Akiko Kishi Svensson

Background: Several studies have investigated the association between metabolic syndrome (MetS) and physical activity (PA). However, further research is needed using objective measures of PA in free-living conditions, while also accounting for the time-variant nature of MetS. This study aimed to: (1) investigate the association between wearable device-measured step count and 5-year MetS incidence in generally healthy Japanese participants, using annual health check-up (AHC) data and interval-censored survival analysis; and (2) assess the current, and recently revised, reference value (8,000 steps/day) of Japan’s Ministry of Health, Labour and Welfare (MHLW), and (3) investigate the possibility of non-linear associations between daily step count and MetS. Methods: This longitudinal prospective observation study identified average daily step count per year/person as the main exposure, and MetS incidence, defined according to Japanese guidelines, as the main outcome. The main analysis included 730 participants without MetS or pre-MetS at baseline. An interval-censored Cox model was applied to assess MetS incidence using time-to-event data. Results: Every 1,000 steps added to the average step count was significantly and inversely associated with incident MetS in adjusted models [Model 1: HR = 0.92; 95% CI: 0.85, 1.00; Model 2: HR = 0.91; 95% CI: 0.83, 0.99] (p < 0.05). The current reference value (8,000 steps/day) also indicated a significant inverse association [Model 1: HR = 0.32; 95% CI: 0.14, 0.71; Model 2: HR = 0.26; 95% CI: 0.11, 0.61] (p < 0.01). Higher daily step counts lowered the risk of MetS incidence according to the amount of steps up to a step count of 12,000, however, the dose-response effect was attenuated beyond 12,000 steps/day. Conclusions: Each additional 1,000 daily steps was associated with a 9% reduction in the risk of developing MetS among healthy participants. The reference value (8,000 steps/day) was associated with a 74% reduction in MetS risk.

Sleep disturbances are prevalent among elderly populations and are linked to various health complications. Understanding the underlying biological mechanisms contributing to sleep disorders is crucial for developing targeted interventions. In this study, we measured 355 plasma proteins in an elderly Japanese cohort (n=77) using a high-throughput proteomic platform. Additionally, we collected over 25,000 person-days of physical activity and sleep behavior data from wrist-worn wearable devices, focusing on total sleep time (TST) across 24 h and daytime sleep. Fragmented sleep was observed as one of the most prevalent sleep disturbances in this population. In protein expression analysis, we identified 9 protein biomarkers associated with increased secondary sleep TST, defined as additional sleep episodes outside of the main sleep episode within 24 h. These findings may suggest disruptions in circadian rhythms or underlying health conditions. Functional analysis revealed that biological processes related to inflammation play a significant role in regulating sleep behavior. Further analysis showed an association of 12 proteins with daytime sleep and 5 proteins with afternoon sleep. Overall, this study identified inflammatory biomarkers and biological processes associated with sleep behavior in the elderly, presenting promising opportunities for developing diagnostic tools and targeted clinical interventions.

Objectives: Many Japanese adults experience poor sleep quality. Prolonged sedentary behavior negatively affects health, including sleep-associated problems. We investigated the association between sedentary behavior and sleep quality among urban white-collar workers in Japan with metabolic syndrome or at risk of metabolic syndrome, a population thought to have particularly prolonged sedentary time. Methods: This study was conducted using a prospective cohort study design as a secondary analysis of participants’ data in a lifestyle intervention program using a smartphone application in 5 different companies (>1000 employees) in Tokyo from December 2018 to March 2019. The primary outcome was sleep quality, obtained as the daily score of relevant items in the St Mary’s Hospital Sleep Questionnaire (SMHSQ) during the study period. The primary exposure was daily sedentary time (hours), which was objectively measured using a wristwatch-type wearable device (Fitbit Versa). Multilevel linear mixed effects analysis adjusted for covariates was used to estimate between-person and within-person associations. Sensitivity analyses were used to investigate between-/within-person associations between sedentary time and the scores of each item related to sleep quality on the SMHSQ. Results: A longer sedentary time was inversely associated with the SMHSQ scores related to sleep quality in multivariable-adjusted models at the individual level (β coefficient −.046; SE 0.016; 95% CI, −.077 to −.015). Longer sedentary time was also inversely associated with each sleep satisfaction-related SMHSQ item. However, no significant associations were observed at the population level. Conclusions: Longer sedentary time was inversely associated with sleep quality at the individual level.

Cardiovascular diseases (CVDs) remain a leading cause of death globally. According to the American Heart Association, approximately 19.1 million deaths were attributed to CVDs in 2020, in particular, ischemic heart disease and stroke. Several known risk factors for CVDs include smoking, alcohol consumption, lack of regular physical activity, and diabetes. The last decade has been characterized by widespread diffusion in the use of wristband-style wearable devices which can monitor and collect heart rate data, among other information. Wearable devices allow the analysis and interpretation of physiological and activity data obtained from the wearer and can therefore be used to monitor and prevent potential CVDs. However, these data are often provided in a manner that does not allow the general user to immediately comprehend possible health risks, and often require further analytics to draw meaningful conclusions. In this paper, we propose a disentangled variational autoencoder (β-VAE) with a bidirectional long short-term memory network (BiLSTM) backend to detect in an unsupervised manner anomalies in heart rate data collected during sleep time with a wearable device from eight heterogeneous participants. Testing was performed on the mean heart rate sampled both at 30 s and 1 min intervals. We compared the performance of our model with other well-known anomaly detection algorithms, and we found that our model outperformed them in almost all considered scenarios and for all considered participants. We also suggest that wearable devices may benefit from the integration of anomaly detection algorithms, in an effort to provide users more processed and straightforward information.

(1) Background: This study examined the cross-sectional association between metabolic syndrome (MetS) status classified into three groups and daily physical activity (PA; step count and active minutes) using a wearable device in Japanese office workers. (2) Methods: This secondary analysis used data from 179 participants in the intervention group of a randomized controlled trial for 3 months. Individuals who had received an annual health check-up and had MetS or were at a high risk of MetS based on Japanese guidelines were asked to use a wearable device and answer questionnaires regarding their daily life for the entire study period. Multilevel mixed-effects logistic regression models adjusted for covariates associated with MetS and PA were used to estimate associations. A sensitivity analysis investigated the associations between MetS status and PA level according to the day of the week. (3) Results: Compared to those with no MetS, those with MetS were not significantly associated with PA, while those with pre-MetS were inversely associated with PA [step count Model 3: OR = 0.60; 95% CI: 0.36, 0.99; active minutes Model 3: OR = 0.62; 95% CI: 0.40, 0.96]. In the sensitivity analysis, day of the week was an effect modifier for both PA (p < 0.001). (4) Conclusions: Compared to those with no MetS, those with pre-MetS, but not MetS, showed significantly lower odds of reaching their daily recommended PA level. Our findings suggest that the day of the week could be a modifier for the association between MetS and PA. Further research with longer study periods and larger sample sizes are needed to confirm our results.

The association between obesity and psychological stress is ambiguous. The aim is to investigate the association between metabolic syndrome (MetS) and body mass index (BMI), respectively, with occupational stress among Japanese office workers. The study is a secondary analysis of the intervention group from a randomized controlled trial. There are 167 participants included in the analysis. Occupational stress is self-reported using the Brief Job Stress Questionnaire (BJSQ). BMI and the classification of MetS/pre-MetS was based on the participants’ annual health check-up data. The primary exposure is divided into three groups: no MetS, pre-MetS, and MetS in accordance with Japanese guidelines. The secondary exposure, BMI, remains as a continuous variable. Multiple linear regression is implemented. Sensitivity analyses are stratified by sleep satisfaction. Pre-MetS is significantly associated with occupational stress (7.84 points; 95% CI: 0.17, 15.51). Among participants with low sleep satisfaction, pre-MetS (14.09 points; 95% CI: 1.71, 26.48), MetS (14.72 points; 95% CI: 0.93, 28.51), and BMI (2.54 points; 95% CI: 0.05, 4.99) are all significantly associated with occupational stress. No significant associations are observed in participants with high sleep satisfaction. The findings of this study indicate that sleep satisfaction may modify the association between MetS and BMI, respectively, and occupational stress.

Background: Reducing the number of items in a questionnaire while maintaining relevant information is important as it is associated with advantages such as higher respondent engagement and reduced response error. However, in health care, after the original design, an a posteriori check of the included items in a questionnaire is often overlooked or considered to be of minor importance. When conducted, this is often based on a single selected method. We argue that before finalizing any lifestyle questionnaire, a posteriori validation should always be conducted using multiple approaches to ensure the robustness of the results. Objective: The objectives of this study are to compare the results of two statistical methods for item reduction (variance inflation factor [VIF] and factor analysis [FA]) in a lifestyle questionnaire constructed by combining items from different sources and analyze the different results obtained from the 2 methods and the conclusions that can be made about the original items. Methods: Data were collected from 79 participants (heterogeneous in age and sex) with a high risk of metabolic syndrome working in a financial company based in Tokyo. The lifestyle questionnaire was constructed by combining items (asked with daily, weekly, and monthly frequency) from multiple validated questionnaires and other selected questions. Item reduction was conducted using VIF and exploratory FA. Adequacy tests were used to check the data distribution and sampling adequacy. Results: Among the daily and weekly questions, both VIF and FA identified redundancies in sleep-related items. Among the monthly questions, both approaches identified redundancies in stress-related items. However, the number of items suggested for reduction often differed: VIF suggested larger reductions than FA for daily questions but fewer reductions for weekly questions. Adequacy tests always confirmed that the structural detection was adequate for the considered items. Conclusions: As expected, our analyses showed that VIF and FA produced both similar and different findings, suggesting that questionnaire designers should consider using multiple methods for item reduction. Our findings using both methods indicate that many questions, especially those related to sleep, are redundant, indicating that the considered lifestyle questionnaire can be shortened.

Physiological time series are affected by many factors, making them highly nonlinear and nonstationary. As a consequence, heart rate time series are often considered difficult to predict and handle. However, heart rate behavior can indicate underlying cardiovascular and respiratory diseases as well as mood disorders. Given the importance of accurate modeling and reliable predictions of heart rate fluctuations for the prevention and control of certain diseases, it is paramount to identify models with the best performance in such tasks. The objectives of this study were to compare the results of three different forecasting models (Autoregressive Model, Long Short-Term Memory Network, and Convolutional Long Short-Term Memory Network) trained and tested on heart rate beats per minute data obtained from twelve heterogeneous participants and to identify the architecture with the best performance in terms of modeling and forecasting heart rate behavior. Heart rate beats per minute data were collected using a wearable device over a period of 10 days from twelve different participants who were heterogeneous in age, sex, medical history, and lifestyle behaviors. The goodness of the results produced by the models was measured using both the mean absolute error and the root mean square error as error metrics. Despite the three models showing similar performance, the Autoregressive Model gave the best results in all settings examined. For example, considering one of the participants, the Autoregressive Model gave a mean absolute error of 2.069 (compared to 2.173 of the Long Short-Term Memory Network and 2.138 of the Convolutional Long Short-Term Memory Network), achieving an improvement of 5.027% and 3.335%, respectively. Similar results can be observed for the other participants. The findings of the study suggest that regardless of an individual’s age, sex, and lifestyle behaviors, their heart rate largely depends on the pattern observed in the previous few minutes, suggesting that heart rate can be reasonably regarded as an autoregressive process. The findings also suggest that minute-by-minute heart rate prediction can be accurately performed using a linear model, at least in individuals without pathologies that cause heartbeat irregularities. The findings also suggest many possible applications for the Autoregressive Model, in principle in any context where minute-by-minute heart rate prediction is required (arrhythmia detection and analysis of the response to training, among others).

Identification of medical conditions using claims data is generally conducted with algorithms based on subject-matter knowledge. However, these claims-based algorithms (CBAs) are highly dependent on the knowledge level and not necessarily optimized for target conditions. We investigated whether machine learning methods can supplement researchers’ knowledge of target conditions in building CBAs. Retrospective cohort study using a claims database combined with annual health check-up results of employees’ health insurance programs for fiscal year 2016–17 in Japan (study population for hypertension, N = 631,289; diabetes, N = 152,368; dyslipidemia, N = 614,434). We constructed CBAs with logistic regression, k-nearest neighbor, support vector machine, penalized logistic regression, tree-based model, and neural network for identifying patients with three common chronic conditions: hypertension, diabetes, and dyslipidemia. We then compared their association measures using a completely hold-out test set (25% of the study population). Among the test cohorts of 157,822, 38,092, and 153,608 enrollees for hypertension, diabetes, and dyslipidemia, 25.4%, 8.4%, and 38.7% of them had a diagnosis of the corresponding condition. The areas under the receiver operating characteristic curve (AUCs) of the logistic regression with/without subject-matter knowledge about the target condition were .923/.921 for hypertension, .957/.938 for diabetes, and .739/.747 for dyslipidemia. The logistic lasso, logistic elastic-net, and tree-based methods yielded AUCs comparable to those of the logistic regression with subject-matter knowledge: .923-.931 for hypertension; .958-.966 for diabetes; .747-.773 for dyslipidemia. We found that machine learning methods can attain AUCs comparable to the conventional knowledge-based method in building CBAs.