An interactive tool to personalise 24-hour activity, sitting and sleep prescription for optimal health outcomes

An interactive tool to personalise 24-hour activity, sitting and sleep prescription for optimal health outcomes | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article An interactive tool to personalise 24-hour activity, sitting and sleep prescription for optimal health outcomes Maddison L Mellow, Tyman E Stanford, Timothy Olds, Aaron Miatke, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6897341/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 17 Mar, 2026 Read the published version in npj Digital Medicine → Version 1 posted 9 You are reading this latest preprint version Abstract Personalised interventions which optimise the balance of physical activity (PA), sleep and sedentary behaviour (i.e., time use) in the 24-hr day may be more effective than one-size-fits-all approaches. We present an interactive app to personalise 24-hr time use based on individuals’ health and sociodemographic characteristics. Analyses used cross-sectional data from 53,057 UK Biobank participants. Average daily time use was measured using 7-day accelerometry data and expressed as a 24-hr composition using isometric log-ratio transformation. Five cognitive composites were derived from web-based tests. Regularized linear regression examined the relationship between 24-hr time-use composition and cognition, with sociodemographic and health characteristics as additional predictors. Model estimates were used to estimate optimized cognition based on the interaction of 24-hr time-use composition and personal characteristics. Our ‘ideal day’ app delivers personalised 24-hr time-use recommendations tailored to individual characteristics. We demonstrate that personalisation of time-use interventions can be achieved in real time using open-source software. Health sciences/Risk factors Health sciences/Health care/Disease prevention/Lifestyle modification Health sciences/Health care/Disease prevention/Preventive medicine Health sciences/Risk factors Health sciences/Health care/Disease prevention/Lifestyle modification Health sciences/Health care/Disease prevention/Preventive medicine Health sciences/Risk factors Health sciences/Health care/Disease prevention/Lifestyle modification Health sciences/Health care/Disease prevention/Preventive medicine time use personalised medicine digital tool cognition Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Personalised (or ‘precision’) medicine approaches to healthcare move beyond the ‘one size fits all’ approach towards individualised prevention and treatment of illness 1 . Rather than using population averages to inform therapy, treatment and behaviour change, personalised approaches consider individual characteristics beyond age and sex, including genetic, environmental and lifestyle factors 1 . Whilst this approach is increasingly projected to be ‘the future’ of drug development (including applications in cancer treatment, pharmacogenomics and immunotherapy), lifestyle interventions which aim to optimise the balance of physical activity (PA), sleep and sedentary behaviours (SB) in the 24-hr day are seldom personalised to the individual. Instead, most approaches administer a universal “dose” (e.g., perform 20 minutes of walking, 3 times per week), or attempt to move all individuals closer to PA, sleep and/or SB guidelines which are also generalised. An important but overlooked consideration is that the ideal 24-hr time-use composition (i.e., the optimal balance of PA, sleep and SB in the 24-hr day) for health outcomes may vary between individuals based on their current time use, health status, or sociodemographic “profile”. That is, the unique combination of factors such as age, sex, mobility, health conditions, and socioeconomic status, as well as their current levels of PA, sleep and SB, may already have participants closer to, or further away from, their optimal day at the beginning of an intervention. Consequently, administering a universal intervention to a sample of individuals may not result in uniform intervention effects. Several key theories and methodologies have highlighted the potential value of personalising 24-hr time-use interventions for health. For example, the “Sweet-Spot” hypothesis postulated by Holtermann and colleagues 2 proposed that the optimal combination of PA, SB, and sleep for health likely differs between adults in physically active versus sedentary occupations. They argue that the generalised advice to “sit less and move more” may actually take adults who work in physically active occupations further away from their optimal 24-hr day, as their optimal day requires more time for recovery (SBs and sleep) than it does for those in sedentary occupations 2 . Similarly, the “Goldilocks Day” approach by Dumuid et al. 3 demonstrates that different health outcomes have different optimal days, thus determining someone’s optimal day will require personalisation to their health priorities. For example, the optimal balance of PA, sleep, and SB looks different for cognitive health than it does for physical or mental health, meaning the optimal day is also dependent on how someone ranks these health domains (i.e., whether cognitive, physical or mental health are of higher importance) 3 , 4 . Taken together, the Sweet-Spot and Goldilocks Day principles may also apply to the current health and sociodemographic profile of the individual. Society today faces several pressing health challenges which may benefit from personalised prevention approaches, including dementia 5 . There are at least 14 established modifiable risk factors for dementia, many of which are potentially bi-directionally related PA, sleep and SB, such as obesity and depression. Additionally, physical inactivity is a risk factor for dementia 6 . It is likely that the optimal 24-hr time-use composition for minimising dementia risk for a 50-year-old male with high school-level education, hypertension, and a history of smoking, all of which have been identified as modifiable risk factors for dementia, looks different in comparison to another 50-year-old male with tertiary education, hypertension, and no smoking history. Despite the two individuals belonging to the same age group and both having a diagnosis of hypertension, the additional differences in modifiable dementia risk factors (education and smoking history) may change the relative impact that time-use behaviours have on dementia risk. That is, a higher amount of PA or different balance of time-use behaviours may be needed to offset the negative impacts of smoking and low education. It is possible to compute the optimal 24-hr time-use composition for an outcome of interest (e.g., cognitive VS physical health) specifically for the average individual with given personal characteristics (e.g., age, sex) based on statistical models 7 . However, the challenge is how to effectively communicate behaviour change recommendations which are tailored to an individual participant in real time, such as in an intervention study. One solution may be via interactive interfaces which allow the user to set their own descriptors (e.g., age, sex, health status, current time use) and view the optimal day for their chosen health outcome. The Your Best Day interactive app developed by Dumuid and colleagues 4 achieves this: users select their sex, age, and current time use, and can then use interactive sliders to visualise the predicted effects of changing their time use on body fat percentage, psychosocial wellbeing, and academic performance in adolescents. Applying the principles of the Your Best Day app, the aim of the current study was to develop an interactive user interface to individualise 24-hr time-use recommendations for the Small Steps Study, a randomized controlled trial aiming to improve 24-hr time use for cognitive health in community-dwelling older adults 8 . Using the UK Biobank 9 as our normative sample, we firstly aimed to identify the optimal 24-hr time-use composition for cognitive function outcomes across a range of sociodemographic and health profiles (i.e., across combinations of a range of modifiable and non-modifiable risk factors for dementia). Secondly, we developed a user interface (housed as an R Shiny app 10 ) which allows users to input information about dementia risk factors (i.e., create their ‘profile’), and view the optimal 24-hr day for cognitive function based on their profile. Methods Study design and participants Data used in this cross-sectional study were from the UK Biobank (application no. 62254), a large prospective cohort study of ~ 500,000 participants aged > 40 years at time of recruitment in 2010 9 . The sample for this study were limited to UK Biobank participants who met the following criteria: accelerometry data were marked as ‘valid’; accelerometry data did not have a data problem indicator; at least four of five cognitive tests included in this study were completed; a diagnosis of dementia, organic amnesic syndrome, delirium, systemic atrophy affecting the central nervous system, extrapyramidal or movement disorder, other degenerative disease of the central nervous system, demyelinating disease of the central nervous system or blindness was not recorded prior to either completing the cognitive tests or wearing the accelerometer; and no indication that participant was unable to walk. A complete list of inclusion and exclusion criteria are displayed in Supplementary File 1. Study measures Sociodemographic and health factors A range of sociodemographic and health variables which have been identified as modifiable dementia risk factors as per the 2020 Lancet commission report 11 were extracted from the assessment centre data (initial visit) for consideration in our model selection procedures. A detailed list of included variables, their UK Biobank field codes and data re-classification protocols (where applicable) is available in Supplementary File 2. As the final behaviour change tool (detailed below) was initially intended for use in an Australian context, where necessary some variables (e.g., education/highest qualification) were re-levelled to align with Australian classifications. Final variables included age, sex, ethnicity (levels = “White”, ”non-White”), highest qualification (levels= “high school”, “Certificate III/Diploma”, “College/University”, “Other professional qualification”), alcohol consumption (levels= “sometimes/never”, “often/very often”), smoking status (levels=“never”, “previous”, “current”), history of depression (levels=“yes”, “no”), hearing difficulty (levels=“yes”, “no”), history of hypertension (levels=“yes”, “no”), history of traumatic brain injury (levels=”yes”, “no”), type 2 diabetes status (levels=“yes”, “no”), body mass index (BMI; levels=”overweight”, “not overweight”), and social isolation (“yes”, “no”). For all variables, missing data were recoded as ‘unknown’ (i.e., some dichotomous variables had three levels: yes, no, unknown). Twenty-four-hour time-use composition The proportions of daily time spent in sleep, SB, light intensity PA (LPA) and moderate-vigorous intensity PA (MVPA) were obtained from the derived accelerometry data 12 within the UK Biobank (field codes are presented in Supplementary File 2). Zero values in any compositional parts (sleep, SB, LPA, MVPA) were imputed assuming censored data (values below the threshold of detection) using a linear model-based imputation of log ratios of the components, iteratively improved via the Expectation-Maximisation algorithm ( lrEM function in the zCompositions R package 13 , 14 ), with the remaining behaviours reduced by the small, imputed value proportionally. Additionally, to mitigate the undue influence of extreme values, the 0.5% of the empirical distribution component percentiles at both tails were truncated to their respective 0.5 and 99.5 percentile values (with proportional shrinking/expansion of the other components to ensure a 24-hr day). The four compositional parts (sleep, SB, LPA, MVPA) were expressed as a set of three isometric log-ratio ( ilr ) coordinates using a sequential binary partition ilr base representing the following quantities (ignoring normalizing constants): the first ilr represented the log-ratio of one behaviour (e.g., sleep) to the geometric mean of the remaining three behaviours (e.g., SB, LPA, MVPA); the second ilr excluded sleep, and represented the log-ratio of the next behaviour in the set (e.g., SB) to the geometric mean of the remaining two behaviours (LPA and MVPA); and the final (third) ilr represented the log-ratio of LPA to MVPA). Together, the three ilr coordinates represent 24-hr time-use composition as a set of linearly independent predictors in regression models. Cognitive function Several web-based cognitive tests collected during the UK Biobank online follow-up were included in this study, including Numeric Memory, Pairs Matching, Fluid Intelligence, Trail Making (A and B), and Symbol Digit Substitution. Supplementary File 2 presents the individual outcome measures extracted from these tests and how they were combined into composite scores. We created four cognitive domain composites (memory, reasoning, executive function and processing speed) using groupings based on a previous UK Biobank study 15 . To create the composites, we first reverse-scored Trail Making A, Trail Making B and Pairs Matching outcomes so that for all included measures, higher scores represented better performance. We then undertook the following steps to normalise the distributions of the measures: Trail Making A and B scores were truncated at 300 seconds and log-transformed; Pairs Matching scores were truncated at -7 seconds; and Symbol Digit Substitution scores were truncated at 31. Individual measures were z-scored using age and sex standardisation (age = 65 years; sex = male, female). Finally, composite scores were created by averaging z-scores (where applicable), and an overall composite score (global cognition) was created by averaging the four z-scores. Supplementary File 2 outlines test outcomes and their corresponding cognitive domains. Data analysis All analyses were conducted in R Statistical Software (version 4.3.1 16 ). Full R codes are published on GitHub ( https://github.com/tystan/ukbb-cog-lasso ). Model selection by regularization Regularized linear regression models were fit with each of the five cognitive function measures as the outcome variable, with the following candidate predictors prior to shrinkage: continuous and discrete main effect predictors (time-use log-ratios and all sociodemographic and health factors); all pairwise main effect interactions; and additional squared continuous main effects (i.e., polynomial terms degree 2). To reduce non-informative candidate predictors and avoid overfitting, Least Absolute Shrinkage and Selection Operator (LASSO) regularization was used. LASSO regression is a coefficient shrinkage method which aims to produce a parsimonious model based on a subset of the potential predictors that are interpretable and related to the outcome, by penalising the sum of the absolute values of a model’s variable coefficients, forcing some coefficients to exactly zero and therefore to become omitted from the model 17 . However, because the ilr variables were required to be treated as analytically inseparable variables, in addition to having many categorical variables with more than two levels, we specifically implemented group LASSO models 18 using the grpreg R package 19 . The group (or “block”) extension of LASSO regression allows multiple predictor variables to be treated as an inseparable group – the shrinkage is applied block-wise, instead of individually, on the prespecified groupings of variables. Such an approach is required to ensure the invariance of the specific ilr basis chosen, but can also be useful to regularize categorical variables with three or more levels (resulting in two or more contrast dummy coded variables) as an inseparable group that the model would otherwise not know are intrinsically related. By applying a group-specific penalty, the coefficients in the group are both shrunk at a group level (e.g., the group of categorical level contrasts) as well as individually 20 . To this end, the ilrs were also treated as a group after centring and an angle preserving rotation was applied (to mitigate the issue of computing different shrinkage results dependent on the ilr basis chosen 21 ). Similarly, the higher order terms were specified as groups, incorporating each crossed (or squared) sub-term in the case at least one categorical variable or ilr in the interaction (or squared) term was retained after shrinkage. Despite group-specific penalties being sought, the model formulation can be equivalently re-expressed with a single penalisation parameter, λ, with the associated sum of the product of the square root of the group size and the Euclidean norm of the group variable’s coefficients 18 . To determine an empirically derived optimal penalty value, λ*, of the shrinkage procedure, we performed a ten-fold cross-validation over a grid of potential λ values seeking the one that minimises the model’s predictive mean squared error (MSE) of the model predicted values compared to observed values. Supplementary File 3 displays the derived λ* and corresponding MSE for each of the cognitive variables. Identifying ‘optimal’ days for cognition Due to the known mediating effects between ilrs (time use) and age, sex and BMI, the data were stratified into 8 mutually exclusive groups based on the following categories: age (< 65 years, ≥ 65 years), sex (male, female), and BMI (with obesity = ≥ 30, without obesity = < 30). Extending the methods outlined in Dumuid et al 3 , the models described previously were used to predict cognition for a “time-use footprint” that was considered feasible and realistic for each of the 8 population strata. To restrict the time-use footprint to realistic values, and to avoid extrapolating from the highest density of sampled time-use data, we used a grid of all possible time-use compositions with 5-minutes spacings constrained within the empirical distributional (multivariate Gaussian) quantiles of the strata-specific sampled data in the ilr -space (“constrained ellipsoid fencing”, a novel method for which codes are available at https://github.com/tystan/ukbb-cog-lasso ). In brief, the empirically estimated multivariate normal 80th percentile contour limits were used after assessment of marginal (univariate), pairwise, and multivariate normality by way of visual checks (where possible in lower dimensions) and cumulative quantile plots. Figure 1 demonstrates, for one of the stratum, the relationships between the sampled time-use compositions, where points are either classified as within or outside the ilr -derived ellipsoid fencing when transformed back to the compositional scale and presented in the 4-simplex tetrahedron (constrained, constant sum space) in which they reside. Approximately 80% of the points are within the ellipsoid fencing as is expected if the ( ilr ) data are from an approximately multivariate Gaussian distribution. Optimal cognitive function was operationalized as the top 5% of the cognitive scores predicted by the models over the given constrained grid and specific person inputs 7 . The optimal 24-hr day for cognitive function was conceptualised as the compositional mean (geometric mean, adjusted to sum to 1440 minutes) of the time-use compositions (sleep, sedentary time, light physical activity, MVPA) associated with the top 5% of predictions 7 . As models for the relationship between time-use composition and cognitive function also included interactions between various sociodemographic and health factors, the estimated optimal day varied depending on the values on these covariates. The optimal day for the “mean” or “average” person in each of the strata is presented in the main manuscript, but for further personalisation of the optimal day estimate according to the value of the covariates, real-time estimation applying the specified covariate values to the model coefficients is made possible via our Shiny app interface (detailed below). Developing the interactive user interface The R “Shiny” package 10 was used to program an interactive interface (app) that can be freely accessed in a web browser via a URL, without the need to install R or any other additional software. Our Shiny app has three components (i.e. R scripts) that communicate with each other: (1) the user interface (ui.R), which determines the appearance of the app and how the user enters information; (2) the server (server.R), which takes input provided by the user interface, sends it for computation, and then returns results back to the user interface; and (3) the global script (global.R), which defines the variables and functions accessible for both the user interface and server. The R scripts used for our Shiny app can be found in a separate GitHub repository ( https://github.com/tystan/ideal-day ). The LASSO regression coefficients from the compositional models were extracted for the app’s global script. To determine the optimal day for user-specified covariates (i.e., their ‘profile’), the app’s user interface requests the user’s sociodemographic and health details. From this, the app matches the user to one of the eight (age, sex, and BMI) defined strata. It then uses the user’s remaining inputted sociodemographic and health variables, and every possible time-use composition within the feasible range for the user’s stratum, to predict cognition. The app then extracts the top 5% of predictions and calculates the mean time-use composition associated with these top 5% of predictions, i.e., the optimal day. The workflow underpinning our Shiny app is displayed in Supplementary File 4. The user is required to enter their current 24-hr time-use composition. In the case of the Small Steps study 8 for which this app was created, these values are derived from wrist-worn accelerometers worn by the participant prior to their initial visit (i.e., 7-day average of time spent in sleep, SB, LPA and MVPA from Fitbit watch). Following the user input of their current day, a bar plot is generated which displays the user’s current 24-hr day (mins/day of sleep, SB, LPA and MVPA) compared to their ‘optimal’ 24-hr day. Feedback is provided about the changes required to reach that day from their current day (e.g., -10 minutes sleep, -20 minutes sedentary behaviour, 0 minutes light physical activity (i.e., no change), + 30 minutes MVPA). The appearance and functionality (e.g., colours, font sizing, wording) of the Shiny app was informed by older adults from the general population (n = 8) through a series of co-design workshops conducted as part of the Small Steps study (Ethics ID 205377). A full description of the co-design process can be viewed elsewhere 22 . Results Participant demographics A sample of 53,057 participants from the UK Biobank were included in this study (Table 1 ). Participants were 62 ± 8 years of age, mostly female (57%), White (97%) and had predominantly college/university (47%) qualifications. The health profile of participants varied across modifiable dementia risk factors, with the most common risk factors including history of hypertension (24%), hearing difficulty (24%), regular alcohol consumption (23%) and history of depression (20%). On average, participants spent most of their 24-hr day in SB (9.6hrs, 40.0%) or sleep (9.0hrs, 37.5%), whereas active behaviours made up < 25% of the day (LPA: 4.8hrs, 20.0%; MVPA: 0.5hrs, 2%). The number of participants per strata ranged from 2064 to 15303 (Supplementary File 5). Table 1 Descriptive characteristics of sample Characteristic Level Overall sample N 53,057 Age (years) 62 (8) Sex (%) Female 30,091 (57%) Male 22,966 (43%) 24-hr time-use composition Sleep 542.7 SB 576.9 LPA 291.2 MVPA 29.2 BMI 26.6 (4.5) Highest qualification High school 10,637 (20%) Certificate III or Diploma 5,805 (11%) College/University 24,742 (47%) Other professional qualification 8,092 (15%) Unknown 3,781 (7.1%) Depression Yes 10,734 (20%) No 9,141 (17%) Unknown 33,182 (63%) Diabetes Yes 1,665 (3.2%) No 51,295 (97%) Unknown 97 (0.2%) Hearing difficulty Yes 12,713 (24%) No 38,221 (72%) Unknown 2,123 (4.0%) Hypertension Yes 12,595 (24%) No 40,390 (76%) Unknown 72 (0.1%) Social isolation Yes 7,851 (15%) No 44,584 (84%) Unknown 622 (1.2%) Smoking status Current 3,447 (6.5%) Previous 18,960 (36%) Never 30,532 (58%) Unknown 118 (0.2%) Alcohol consumption (frequency) Often/very often 12,415 (23%) Sometimes/never 26,641 (50%) Unknown 14,001 (26%) History of traumatic brain injury 106 (0.2%) Ethnicity White 51,577 (97%) Non-white 1,480 (2.8%) Days between accelerometry and cognitive testing 52 (232) Note. Mean and standard deviation (SD) are presented for continuous variables, and count and proportion (n(%)) are presented for categorical variables for the overall sample. BMI = body mass index; SB = sedentary behaviour; LPA = light intensity physical activity; MVPA = moderate-to-vigorous intensity physical activity. A positive value for days between accelerometry and cognitive testing indicates that accelerometry measures occurred later than cognitive testing. ***INSERT Table 1 NEAR HERE (at bottom of manuscript)*** Model selection Model coefficients for terms containing ilrs (i.e., main effects, higher order terms or interactions with time use) are displayed in Fig. 2 . Across all five cognitive outcomes, main effects were retained for ilr s, as well as interactions between ilr s and ethnicity, and ilr s and alcohol consumption. The remaining interaction terms containing ilrs were retained non-uniformly across cognitive outcomes. We note that predictor variables in group LASSO regression, and regularized regression more generally, are block standardised to zero vector mean and identity unit covariance, so when used to predict the (unit variance scaled) outcome, the predictors’ strength of association with the outcome are directly comparable by magnitude of their associated coefficient. The complete model coefficient output (main effects, higher order terms and interaction terms for all variables) can be viewed in our GitHub repository ( https://github.com/tystan/ukbb-cog-lasso ). Optimal days for cognitive outcomes The optimal 24-hour day for cognitive function varied across strata, across cognitive outcomes, and was further modified by health and sociodemographic characteristics. Optimal 24-hr time use varies between age, sex, and BMI-defined strata Figure 3 displays the estimated optimal 24-hr time-use compositions for each cognitive domain (across four compositional parts), and corresponding cognitive responses (cognitive domain z-score) across the eight age, sex, and BMI-defined strata. Across almost all groups, the optimal amount of SB and MVPA for cognitive function was greater than the strata mean, whereas the optimal amount of sleep and LPA was lower than the strata mean. For example, the average time-use composition for females aged < 65 years without obesity (top row of figure) was 542 minutes of sleep (~ 9hrs), 549 minutes of SB (~ 9.2hr), 318 minutes of LPA (~ 5.3hr), and 30 min of MVPA. Comparatively, the optimal day for global cognition within this stratum was 480 minutes of sleep (~ 8hrs), 705 minutes of SB (~ 11.5hrs), 214 minutes of LPA (~ 3.5hrs), and 39 minutes of MVPA. Thus, to achieve an optimal 24-hr day for global cognition, on average, this stratum required approximately 60 minutes less sleep, 100 minutes less LPA, 150 minutes more SB, and 10 minutes more MVPA than the stratum’s mean composition. Across strata, the optimal amount of MVPA for cognitive function varied the most by obesity status, whereby strata with obesity consistently required lower MVPA for their optimal day composition compared to those without obesity. Within the same age and sex classifications, differences in optimal MVPA by obesity status for global cognition ranged from ~ 12 minutes (females aged > 65 yrs) to ~ 18 minutes (males aged > 65 years). We note that the variability by obesity status is related to the constrained (feasible) time-use footprint containing lower MVPA in strata with VS without obesity – in other words, rather than individuals with obesity needing less MVPA to benefit cognition, their feasible limits of MVPA are lower due to the relationship between MVPA and BMI (see Supplementary File 6 for comparison of feasible limits of time-use behaviours across strata). Less consistent patterns were observed across age and sex classifications, and across other time-use behaviours. Optimal 24-hr time use for cognition varies between cognitive domains The optimal day for cognitive function varied considerably across cognitive domains (Fig. 3 ). For example, within one stratum (e.g., females, aged < 65 years, without obesity) the optimal amount of each time-use behaviour for different cognitive domains varied as follows: optimal sleep varied from 466 minutes (7.8 hrs, memory) to 492 minutes (8.2 hrs, processing speed); optimal SB varied from 695 minutes (11.6 hrs, reasoning) to 705 minutes (11.8 hrs, global cognition); optimal LPA varied from 206 minutes (3.4 hrs, processing speed) to 227 minutes (3.8hrs, memory); and optimal MVPA varied from 35 minutes (processing speed) to 51 minutes (reasoning). Optimal 24-hr time use is further modified by user inputs In addition to age, sex, and BMI, other health and sociodemographic factors altered the optimal day for cognitive function to varying extents. Notably, presence/absence of TBI history had the strongest influence on optimal days, whereas additional health and sociodemographic factors (e.g., hypertension, smoking, alcohol consumption) had less impact on the optimal day within strata. To demonstrate using a practical example, consider ‘Person A’ who has the following characteristics: female, aged < 65 years, without obesity, college/university education, and no history of any health and sociodemographic factors (i.e., no hypertension, smoking, diabetes, depression, TBI, hearing loss, social isolation, or high alcohol consumption), with a current time-use composition of 7.8hrs sleep, 13.0hrs SB, 3.0hrs LPA, and 0.2hrs MVPA. To achieve their best day for global cognition, Person A would be recommended to increase their sleep (+ 0.2hr), LPA (+ 0.6hrs), and MVPA (+ 0.5hrs), and decrease their SB (-1.2hr). In comparison, Person B has the same baseline time use and belongs to the same stratum, but instead has high school education, history of hypertension, current smoking, diabetes, depressive symptoms, hearing loss, social isolation and frequent alcohol consumption. Despite a vastly different profile of health and sociodemographic characteristics, the recommended changes for Person B are very similar to Person A: increase their sleep (+ 0.3 hr), LPA (+ 0.5hr), and MVPA (+ 0.5hr), and decrease their SB (-1.3hrs). Notably, changing Person B’s characteristics to also include a history of TBI dramatically changes the recommendations: increase LPA (+ 3.3hrs) and MVPA (+ 0.2hrs) and decrease sleep (-0.4hrs) and SB (-3.1hrs). It is important to note that only 0.2% of the entire sample reported history of TBI (using the self-report variable), which may reduce confidence in these relationships. Interactive user interface for ‘ideal day’ personalisation The ‘ideal day’ interactive personalisation tool can be viewed at https://arena2024.shinyapps.io/ideal-day and the underlying code freely accessed on GitHub ( https://github.com/tystan/ideal-day ). The tool is divided stepwise into six main tabs. First, participants can select the cognitive outcome they are interested in predicting for (global cognition, memory, processing speed, executive function, or reasoning). In the ‘Demographics’ tab, the user is asked to input their current age (years), current weight (kg) and height (cm) using free-text boxes, as well as their sex and highest qualification from pre-defined options. The ‘Health’ tab then asks the user a series of multiple-choice questions (with mostly ‘yes’, ‘no’, or ‘unknown’ response options) about modifiable dementia risk factors including history of hypertension, type 2 diabetes, depression, social isolation, hearing loss, TBI, alcohol consumption and tobacco smoking. Finally, users are required to enter their current time use (‘Time-use’ tab) using free text boxes, entering the number of hours per day they spend in sleep, sitting, light physical activity, and moderate-vigorous physical activity. Using these inputs, the tool then displays the user’s current day (i.e., current time-use composition) next to their ‘optimal day’ in the ‘Ideal day’ tab. We note that the use of ‘ideal’ rather than ‘optimal’ was chosen for the Small Steps study app, in response to preferences indicated during the co-design process 22 . In the example displayed in Fig. 4 , the user is interested in determining the ideal day for their global cognition. Their current time use is 7.8 h sleep, 13 h sedentary behaviour, 3 h LPA and 0.2 h MVPA per day, and based on the optimisation analysis, to achieve their personalised ‘ideal day’ the user is advised to increase time in sleep, LPA and MVPA, and decrease time in sitting. As the Small Steps intervention aims to help participants make small, beneficial changes in behaviour which suit their preferences, needs and constraints, the final component of the Shiny app (the ‘Small Steps’ tab) dynamically reports on the whether the relative direction from current time-use to the selected change in time-use (e.g., increasing 10 min of MVPA –and decreasing sleep by 10 min of sleep) is aligned with the theoretical direction of current time-use to the optimal time-use. This feedback was operationalised using a traffic light system, where green lights indicated the proposed time-use change is moving towards the ideal day, and red lights indicated moving away from the ideal day. The methods underpinning this feature are beyond the scope of the current paper, and will be described elsewhere. Discussion This study described the development of a novel interactive user interface which could be used to generate personalised behaviour change recommendations for individuals in real time. In our proof-of-concept example, we showed that the ‘ideal day’ (i.e., the optimal balance of sleep, SB and PA in the 24-hr day, or ’24-hr time-use composition’) for cognitive outcomes differed depending on the health and sociodemographic profile (i.e., modifiable dementia risk factor profile) of participants in a sub-sample from the UK Biobank. We found that 24-hr time-use composition was associated with all five cognitive outcomes (global cognition, memory, processing speed, executive function and reasoning), and that the optimal day for cognitive function varied across strata, and across cognitive domains. Within each of the stratum, the relationship between 24-hr time-use composition and cognitive performance was altered by other health and sociodemographic characteristics. The optimal day (target durations of daily activities) generated by our interactive app varied depending on user inputs regarding their demographic and health characteristics, These potentially substantial differences in optimal day predictions demonstrate the importance of personalising 24-hr time-use interventions for health outcomes (e.g., cognitive health in older adults). Comparison with previous methodological approaches Our study builds on the formative work of recent studies that have published “ideal”, “optimal”, or “Goldilocks” days for a single health outcome, or population 3,7,23 . We extend this work in five important ways, made possible by the large underpinning dataset. First, we developed a sophisticated data-driven model-selection procedure using group LASSO regression that considered all main effect candidate predictors, interactions and second-order polynomial terms. This allowed us to consider more complicated multiplicative relationships between time use, health and sociodemographic characteristics (e.g., modifiable dementia risk factors) without model overfitting. The group LASSO method overcomes the violation of the assumption of invariance under the choice of ilr transformation, which would occur if a single log-ratio is retained while others are discarded in standard LASSO regression 24 . Second, to address the same issue of invariance under alternate log-ratio transformations, we implemented a multivariate scaling method for the log-ratios. Third, we stratified our analyses by key characteristics (age, sex, and BMI) which showed strong interactions with the time-use log-ratios, ensuring differences in associations between these strata would be reflected in our final optimal day predictions. Fourth, to ensure our models best represented the empirical relationships present in the data, we tested for all pairwise interactions between included predictors, and for non-linear relationships. Fifth, we present a new method (constrained ellipsoid fencing) to improve how the empirical time-use footprint is selected for prediction. Previous methods applied univariate constraints (e.g., restricting at the 3 rd standard deviation) to each behaviour separately 3 which is incongruent with the compositional approach, and results in extrapolations into unsampled territory where there are few or no empirical data points. Taken together, we present a novel time-use optimisation pipeline which can be replicated for alternative outcomes, populations and predictor variables using our published code (available on GitHub). Comparison with previous 24-hr time use and cognition research We contribute new findings to a growing literature regarding the associations between 24-hr time-use composition and cognitive function in late adulthood. We found that 24-hr time-use composition was associated with cognitive function in older adults, which is congruent with some previous compositional studies 25-27 , but not all 28-31 . Patterns across the predicted optimal days in our study suggest that, compared to the “average” individual in each of the 8 age/sex/BMI strata, better cognition was associated with more time in MVPA and SB, and less time in LPA and sleep (relative to the mean time-use composition of the strata). Our findings for MVPA and LPA are consistent with many previous non-compositional and compositional studies 32-34 , which report cognitive benefits of higher intensity PA, and negative albeit largely inconclusive association between LPA and cognition. We provide evidence that light intensities of PA may not be sufficient to provide cognitive benefits, as these take time from more beneficial activities such as MVPA. Our analyses show that SB also has beneficial associations, up to an optimal duration, after which benefits appear to wane. Thus, when aiming to optimise cognition, MVPA and SB may compete for time-shares within the 24-hr time window. Our findings for SB contribute to a mixed literature, whereby some studies have reported beneficial associations between SB and cognition 26 , whilst others have found negative (or no) associations 35,36 . It is increasingly recognised that the type and context of SBs may alter their association with cognitive outcomes 35,37 . For example, our recent work among older adults demonstrated that cognitively engaging SBs (e.g., reading, computer use) are beneficially associated, whereas cognitively passive SBs (e.g., TV watching) are detrimentally associated with cognition 26 . As our SB variable was derived from accelerometry in the current study, we were unable to differentiate between cognitively engaging and passive sedentary time. Notably, the mean SB time in our UK Biobank sample was considerably lower (~9.6hrs/day) than other samples exploring similar relationships in older adults (e.g., 11hrs 29 -12hrs 25 ). It is likely that this contributed to the finding that optimal days required an increase in SB across all strata. Finally, for all strata, we found that optimal sleep duration was lower than the mean sleep duration. The extant literature suggests there is an inverted U-shaped relationship between sleep and cognition, whereby long or short sleep duration (e.g., 9 hr) is associated with reduced cognitive performance 38-40 . In the current sample, mean sleep duration ranged between 8.8 and 9.2 hr across strata, which is close to the upper bound of the range of sleep durations associated with better cognition in aforementioned studies. However, it is important to note that the Biobank sleep measure did not account for nighttime awakenings, or time awake in bed. Thus, longer sleep durations may reflect poorer sleep efficiency, where relatively longer durations of the “sleep” variable are spent lying awake in bed rather than asleep. Moreover, the derived time-use behaviours may have subsequently overestimated time in sleep and underestimated time in SB (which ranged between 9.2 and 10.5 hr/day in this sample). Implications for time-use personalization Our findings provide evidence that different population sub-groups require different balances of sleep, SB, LPA and MVPA in the 24-hr day for cognition, and unique combinations of sociodemographic and health characteristics may further adjust the ‘ideal day’ for cognition. This supports the Sweet Spot Hypothesis, and suggests that ‘one-size-fits-all’ approaches to time-use recommendations may not confer equitable benefits across a sample of participants. In our study, the patterns of associations were relatively consistent between the strata (i.e., for all strata, the optimal days had more MVPA and SB, and less LPA and sleep than the strata average). The differences between strata were in the recommended durations of the behaviours, which was directly linked to the bounds of the time-use footprint considered “feasible” for each of the strata. For example, the maximum duration of MVPA considered feasible for females aged >65 years with obesity was 110 minutes (minimum = 5 minutes), compared to almost double (205 minutes; minimum = 10 minutes) among males aged <65 years without obesity. Constraining the time-use footprint to a feasible range is crucial to producing behaviour change recommendations that are meaningful and achievable for the target population. Our study describes new methods for computing personalised 24-hr time-use recommendations via an interactive interface. Time-use recommendations have, over the last two decades, moved from a focus on individual behaviours (PA, SB, sleep) towards recommendations encompassing the whole 24-hr day (e.g., Canadian 24-Hour Movement Guidelines for Adults 41 ). However, even the most recent guidelines are merely composites of separate sets of recommendations for each of the movement behaviours. Furthermore, the guidelines apply to all people within the designated age bands, agnostic to other personal characteristics and current behaviours. Understanding the optimal balance and/or trade-offs across the 24-h day needed to maintain health has been the focus of an increasing number of studies in the past decade 42 . The app presented here goes beyond the one-size-fits-all approach to optimising the balance of PA, sleep and SB in a true 24-hr day, and may be an important next step towards personalised approaches to chronic disease prevention. With relevance to the cognitive aging and dementia prevention field, interventions (including multi-domain trials) which have incorporated a component of PA have yielded mixed findings 43 . It is possible that these mixed findings may be, at least in part, due to the lack of consideration of trade-offs being made with other components of the day (sleep and SB), or the lack of personalisation based on previous exposures to other modifiable risk factors for dementia. Above all, calculating and then communicating optimal days which are personalised to the individual’s sociodemographic and health characteristics is complex. Our app presents a potential solution by providing an accessible interface, co-designed with consumers, to generate and translate 24-hr activity interventions that are tailored to the individual. Strengths and limitations This study has a number of strengths. We used robust model selection procedures to avoid over-fitting models, and achieve a balance between complexity and understanding. This study included a large population-based sample, and 24-hr activity and sleep data were collected using device-based methods (accelerometry) which may be less susceptible to recall bias or inaccuracy in older populations. We explored several cognitive outcomes, which strengthens evidence of domain-specific associations between 24-hr time-use composition and cognitive function in older adult populations. Finally, the user interface of the app was co-designed with community-dwelling older adults for whom the app was initially intended 22 . As a result, the user interface is accessible and easily to interpret, avoiding complex language. There are limitations which must also be acknowledged. Cross-sectional data were used to estimate relationships between 24-hr time-use composition and cognitive outcomes. Extending the methods presented here by using longitudinal data would also allow researchers to explicitly consider the effect of within-person changes in time use in addition to the between-person differences that the current recommendations are based on. Twenty-four-hour time-use data were measured using wrist-worn accelerometers which are not considered the gold standard for measuring postural changes (i.e., differentiating between sitting and sleeping). It is possible that some SBs (e.g., time awake in bed) were classified as sleep, resulting in the over-estimation of time in sleep. This potential discrepancy may have contributed to the findings that 1) participants engaged in a lower-than-expected amount of SB per day (9hrs) compared to other cohorts of similar age, and 2) more time in SB was associated with better cognitive function. Finally, the UK Biobank sample used in this study are relatively homogenous in their characteristics and don’t reflect the most at-risk groups for cognitive decline and dementia (particularly Alzheimer’s disease). Thus, this approach to personalising 24-hr time-use interventions should be explored in datasets whereby the population are more representative of at-risk populations (e.g., those from low-to-middle income countries, with greater ethnic diversity) 6,11 . Future directions This first-of-its-kind, proof-of-concept study provides important foundations for future time-use personalisation research and intervention studies to build upon. We anticipate several key future directions for this work. First, as this analysis pipeline is able to be replicated for alternative study types (e.g., time-to-event analyses), future studies should explore the utility of time-use personalisation for clinical outcomes, such as onset of Alzheimer’s disease or other chronic diseases. Second, with relevance to the association between time-use composition and cognitive performance, these relationships should be explored longitudinally and account for additional factors which may have confounding effects, such as genetics (e.g., carriage of dementia risk genes such as apolipoprotein E ε4). Third, the pipeline can be extended to create personalised ‘optimal days’ for multiple response variables concurrently (i.e., the Goldilocks method 3 ). Fourth, future studies should consider the context in which activity occurs (e.g., cognitively active vs. cognitively passive SB; MVPA occurring in work vs leisure) to further personalise optimal days for health. Declarations Data availability The data that support the findings of this study are available from the UK Biobank but restrictions apply to the availability of these data, which were used under licence for the current study, and so are not publicly available. Code availability The underlying code for the data preparation, cleaning, analysis and diagnostics for this study is freely accessible on GitHub and can be accessed via this link (https://github.com/tystan/ukbb-cog-lasso ) in addition to the ‘ideal day’ interactive personalisation tool hosted at https://arena2024.shinyapps.io/ideal-day/with the associated underlying code also freely accessed on GitHub too ( https://github.com/tystan/ideal-day) . Competing interests: All authors declare no financial or non-financial competing interests. Author Contribution MM conceptualized the study, supported data analysis, and prepared the manuscript. TS conceptualized the study, conducted data analysis and contributed to manuscript development. TO and AM contributed to manuscript development. AS led the co-design of the app interface, conceptualized the study, and contributed to manuscript development. DD conceptualized the study, supported data analysis, and prepared the manuscript. All authors read and approved the final manuscript. Acknowledgement This research has been conducted using the UK Biobank Resource under Application Number 62254. We wish to acknowledge the wider Small Steps team for their contribution to the co-design of the digital interface, detailed elsewhere. References Harvey, A. et al. The future of technologies for personalised medicine. New Biotechnology 29, 625–633, doi: https://doi.org/10.1016/j.nbt.2012.03.009 (2012). Holtermann, A. et al. 24-Hour Physical Behavior Balance for Better Health for All: “The Sweet-Spot Hypothesis”. Sports Medicine - Open 7, 98, doi: 10.1186/s40798-021-00394-8 (2021). Dumuid, D. et al. Goldilocks Days: optimising children’s time use for health and well-being. Journal of Epidemiology and Community Health 76, 301, doi: 10.1136/jech-2021-216686 (2022). Dumuid, D. et al. Your best day: An interactive app to translate how time reallocations within a 24-hour day are associated with health measures. PLOS ONE 17, e0272343, doi: 10.1371/journal.pone.0272343 (2022). Sommerlad, A. & Mukadam, N. Evaluating risk of dementia in older people: a pathway to personalized prevention? European Heart Journal 41, 4034–4036, doi: 10.1093/eurheartj/ehaa691 (2020). Livingston, G. et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. The Lancet 404, 572–628, doi: 10.1016/S0140-6736(24)01296-0 (2024). Dumuid, D. et al. The “Goldilocks Day” for Children's Skeletal Health: Compositional Data Analysis of 24-Hour Activity Behaviors. Journal of Bone and Mineral Research 35, 2393–2403, doi: 10.1002/jbmr.4143 (2020). Mellow, M. L. et al. ‘Small Steps’ towards improving 24-hour time-use behaviours to decrease the risk of dementia: protocol for a personalised, web-based randomised controlled trial in community-dwelling older adults. medRxiv , 2025.2005.2026.25328336, doi: 10.1101/2025.05.26.25328336 (2025). Sudlow, C. et al. UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLOS Medicine 12, e1001779, doi: 10.1371/journal.pmed.1001779 (2015). shiny: Web Application Framework for R (2025). Livingston, G. et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. The Lancet 396, 413–446, doi: 10.1016/S0140-6736(20)30367-6 (2020). Walmsley, R. et al. Reallocation of time between device-measured movement behaviours and risk of incident cardiovascular disease. British Journal of Sports Medicine 56, 1008, doi: 10.1136/bjsports-2021-104050 (2022). Palarea-Albaladejo, J. & Martín-Fernández, J. A. zCompositions — R package for multivariate imputation of left-censored data under a compositional approach. Chemometrics and Intelligent Laboratory Systems 143, 85–96, doi: https://doi.org/10.1016/j.chemolab.2015.02.019 (2015). Martín-Fernández, J. A., Hron, K., Templ, M., Filzmoser, P. & Palarea-Albaladejo, J. Model-based replacement of rounded zeros in compositional data: Classical and robust approaches. Computational Statistics & Data Analysis 56, 2688–2704, doi: https://doi.org/10.1016/j.csda.2012.02.012 (2012). Campbell, T. & Cullen, B. Estimating the effect of physical activity on cognitive function within the UK Biobank cohort. International Journal of Epidemiology 52, 1592–1611, doi: 10.1093/ije/dyad009 (2023). R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria, 2023). Friedman, J. H., Hastie, T. & Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. Journal of Statistical Software 33, 1–22, doi: 10.18637/jss.v033.i01 (2010). Yuan, M. & Lin, Y. Model Selection and Estimation in Regression with Grouped Variables. Journal of the Royal Statistical Society Series B: Statistical Methodology 68, 49–67, doi: 10.1111/j.1467-9868.2005.00532.x (2006). Breheny, P. & Huang, J. Group descent algorithms for nonconvex penalized linear and logistic regression models with grouped predictors. Statistics and Computing 25, 173–187, doi: 10.1007/s11222-013-9424-2 (2015). Hastie, T., Tibshirani, R. & Friedman, J. H. The Elements of Statistical Learning: Data Mining, Inference, and Prediction . 2nd edn, (Springer, 2009). Martín-Fernández, J.-A., Daunis i Estadella, J. & Mateu-Figueras, G. On the interpretation of differences between groups for compositional data. SORT: statistics and operations research transactions 39, 0231–0252 (2015). Blake, H. T. et al. Co-design of a digital 24-hour time-use intervention with older adults and allied health professionals. Frontiers in Digital Health Volume 7–2025, doi: 10.3389/fdgth.2025.1544489 (2025). Zahran, S. et al. Optimal levels of sleep, sedentary behaviour, and physical activity needed to support cognitive function in children of the early years. BMC Pediatrics 24, 735, doi: 10.1186/s12887-024-05186-z (2024). Tolosana-Delgado, R., Arya, E., Khodadadzadeh, M. & Boogaart, K. On machine learning algorithms and compositional data . (2019). Collins, A. M. et al. 24-Hour time use and cognitive performance in late adulthood: results from the Investigating Gains in Neurocognition in an Intervention Trial of Exercise (IGNITE) study. Age and Ageing 54, afaf072, doi: 10.1093/ageing/afaf072 (2025). Mellow, M. L. et al. Should We Work Smarter or Harder for Our Health? A Comparison of Intensity and Domain-Based Time-Use Compositions and Their Associations With Cognitive and Cardiometabolic Health. The Journals of Gerontology: Series A 79, glae233, doi: 10.1093/gerona/glae233 (2024). Dumuid, D. et al. Does APOE ɛ4 Status Change How 24-Hour Time-Use Composition Is Associated with Cognitive Function? An Exploratory Analysis Among Middle-to-Older Adults. Journal of Alzheimer’s Disease 88, 1157–1165, doi: 10.3233/JAD-220181 (2022). Mellow, M. L. et al. Cross-sectional associations between 24-hour time-use composition, grey matter volume and cognitive function in healthy older adults. International Journal of Behavioral Nutrition and Physical Activity 21, 11, doi: 10.1186/s12966-023-01557-4 (2024). Mellow, M. L. et al. Twenty-four-hour time-use composition and cognitive function in older adults: cross-sectional findings of the ACTIVate study. Frontiers in Human Neuroscience Volume 16–2022, doi: 10.3389/fnhum.2022.1051793 (2022). Wu, Y. et al. Analysis of the 24-h activity cycle: An illustration examining the association with cognitive function in the Adult Changes in Thought study. Frontiers in Psychology Volume 14–2023, doi: 10.3389/fpsyg.2023.1083344 (2023). Balbim, G. M. et al. The Association of the 24-Hour Activity Cycle Profiles With Cognition in Older Adults With Mild Cognitive Impairment: A Cross-Sectional Study. The Journals of Gerontology: Series A 79, glae099, doi: 10.1093/gerona/glae099 (2024). Mellow, M. L. et al. How are combinations of physical activity, sedentary behaviour and sleep related to cognitive function in older adults? A systematic review. Experimental Gerontology 159, 111698, doi: https://doi.org/10.1016/j.exger.2022.111698 (2022). Erickson, K. I., Donofry, S. D., Sewell, K. R., Brown, B. M. & Stillman, C. M. Cognitive Aging and the Promise of Physical Activity. Annual Review of Clinical Psychology 18, 417–442, doi: https://doi.org/10.1146/annurev-clinpsy-072720-014213 (2022). Singh, B. et al. Effectiveness of exercise for improving cognition, memory and executive function: a systematic umbrella review and meta-meta-analysis. British Journal of Sports Medicine (2025). Coelho, L. et al. The association between sedentary behavior and cognitive ability in older adults. Aging Clinical and Experimental Research 32, 2339–2347, doi: 10.1007/s40520-019-01460-8 (2020). Maasakkers, C. M. et al. The Association of Sedentary Behaviour and Cognitive Function in People Without Dementia: A Coordinated Analysis Across Five Cohort Studies from COSMIC. Sports Medicine 50, 403–413, doi: 10.1007/s40279-019-01186-7 (2020). Olanrewaju, O., Stockwell, S., Stubbs, B. & Smith, L. Sedentary behaviours, cognitive function, and possible mechanisms in older adults: a systematic review. Aging Clinical and Experimental Research 32, 969–984, doi: 10.1007/s40520-019-01457-3 (2020). Lo, J. C., Groeger, J. A., Cheng, G. H., Dijk, D.-J. & Chee, M. W. L. Self-reported sleep duration and cognitive performance in older adults: a systematic review and meta-analysis. Sleep Medicine 17, 87–98, doi: https://doi.org/10.1016/j.sleep.2015.08.021 (2016). Gildner, T. E., Liebert, M. A., Kowal, P., Chatterji, S. & Snodgrass, J. J. Associations between sleep duration, sleep quality, and cognitive test performance among older adults from six middle income countries: results from the Study on Global Ageing and Adult Health (SAGE). Journal of clinical sleep medicine 10, 613–621 (2014). Devore, E. E., Grodstein, F. & Schernhammer, E. S. Sleep duration in relation to cognitive function among older adults: a systematic review of observational studies. Neuroepidemiology 46, 57–78 (2016). Ross, R. et al. Canadian 24-Hour Movement Guidelines for Adults aged 18–64 years and Adults aged 65 years or older: an integration of physical activity, sedentary behaviour, and sleep. Applied Physiology, Nutrition, and Metabolism 45, S57-S102, doi: 10.1139/apnm-2020-0467 (2020). Miatke, A. et al. The association between reallocations of time and health using compositional data analysis: a systematic scoping review with an interactive data exploration interface. International Journal of Behavioral Nutrition and Physical Activity 20, 127, doi: 10.1186/s12966-023-01526-x (2023). Castro, C. B. et al. Multi-Domain Interventions for Dementia Prevention–A Systematic Review. The Journal of nutrition, health and aging 27, 1271–1280, doi: https://doi.org/10.1007/s12603-023-2046-2 (2023). Additional Declarations No competing interests reported. Supplementary Files UKBSupplementaryFilescombinedFINAL.pdf Cite Share Download PDF Status: Published Journal Publication published 17 Mar, 2026 Read the published version in npj Digital Medicine → Version 1 posted Editorial decision: Revision requested 15 Jul, 2025 Reviews received at journal 15 Jul, 2025 Reviews received at journal 14 Jul, 2025 Reviewers agreed at journal 23 Jun, 2025 Reviewers agreed at journal 23 Jun, 2025 Reviewers invited by journal 22 Jun, 2025 Editor assigned by journal 18 Jun, 2025 Submission checks completed at journal 18 Jun, 2025 First submitted to journal 15 Jun, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6897341","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":475259627,"identity":"8c4e6299-8ea8-494f-8e16-99974c19d626","order_by":0,"name":"Maddison L Mellow","email":"data:image/png;base64,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","orcid":"","institution":"University of South Australia","correspondingAuthor":true,"prefix":"","firstName":"Maddison","middleName":"L","lastName":"Mellow","suffix":""},{"id":475259628,"identity":"25496ccf-c7ba-4b9e-8696-3136c652557f","order_by":1,"name":"Tyman E Stanford","email":"","orcid":"","institution":"University of South Australia","correspondingAuthor":false,"prefix":"","firstName":"Tyman","middleName":"E","lastName":"Stanford","suffix":""},{"id":475259629,"identity":"fcef8316-1a0b-48ac-951c-450791702249","order_by":2,"name":"Timothy Olds","email":"","orcid":"","institution":"University of South Australia","correspondingAuthor":false,"prefix":"","firstName":"Timothy","middleName":"","lastName":"Olds","suffix":""},{"id":475259630,"identity":"1156f316-4058-4e19-b9aa-3eee3bfaa2ac","order_by":3,"name":"Aaron Miatke","email":"","orcid":"","institution":"University of South Australia","correspondingAuthor":false,"prefix":"","firstName":"Aaron","middleName":"","lastName":"Miatke","suffix":""},{"id":475259631,"identity":"f2de8ac5-f470-4860-9fce-b3327b920846","order_by":4,"name":"Ashleigh E Smith","email":"","orcid":"","institution":"University of South Australia","correspondingAuthor":false,"prefix":"","firstName":"Ashleigh","middleName":"E","lastName":"Smith","suffix":""},{"id":475259632,"identity":"f140ec31-c6b0-4e9d-8850-b2ab9e03d61d","order_by":5,"name":"Dorothea Dumuid","email":"","orcid":"","institution":"University of South Australia","correspondingAuthor":false,"prefix":"","firstName":"Dorothea","middleName":"","lastName":"Dumuid","suffix":""}],"badges":[],"createdAt":"2025-06-15 08:53:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6897341/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6897341/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41746-026-02542-4","type":"published","date":"2026-03-17T15:59:24+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":85649569,"identity":"fd3637ab-e9d1-4751-99b1-837e2623abbc","added_by":"auto","created_at":"2025-06-30 09:00:32","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":135498,"visible":true,"origin":"","legend":"\u003cp\u003eStatic representations at different viewing angles of the 4-simplex tetrahedron containing all theoretically possible time-use compositions with a random sub-sample (for viewability) of the female/aged \u0026lt;65 years/without obesity stratum as points within the tetrahedron. The 80\u003csup\u003eth\u003c/sup\u003e percentile ellipsoid fencing transformed back to the compositional scale is shown as a semi-transparent purple surface. Sampled time-use compositions are coloured orange if they are outside the fencing, and purple if within the fencing. For the optimisation procedure, the model predictions on a grid of compositions (5-minute spacings) that strictly lie within the ellipsoid fencing are used to restrict possible time-use footprints to realistic values.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6897341/v1/7d8f5053b5b05d17c7d7792f.png"},{"id":85651060,"identity":"f58327ed-51f9-45ef-a7c6-eebf342bf593","added_by":"auto","created_at":"2025-06-30 09:16:32","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":426589,"visible":true,"origin":"","legend":"\u003cp\u003eIsometric log-ratio (\u003cem\u003eilr\u003c/em\u003e) main, squared, and interaction (between health/sociodemographic characteristics and ilrs) coefficients (y-axis) in the group LASSO model fits for each outcome (x-axis) are presented by text and coloured cells. Blank cells (white spaces) indicate model terms that were removed from the model following LASSO procedures (i.e., 0 coefficients). Coefficient names are structured such that the overall variable is listed first, followed by the level (for categorical variables) in square brackets (e.g., Smoking[Current]). The magnitude of estimates ranges from -0.1 to 0.1, indicating the strength of the association with the cognitive outcome variable.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6897341/v1/6a6e2ce1fda3d6e87746cf4a.png"},{"id":85648027,"identity":"13abd2d7-3aa8-401a-a5fa-f085db3110b6","added_by":"auto","created_at":"2025-06-30 08:52:32","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":274641,"visible":true,"origin":"","legend":"\u003cp\u003eOptimal time-use compositions for cognitive variables (differentiated by colour) across age/sex/BMI strata. The coloured dots in the four left columns show optimal durations of activities for the ‘average person’ in each of the strata, alongside the mean activity duration observed in the respective strata (gray) and overall sample (black) mean. The right-most column shows the predicted cognitive responses (z-scores) for the compositions depicted on the left. Supplementary File 5 provides the values/levels of covariates used to define the ‘average person’ in each stratum. BMI = body mass index, SB = sedentary behaviour, LPA = light physical activity, MVPA = moderate-to-vigorous physical activity, SD = standard deviation; ‘w/ obesity’ = with obesity; ‘w/out obesity’ = without obesity.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6897341/v1/96ddfbf719912667e94db2bd.png"},{"id":85648029,"identity":"8df1ec48-e21d-4999-a566-4ee4327c7737","added_by":"auto","created_at":"2025-06-30 08:52:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":128254,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eNote. \u003c/em\u003eOptimisation displayed in Figure 4 is based on a user with the following characteristics: sex=female; age=60 years; highest qualification=University/college; weight=80 kg; height=170 cm; high blood pressure=yes; smoking=current; alcohol consumption=≥3 standard drinks per week; type 2 diabetes=no; depression=no; social isolation=no; hearing difficulty or loss=no; traumatic brain injury=no; sleep=8 h, sedentary behaviour =13 h, LPA = 3 h; MVPA= 0.2 h.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6897341/v1/ea603e9e6beb49157706aecd.png"},{"id":105223506,"identity":"eef5d176-3de9-404a-9a4c-3d2fd5a46144","added_by":"auto","created_at":"2026-03-23 16:07:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1577375,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6897341/v1/f57399b3-2623-459b-927f-d32039e31d34.pdf"},{"id":85649570,"identity":"605b2de6-8cc9-4eab-9298-fb99de2beb93","added_by":"auto","created_at":"2025-06-30 09:00:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":933191,"visible":true,"origin":"","legend":"","description":"","filename":"UKBSupplementaryFilescombinedFINAL.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6897341/v1/7879b9514cd0c0f006c1ed6a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eAn interactive tool to personalise 24-hour activity, sitting and sleep prescription for optimal health outcomes\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003ePersonalised (or \u0026lsquo;precision\u0026rsquo;) medicine approaches to healthcare move beyond the \u0026lsquo;one size fits all\u0026rsquo; approach towards individualised prevention and treatment of illness \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Rather than using population averages to inform therapy, treatment and behaviour change, personalised approaches consider individual characteristics beyond age and sex, including genetic, environmental and lifestyle factors \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e. Whilst this approach is increasingly projected to be \u0026lsquo;the future\u0026rsquo; of drug development (including applications in cancer treatment, pharmacogenomics and immunotherapy), lifestyle interventions which aim to optimise the balance of physical activity (PA), sleep and sedentary behaviours (SB) in the 24-hr day are seldom personalised to the individual. Instead, most approaches administer a universal \u0026ldquo;dose\u0026rdquo; (e.g., perform 20 minutes of walking, 3 times per week), or attempt to move all individuals closer to PA, sleep and/or SB guidelines which are also generalised.\u003c/p\u003e \u003cp\u003eAn important but overlooked consideration is that the ideal 24-hr time-use composition (i.e., the optimal balance of PA, sleep and SB in the 24-hr day) for health outcomes may vary between individuals based on their current time use, health status, or sociodemographic \u0026ldquo;profile\u0026rdquo;. That is, the unique combination of factors such as age, sex, mobility, health conditions, and socioeconomic status, as well as their current levels of PA, sleep and SB, may already have participants closer to, or further away from, their optimal day at the beginning of an intervention. Consequently, administering a universal intervention to a sample of individuals may not result in uniform intervention effects. Several key theories and methodologies have highlighted the potential value of personalising 24-hr time-use interventions for health. For example, the \u0026ldquo;Sweet-Spot\u0026rdquo; hypothesis postulated by Holtermann and colleagues \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e proposed that the optimal combination of PA, SB, and sleep for health likely differs between adults in physically active versus sedentary occupations. They argue that the generalised advice to \u0026ldquo;sit less and move more\u0026rdquo; may actually take adults who work in physically active occupations further away from their optimal 24-hr day, as their optimal day requires more time for recovery (SBs and sleep) than it does for those in sedentary occupations \u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Similarly, the \u0026ldquo;Goldilocks Day\u0026rdquo; approach by Dumuid et al. \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e demonstrates that different health outcomes have different optimal days, thus determining someone\u0026rsquo;s optimal day will require personalisation to their health priorities. For example, the optimal balance of PA, sleep, and SB looks different for cognitive health than it does for physical or mental health, meaning the optimal day is also dependent on how someone ranks these health domains (i.e., whether cognitive, physical or mental health are of higher importance) \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Taken together, the Sweet-Spot and Goldilocks Day principles may also apply to the current health and sociodemographic profile of the individual.\u003c/p\u003e \u003cp\u003eSociety today faces several pressing health challenges which may benefit from personalised prevention approaches, including dementia \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. There are at least 14 established modifiable risk factors for dementia, many of which are potentially bi-directionally related PA, sleep and SB, such as obesity and depression. Additionally, physical inactivity is a risk factor for dementia \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. It is likely that the optimal 24-hr time-use composition for minimising dementia risk for a 50-year-old male with high school-level education, hypertension, and a history of smoking, all of which have been identified as modifiable risk factors for dementia, looks different in comparison to another 50-year-old male with tertiary education, hypertension, and no smoking history. Despite the two individuals belonging to the same age group and both having a diagnosis of hypertension, the additional differences in modifiable dementia risk factors (education and smoking history) may change the relative impact that time-use behaviours have on dementia risk. That is, a higher amount of PA or different balance of time-use behaviours may be needed to offset the negative impacts of smoking and low education.\u003c/p\u003e \u003cp\u003eIt is possible to compute the optimal 24-hr time-use composition for an outcome of interest (e.g., cognitive VS physical health) specifically for the average individual with given personal characteristics (e.g., age, sex) based on statistical models \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. However, the challenge is how to effectively communicate behaviour change recommendations which are tailored to an individual participant in real time, such as in an intervention study. One solution may be \u003cem\u003evia\u003c/em\u003e interactive interfaces which allow the user to set their own descriptors (e.g., age, sex, health status, current time use) and view the optimal day for their chosen health outcome. The Your Best Day interactive app developed by Dumuid and colleagues \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e achieves this: users select their sex, age, and current time use, and can then use interactive sliders to visualise the predicted effects of changing their time use on body fat percentage, psychosocial wellbeing, and academic performance in adolescents.\u003c/p\u003e \u003cp\u003eApplying the principles of the Your Best Day app, the aim of the current study was to develop an interactive user interface to individualise 24-hr time-use recommendations for the \u003cem\u003eSmall Steps\u003c/em\u003e Study, a randomized controlled trial aiming to improve 24-hr time use for cognitive health in community-dwelling older adults \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Using the UK Biobank \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e as our normative sample, we firstly aimed to identify the optimal 24-hr time-use composition for cognitive function outcomes across a range of sociodemographic and health profiles (i.e., across combinations of a range of modifiable and non-modifiable risk factors for dementia). Secondly, we developed a user interface (housed as an R Shiny app \u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e) which allows users to input information about dementia risk factors (i.e., create their \u0026lsquo;profile\u0026rsquo;), and view the optimal 24-hr day for cognitive function based on their profile.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eStudy design and participants\u003c/p\u003e \u003cp\u003eData used in this cross-sectional study were from the UK Biobank (application no. 62254), a large prospective cohort study of ~\u0026thinsp;500,000 participants aged\u0026thinsp;\u0026gt;\u0026thinsp;40 years at time of recruitment in 2010 \u003csup\u003e9\u003c/sup\u003e. The sample for this study were limited to UK Biobank participants who met the following criteria: accelerometry data were marked as \u0026lsquo;valid\u0026rsquo;; accelerometry data did not have a data problem indicator; at least four of five cognitive tests included in this study were completed; a diagnosis of dementia, organic amnesic syndrome, delirium, systemic atrophy affecting the central nervous system, extrapyramidal or movement disorder, other degenerative disease of the central nervous system, demyelinating disease of the central nervous system or blindness was not recorded prior to either completing the cognitive tests or wearing the accelerometer; and no indication that participant was unable to walk. A complete list of inclusion and exclusion criteria are displayed in Supplementary File 1.\u003c/p\u003e \u003cp\u003eStudy measures\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eSociodemographic and health factors\u003c/h2\u003e \u003cp\u003eA range of sociodemographic and health variables which have been identified as modifiable dementia risk factors as per the 2020 Lancet commission report \u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e were extracted from the assessment centre data (initial visit) for consideration in our model selection procedures. A detailed list of included variables, their UK Biobank field codes and data re-classification protocols (where applicable) is available in Supplementary File 2. As the final behaviour change tool (detailed below) was initially intended for use in an Australian context, where necessary some variables (e.g., education/highest qualification) were re-levelled to align with Australian classifications. Final variables included age, sex, ethnicity (levels = \u0026ldquo;White\u0026rdquo;, \u0026rdquo;non-White\u0026rdquo;), highest qualification (levels= \u0026ldquo;high school\u0026rdquo;, \u0026ldquo;Certificate III/Diploma\u0026rdquo;, \u0026ldquo;College/University\u0026rdquo;, \u0026ldquo;Other professional qualification\u0026rdquo;), alcohol consumption (levels= \u0026ldquo;sometimes/never\u0026rdquo;, \u0026ldquo;often/very often\u0026rdquo;), smoking status (levels=\u0026ldquo;never\u0026rdquo;, \u0026ldquo;previous\u0026rdquo;, \u0026ldquo;current\u0026rdquo;), history of depression (levels=\u0026ldquo;yes\u0026rdquo;, \u0026ldquo;no\u0026rdquo;), hearing difficulty (levels=\u0026ldquo;yes\u0026rdquo;, \u0026ldquo;no\u0026rdquo;), history of hypertension (levels=\u0026ldquo;yes\u0026rdquo;, \u0026ldquo;no\u0026rdquo;), history of traumatic brain injury (levels=\u0026rdquo;yes\u0026rdquo;, \u0026ldquo;no\u0026rdquo;), type 2 diabetes status (levels=\u0026ldquo;yes\u0026rdquo;, \u0026ldquo;no\u0026rdquo;), body mass index (BMI; levels=\u0026rdquo;overweight\u0026rdquo;, \u0026ldquo;not overweight\u0026rdquo;), and social isolation (\u0026ldquo;yes\u0026rdquo;, \u0026ldquo;no\u0026rdquo;). For all variables, missing data were recoded as \u0026lsquo;unknown\u0026rsquo; (i.e., some dichotomous variables had three levels: yes, no, unknown).\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eTwenty-four-hour time-use composition\u003c/h3\u003e\n\u003cp\u003eThe proportions of daily time spent in sleep, SB, light intensity PA (LPA) and moderate-vigorous intensity PA (MVPA) were obtained from the derived accelerometry data \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e within the UK Biobank (field codes are presented in Supplementary File 2). Zero values in any compositional parts (sleep, SB, LPA, MVPA) were imputed assuming censored data (values below the threshold of detection) using a linear model-based imputation of log ratios of the components, iteratively improved via the Expectation-Maximisation algorithm (\u003cem\u003elrEM\u003c/em\u003e function in the zCompositions R package \u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e,\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e), with the remaining behaviours reduced by the small, imputed value proportionally. Additionally, to mitigate the undue influence of extreme values, the 0.5% of the empirical distribution component percentiles at both tails were truncated to their respective 0.5 and 99.5 percentile values (with proportional shrinking/expansion of the other components to ensure a 24-hr day). The four \u003cem\u003ecompositional parts\u003c/em\u003e (sleep, SB, LPA, MVPA) were expressed as a set of three isometric log-ratio (\u003cem\u003eilr\u003c/em\u003e) coordinates using a sequential binary partition \u003cem\u003eilr\u003c/em\u003e base representing the following quantities (ignoring normalizing constants): the first \u003cem\u003eilr\u003c/em\u003e represented the log-ratio of one behaviour (e.g., sleep) to the geometric mean of the remaining three behaviours (e.g., SB, LPA, MVPA); the second \u003cem\u003eilr\u003c/em\u003e excluded sleep, and represented the log-ratio of the next behaviour in the set (e.g., SB) to the geometric mean of the remaining two behaviours (LPA and MVPA); and the final (third) \u003cem\u003eilr\u003c/em\u003e represented the log-ratio of LPA to MVPA). Together, the three \u003cem\u003eilr\u003c/em\u003e coordinates represent 24-hr time-use composition as a set of linearly independent predictors in regression models.\u003c/p\u003e\n\u003ch3\u003eCognitive function\u003c/h3\u003e\n\u003cp\u003eSeveral web-based cognitive tests collected during the UK Biobank online follow-up were included in this study, including Numeric Memory, Pairs Matching, Fluid Intelligence, Trail Making (A and B), and Symbol Digit Substitution. Supplementary File 2 presents the individual outcome measures extracted from these tests and how they were combined into composite scores. We created four cognitive domain composites (memory, reasoning, executive function and processing speed) using groupings based on a previous UK Biobank study \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e. To create the composites, we first reverse-scored Trail Making A, Trail Making B and Pairs Matching outcomes so that for all included measures, higher scores represented better performance. We then undertook the following steps to normalise the distributions of the measures: Trail Making A and B scores were truncated at 300 seconds and log-transformed; Pairs Matching scores were truncated at -7 seconds; and Symbol Digit Substitution scores were truncated at 31. Individual measures were z-scored using age and sex standardisation (age\u0026thinsp;=\u0026thinsp;\u0026lt;\u0026thinsp;65 years, \u0026gt;\u0026thinsp;65 years; sex\u0026thinsp;=\u0026thinsp;male, female). Finally, composite scores were created by averaging z-scores (where applicable), and an overall composite score (global cognition) was created by averaging the four z-scores. Supplementary File 2 outlines test outcomes and their corresponding cognitive domains.\u003c/p\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003eData analysis\u003c/h2\u003e \u003cp\u003eAll analyses were conducted in R Statistical Software (version 4.3.1 \u003csup\u003e16\u003c/sup\u003e). Full R codes are published on GitHub (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/tystan/ukbb-cog-lasso\u003c/span\u003e\u003cspan address=\"https://github.com/tystan/ukbb-cog-lasso\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e\u003c/p\u003e \u003cp\u003eModel selection by regularization\u003c/p\u003e \u003cp\u003eRegularized linear regression models were fit with each of the five cognitive function measures as the outcome variable, with the following candidate predictors prior to shrinkage: continuous and discrete main effect predictors (time-use log-ratios and all sociodemographic and health factors); all pairwise main effect interactions; and additional squared continuous main effects (i.e., polynomial terms degree 2). To reduce non-informative candidate predictors and avoid overfitting, Least Absolute Shrinkage and Selection Operator (LASSO) regularization was used. LASSO regression is a coefficient shrinkage method which aims to produce a parsimonious model based on a subset of the potential predictors that are interpretable and related to the outcome, by penalising the sum of the absolute values of a model\u0026rsquo;s variable coefficients, forcing some coefficients to exactly zero and therefore to become omitted from the model \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e. However, because the \u003cem\u003eilr\u003c/em\u003e variables were required to be treated as analytically inseparable variables, in addition to having many categorical variables with more than two levels, we specifically implemented group LASSO models \u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e using the grpreg R package \u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e. The group (or \u0026ldquo;block\u0026rdquo;) extension of LASSO regression allows multiple predictor variables to be treated as an inseparable group \u0026ndash; the shrinkage is applied block-wise, instead of individually, on the prespecified groupings of variables. Such an approach is required to ensure the invariance of the specific \u003cem\u003eilr\u003c/em\u003e basis chosen, but can also be useful to regularize categorical variables with three or more levels (resulting in two or more contrast dummy coded variables) as an inseparable group that the model would otherwise not know are intrinsically related. By applying a group-specific penalty, the coefficients in the group are both shrunk at a group level (e.g., the group of categorical level contrasts) as well as individually \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e. To this end, the \u003cem\u003eilrs\u003c/em\u003e were also treated as a group after centring and an angle preserving rotation was applied (to mitigate the issue of computing different shrinkage results dependent on the \u003cem\u003eilr\u003c/em\u003e basis chosen \u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e). Similarly, the higher order terms were specified as groups, incorporating each crossed (or squared) sub-term in the case at least one categorical variable or \u003cem\u003eilr\u003c/em\u003e in the interaction (or squared) term was retained after shrinkage.\u003c/p\u003e \u003cp\u003eDespite group-specific penalties being sought, the model formulation can be equivalently re-expressed with a single penalisation parameter, λ, with the associated sum of the product of the square root of the group size and the Euclidean norm of the group variable\u0026rsquo;s coefficients \u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. To determine an empirically derived optimal penalty value, λ*, of the shrinkage procedure, we performed a ten-fold cross-validation over a grid of potential λ values seeking the one that minimises the model\u0026rsquo;s predictive mean squared error (MSE) of the model predicted values compared to observed values. Supplementary File 3 displays the derived λ* and corresponding MSE for each of the cognitive variables.\u003c/p\u003e \u003cp\u003eIdentifying \u0026lsquo;optimal\u0026rsquo; days for cognition\u003c/p\u003e \u003cp\u003eDue to the known mediating effects between \u003cem\u003eilrs\u003c/em\u003e (time use) and age, sex and BMI, the data were stratified into 8 mutually exclusive groups based on the following categories: age (\u0026lt;\u0026thinsp;65 years, \u0026ge;\u0026thinsp;65 years), sex (male, female), and BMI (with obesity\u0026thinsp;=\u0026thinsp;\u0026ge;\u0026thinsp;30, without obesity\u0026thinsp;=\u0026thinsp;\u0026lt;\u0026thinsp;30). Extending the methods outlined in Dumuid et al \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e, the models described previously were used to predict cognition for a \u0026ldquo;time-use footprint\u0026rdquo; that was considered feasible and realistic for each of the 8 population strata. To restrict the time-use footprint to realistic values, and to avoid extrapolating from the highest density of sampled time-use data, we used a grid of all possible time-use compositions with 5-minutes spacings constrained within the empirical distributional (multivariate Gaussian) quantiles of the strata-specific sampled data in the \u003cem\u003eilr\u003c/em\u003e-space (\u0026ldquo;constrained ellipsoid fencing\u0026rdquo;, a novel method for which codes are available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/tystan/ukbb-cog-lasso\u003c/span\u003e\u003cspan address=\"https://github.com/tystan/ukbb-cog-lasso\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e In brief, the empirically estimated multivariate normal 80th percentile contour limits were used after assessment of marginal (univariate), pairwise, and multivariate normality by way of visual checks (where possible in lower dimensions) and cumulative quantile plots. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e demonstrates, for one of the stratum, the relationships between the sampled time-use compositions, where points are either classified as within or outside the \u003cem\u003eilr\u003c/em\u003e-derived ellipsoid fencing when transformed back to the compositional scale and presented in the 4-simplex tetrahedron (constrained, constant sum space) in which they reside. Approximately 80% of the points are within the ellipsoid fencing as is expected if the (\u003cem\u003eilr\u003c/em\u003e) data are from an approximately multivariate Gaussian distribution.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eOptimal cognitive function was operationalized as the top 5% of the cognitive scores predicted by the models over the given constrained grid and specific person inputs \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. The optimal 24-hr day for cognitive function was conceptualised as the compositional mean (geometric mean, adjusted to sum to 1440 minutes) of the time-use compositions (sleep, sedentary time, light physical activity, MVPA) associated with the top 5% of predictions \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e. As models for the relationship between time-use composition and cognitive function also included interactions between various sociodemographic and health factors, the estimated optimal day varied depending on the values on these covariates. The optimal day for the \u0026ldquo;mean\u0026rdquo; or \u0026ldquo;average\u0026rdquo; person in each of the strata is presented in the main manuscript, but for further personalisation of the optimal day estimate according to the value of the covariates, real-time estimation applying the specified covariate values to the model coefficients is made possible via our Shiny app interface (detailed below).\u003c/p\u003e \u003cp\u003eDeveloping the interactive user interface\u003c/p\u003e \u003cp\u003eThe R \u0026ldquo;Shiny\u0026rdquo; package \u003csup\u003e\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e was used to program an interactive interface (app) that can be freely accessed in a web browser via a URL, without the need to install R or any other additional software. Our Shiny app has three components (i.e. R scripts) that communicate with each other: (1) the user interface (ui.R), which determines the appearance of the app and how the user enters information; (2) the server (server.R), which takes input provided by the user interface, sends it for computation, and then returns results back to the user interface; and (3) the global script (global.R), which defines the variables and functions accessible for both the user interface and server. The R scripts used for our Shiny app can be found in a separate GitHub repository (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/tystan/ideal-day\u003c/span\u003e\u003cspan address=\"https://github.com/tystan/ideal-day\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e The LASSO regression coefficients from the compositional models were extracted for the app\u0026rsquo;s global script. To determine the optimal day for user-specified covariates (i.e., their \u0026lsquo;profile\u0026rsquo;), the app\u0026rsquo;s user interface requests the user\u0026rsquo;s sociodemographic and health details. From this, the app matches the user to one of the eight (age, sex, and BMI) defined strata. It then uses the user\u0026rsquo;s remaining inputted sociodemographic and health variables, and every possible time-use composition within the feasible range for the user\u0026rsquo;s stratum, to predict cognition. The app then extracts the top 5% of predictions and calculates the mean time-use composition associated with these top 5% of predictions, i.e., the optimal day. The workflow underpinning our Shiny app is displayed in Supplementary File 4.\u003c/p\u003e \u003cp\u003eThe user is required to enter their current 24-hr time-use composition. In the case of the \u003cem\u003eSmall Steps\u003c/em\u003e study \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e for which this app was created, these values are derived from wrist-worn accelerometers worn by the participant prior to their initial visit (i.e., 7-day average of time spent in sleep, SB, LPA and MVPA from Fitbit watch). Following the user input of their current day, a bar plot is generated which displays the user\u0026rsquo;s current 24-hr day (mins/day of sleep, SB, LPA and MVPA) compared to their \u0026lsquo;optimal\u0026rsquo; 24-hr day. Feedback is provided about the changes required to reach that day from their current day (e.g., -10 minutes sleep, -20 minutes sedentary behaviour, 0 minutes light physical activity (i.e., no change), +\u0026thinsp;30 minutes MVPA).\u003c/p\u003e \u003cp\u003eThe appearance and functionality (e.g., colours, font sizing, wording) of the Shiny app was informed by older adults from the general population (n\u0026thinsp;=\u0026thinsp;8) through a series of co-design workshops conducted as part of the \u003cem\u003eSmall Steps\u003c/em\u003e study (Ethics ID 205377). A full description of the co-design process can be viewed elsewhere \u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eParticipant demographics\u003c/p\u003e \u003cp\u003eA sample of 53,057 participants from the UK Biobank were included in this study (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Participants were 62\u0026thinsp;\u0026plusmn;\u0026thinsp;8 years of age, mostly female (57%), White (97%) and had predominantly college/university (47%) qualifications. The health profile of participants varied across modifiable dementia risk factors, with the most common risk factors including history of hypertension (24%), hearing difficulty (24%), regular alcohol consumption (23%) and history of depression (20%). On average, participants spent most of their 24-hr day in SB (9.6hrs, 40.0%) or sleep (9.0hrs, 37.5%), whereas active behaviours made up \u0026lt;\u0026thinsp;25% of the day (LPA: 4.8hrs, 20.0%; MVPA: 0.5hrs, 2%). The number of participants per strata ranged from 2064 to 15303 (Supplementary File 5).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescriptive characteristics of sample\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLevel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOverall sample\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eN\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e53,057\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge (years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e62 (8)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eSex (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30,091 (57%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e22,966 (43%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003e24-hr time-use composition\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSleep\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e542.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSB\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e576.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLPA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e291.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMVPA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBMI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26.6 (4.5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"4\" rowspan=\"5\"\u003e \u003cp\u003eHighest qualification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHigh school\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10,637 (20%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCertificate III or Diploma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5,805 (11%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCollege/University\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24,742 (47%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOther professional qualification\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e8,092 (15%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,781 (7.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eDepression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10,734 (20%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9,141 (17%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33,182 (63%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eDiabetes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,665 (3.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e51,295 (97%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e97 (0.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eHearing difficulty\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12,713 (24%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e38,221 (72%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2,123 (4.0%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eHypertension\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12,595 (24%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e40,390 (76%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e72 (0.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eSocial isolation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7,851 (15%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e44,584 (84%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e622 (1.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eSmoking status\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCurrent\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,447 (6.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrevious\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18,960 (36%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNever\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e30,532 (58%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e118 (0.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eAlcohol consumption (frequency)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOften/very often\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12,415 (23%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSometimes/never\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26,641 (50%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eUnknown\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14,001 (26%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHistory of traumatic brain injury\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e106 (0.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eEthnicity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eWhite\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e51,577 (97%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNon-white\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,480 (2.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDays between accelerometry and cognitive testing\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e52 (232)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"3\"\u003e\u003cem\u003eNote.\u003c/em\u003e Mean and standard deviation (SD) are presented for continuous variables, and count and proportion (n(%)) are presented for categorical variables for the overall sample. BMI\u0026thinsp;=\u0026thinsp;body mass index; SB\u0026thinsp;=\u0026thinsp;sedentary behaviour; LPA\u0026thinsp;=\u0026thinsp;light intensity physical activity; MVPA\u0026thinsp;=\u0026thinsp;moderate-to-vigorous intensity physical activity. A positive value for days between accelerometry and cognitive testing indicates that accelerometry measures occurred later than cognitive testing.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e***INSERT Table \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e NEAR HERE (at bottom of manuscript)***\u003c/h2\u003e \u003cp\u003eModel selection\u003c/p\u003e \u003cp\u003eModel coefficients for terms containing \u003cem\u003eilrs\u003c/em\u003e (i.e., main effects, higher order terms or interactions with time use) are displayed in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Across all five cognitive outcomes, main effects were retained for \u003cem\u003eilr\u003c/em\u003es, as well as interactions between \u003cem\u003eilr\u003c/em\u003es and ethnicity, and \u003cem\u003eilr\u003c/em\u003es and alcohol consumption. The remaining interaction terms containing \u003cem\u003eilrs\u003c/em\u003e were retained non-uniformly across cognitive outcomes. We note that predictor variables in group LASSO regression, and regularized regression more generally, are block standardised to zero vector mean and identity unit covariance, so when used to predict the (unit variance scaled) outcome, the predictors\u0026rsquo; strength of association with the outcome are directly comparable by magnitude of their associated coefficient. The complete model coefficient output (main effects, higher order terms and interaction terms for all variables) can be viewed in our GitHub repository (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/tystan/ukbb-cog-lasso\u003c/span\u003e\u003cspan address=\"https://github.com/tystan/ukbb-cog-lasso\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eOptimal days for cognitive outcomes\u003c/p\u003e \u003cp\u003eThe optimal 24-hour day for cognitive function varied across strata, across cognitive outcomes, and was further modified by health and sociodemographic characteristics.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eOptimal 24-hr time use varies between age, sex, and BMI-defined strata\u003c/h3\u003e\n\u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e displays the estimated optimal 24-hr time-use compositions for each cognitive domain (across four compositional parts), and corresponding cognitive responses (cognitive domain z-score) across the eight age, sex, and BMI-defined strata. Across almost all groups, the optimal amount of SB and MVPA for cognitive function was greater than the strata mean, whereas the optimal amount of sleep and LPA was lower than the strata mean. For example, the average time-use composition for females aged\u0026thinsp;\u0026lt;\u0026thinsp;65 years without obesity (top row of figure) was 542 minutes of sleep (~\u0026thinsp;9hrs), 549 minutes of SB (~\u0026thinsp;9.2hr), 318 minutes of LPA (~\u0026thinsp;5.3hr), and 30 min of MVPA. Comparatively, the \u003cem\u003eoptimal\u003c/em\u003e day for global cognition within this stratum was 480 minutes of sleep (~\u0026thinsp;8hrs), 705 minutes of SB (~\u0026thinsp;11.5hrs), 214 minutes of LPA (~\u0026thinsp;3.5hrs), and 39 minutes of MVPA. Thus, to achieve an optimal 24-hr day for global cognition, on average, this stratum required approximately 60 minutes less sleep, 100 minutes less LPA, 150 minutes more SB, and 10 minutes more MVPA than the stratum\u0026rsquo;s mean composition.\u003c/p\u003e \u003cp\u003eAcross strata, the optimal amount of MVPA for cognitive function varied the most by obesity status, whereby strata with obesity consistently required lower MVPA for their optimal day composition compared to those without obesity. Within the same age and sex classifications, differences in optimal MVPA by obesity status for global cognition ranged from ~\u0026thinsp;12 minutes (females aged\u0026thinsp;\u0026gt;\u0026thinsp;65 yrs) to ~\u0026thinsp;18 minutes (males aged\u0026thinsp;\u0026gt;\u0026thinsp;65 years). We note that the variability by obesity status is related to the constrained (feasible) time-use footprint containing lower MVPA in strata with VS without obesity \u0026ndash; in other words, rather than individuals with obesity needing less MVPA to benefit cognition, their feasible limits of MVPA are lower due to the relationship between MVPA and BMI (see Supplementary File 6 for comparison of feasible limits of time-use behaviours across strata). Less consistent patterns were observed across age and sex classifications, and across other time-use behaviours.\u003c/p\u003e\n\u003ch3\u003eOptimal 24-hr time use for cognition varies between cognitive domains\u003c/h3\u003e\n\u003cp\u003eThe optimal day for cognitive function varied considerably across cognitive domains (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For example, within one stratum (e.g., females, aged\u0026thinsp;\u0026lt;\u0026thinsp;65 years, without obesity) the optimal amount of each time-use behaviour for different cognitive domains varied as follows: optimal sleep varied from 466 minutes (7.8 hrs, memory) to 492 minutes (8.2 hrs, processing speed); optimal SB varied from 695 minutes (11.6 hrs, reasoning) to 705 minutes (11.8 hrs, global cognition); optimal LPA varied from 206 minutes (3.4 hrs, processing speed) to 227 minutes (3.8hrs, memory); and optimal MVPA varied from 35 minutes (processing speed) to 51 minutes (reasoning).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eOptimal 24-hr time use is further modified by user inputs\u003c/h2\u003e \u003cp\u003eIn addition to age, sex, and BMI, other health and sociodemographic factors altered the optimal day for cognitive function to varying extents. Notably, presence/absence of TBI history had the strongest influence on optimal days, whereas additional health and sociodemographic factors (e.g., hypertension, smoking, alcohol consumption) had less impact on the optimal day within strata.\u003c/p\u003e \u003cp\u003eTo demonstrate using a practical example, consider \u0026lsquo;Person A\u0026rsquo; who has the following characteristics: female, aged\u0026thinsp;\u0026lt;\u0026thinsp;65 years, without obesity, college/university education, and no history of any health and sociodemographic factors (i.e., no hypertension, smoking, diabetes, depression, TBI, hearing loss, social isolation, or high alcohol consumption), with a current time-use composition of 7.8hrs sleep, 13.0hrs SB, 3.0hrs LPA, and 0.2hrs MVPA. To achieve their best day for global cognition, Person A would be recommended to increase their sleep (+\u0026thinsp;0.2hr), LPA (+\u0026thinsp;0.6hrs), and MVPA (+\u0026thinsp;0.5hrs), and decrease their SB (-1.2hr). In comparison, Person B has the same baseline time use and belongs to the same stratum, but instead has high school education, history of hypertension, current smoking, diabetes, depressive symptoms, hearing loss, social isolation and frequent alcohol consumption. Despite a vastly different profile of health and sociodemographic characteristics, the recommended changes for Person B are very similar to Person A: increase their sleep (+\u0026thinsp;0.3 hr), LPA (+\u0026thinsp;0.5hr), and MVPA (+\u0026thinsp;0.5hr), and decrease their SB (-1.3hrs). Notably, changing Person B\u0026rsquo;s characteristics to also include a history of TBI dramatically changes the recommendations: increase LPA (+\u0026thinsp;3.3hrs) and MVPA (+\u0026thinsp;0.2hrs) and decrease sleep (-0.4hrs) and SB (-3.1hrs). It is important to note that only 0.2% of the entire sample reported history of TBI (using the self-report variable), which may reduce confidence in these relationships.\u003c/p\u003e \u003cp\u003eInteractive user interface for \u0026lsquo;ideal day\u0026rsquo; personalisation\u003c/p\u003e \u003cp\u003eThe \u0026lsquo;ideal day\u0026rsquo; interactive personalisation tool can be viewed at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://arena2024.shinyapps.io/ideal-day\u003c/span\u003e\u003cspan address=\"https://arena2024.shinyapps.io/ideal-day\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e and the underlying code freely accessed on GitHub (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/tystan/ideal-day\u003c/span\u003e\u003cspan address=\"https://github.com/tystan/ideal-day\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The tool is divided stepwise into six main tabs. First, participants can select the cognitive outcome they are interested in predicting for (global cognition, memory, processing speed, executive function, or reasoning). In the \u0026lsquo;Demographics\u0026rsquo; tab, the user is asked to input their current age (years), current weight (kg) and height (cm) using free-text boxes, as well as their sex and highest qualification from pre-defined options. The \u0026lsquo;Health\u0026rsquo; tab then asks the user a series of multiple-choice questions (with mostly \u0026lsquo;yes\u0026rsquo;, \u0026lsquo;no\u0026rsquo;, or \u0026lsquo;unknown\u0026rsquo; response options) about modifiable dementia risk factors including history of hypertension, type 2 diabetes, depression, social isolation, hearing loss, TBI, alcohol consumption and tobacco smoking. Finally, users are required to enter their current time use (\u0026lsquo;Time-use\u0026rsquo; tab) using free text boxes, entering the number of hours per day they spend in sleep, sitting, light physical activity, and moderate-vigorous physical activity.\u003c/p\u003e \u003cp\u003eUsing these inputs, the tool then displays the user\u0026rsquo;s current day (i.e., current time-use composition) next to their \u0026lsquo;optimal day\u0026rsquo; in the \u0026lsquo;Ideal day\u0026rsquo; tab. We note that the use of \u0026lsquo;ideal\u0026rsquo; rather than \u0026lsquo;optimal\u0026rsquo; was chosen for the \u003cem\u003eSmall Steps\u003c/em\u003e study app, in response to preferences indicated during the co-design process\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn the example displayed in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, the user is interested in determining the ideal day for their global cognition. Their current time use is 7.8 h sleep, 13 h sedentary behaviour, 3 h LPA and 0.2 h MVPA per day, and based on the optimisation analysis, to achieve their personalised \u0026lsquo;ideal day\u0026rsquo; the user is advised to increase time in sleep, LPA and MVPA, and decrease time in sitting. As the Small Steps intervention aims to help participants make small, beneficial changes in behaviour which suit their preferences, needs and constraints, the final component of the Shiny app (the \u0026lsquo;Small Steps\u0026rsquo; tab) dynamically reports on the whether the relative direction from current time-use to the selected change in time-use (e.g., increasing 10 min of MVPA \u0026ndash;and decreasing sleep by 10 min of sleep) is aligned with the theoretical direction of current time-use to the optimal time-use. This feedback was operationalised using a traffic light system, where green lights indicated the proposed time-use change is moving towards the ideal day, and red lights indicated moving away from the ideal day. The methods underpinning this feature are beyond the scope of the current paper, and will be described elsewhere.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study described the development of a novel interactive user interface which could be used to generate personalised behaviour change recommendations for individuals in real time. In our proof-of-concept example, we showed that the ‘ideal day’ (i.e., the optimal balance of sleep, SB and PA in the 24-hr day, or ’24-hr time-use composition’) for cognitive outcomes differed depending on the health and sociodemographic profile (i.e., modifiable dementia risk factor profile) of participants in a sub-sample from the UK Biobank. We found that 24-hr time-use composition was associated with all five cognitive outcomes (global cognition, memory, processing speed, executive function and reasoning), and that the optimal day for cognitive function varied across strata, and across cognitive domains. Within each of the stratum, the relationship between 24-hr time-use composition and cognitive performance was altered by other health and sociodemographic characteristics. The optimal day (target durations of daily activities) generated by our interactive app varied depending on user inputs regarding their demographic and health characteristics, These potentially substantial differences in optimal day predictions demonstrate the importance of personalising 24-hr time-use interventions for health outcomes (e.g., cognitive health in older adults).\u003c/p\u003e\n\u003cp\u003eComparison with previous methodological approaches \u003c/p\u003e\n\u003cp\u003eOur study builds on the formative work of recent studies that have published “ideal”, “optimal”, or “Goldilocks” days for a single health outcome, or population \u003csup\u003e3,7,23\u003c/sup\u003e. We extend this work in five important ways, made possible by the large underpinning dataset. First, we developed a sophisticated data-driven model-selection procedure using group LASSO regression that considered all main effect candidate predictors, interactions and second-order polynomial terms. This allowed us to consider more complicated multiplicative relationships between time use, health and sociodemographic characteristics (e.g., modifiable dementia risk factors) without model overfitting. The group LASSO method overcomes the violation of the assumption of invariance under the choice of \u003cem\u003eilr \u003c/em\u003etransformation, which would occur if a single log-ratio is retained while others are discarded in standard LASSO regression \u003csup\u003e24\u003c/sup\u003e. Second, to address the same issue of invariance under alternate log-ratio transformations, we implemented a multivariate scaling method for the log-ratios. Third, we stratified our analyses by key characteristics (age, sex, and BMI) which showed strong interactions with the time-use log-ratios, ensuring differences in associations between these strata would be reflected in our final optimal day predictions. Fourth, to ensure our models best represented the empirical relationships present in the data, we tested for all pairwise interactions between included predictors, and for non-linear relationships. Fifth, we present a new method (constrained ellipsoid fencing) to improve how the empirical time-use footprint is selected for prediction. Previous methods applied univariate constraints (e.g., restricting at the 3\u003csup\u003erd\u003c/sup\u003e standard deviation) to each behaviour separately \u003csup\u003e3\u003c/sup\u003e which is incongruent with the compositional approach, and results in extrapolations into unsampled territory where there are few or no empirical data points. Taken together, we present a novel time-use optimisation pipeline which can be replicated for alternative outcomes, populations and predictor variables using our published code (available on GitHub). \u003c/p\u003e\n\u003cp\u003eComparison with previous 24-hr time use and cognition research\u003c/p\u003e\n\u003cp\u003eWe contribute new findings to a growing literature regarding the associations between 24-hr time-use composition and cognitive function in late adulthood. We found that 24-hr time-use composition was associated with cognitive function in older adults, which is congruent with some previous compositional studies \u003csup\u003e25-27\u003c/sup\u003e, but not all \u003csup\u003e28-31\u003c/sup\u003e. Patterns across the predicted optimal days in our study suggest that, compared to the “average” individual in each of the 8 age/sex/BMI strata, better cognition was associated with more time in MVPA and SB, and less time in LPA and sleep (relative to the mean time-use composition of the strata). \u003c/p\u003e\n\u003cp\u003eOur findings for MVPA and LPA are consistent with many previous non-compositional and compositional studies \u003csup\u003e32-34\u003c/sup\u003e, which report cognitive benefits of higher intensity PA, and negative albeit largely inconclusive association between LPA and cognition. We provide evidence that light intensities of PA may not be sufficient to provide cognitive benefits, as these take time from more beneficial activities such as MVPA. Our analyses show that SB also has beneficial associations, up to an optimal duration, after which benefits appear to wane. Thus, when aiming to optimise cognition, MVPA and SB may compete for time-shares within the 24-hr time window. Our findings for SB contribute to a mixed literature, whereby some studies have reported beneficial associations between SB and cognition \u003csup\u003e26\u003c/sup\u003e, whilst others have found negative (or no) associations \u003csup\u003e35,36\u003c/sup\u003e. It is increasingly recognised that the type and context of SBs may alter their association with cognitive outcomes \u003csup\u003e35,37\u003c/sup\u003e. For example, our recent work among older adults demonstrated that cognitively engaging SBs (e.g., reading, computer use) are beneficially associated, whereas cognitively passive SBs (e.g., TV watching) are detrimentally associated with cognition \u003csup\u003e26\u003c/sup\u003e. As our SB variable was derived from accelerometry in the current study, we were unable to differentiate between cognitively engaging and passive sedentary time. Notably, the mean SB time in our UK Biobank sample was considerably lower (~9.6hrs/day) than other samples exploring similar relationships in older adults (e.g., 11hrs \u003csup\u003e29\u003c/sup\u003e-12hrs \u003csup\u003e25\u003c/sup\u003e). It is likely that this contributed to the finding that optimal days required an increase in SB across all strata. Finally, for all strata, we found that optimal sleep duration was lower than the mean sleep duration. The extant literature suggests there is an inverted U-shaped relationship between sleep and cognition, whereby long or short sleep duration (e.g., \u0026lt;6 or \u0026gt;9 hr) is associated with reduced cognitive performance \u003csup\u003e38-40\u003c/sup\u003e. In the current sample, mean sleep duration ranged between 8.8 and 9.2 hr across strata, which is close to the upper bound of the range of sleep durations associated with better cognition in aforementioned studies. However, it is important to note that the Biobank sleep measure did not account for nighttime awakenings, or time awake in bed. Thus, longer sleep durations may reflect poorer sleep efficiency, where relatively longer durations of the “sleep” variable are spent lying awake in bed rather than asleep. Moreover, the derived time-use behaviours may have subsequently overestimated time in sleep and underestimated time in SB (which ranged between 9.2 and 10.5 hr/day in this sample). \u003c/p\u003e\n\u003cp\u003eImplications for time-use personalization\u003c/p\u003e\n\u003cp\u003eOur findings provide evidence that different population sub-groups require different balances of sleep, SB, LPA and MVPA in the 24-hr day for cognition, and unique combinations of sociodemographic and health characteristics may further adjust the ‘ideal day’ for cognition. This supports the Sweet Spot Hypothesis, and suggests that ‘one-size-fits-all’ approaches to time-use recommendations may not confer equitable benefits across a sample of participants. In our study, the patterns of associations were relatively consistent between the strata (i.e., for all strata, the optimal days had more MVPA and SB, and less LPA and sleep than the strata average). The differences between strata were in the recommended durations of the behaviours, which was directly linked to the bounds of the time-use footprint considered “feasible” for each of the strata. For example, the maximum duration of MVPA considered feasible for females aged \u0026gt;65 years with obesity was 110 minutes (minimum = 5 minutes), compared to almost double (205 minutes; minimum = 10 minutes) among males aged \u0026lt;65 years without obesity. Constraining the time-use footprint to a feasible range is crucial to producing behaviour change recommendations that are meaningful and achievable for the target population. \u003c/p\u003e\n\u003cp\u003eOur study describes new methods for computing personalised 24-hr time-use recommendations via an interactive interface. Time-use recommendations have, over the last two decades, moved from a focus on individual behaviours (PA, SB, sleep) towards recommendations encompassing the whole 24-hr day (e.g., Canadian 24-Hour Movement Guidelines for Adults \u003csup\u003e41\u003c/sup\u003e). However, even the most recent guidelines are merely composites of separate sets of recommendations for each of the movement behaviours. Furthermore, the guidelines apply to all people within the designated age bands, agnostic to other personal characteristics and current behaviours. Understanding the optimal balance and/or trade-offs across the 24-h day needed to maintain health has been the focus of an increasing number of studies in the past decade \u003csup\u003e42\u003c/sup\u003e. The app presented here goes beyond the one-size-fits-all approach to optimising the balance of PA, sleep and SB in a true 24-hr day, and may be an important next step towards personalised approaches to chronic disease prevention. \u003c/p\u003e\n\u003cp\u003eWith relevance to the cognitive aging and dementia prevention field, interventions (including multi-domain trials) which have incorporated a component of PA have yielded mixed findings \u003csup\u003e43\u003c/sup\u003e. It is possible that these mixed findings may be, at least in part, due to the lack of consideration of trade-offs being made with other components of the day (sleep and SB), or the lack of personalisation based on previous exposures to other modifiable risk factors for dementia. Above all, calculating and then communicating optimal days which are personalised to the individual’s sociodemographic and health characteristics is complex. Our app presents a potential solution by providing an accessible interface, co-designed with consumers, to generate and translate 24-hr activity interventions that are tailored to the individual.\u003c/p\u003e\n\u003cp\u003eStrengths and limitations\u003c/p\u003e\n\u003cp\u003eThis study has a number of strengths. We used robust model selection procedures to avoid over-fitting models, and achieve a balance between complexity and understanding. This study included a large population-based sample, and 24-hr activity and sleep data were collected using device-based methods (accelerometry) which may be less susceptible to recall bias or inaccuracy in older populations. We explored several cognitive outcomes, which strengthens evidence of domain-specific associations between 24-hr time-use composition and cognitive function in older adult populations. Finally, the user interface of the app was co-designed with community-dwelling older adults for whom the app was initially intended \u003csup\u003e22\u003c/sup\u003e. As a result, the user interface is accessible and easily to interpret, avoiding complex language. There are limitations which must also be acknowledged. Cross-sectional data were used to estimate relationships between 24-hr time-use composition and cognitive outcomes. Extending the methods presented here by using longitudinal data would also allow researchers to explicitly consider the effect of within-person changes in time use in addition to the between-person differences that the current recommendations are based on. Twenty-four-hour time-use data were measured using wrist-worn accelerometers which are not considered the gold standard for measuring postural changes (i.e., differentiating between sitting and sleeping). It is possible that some SBs (e.g., time awake in bed) were classified as sleep, resulting in the over-estimation of time in sleep. This potential discrepancy may have contributed to the findings that 1) participants engaged in a lower-than-expected amount of SB per day (9hrs) compared to other cohorts of similar age, and 2) more time in SB was associated with better cognitive function. Finally, the UK Biobank sample used in this study are relatively homogenous in their characteristics and don’t reflect the most at-risk groups for cognitive decline and dementia (particularly Alzheimer’s disease). Thus, this approach to personalising 24-hr time-use interventions should be explored in datasets whereby the population are more representative of at-risk populations (e.g., those from low-to-middle income countries, with greater ethnic diversity) \u003csup\u003e6,11\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFuture directions\u003c/p\u003e\n\u003cp\u003eThis first-of-its-kind, proof-of-concept study provides important foundations for future time-use personalisation research and intervention studies to build upon. We anticipate several key future directions for this work. First, as this analysis pipeline is able to be replicated for alternative study types (e.g., time-to-event analyses), future studies should explore the utility of time-use personalisation for clinical outcomes, such as onset of Alzheimer’s disease or other chronic diseases. Second, with relevance to the association between time-use composition and cognitive performance, these relationships should be explored longitudinally and account for additional factors which may have confounding effects, such as genetics (e.g., carriage of dementia risk genes such as apolipoprotein E ε4). Third, the pipeline can be extended to create personalised ‘optimal days’ for multiple response variables concurrently (i.e., the Goldilocks method \u003csup\u003e3\u003c/sup\u003e). Fourth, future studies should consider the context in which activity occurs (e.g., cognitively active vs. cognitively passive SB; MVPA occurring in work vs leisure) to further personalise optimal days for health.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available from the UK Biobank but restrictions apply to the availability of these data, which were used under licence for the current study, and so are not publicly available.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe underlying code for the data preparation, cleaning, analysis and diagnostics for this study is freely accessible on GitHub and can be accessed via this link (https://github.com/tystan/ukbb-cog-lasso\u003cu\u003e)\u0026nbsp;\u003c/u\u003ein addition to the \u0026lsquo;ideal day\u0026rsquo; interactive personalisation tool hosted at https://arena2024.shinyapps.io/ideal-day/with the associated underlying code also freely accessed on GitHub too (\u003cu\u003ehttps://github.com/tystan/ideal-day)\u003c/u\u003e.\u003c/p\u003e\u003cp\u003e \u003ch2\u003eCompeting interests:\u003c/h2\u003e \u003cp\u003eAll authors declare no financial or non-financial competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eMM conceptualized the study, supported data analysis, and prepared the manuscript. TS conceptualized the study, conducted data analysis and contributed to manuscript development. TO and AM contributed to manuscript development. AS led the co-design of the app interface, conceptualized the study, and contributed to manuscript development. DD conceptualized the study, supported data analysis, and prepared the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThis research has been conducted using the UK Biobank Resource under Application Number 62254. We wish to acknowledge the wider Small Steps team for their contribution to the co-design of the digital interface, detailed elsewhere.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eHarvey, A. \u003cem\u003eet al.\u003c/em\u003e The future of technologies for personalised medicine. \u003cem\u003eNew Biotechnology\u003c/em\u003e 29, 625\u0026ndash;633, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.nbt.2012.03.009\u003c/span\u003e\u003cspan address=\"10.1016/j.nbt.2012.03.009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoltermann, A. \u003cem\u003eet al.\u003c/em\u003e 24-Hour Physical Behavior Balance for Better Health for All: \u0026ldquo;The Sweet-Spot Hypothesis\u0026rdquo;. \u003cem\u003eSports Medicine - Open\u003c/em\u003e 7, 98, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s40798-021-00394-8\u003c/span\u003e\u003cspan address=\"10.1186/s40798-021-00394-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumuid, D. \u003cem\u003eet al.\u003c/em\u003e Goldilocks Days: optimising children\u0026rsquo;s time use for health and well-being. \u003cem\u003eJournal of Epidemiology and Community Health\u003c/em\u003e 76, 301, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/jech-2021-216686\u003c/span\u003e\u003cspan address=\"10.1136/jech-2021-216686\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumuid, D. \u003cem\u003eet al.\u003c/em\u003e Your best day: An interactive app to translate how time reallocations within a 24-hour day are associated with health measures. \u003cem\u003ePLOS ONE\u003c/em\u003e 17, e0272343, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pone.0272343\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0272343\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSommerlad, A. \u0026amp; Mukadam, N. Evaluating risk of dementia in older people: a pathway to personalized prevention? \u003cem\u003eEuropean Heart Journal\u003c/em\u003e 41, 4034\u0026ndash;4036, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/eurheartj/ehaa691\u003c/span\u003e\u003cspan address=\"10.1093/eurheartj/ehaa691\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLivingston, G. \u003cem\u003eet al.\u003c/em\u003e Dementia prevention, intervention, and care: 2024 report of the \u0026lt;\u0026thinsp;em\u0026thinsp;\u0026gt;\u0026thinsp;Lancet\u0026lt;/em\u0026thinsp;\u0026gt;\u0026thinsp;standing Commission. \u003cem\u003eThe Lancet\u003c/em\u003e 404, 572\u0026ndash;628, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(24)01296-0\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(24)01296-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumuid, D. \u003cem\u003eet al.\u003c/em\u003e The \u0026ldquo;Goldilocks Day\u0026rdquo; for Children's Skeletal Health: Compositional Data Analysis of 24-Hour Activity Behaviors. \u003cem\u003eJournal of Bone and Mineral Research\u003c/em\u003e 35, 2393\u0026ndash;2403, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/jbmr.4143\u003c/span\u003e\u003cspan address=\"10.1002/jbmr.4143\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMellow, M. L. \u003cem\u003eet al.\u003c/em\u003e \u0026lsquo;Small Steps\u0026rsquo; towards improving 24-hour time-use behaviours to decrease the risk of dementia: protocol for a personalised, web-based randomised controlled trial in community-dwelling older adults. \u003cem\u003emedRxiv\u003c/em\u003e, 2025.2005.2026.25328336, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1101/2025.05.26.25328336\u003c/span\u003e\u003cspan address=\"10.1101/2025.05.26.25328336\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSudlow, C. \u003cem\u003eet al.\u003c/em\u003e UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. \u003cem\u003ePLOS Medicine\u003c/em\u003e 12, e1001779, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1371/journal.pmed.1001779\u003c/span\u003e\u003cspan address=\"10.1371/journal.pmed.1001779\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eshiny: Web Application Framework for R (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLivingston, G. \u003cem\u003eet al.\u003c/em\u003e Dementia prevention, intervention, and care: 2020 report of the \u0026lt;\u0026thinsp;em\u0026thinsp;\u0026gt;\u0026thinsp;Lancet\u0026lt;/em\u0026thinsp;\u0026gt;\u0026thinsp;Commission. \u003cem\u003eThe Lancet\u003c/em\u003e 396, 413\u0026ndash;446, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S0140-6736(20)30367-6\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(20)30367-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWalmsley, R. \u003cem\u003eet al.\u003c/em\u003e Reallocation of time between device-measured movement behaviours and risk of incident cardiovascular disease. \u003cem\u003eBritish Journal of Sports Medicine\u003c/em\u003e 56, 1008, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bjsports-2021-104050\u003c/span\u003e\u003cspan address=\"10.1136/bjsports-2021-104050\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePalarea-Albaladejo, J. \u0026amp; Mart\u0026iacute;n-Fern\u0026aacute;ndez, J. A. zCompositions \u0026mdash; R package for multivariate imputation of left-censored data under a compositional approach. \u003cem\u003eChemometrics and Intelligent Laboratory Systems\u003c/em\u003e 143, 85\u0026ndash;96, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chemolab.2015.02.019\u003c/span\u003e\u003cspan address=\"10.1016/j.chemolab.2015.02.019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMart\u0026iacute;n-Fern\u0026aacute;ndez, J. A., Hron, K., Templ, M., Filzmoser, P. \u0026amp; Palarea-Albaladejo, J. Model-based replacement of rounded zeros in compositional data: Classical and robust approaches. \u003cem\u003eComputational Statistics \u0026amp; Data Analysis\u003c/em\u003e 56, 2688\u0026ndash;2704, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.csda.2012.02.012\u003c/span\u003e\u003cspan address=\"10.1016/j.csda.2012.02.012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCampbell, T. \u0026amp; Cullen, B. Estimating the effect of physical activity on cognitive function within the UK Biobank cohort. \u003cem\u003eInternational Journal of Epidemiology\u003c/em\u003e 52, 1592\u0026ndash;1611, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/ije/dyad009\u003c/span\u003e\u003cspan address=\"10.1093/ije/dyad009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eR: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria, 2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFriedman, J. H., Hastie, T. \u0026amp; Tibshirani, R. Regularization Paths for Generalized Linear Models via Coordinate Descent. \u003cem\u003eJournal of Statistical Software\u003c/em\u003e 33, 1\u0026ndash;22, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.18637/jss.v033.i01\u003c/span\u003e\u003cspan address=\"10.18637/jss.v033.i01\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuan, M. \u0026amp; Lin, Y. Model Selection and Estimation in Regression with Grouped Variables. \u003cem\u003eJournal of the Royal Statistical Society Series B: Statistical Methodology\u003c/em\u003e 68, 49\u0026ndash;67, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/j.1467-9868.2005.00532.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1467-9868.2005.00532.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBreheny, P. \u0026amp; Huang, J. Group descent algorithms for nonconvex penalized linear and logistic regression models with grouped predictors. \u003cem\u003eStatistics and Computing\u003c/em\u003e 25, 173\u0026ndash;187, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11222-013-9424-2\u003c/span\u003e\u003cspan address=\"10.1007/s11222-013-9424-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHastie, T., Tibshirani, R. \u0026amp; Friedman, J. H. \u003cem\u003eThe Elements of Statistical Learning: Data Mining, Inference, and Prediction\u003c/em\u003e. 2nd edn, (Springer, 2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMart\u0026iacute;n-Fern\u0026aacute;ndez, J.-A., Daunis i Estadella, J. \u0026amp; Mateu-Figueras, G. On the interpretation of differences between groups for compositional data. \u003cem\u003eSORT: statistics and operations research transactions\u003c/em\u003e 39, 0231\u0026ndash;0252 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlake, H. T. \u003cem\u003eet al.\u003c/em\u003e Co-design of a digital 24-hour time-use intervention with older adults and allied health professionals. \u003cem\u003eFrontiers in Digital Health\u003c/em\u003e Volume 7\u0026ndash;2025, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fdgth.2025.1544489\u003c/span\u003e\u003cspan address=\"10.3389/fdgth.2025.1544489\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZahran, S. \u003cem\u003eet al.\u003c/em\u003e Optimal levels of sleep, sedentary behaviour, and physical activity needed to support cognitive function in children of the early years. \u003cem\u003eBMC Pediatrics\u003c/em\u003e 24, 735, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12887-024-05186-z\u003c/span\u003e\u003cspan address=\"10.1186/s12887-024-05186-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTolosana-Delgado, R., Arya, E., Khodadadzadeh, M. \u0026amp; Boogaart, K. \u003cem\u003eOn machine learning algorithms and compositional data\u003c/em\u003e. (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCollins, A. M. \u003cem\u003eet al.\u003c/em\u003e 24-Hour time use and cognitive performance in late adulthood: results from the Investigating Gains in Neurocognition in an Intervention Trial of Exercise (IGNITE) study. \u003cem\u003eAge and Ageing\u003c/em\u003e 54, afaf072, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/ageing/afaf072\u003c/span\u003e\u003cspan address=\"10.1093/ageing/afaf072\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMellow, M. L. \u003cem\u003eet al.\u003c/em\u003e Should We Work Smarter or Harder for Our Health? A Comparison of Intensity and Domain-Based Time-Use Compositions and Their Associations With Cognitive and Cardiometabolic Health. \u003cem\u003eThe Journals of Gerontology: Series A\u003c/em\u003e 79, glae233, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/gerona/glae233\u003c/span\u003e\u003cspan address=\"10.1093/gerona/glae233\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDumuid, D. \u003cem\u003eet al.\u003c/em\u003e Does APOE ɛ4 Status Change How 24-Hour Time-Use Composition Is Associated with Cognitive Function? An Exploratory Analysis Among Middle-to-Older Adults. \u003cem\u003eJournal of Alzheimer\u0026rsquo;s Disease\u003c/em\u003e 88, 1157\u0026ndash;1165, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3233/JAD-220181\u003c/span\u003e\u003cspan address=\"10.3233/JAD-220181\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMellow, M. L. \u003cem\u003eet al.\u003c/em\u003e Cross-sectional associations between 24-hour time-use composition, grey matter volume and cognitive function in healthy older adults. \u003cem\u003eInternational Journal of Behavioral Nutrition and Physical Activity\u003c/em\u003e 21, 11, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12966-023-01557-4\u003c/span\u003e\u003cspan address=\"10.1186/s12966-023-01557-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMellow, M. L. \u003cem\u003eet al.\u003c/em\u003e Twenty-four-hour time-use composition and cognitive function in older adults: cross-sectional findings of the ACTIVate study. \u003cem\u003eFrontiers in Human Neuroscience\u003c/em\u003e Volume 16\u0026ndash;2022, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fnhum.2022.1051793\u003c/span\u003e\u003cspan address=\"10.3389/fnhum.2022.1051793\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWu, Y. \u003cem\u003eet al.\u003c/em\u003e Analysis of the 24-h activity cycle: An illustration examining the association with cognitive function in the Adult Changes in Thought study. \u003cem\u003eFrontiers in Psychology\u003c/em\u003e Volume 14\u0026ndash;2023, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fpsyg.2023.1083344\u003c/span\u003e\u003cspan address=\"10.3389/fpsyg.2023.1083344\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBalbim, G. M. \u003cem\u003eet al.\u003c/em\u003e The Association of the 24-Hour Activity Cycle Profiles With Cognition in Older Adults With Mild Cognitive Impairment: A Cross-Sectional Study. \u003cem\u003eThe Journals of Gerontology: Series A\u003c/em\u003e 79, glae099, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/gerona/glae099\u003c/span\u003e\u003cspan address=\"10.1093/gerona/glae099\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMellow, M. L. \u003cem\u003eet al.\u003c/em\u003e How are combinations of physical activity, sedentary behaviour and sleep related to cognitive function in older adults? A systematic review. \u003cem\u003eExperimental Gerontology\u003c/em\u003e 159, 111698, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.exger.2022.111698\u003c/span\u003e\u003cspan address=\"10.1016/j.exger.2022.111698\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErickson, K. I., Donofry, S. D., Sewell, K. R., Brown, B. M. \u0026amp; Stillman, C. M. Cognitive Aging and the Promise of Physical Activity. \u003cem\u003eAnnual Review of Clinical Psychology\u003c/em\u003e 18, 417\u0026ndash;442, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1146/annurev-clinpsy-072720-014213\u003c/span\u003e\u003cspan address=\"10.1146/annurev-clinpsy-072720-014213\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSingh, B. \u003cem\u003eet al.\u003c/em\u003e Effectiveness of exercise for improving cognition, memory and executive function: a systematic umbrella review and meta-meta-analysis. \u003cem\u003eBritish Journal of Sports Medicine\u003c/em\u003e (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCoelho, L. \u003cem\u003eet al.\u003c/em\u003e The association between sedentary behavior and cognitive ability in older adults. \u003cem\u003eAging Clinical and Experimental Research\u003c/em\u003e 32, 2339\u0026ndash;2347, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s40520-019-01460-8\u003c/span\u003e\u003cspan address=\"10.1007/s40520-019-01460-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaasakkers, C. M. \u003cem\u003eet al.\u003c/em\u003e The Association of Sedentary Behaviour and Cognitive Function in People Without Dementia: A Coordinated Analysis Across Five Cohort Studies from COSMIC. \u003cem\u003eSports Medicine\u003c/em\u003e 50, 403\u0026ndash;413, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s40279-019-01186-7\u003c/span\u003e\u003cspan address=\"10.1007/s40279-019-01186-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOlanrewaju, O., Stockwell, S., Stubbs, B. \u0026amp; Smith, L. Sedentary behaviours, cognitive function, and possible mechanisms in older adults: a systematic review. \u003cem\u003eAging Clinical and Experimental Research\u003c/em\u003e 32, 969\u0026ndash;984, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s40520-019-01457-3\u003c/span\u003e\u003cspan address=\"10.1007/s40520-019-01457-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLo, J. C., Groeger, J. A., Cheng, G. H., Dijk, D.-J. \u0026amp; Chee, M. W. L. Self-reported sleep duration and cognitive performance in older adults: a systematic review and meta-analysis. \u003cem\u003eSleep Medicine\u003c/em\u003e 17, 87\u0026ndash;98, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.sleep.2015.08.021\u003c/span\u003e\u003cspan address=\"10.1016/j.sleep.2015.08.021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGildner, T. E., Liebert, M. A., Kowal, P., Chatterji, S. \u0026amp; Snodgrass, J. J. Associations between sleep duration, sleep quality, and cognitive test performance among older adults from six middle income countries: results from the Study on Global Ageing and Adult Health (SAGE). \u003cem\u003eJournal of clinical sleep medicine\u003c/em\u003e 10, 613\u0026ndash;621 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDevore, E. E., Grodstein, F. \u0026amp; Schernhammer, E. S. Sleep duration in relation to cognitive function among older adults: a systematic review of observational studies. \u003cem\u003eNeuroepidemiology\u003c/em\u003e 46, 57\u0026ndash;78 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoss, R. \u003cem\u003eet al.\u003c/em\u003e Canadian 24-Hour Movement Guidelines for Adults aged 18\u0026ndash;64 years and Adults aged 65 years or older: an integration of physical activity, sedentary behaviour, and sleep. \u003cem\u003eApplied Physiology, Nutrition, and Metabolism\u003c/em\u003e 45, S57-S102, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1139/apnm-2020-0467\u003c/span\u003e\u003cspan address=\"10.1139/apnm-2020-0467\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMiatke, A. \u003cem\u003eet al.\u003c/em\u003e The association between reallocations of time and health using compositional data analysis: a systematic scoping review with an interactive data exploration interface. \u003cem\u003eInternational Journal of Behavioral Nutrition and Physical Activity\u003c/em\u003e 20, 127, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12966-023-01526-x\u003c/span\u003e\u003cspan address=\"10.1186/s12966-023-01526-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCastro, C. B. \u003cem\u003eet al.\u003c/em\u003e Multi-Domain Interventions for Dementia Prevention\u0026ndash;A Systematic Review. \u003cem\u003eThe Journal of nutrition, health and aging\u003c/em\u003e 27, 1271\u0026ndash;1280, doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12603-023-2046-2\u003c/span\u003e\u003cspan address=\"10.1007/s12603-023-2046-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2023).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"time use, personalised medicine, digital tool, cognition","lastPublishedDoi":"10.21203/rs.3.rs-6897341/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6897341/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePersonalised interventions which optimise the balance of physical activity (PA), sleep and sedentary behaviour (i.e., time use) in the 24-hr day may be more effective than one-size-fits-all approaches. We present an interactive app to personalise 24-hr time use based on individuals\u0026rsquo; health and sociodemographic characteristics. Analyses used cross-sectional data from 53,057 UK Biobank participants. Average daily time use was measured using 7-day accelerometry data and expressed as a 24-hr composition using isometric log-ratio transformation. Five cognitive composites were derived from web-based tests. Regularized linear regression examined the relationship between 24-hr time-use composition and cognition, with sociodemographic and health characteristics as additional predictors. Model estimates were used to estimate optimized cognition based on the interaction of 24-hr time-use composition and personal characteristics. Our \u0026lsquo;ideal day\u0026rsquo; app delivers personalised 24-hr time-use recommendations tailored to individual characteristics. We demonstrate that personalisation of time-use interventions can be achieved in real time using open-source software.\u003c/p\u003e","manuscriptTitle":"An interactive tool to personalise 24-hour activity, sitting and sleep prescription for optimal health outcomes","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-30 08:52:27","doi":"10.21203/rs.3.rs-6897341/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-07-15T09:12:09+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-15T05:13:00+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-07-14T12:02:12+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"239265768848139960031318939792508201300","date":"2025-06-23T14:48:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"169667338116651136727101229288671387006","date":"2025-06-23T09:36:54+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-06-22T13:42:50+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-18T23:58:55+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-06-18T04:45:31+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Digital Medicine","date":"2025-06-15T08:51:31+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-digital-medicine","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"npjdigitalmed","sideBox":"Learn more about [npj Digital Medicine](http://www.nature.com/npjdigitalmed/)","snPcode":"41746","submissionUrl":"https://submission.springernature.com/new-submission/41746/3","title":"npj Digital Medicine","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1b1d7bf1-4c6a-41ce-bcfa-44b9c6b386ff","owner":[],"postedDate":"June 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":50596905,"name":"Health sciences/Risk factors"},{"id":50596906,"name":"Health sciences/Health care/Disease prevention/Lifestyle modification"},{"id":50596907,"name":"Health sciences/Health care/Disease prevention/Preventive medicine"},{"id":50596908,"name":"Health sciences/Risk factors"},{"id":50596909,"name":"Health sciences/Health care/Disease prevention/Lifestyle modification"},{"id":50596910,"name":"Health sciences/Health care/Disease prevention/Preventive medicine"},{"id":50596911,"name":"Health sciences/Risk factors"},{"id":50596912,"name":"Health sciences/Health care/Disease prevention/Lifestyle modification"},{"id":50596913,"name":"Health sciences/Health care/Disease prevention/Preventive medicine"}],"tags":[],"updatedAt":"2026-03-23T16:04:18+00:00","versionOfRecord":{"articleIdentity":"rs-6897341","link":"https://doi.org/10.1038/s41746-026-02542-4","journal":{"identity":"npj-digital-medicine","isVorOnly":false,"title":"npj Digital Medicine"},"publishedOn":"2026-03-17 15:59:24","publishedOnDateReadable":"March 17th, 2026"},"versionCreatedAt":"2025-06-30 08:52:27","video":"","vorDoi":"10.1038/s41746-026-02542-4","vorDoiUrl":"https://doi.org/10.1038/s41746-026-02542-4","workflowStages":[]},"version":"v1","identity":"rs-6897341","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6897341","identity":"rs-6897341","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}