Definition, indications, and exclusions
A circadian rhythm disorder digital therapeutic is a software treatment function for a persistent mismatch between biological timing and the required sleep–wake schedule. Through structured assessment, time-sensitive behavioral plans, scheduled light and darkness, management of sleep opportunity, and professional follow-up, it aims to improve insomnia, sleepiness, or impaired function caused by a specified circadian disorder. Its core task is not to demand an early bedtime; it is to determine the direction of misalignment, the real-world time requirement, and what can be changed safely.
Delayed sleep–wake phase disorder, advanced sleep–wake phase disorder, non-24-hour sleep–wake rhythm disorder, and irregular sleep–wake rhythm disorder have different patterns. Shift work disorder and jet lag also involve external timing demands. The same complaint of difficulty falling asleep can require light or light avoidance at opposite times. Late bedtimes, night work, weekend catch-up sleep, self-described morningness or eveningness, or one wearable record cannot diagnose a circadian disorder or exclude insomnia, insufficient sleep, OSA, medication effects, or a mood condition.
The therapeutic product should specify each disorder, age group, visual status, diagnostic prerequisite, target outcome, human-support model, and use environment it covers. A calendar, alarm, generic sleep-hygiene tool, jet-lag calculator, or blue-light filter may assist a plan. Without a validated therapeutic mechanism, risk controls, and product-level clinical evidence, however, a personalized timetable alone does not justify a disease-treatment claim.
Assessment, differential diagnosis, and phase model
Before activation, a sleep diary should span workdays, free days, or a complete shift cycle and be reconciled with sleep opportunity, actual bed and rise times, sleep onset and waking, naps, light exposure, work or school, caregiving, travel, and medication or substance use. Actigraphy can help reveal a longitudinal pattern, but it infers sleep from movement and does not directly measure the biological clock. Phone activity, heart rate, skin temperature, and environmental light are also proxies rather than phase itself.
In an appropriate setting, professional assessment may use validated actigraphy or a physiologic phase marker. Collection of dim-light melatonin onset depends on controlled light, sampling, and analysis. A home sample affected by bright light, missed specimens, food, clock error, or transport cannot be presented as precise ground truth. Consumer-wearable models may predict phase in some populations, but their average error and complete failures must be disclosed; a population model must not produce minute-level prescriptions with false certainty.
Differential assessment should cover chronic insomnia, insufficient sleep, OSA, central hypersomnolence, depression or mania, neurodevelopmental or neurodegenerative conditions, visual and ocular disease, epilepsy, pain, medications, and substances. Software may identify a pattern and uncertainty, but diagnosis integrates symptoms, persistence, timing records, functional impairment, and alternative explanations. Conflicting data, a continuously drifting rhythm, or a high-risk context should pause autonomous planning and trigger qualified review.
Time-sensitive intervention and therapeutic dose
Light can shift circadian phase and acutely affect alertness. Its direction depends on timing relative to the individual's rhythm, while actual dose also depends on spectrum, illuminance, duration, distance, gaze, ocular exposure, environment, and prior light history. The software should say whether a task is intended to advance phase, delay phase, stabilize timing, or improve immediate alertness, and should record the gap between intended and actual exposure rather than merely confirming that a lamp was switched on.
Light avoidance is likewise bounded in time. It may involve dimming an environment, reducing strong light during a defined interval, or using an assessed light-reduction method. It does not mean remaining in darkness all day, wearing dark lenses in an unsafe setting, or treating every screen as an identical light source. Weather, season, latitude, window glass, indoor position, and device spectrum change whether daylight and a lamp are interchangeable; the program should show uncertainty rather than a rigid timetable to the minute.
Sleep and rise timing, activity, meals, and other time cues can be combined only with evidence for each component. If a plan involves melatonin, a receptor agonist, a wake-promoting agent, or a sleep medication, software may implement a professionally confirmed plan through a lawful pathway but must not independently provide a personal dose, advise medication changes, or imply that a supplement is risk-free. Dose also includes task delivery, verified completion, actual sleep opportunity, and follow-up; pressing Done is not evidence that physiologic phase shifted.
Shift work, travel, and sleepiness safety
Not every shift worker has shift work disorder, and frequent rotation often makes full adaptation to one night schedule unrealistic or undesirable. Goals may be to protect sufficient sleep, reduce sleepiness at the highest-risk times, arrange feasible light–dark exposure, and improve between-shift recovery rather than maximize phase shift. Fixed nights, rapid rotation, early starts, on-call duty, and transmeridian travel require different models. A plan must not keep running silently after an employer changes the roster.
Naps, caffeine, light, and sleep scheduling may improve alertness or sleep for some people, but evidence, timing, and tradeoffs differ. Sleep inertia can follow a nap, caffeine can impair subsequent sleep, and melatonin or another sleep-promoting intervention may transiently reduce alertness. The software must not bundle these into an automatic package for every occupation or replace the fatigue-management rules of aviation, road transport, clinical duty, or another safety-critical industry.
A person who is markedly sleepy should not drive, operate machinery, or perform a safety-critical task. The plan should prioritize stopping the hazardous activity, rest, alternative transport, and timely assessment over a light task or check-in. Chest pain, severe breathing difficulty, altered consciousness, acute neurologic symptoms, or an acute change involving risk of harm, mania, or psychosis belongs in the applicable local emergency pathway rather than an asynchronous app queue.
Clock, sensor, and plan integrity
Time is the dose in circadian treatment. The system should retain an event's absolute time, local time, time zone, daylight-saving state, and the reference clock used by the plan. It must handle midnight boundaries, the international date line, daylight-saving transitions, manual clock changes, and operating-system time correction. Pretravel, in-transit, and postarrival plans need explicit transition points. When a clock conflict cannot be resolved, the product should suspend high-impact recommendations.
The location, orientation, occlusion, clothing, indoor–outdoor transition, and sampling rate of a light sensor affect its estimate of light reaching the eyes; a phone in a pocket does not represent ocular exposure. An activity sensor may label quiet wake as sleep or motion in a vehicle as wake. The record should retain source, device, firmware, algorithm version, wear state, missingness reason, and manual correction. Yesterday's value or a population average must not fill a critical phase input.
The planning engine should display input coverage, the interval around the phase estimate, the reason for the last recalculation, and the next review point, while distinguishing measurement, model inference, patient report, and clinical decision. When a diary, sensor, and roster disagree, it must not silently choose the most convenient source. Low model confidence, device replacement, or an interface-semantic change should cause safe degradation, human review, or a return to lower-risk general support.
Clinical evidence, product validation, and outcomes
Guideline recommendations vary across circadian disorders, populations, and interventions, and many are conditional with limited certainty. Shift-work studies also often use simulated schedules, workers without a formal diagnosis, or short-term symptom outcomes. A digital calendar and reminders may make a regimen easier to execute, but they do not by themselves establish improvement in phase, sleepiness, or function. Usability studies, engagement, and uncontrolled before–after changes are early evidence, not substitutes for appropriately controlled product evaluation.
Studies should specify diagnosis, direction of phase, shift pattern or visual status, comorbidity, comparator services, actual light–dark exposure, human support, algorithm, and software version. Outcomes should distinguish sleep timing, objective or physiologic phase, insomnia and sleepiness, alertness or performance, attendance, safety events, quality of life, and adverse effects. A phase shift is not necessarily a functional benefit, and short-term alertness does not prove reduced long-term health risk.
Technical validation should cover phase-model error, abstention, time-zone and date boundaries, missing data, task delivery, light-dose estimation, outliers, equipment replacement, and version upgrades, with training and testing separated by person rather than fragments of the same record. Subgroup evaluation should include age, sex and gender, visual status, skin-related sensor differences, language, geography, season, latitude, roster, and device access. A rule or model update that changes timing, dose, population, or a safety boundary requires renewed evidence and risk review.
Human factors, privacy, and lifecycle governance
A therapeutic plan should accommodate work, school, caregiving, religious activity, the household light environment, and available equipment, and allow the user to record why a task was not feasible. The interface should display phase-shift direction, plan time zone, task window, and conflict priority clearly and undergo realistic testing with people who have low vision, color-vision differences, cognitive burden, different languages, or assistive technologies. Excess reminders, a bright screen at night, or shaming streaks can themselves disturb sleep or drive disengagement.
Rosters, location, light, activity, sleep, visual status, and medication data can reveal a workplace, home routine, health condition, and safety-sensitive role. The product should minimize collection, separate clinical from employment use, enforce role-based access, encrypt data, log access, and define retention and deletion. Employers and insurers should not receive individual trajectories beyond a specific authorized purpose, and a patient who declines nonessential sharing, marketing, or location access still needs a feasible treatment path.
Connected lamps, wearables, phones, cloud services, and clinical systems form a dependency chain and require component inventories, vulnerability response, signed updates, backup recovery, and outage exercises. Organizations should monitor wrong-phase recommendations, sleepiness events, mood change, ocular discomfort, alert fatigue, model drift, and differential dropout. When a service or vendor exits, the patient needs a plan-termination notice, data export, transfer of accountability, and alternative care so that an obsolete timing regimen does not continue without monitoring.