Analysis Of Digital Health Data

Analysis Of Digital Health Data: From Research Lab To App Store Presented by: Prof. Mulvenna Date: 12 May 2021 Outline - Inspired reminiscence study - Inspired reminiscence app - Discussion InspireD Reminiscence Study - Cluster 1: The hooked adopters. A single user (3.6%) who fully adopted the system. They had 7.2 times more interactions than their carer. The person with dementia obsessively used the app, the carer showed a normal amount of usage. - Cluster 2: Typical users. The plurality of users (43%) fall into this cluster. This user has 1.7 times more interactions with the app than their carer. This dyad uses the app 15% of days in a month. - Cluster 3: Disengaged irregular users. 25% of users. The person with dementia had 25% and 25% fewer interactions than their carer, while the carer has fewer interactions than other carers in all other clusters - Cluster 4: The well supported users, 29% of users. These users had 36% fewer interactions with the app than their carers. The carers were enthusiastic and had more interactions than other carers in all other clusters but struggled get people with dementia to the same engagement level. Ecological momentary assessment for Inspired study - Also called Experience Sampling Method - Capture real-time data in an individual's natural environment through repeated sampling, using: Psychometric scales, Open-ended questions, or anything else used to assess an individual's condition in that place and time. - Provides a high degree of ecological validity as they study people as they go about day-to day life. - EMA requires participants to respond to questions 'in the moment', so it avoids recall bias which makes it a useful tool for people living with dementia where their memory is impaired. - A total of 5 different EMA questions from the mutuality scale were presented to people living with dementia and their carers Number Q1 Q2 Q3 Q4 Q5 Question How attached are you (partners name}? How much do the two of you laugh together? How much do you confide in {partners name)? How much do you enjoy sharing past experiences with (partners name)? How much do you like to sit and talk with {partners name}? Ecological Momentary Assessment Questions Presented And Dismissed - Overall dismissal rate for questions asked during the trial period was 30.9%. Hence, almost 70% of questions were answered - People living with dementia used the app more in the trial period than carers but despite this had a lower dismissal rate for questions (26.8%) compared to their carers (36.8%) Question Q1 Q2 Q3 Q4 Q5 Presented n (%) 197 (26.09) 131 (17.35) 116 (15.36) 165 (21.85) 146 (19.34) Dismissed n (%) 48 (20.60) 39 (16.74) 38 (16.74) 54 (23.18) 53 (22.75) p-value <0.001 <0.001 <0.001 <0.001 <0.001 Inspired Reminiscence Study - Future design recommendations for digital health apps incorporating EMA Aspect of Recommendation EMA Questions The actual EMA questions which are presented should be informed by user asked group co-design (e.g., user group made up of people living with dementia and their carers). First EMA questions should only be presented after a period of 2-3 weeks to give question users time to get comfortable using the app. Frequency No more than one question presented to users per day. Hour of day Questions presented at times when users were more likely to answer rather than dismiss questions i.e., after 8pm in the evening. If user is trying to complete a task within the app such as uploading content to the app, then a question should not be presented. Current task InspireD Reminiscence App Question "How much are you enjoying reminiscing today?" "How do you feel today while reminiscing with {carer name/ person with dementia name}?" "Do you find reminiscing with this app useful?" Response options Frequency of presentation Scale from 1 (not at all)-5 (a Every 2 weeks great deal) or dismiss Happy, neutral sad, frustrated Every 4 weeks or dismiss Extremely, very, moderately, Every 6 months slightly, not at all or dismiss New & Emerging Concepts In Digital Health - Data in the wild - 'found' data - Real world data (RWD) / Real world evidence - Ecological momentary assessment (EMA) - Dynamic consent - Ethical gamification - Digital phenotyping Issues In Digital Health - Blending services - Waiting list 'optimization' - Data ownership - Access by healthcare professionals - Managing safety - Digital exclusion In Summary - Use of data and analysis using artificial intelligence and machine learning is growing - Many national public health organizations are exploring how to use digital technologies such as health apps and cloud-based services for self-management and/or treatment - Logging users' interactions allows for greater insight into user needs and provides ideas for improving these digital interventions, for example through enhanced personalization - Risks in using this data are manifold - However, on balance it may be argued that public health (services) benefit since data can be automatically and hence cost-effectively collected - Such data may facilitate new ways for digital epidemiological analyses and provide data to inform health policies - Public health organizations can and do promote health apps - Digital phenotyping analysis using machine learning and artificial intelligence may be taken up by these organizations - Clear need for guidelines for ownership of digital phenotype data and information derived from these data