing that can be customized to each patient’s unique condition and abilities. These same digital technologies can also capture novel environmental data, including information on mobility, social connections, sleep patterns, and exercise. Linking the assessment and environmental data from these digital technologies will offer a new window into neurologic and behavioral manifestations. While machine learning algorithms may soon help spot early risk or warning signs, even today clinicians can learn much about how exercise, sleep, and activity influence clinical presentations.
This smartphone application collects active (cognitive tests) and passive (environmental surroundings) data. The results from the tests will be used to establish a database of performance results of patients who fall within the normal, MCI, AD, and non-AD categories of dementia in their personal, natural surroundings. The analyses will identify which tests best evaluate mental status. Patients will be scored on a scale based on their performance that reflects their level of cognition. Any redundant testing will be omitted. This information can provide insights that will help create an adaptive program that relies on the most efficient series of tests for identifying dementia and detecting decline over time.