# Daily Mood, Screen Time, and Steps in Offspring of Parents with Bipolar Disorder **URL:** \ **Category:** Use Cases\ **Created:** [November 7, 2026, 11:17am UTC](https://forum.avicennaresearch.com/t/daily-mood-screen-time-and-steps-in-offspring-of-parents-with-bipolar-disorder/2248 "2026-11-07T11:17:08Z")\ **Posts on this page:** 1\ **Page:** 1
**Author:** ![zahra](https://forum.avicennaresearch.com/letter_avatar_proxy/v4/letter/z/aeb1de/32.png) [@zahra](https://forum.avicennaresearch.com/u/zahra)\ **Post date:** [November 7, 2026, 11:17am UTC](https://forum.avicennaresearch.com/t/daily-mood-screen-time-and-steps-in-offspring-of-parents-with-bipolar-disorder/2248/1 "2026-11-07T11:17:08Z")
Do the things people do in a day change how they feel a few hours later, or does the feeling come first? Physical activity is often associated with better mood, while smartphone use has been linked to poorer mental health. But few studies have examined how these behaviors and feelings predict each other throughout the day, particularly in people with a family history of mood disorders. This study measured both sides as the day went on. For 14 days, 128 adults answered five short mood surveys a day in the Avicenna app (then Ethica). In the background, the app recorded their step count and when their phone screen was on. Because each survey could be matched with the hours of activity just before it, the researchers could ask whether a more active stretch was followed by a better mood, and whether a stretch of heavy phone use was followed by a worse one. Most participants were adult children of parents with bipolar disorder, followed since 1997 in the **Dutch Bipolar Offspring Study** at **Erasmus MC Sophia Children’s Hospital** in Rotterdam. The analysis was carried out with co-authors at Tilburg University and the University of Michigan. > When you pair surveys with passive sensing, decide in advance which stretch of sensor data belongs to each survey, and record how much of that stretch the phone actually captured. ## Study at a Glance | Study feature | Detail | | --- | --- | | **Population** | 82 adult offspring of parents with bipolar disorder and 46 control participants (mean age 37.8; 57% female) | | **Cohort** | Dutch Bipolar Offspring Study, a cohort of 140 offspring followed from a mean age of 16 to 38 across five assessments since 1997 | | **Data collection** | 14 days, between September and December 2021 | | **Surveys** | Five short surveys a day (about 2 minutes each), one in each of five fixed time blocks between 7:30 a.m. and 9:30 p.m. | | **Passive sensing** | Step count and screen state, collected in the background from participants’ phones | | **Step count analyses** | Based on the 67 participants who used iPhones | | **Payment** | 50 euros for cohort participants (interview, online questionnaires, and smartphone study); 25 euros for controls | | **Ethics** | Medical Ethics Committee of Erasmus MC, Rotterdam | | **Funding** | ZonMw, the Dutch Research Council (NWO), the European Commission, the Stanley Medical Research Institute, and others | ## How the Study Worked ### A Cohort Followed for 22 Years The Dutch Bipolar Offspring Study began in 1997 with 140 adolescents who had a parent with bipolar disorder. The cohort was assessed at baseline and after 1, 5, 12, and 22 years, each time with a structured psychiatric interview. At the 22-year follow-up, 100 of the original participants took part (71% retention), and 82 of them also completed the smartphone study. The smartphone study was one part of that follow-up, alongside the interview and a set of online questionnaires. The lifetime diagnoses from all five interviews let the researchers group participants by their history: recurrent mood episodes, a single mood episode or anxiety disorder, or neither. The researchers also recruited a control group of the same age range, through online advertisements and through friends and acquaintances of the cohort participants. Controls had no personal history of psychiatric disorders and no parent with bipolar disorder, depression, or psychosis. ### Five Surveys a Day Each day was divided into five two-hour blocks: 7:30 to 9:30 a.m., 10:30 a.m. to 12:30 p.m., 1:30 to 3:30 p.m., 4:30 to 6:30 p.m., and 7:30 to 9:30 p.m. The app prompted one survey at a random time within each block, with times weighted toward the middle of the block. Each survey asked participants to rate 17 feelings on a scale from 1 to 7: - **Positive affect:** content, relaxed, calm, excited, energetic, and enthusiastic - **Negative affect:** upset, irritable, nervous, anxious, guilty, sad, bored, ruminating, sluggish, tired, and insecure On the first day, participants also completed a self-rated questionnaire on current depressive symptoms. ### Steps and Screen Time in the Background The app collected two kinds of data without any action from participants: - **Step count:** steps counted by the phone’s own step counter, as a proxy for physical activity - **Screen state:** when the screen turned on and off, as a proxy for phone use. The content of the screen was not recorded. ### Matching Sensor Data to Each Survey Sensor data arrive continuously, but surveys arrive five times a day. Before analyzing anything, the researchers had to decide which steps and which minutes of screen time belonged to which survey. **Each survey was matched with the hours before it.** The researchers added up the steps and screen-on time between one survey and the next, and paired that total with the later survey. A survey at 2 p.m. was matched with the activity since the late-morning survey. The analysis could then look at both directions: whether the activity before a survey predicted that survey’s mood, and whether a survey’s mood predicted the activity before the next one. **The first questionnaire in the monitoring period was excluded.** The researchers also excluded the first and last sensor-data bins. This left the observations needed to examine the relationships between behavior and mood across consecutive measurement intervals. **Coverage was recorded for every window.** A phone can stop recording for a while, and a window with no screen-on time could mean either an unused phone or a gap in the data. So for each window, the researchers calculated the share of time for which the app had recorded screen state, and included that share in every model. Coverage was high: 82.9% of windows were fully covered, and only 4.4% were less than half covered. **The proxies have limits.** A phone’s step counter misses steps when the phone is left on a desk or in a bag, and activities such as swimming or team sports. The screen also turns on when a notification arrives, so screen-on time may be higher for people who receive many notifications. The authors list both among the study’s limitations. ## What the Data Showed The analyses looked at how each person’s mood and behavior changed from one survey to the next, compared with that person’s own usual levels. **Steps and positive affect predicted each other.** When participants took more steps than usual, they reported more positive affect at the next survey. When they felt more positive than usual, they took more steps before the next one. **More screen-on time predicted higher negative affect.** More screen-on time than usual was followed by more negative affect at the next survey. Negative affect did not predict later screen-on time. **Positive affect predicted less screen-on time.** When participants felt more positive than usual, their screen-on time before the next survey was lower. **Psychiatric history was associated with differences in average mood.** Participants with a history of single or recurrent mood episodes had lower average positive affect than those in the other/no psychopathology group. The single mood episode/anxiety group also had higher average negative affect. However, these groups did not differ in the temporal relationships between mood and behavior. All of these effects were small. The authors interpret them as predictive links within a person’s day, not as proof that one causes the other, and note that sleep, stress, or social contact could explain some of them. They suggest that future interventions could explore ways to promote positive affect and physical activity, rather than focusing only on reducing screen time. The sample also limits how broadly the findings can be applied. Participants were predominantly White and had relatively high socioeconomic status. The control group had no personal or parental history of psychiatric disorders, and the small number of participants with recurrent mood disorders limited comparisons between specific diagnostic groups. ## What Could Researchers Do Differently Today? Researchers could use wearable integrations such as [Garmin Health SDK](https://forum.avicennaresearch.com/t/garmin-health-sdk/1261), [Google Health](https://forum.avicennaresearch.com/t/google-health/2192), or [Apple HealthKit](https://forum.avicennaresearch.com/t/healthkit/1384) to capture steps even when participants aren’t carrying their phones. They could also use [App Usage](https://forum.avicennaresearch.com/t/app-usage/1247) to go beyond screen-on time and examine which apps participants use. Android provides app activity timestamps, while iOS supports hourly Screen Time reports through a Custom App. ## Research Team **Fleur Helmink** First author Erasmus MC Sophia Children’s Hospital, Rotterdam **Esther Mesman** Corresponding author and shared last author Erasmus MC Sophia Children’s Hospital, Rotterdam **Sarah Sperry** Shared last author University of Michigan ## Publications - [Temporal dynamics of affect, screen time, and step count in individuals at risk for mood disorders](https://doi.org/10.1017/S0033291726105297). Helmink, F., van Roekel, E., Hillegers, M., Mesman, E., & Sperry, S. (2026). _Psychological Medicine_, 56, e291. - [Beyond the window of risk? The Dutch Bipolar Offspring Study: 22-year follow-up](https://doi.org/10.1016/j.jaac.2024.05.024). Helmink, F. G. L., Mesman, E., & Hillegers, M. H. J. (2025). _Journal of the American Academy of Child & Adolescent Psychiatry_, 64(5), 593–601. This paper reports the psychiatric outcomes of the same follow-up, without the smartphone data. ## Planning a Similar Study? Combining repeated surveys with passive sensing helps researchers examine how behavior and mood predict each other over time. Define the sensor-data windows for each survey in advance, choose sources with suitable time resolution, and account for gaps in collection. Avicenna supports this approach through [Time Triggering Logic](https://forum.avicennaresearch.com/t/triggering-logics/1238#p-2865-time-tl-5), including the **Period** option for sending surveys at random times within a defined window, using a **Uniform** or **Normal** distribution. Researchers can combine it with [Pedometer](https://forum.avicennaresearch.com/t/motion-sensors/1256#p-2883-pedometer-19), [Screen State](https://forum.avicennaresearch.com/t/screen-state/1257#p-2884-interpretation-3), and [Activity Sessions](https://forum.avicennaresearch.com/t/activity-sessions/1235#p-2862-view-and-manage-activity-sessions-1) to collect behavioral data and monitor survey completion. [Explore Avicenna’s features and integrations](https://forum.avicennaresearch.com/c/reference/17) or talk to our team to see how you can build your study protocol.