Wren

Science

Our science.

Wren is an iPhone app for people living with bipolar. It keeps a record of their days and talks with them about it. This page sets out what it records and why, what research says about keeping a record like it, and how Wren uses AI.

What Wren records, and why

Mood and energy, once a day

The daily check-in is two questions. Mood is marked on seven steps from Depressed to Elevated, and energy on seven from Slowed to Wired. They are separate because in bipolar a change in mood comes with a change in energy or activity, but the two do not always move together: a mixed episode has symptoms of a high and of depression at once (NIMH). Energy can also move first. In a review of 22 studies, increased energy and more goal-directed activity were among the warning signs reported only before a high (Andrade-González, 2020).

Mood runs from Depressed to Elevated rather than from bad to good because feeling very happy and excited is itself a symptom of a high (NHS). It is asked once a day because that is what people keep up: in a review of app-based mood questionnaires for bipolar, adherence was above 70% in all but one study that asked once a day, while studies that asked twice a day ranged from 42% to 95% (Chan, 2021).

The night

Each morning Wren asks how the night went, from Rough to Sound, with its start and end if they are known. People with bipolar in one study named sleep as the most important thing to keep an eye on (Van Til, 2020), and across 17 studies disturbed sleep was the most consistently reported early sign of mania, in a median of 77% of people (Jackson, 2003).

Anything else someone chooses

Other things can be added to the record in a tap or a sentence: racing thoughts, trouble focusing, time with people, spending, alcohol, a walk, or anything in a person’s own words. Medication has its own place: each dose is marked taken, not taken, or left unanswered, and an unanswered dose is never counted as either.

From Apple Health and the weather, if switched on

Wren can read ten things from Apple Health: when each night started and ended, the day’s steps, resting heart rate, exercise, walks, distance walked and run, flights of stairs, standing minutes, time in daylight and average heart rate. Each is its own switch, and Wren never writes to Apple Health. It can also record the day’s sunshine, temperature and how fast daylight is changing.

Why a record over months

One day says little. What a record can show is a person’s own pattern: how their nights, energy and mood usually look, and when they change. In a year-long study in which 29 people recorded sleep and mood every day, changes in sleep came before mood changes for about two-fifths of them, and not for the rest (Ulrichsen, 2026). That is why Wren compares each person with their own usual, not with anyone else.

The research behind it

Keeping a mood record is a long-standing part of bipolar care: almost every CBT programme for bipolar includes it (Özdel, 2021). Research on doing it by phone asks three questions: do people find it useful, does it make them better, and can it see an episode coming.

People find it useful

In a 2023 survey of 362 people with bipolar who keep a long-term electronic mood record, 88% said it helped. They valued seeing their mood over time, spotting patterns and triggers, and noticing early warning signs. A third had shared their mood graphs, and 90% of those found sharing with a health professional helpful (Gordon-Smith, 2023).

Spotting early signs helps when it leads to care

A 1999 trial took 69 people who had relapsed in the past year. Some had seven to twelve sessions with a psychologist on top of their usual care, learning to recognise their own early signs and to get treatment quickly; the rest had usual care alone. Over 18 months the first group had fewer manic relapses and their first came later; there was no difference for depression (Perry, 1999). The benefit came from recognising signs and acting on them with professional help, not from a record alone.

What one self-monitoring trial found

The MONARCA I trial followed 78 people with bipolar for six months, comparing daily smartphone self-monitoring with a clinical feedback loop against a smartphone used normally. It found no significant effect on depressive or manic symptoms, with a tendency towards more sustained depressive symptoms in the self-monitoring group (Faurholt-Jepsen, 2015).

Seeing an episode coming is not yet reliable

A 2021 review of 15 studies that tried to predict episodes from phone and wearable data found results too varied to combine, and concluded that the technology needs more development before it is used in care (Ortiz, 2021). In a study of 20 people who answered questions five times a day, warning signals raised the chance that a change was coming from about a third to just under a half, and the absence of a signal did not mean no change was coming (Bos, 2022).

How Wren uses AI

Wren uses AI in two places.

In conversation

Wren’s replies are written by an AI model from OpenAI, following instructions. They tell it to offer rather than instruct; never to tell someone what state they are in; never to treat a high as good news; and never to advise on medication, including whether to take it. If it notices a change against what someone has told it before, such as several late nights in a week, it may ask about it once, as a question, and lets it go if they say it is nothing.

When Wren writes something into the record from a conversation, it passes the same checks as a tap in the app, and it can be changed or deleted. Wren asks permission before sending any personal or health information to OpenAI; the privacy policy lists exactly what is sent.

Reading the record

A second model, from TypeSafe AI, reads the record and marks stretches of days as high, low, mixed or stable. It also looks at what else changes around those stretches, such as nights starting later before a high. Our code works out the figures, the model judges what they show, and Wren shows how sure it is as a percentage.

Who writes it

Wren is made by Blue Door Labs Ltd.

Sources

  1. National Institute of Mental Health. Bipolar disorder. www.nimh.nih.gov/health/publications/bipolar-disorder
  2. Andrade-González N, Álvarez-Cadenas L, Saiz-Ruiz J, Lahera G. Initial and relapse prodromes in adult patients with episodes of bipolar disorder: a systematic review. European Psychiatry, 2020. doi.org/10.1192/j.eurpsy.2019.18
  3. NHS. Bipolar disorder. www.nhs.uk/mental-health/conditions/bipolar-disorder/
  4. Chan EC, Sun Y, Aitchison KJ, Sivapalan S. Mobile app-based self-report questionnaires for the assessment and monitoring of bipolar disorder: systematic review. JMIR Formative Research, 2021. doi.org/10.2196/13770
  5. Van Til K, McInnis MG, Cochran A. A comparative study of engagement in mobile and wearable health monitoring for bipolar disorder. Bipolar Disorders, 2020. doi.org/10.1111/bdi.12849
  6. Jackson A, Cavanagh J, Scott J. A systematic review of manic and depressive prodromes. Journal of Affective Disorders, 2003. doi.org/10.1016/s0165-0327(02)00266-5
  7. Ulrichsen A, Mühlbauer E, Ludwig VM, et al. Sleep fluctuations precede self-reported mood changes in bipolar disorder: results from the BipoSense study. International Journal of Bipolar Disorders, 2026. doi.org/10.1186/s40345-026-00416-y
  8. Özdel K, Kart A, Türkçapar MH. Cognitive behavioral therapy in treatment of bipolar disorder. Noro Psikiyatri Arsivi, 2021. doi.org/10.29399/npa.27419
  9. Gordon-Smith K, Saunders KEA, Morton T, et al. User perspectives on long-term remote active electronic self-monitoring of mood symptoms in bipolar spectrum disorders. Journal of Affective Disorders, 2023. doi.org/10.1016/j.jad.2022.12.090
  10. Perry A, Tarrier N, Morriss R, McCarthy E, Limb K. Randomised controlled trial of efficacy of teaching patients with bipolar disorder to identify early symptoms of relapse and obtain treatment. BMJ, 1999. doi.org/10.1136/bmj.318.7177.149
  11. Faurholt-Jepsen M, Frost M, Ritz C, et al. Daily electronic self-monitoring in bipolar disorder using smartphones – the MONARCA I trial: a randomized, placebo-controlled, single-blind, parallel group trial. Psychological Medicine, 2015. doi.org/10.1017/S0033291715000410
  12. Ortiz A, Maslej MM, Husain MI, Daskalakis ZJ, Mulsant BH. Apps and gaps in bipolar disorder: a systematic review on electronic monitoring for episode prediction. Journal of Affective Disorders, 2021. doi.org/10.1016/j.jad.2021.08.140
  13. Bos FM, Schreuder MJ, George SV, et al. Anticipating manic and depressive transitions in patients with bipolar disorder using early warning signals. International Journal of Bipolar Disorders, 2022. doi.org/10.1186/s40345-022-00258-4