Skip to main content
Journal Club4 min read

The Sleep Diary That Fights the Sleep-Deprived Brain

A small eye-tracking pilot makes an uncomfortable point: the people asked to keep a precise sleep diary are the ones whose attention the poor sleep has already eroded. The interface is not neutral — but this is a pilot, and it measured strain, not cure.

Dr. Sven Jungmann

Dr. Sven Jungmann

CEO

Editorial collage of a tired person at night lit by a blue phone screen, an erratic teal eye-movement line across a navy rectangle, faint empty diary fields below, and one amber dot in a single field.

An eye-tracker watched the surface; a questionnaire asked about the inside. In a small Taiwanese pilot, those two instruments disagreed in a way worth sitting with. The camera could see where a tired person's gaze jumped across a sleep-diary screen. But the strongest predictor of whether they could remember what they had just entered was not in the gaze at all — it was in what they reported feeling pulled away by. The thing that hurt memory most was, in part, invisible to the device we would most like to trust.

Begin one step back, with what the therapy actually demands. Cognitive behavioural therapy for insomnia turns on the sleep diary: every morning, often after a broken night, the patient reconstructs the previous one with some precision — bedtime, how long until sleep came, how many awakenings, when they finally rose. The instrument the whole treatment depends on is filled in by exactly the people whose attention and working memory the poor sleep has already worn down. The tool leans hardest on the faculty the condition has weakened. That paradox is what this study, in JMIR Human Factors, set out to observe — not whether an app works, but whether its interface quietly taxes the cognition a patient needs to use it.

The design, and what it can carry

Thirty-three adults, aged 20 to 64, each used three deliberately varied sleep-diary interfaces in counterbalanced order so sequence would not skew the result. Interface A paired a dark night mode with a blue scheme, circular fields and slide-in input; Interface B a light day mode, blue scheme and tap input; Interface C a light day mode, green scheme and manual keyboard entry. As participants worked, an eye-tracker logged gaze and counted saccades — the fast jumps of the eye that rise with cognitive load — as an objective proxy for effort. Subjective attention and distraction were captured separately by questionnaire, and short-term memory was tested against the entries just made.

Two constraints are built into a design like this. Thirty-three people is a pilot, not a trial — and a thin one for the target condition: only six met criteria for clinical insomnia, with another thirteen reporting symptoms. The authors say plainly that recruitment was hard and that severe insomnia is under-represented. The sample was also regional and largely of Asian descent — a reasonable first look, not a basis for general claims. And because colour, mode and input method varied together across the three interfaces, the numbers are correlations, not the isolated effect of any one design choice.

Where the signal is coherent

Within those limits the pattern holds together. Poorer sleep quality went with more saccades — more visible strain — on Interface B (r = −0.448, P = .01) and Interface C (r = −0.491, P = .005). Interface A, the calm night-mode design with guided slider input, produced the fewest saccades and the lightest apparent load. Poor sleep and effortful design did not merely coexist; they appeared to compound, the tired eye working hardest against the busiest screen.

Then the more telling result, the one the lead returns to. The questionnaire measures of attention and distraction tracked memory more consistently than the eye-tracking did. On Interface B, the association between difficulty answering quickly and poorer recall reached r = −0.564 (P < .001) — the firmest correlation in the paper. The authors read this as recall being driven more by a person's internal attentional state than by anything their gaze betrays on the surface.

The interface is not neutral. For a patient whose sleep has already eroded their attention, it is a cognitive context that can either steady the task or fragment it.

The line the authors draw

Here a careful reader holds the boundary the authors set themselves. The study measured user experience, cognitive strain and short-term recall during the task. It did not measure whether any interface improved sleep, adherence to therapy, or any clinical outcome; in their own words, the therapeutic effect of these interfaces was outside the study's scope. A diary that is gentler on a tired mind is a reasonable hypothesis for better data and better adherence — it is not evidence of either, and a pilot this size cannot promote a correlation into a design rule.

Digital therapeutics for insomnia are reimbursed in several European systems, Germany among them, and the diary sits at the centre of nearly all of them. The defensible takeaway is modest and worth keeping: the burden of an interface is not a cosmetic question for the design team but a clinical variable — plausibly heaviest for the patients who are most impaired and most in need. Build the diary for the cognitive state of the person using it, keep the interaction guided and sequential rather than effortful, and then test whether that actually helps, in a study large enough to say so. The hypothesis is good. The evidence is a beginning.

Source: Su KC, Chiu HY, Wu KC, Chang CC. Designing App Interfaces to Elicit Specific Emotional Responses and Improve Attention and Short-Term Memory in Patients With Insomnia Undergoing Brief Cognitive Behavioral Therapy. JMIR Human Factors 2026;13:e79883. A within-subject eye-tracking pilot in 33 participants that measured cognitive load and short-term memory, not any therapeutic or sleep outcome; the authors flag the small, regional sample and the under-representation of severe insomnia.

#Journal Club#Digital Health#Human Factors#Insomnia#Evidence-Based Medicine

Keep reading

Editorial collage of an oncologist's hands on a thick claims ledger, with a teal three-column bar chart rising only partway and a single amber accent.
Journal Club

An Explainable Model, Honest Numbers, and a Funder Worth Noticing

An explainable AI model predicted how long myeloma patients would stay on treatment, using twenty years of Japanese claims data and 647 variables. The discrimination is modest and fairly reported. The part that needs a careful eye is who paid, and which finding they got.

Dr. Sven JungmannCEO
Editorial collage of four people mid-conversation arranged around a teal circle with a single amber dot at its centre.
Journal Club

Four Conversations About Clinical AI That Quietly Agree

Four NEJM AI podcast interviews, recorded months apart, keep landing in the same three places: a values vacuum, a bias we taught the machine, and a trust gap that tracks consequence. None of it is evidence. The agreement is still worth an hour.

Dr. Sven JungmannCEO
Editorial collage of a surgeon's gloved hands beside an anaesthesia monitor showing a teal arterial-pressure waveform, with a closed operating-room door suggested behind and a single amber accent.
Journal Club

Surgical AI That Works in the Paper but Not in the Room

A scoping review screened 275 records to find every AI model meant to prevent surgical complications and follow it to the bedside. Of 19 studies, the models were often accurate. Two are in routine use — and the bottleneck is not the algorithm.

Dr. Sven JungmannCEO

This analysis comes from the people behind Visite.

Our weekly newsletter on AI in medicine. Every Friday, rigorously checked.

By signing up you agree to receive Grand Rounds by email. Unsubscribe anytime. More in our privacy policy.

Want to see this in your hospital?

30 minutes. Your questions. Our physician-founder shows you the platform personally.

Book a demo

No commitment. No sales pitch. Physician to physician.