Development of a Screening Algorithm for Timely Identification of Patients with Mild Cognitive Impairment and Early Dementia in Home Healthcare

Mild cognitive impairment (MCI) and early-stage dementia (ED) have a significant impact on elderly patients’ quality of life, healthcare utilization, and cost. Despite nationwide efforts for timely diagnosis of MCI and ED (MCI-ED), more than 50% of patients remained underdiagnosed and undertreated. This is mostly due to patients’ inability to recognize early symptoms, limited availability of biomarkers, and clinicians’ insufficient time to assess patients for MCI-ED, particularly for patients admitted to the home healthcare (HHC) setting. This K99/R00 will address barriers to early identification of MCI-ED via the development of an innovative algorithm built on a combination of multiple data streams, including data extracted from electronic health records (EHRs) and audio-recorded patient-clinician verbal communication during routine encounters. Our primary goal is to utilize the routinely generated data in the HHC setting, including OASIS (Outcome and Assessment Information Set - a federally required assessment of patients admitted to HHC), HHC nurses’ notes, and HHC patient-nurse verbal communication to develop an MCI-ED screening algorithm. The long-term training goal is for Dr. Zolnoori to become an independent investigator conducting a program of research dedicated to mitigating the burden of delayed care for patients with MCI-ED by developing low-cost, effective informatics solutions. The solutions will take advantage of easily accessible data generated in clinical encounters and will be built on novel data science methods, particularly speech analysis, the focus of her postdoctoral work. Using exceptional resources available from Columbia University and the Visiting Nurse Service of New York, the K99 phase of this project will focus on gaining essential competencies and skills in theory and practice of speech analysis and cognitive impairment to quantify properties of MCI-ED patients’ verbal communications in interactions with HHC nurses. The R00 phase will focus on the development of a screening algorithm for the early identification of MCI-ED. The specific aims are to 1) model MCI-ED patients’ verbal communications with HHC nurses using an automated speech analysis system; 2) utilize existing natural language processing algorithms to automatically identify MCI-ED related information, including i) clinical symptoms, ii) lifestyle risk factors, and iii) communication deficits from both HHC clinical notes and patient-nurse verbal communication; and 3) develop a sensitive screening algorithm to identify HHC patients with MCI-ED. To accomplish the research aims and training goals, an interdisciplinary team of scientists with expertise in speech analysis, cognitive impairment, HHC services, biostatistics, and career development mentorship has been assembled. This project is significant because this algorithm will be built on easily accessible data streams generated during routine patient-nurse encounters. The algorithm has a strong potential to be integrated into HHC clinical workflow to raise clinician’s attention to the patient’s cognitive functioning for further evaluation and development of proper interventions to reduce the risk of negative outcomes.