Bipolar Pattern Tracking: How Daily Data Predicts Mood Episodes

The most disruptive thing about bipolar disorder is not the episodes themselves. It is that they tend to arrive without warning.

Or that is how it seems. What decades of clinical research has established is that mood episodes are not random. They have precursors. Sleep changes, energy shifts, altered social rhythms, and subtle circadian disruptions tend to appear days before a full manic or depressive episode unfolds. The window for early intervention exists. The question is whether anyone is capturing the data to see it.

What This Page Covers

• What daily tracking reveals about bipolar disorder that appointments cannot
• The science behind circadian rhythms and episode prediction
• How accurate current tracking models are
• What to track and why each data point matters
• What happens when this data informs a prescriber’s decisions

Why Quarterly Appointments Are Not Enough

Bipolar disorder is characterized by recurrent episodes of mania or hypomania and depression. Most people with bipolar disorder experience a change in symptom severity and mood at least three times per year. The clinical management challenge is that these episodes do not announce themselves at convenient times.

A quarterly appointment captures a 15-minute snapshot of how a patient is doing at a specific moment. It does not capture the three weeks of gradual sleep disruption that preceded the manic episode two months ago. It does not show the pattern of increased energy and reduced need for sleep that reliably signals an approaching high. And it does not catch the subtle shift from stability to prodrome in real time, which is the only window where early intervention is possible.

This is not a failure of clinical judgment. It is a structural limitation of the appointment-based care model. Clinicians making decisions at quarterly intervals are working from incomplete data.

The Sleep-Mood Connection: Where Tracking Starts

The relationship between sleep and mood episodes in bipolar disorder is one of the most robustly documented in psychiatric research. Sleep disruption is both a symptom and a predictor of mood episode onset.

A 2025 study published in Bipolar Disorders that followed patients with bipolar disorder daily for one year found that many patients identified changes in their sleep before an episode began, and that daily sleep logs could help identify these changes, serving as a reliable early warning sign for both manic and depressive episodes.

The mechanism is rooted in circadian biology. Bipolar disorder involves dysregulation of the body’s internal clock, the circadian system that governs sleep-wake cycles, hormone release, and body temperature. A 2024 study in npj Digital Medicine, analyzing longitudinal data from 168 patients with mood disorders across an average of 587 days, found that daily circadian phase shifts were the most significant predictors of mood episodes, with delays in circadian phase linked to depressive episodes and advances linked to manic episodes. Using sleep-wake data alone, the study achieved area under the curve values of 0.80 for predicting depressive episodes, 0.98 for manic episodes, and 0.95 for hypomanic episodes.

That means a daily sleep pattern, measured passively, can predict an impending manic episode with 0.98 AUC in well-trained models. That capability is not available in the quarterly appointment.

What Digital Tracking Research Shows

The evidence on digital phenotyping for bipolar disorder has accumulated rapidly since 2015.

A 2024 study published in Acta Psychiatrica Scandinavica, conducted at Brigham and Women’s Hospital, analyzed longitudinal Fitbit data from 54 adults with bipolar disorder over nine months. Using a personalized machine learning approach trained entirely on passive Fitbit data, the model achieved 86% AUC for detecting depressive symptomatology and 85.2% AUC for detecting hypomanic and manic symptomatology.

A larger 2022 prospective study involving 495 patients across eight hospitals in South Korea, followed for an average of 279.7 days, used wearable devices and smartphone data to develop mood episode prediction algorithms. The prediction accuracy for impending major depressive episodes was 90.1%, for manic episodes 92.6%, and for hypomanic episodes 93.0%, with area under the curve values above 0.93. Phenotypes indicating circadian rhythm misalignment were the strongest predictors of episode recurrence across all mood states.

What these studies collectively establish is that the data needed for early episode prediction already exists in how people sleep and move through their days. The clinical value of that data depends on whether it is being collected, reviewed, and acted on.

What to Track: A Clinical Perspective

Not all data is equally useful for bipolar episode prediction. The highest-signal categories, based on clinical and research evidence, are:

Sleep duration and timing. How long you slept and when your sleep began and ended. Changes of more than an hour in either direction from your baseline, sustained over several consecutive days, are early warning indicators.

Sleep quality. How rested you felt on waking, whether you woke during the night, and whether your sleep felt restorative. A person can sleep eight hours and wake exhausted, which carries different clinical meaning than sleeping six hours and waking refreshed.

Energy levels. Subjective energy, separate from mood. A patient who reports feeling great but also notices they need much less sleep than usual is describing a different clinical picture than one who reports elevated mood and normal sleep.

Mood. Daily self-rated mood on a simple scale provides the clearest longitudinal trend. A single good day or bad day is noise. A direction of travel over five to seven days is signal.

Activity and social rhythms. Changes in how much you are moving, how often you are engaging with other people, and whether your daily schedule is staying regular. Disrupted social rhythms, even positive disruptions like increased socializing, can precede hypomanic episodes.

None of these data points requires specialized equipment. A daily structured check-in of five to ten minutes captures all of them.

Turning Data Into Clinical Action

Tracking data that does not reach a prescriber has limited clinical value. The patient who notices they have been sleeping five hours a night for three consecutive days has relevant clinical information. If their prescriber does not know this until the next quarterly appointment, the intervention window may have closed.

A 2023 systematic review of digital phenotyping for mental health monitoring concluded that data from smartphones and wearable devices can identify digital phenotype patterns in the days preceding a mood episode, supporting early clinical intervention before full relapse. The clinical potential exists. The limiting factor is whether that data is routed to someone who can act on it.

This is why the design of the monitoring system matters as much as the monitoring itself. Daily check-in data that sits in a patient-facing app without prescriber access is better than not tracking. Daily check-in data that feeds directly into a prescriber’s review workflow is qualitatively different.

How SiggyMD Approaches Longitudinal Monitoring

SiggyMD’s daily check-in model is designed around the kind of longitudinal data described in this research. Daily mood, sleep, and energy check-ins build a continuous clinical record that the prescriber reviews as part of ongoing care.

When sleep duration shifts outside a patient’s established baseline, or when mood scores begin trending in either direction over consecutive days, the clinical picture changes in the prescriber’s view before the patient may even recognize it as significant.

SiggyMD’s current clinical scope covers anxiety and depression. Patients with established bipolar diagnoses should discuss how ongoing monitoring integrates with their existing treatment team. For patients in the diagnostic gray area, where depressive symptoms have been present but prior antidepressant response has been incomplete or mood has been cyclical, longitudinal data is precisely what helps a prescriber clarify the clinical picture over time.

“What I see with pattern data is things patients do not always see in the moment,” says Elizabeth Lokenauth, PA-C, of the SiggyMD clinical team. “A patient reports they are doing well. Their sleep data shows they have averaged five hours a night for ten days. Those are two very different pieces of information, and they lead to a different conversation than the one I would have had without the data.”

What Members Are Saying

KM

K.M., 36

Bipolar Disorder and Depression

“My depressive episodes used to come out of nowhere. When I started tracking daily, I noticed that every episode was preceded by two weeks of interrupted sleep and lower energy. Once I saw the pattern, my prescriber and I built in a checkpoint whenever my sleep logged below six hours for three days in a row. I have not had an uninterrupted downward spiral since.”

TW

T.W., 42

Mood Disorder

“I spent years being told I just had depression. The tracking revealed a pattern of elevated mood and reduced sleep alternating with crashes. My prescriber could see in the data what I could not see in the moment. It changed my diagnosis and changed my treatment.”

Member stories reflect real experiences. Names and identifying details have been changed to protect privacy. Results vary. SiggyMD is currently invite-only.

Bottom Line

Bipolar mood episodes have precursors. Sleep changes, circadian disruption, and shifts in energy and social rhythm are measurable and predictable in ways that a quarterly appointment cannot capture. The research is clear: daily tracking data, when analyzed appropriately, can predict impending mood episodes with accuracy well above clinical chance.

The clinical benefit of that prediction depends on whether the data reaches a prescriber who can act on it. Tracking without clinical connection is better than not tracking. Tracking connected to an ongoing prescriber relationship is the system that can actually change outcomes.

If you are in crisis or experiencing suicidal thoughts, call or text 988 for the Suicide and Crisis Lifeline. If you are in immediate danger, call 911.

If you are managing a mood disorder and want to understand how continuous monitoring could support your care, you can read about how AI mood tracking connects to your treatment plan or contact SiggyMD to start an intake and speak with a licensed clinician.

Sources

1. Ulrichsen A, et al. Can Sleep Parameters Predict Upcoming Mood Episodes in Bipolar Disorder? Bipolar Disorders. 2025;27(6):449-460.

2. Song YM, et al. Accurately predicting mood episodes in mood disorder patients using wearable sleep and circadian rhythm features. npj Digital Medicine. 2024.

3. Lipschitz JM, Saghafian S, Pike CK, et al. Digital phenotyping in bipolar disorder: Using longitudinal Fitbit data and personalized machine learning to predict mood symptomatology. Acta Psychiatrica Scandinavica. 2024;151(3):434-447.

4. Lee HJ, Cho CH, Lee T, et al. Prediction of impending mood episode recurrence using real-time digital phenotypes in major depression and bipolar disorders in South Korea. Psychological Medicine. 2022;53(12):5636-5644.

5. Bufano P, Laurino M, Said S, et al. Digital Phenotyping for Monitoring Mental Disorders: Systematic Review. Journal of Medical Internet Research. 2023;25:e46778.

6. National Institute of Mental Health. Bipolar Disorder: Statistics. NIMH. Accessed June 2026.