Know by week four whether the medication is working.
A two-minute weekly check-in on WhatsApp measures each patient’s response with validated scales. Before the visit, the psychiatrist reads one page and decides.
Measurement-based care for public mental health services in Latin America. Pilots in 2026–27.
Marta R., treated by Dr. FuentesPHQ-9, weekly check-ins
- Days taken
- 776777767
- Side effects
- 022000000
Before the week 4 visit
- Symptoms
- PHQ-9 from 18 to 17 in 4 weeks
- Adherence
- 6.8 of 7 days a week
- Side effects
- Nausea, burden 2 of 6, weeks 1 and 2
- She asks
Can I take it at night?
Improvement below the configured threshold at 4 weeksPHQ-9 from 18 at baseline to 17 (down 6%; threshold: down 20%). Review using clinical judgement.
- Dr. Fuentes raises the dose. A new episode starts from its own baseline.
- Week 8: PHQ-9 of 8, a response.
A synthetic patient on sertraline barely improves by week 4. The brief says so, with the rule and the numbers. Her psychiatrist raises the dose, and the new episode reaches a response by week 8. The rule describes the data; the psychiatrist decides.
- 74%of patients reached remission when their treatment was measured, against 29% with standard care.Randomised trial, Guo and colleagues, 2015
- 4 weeksis when guidelines say to act if symptoms have not improved by at least 20%.CANMAT 2023
- 2 mina week for the patient, on WhatsApp. Nothing to install.PHQ-2 or PHQ-9, adherence, side effects
- 1 pagefor the psychiatrist before each visit, every line traceable to the record.Built from templates, not written by AI
Between visits, nobody measures whether the drug is helping.
A psychiatrist starts or changes a medication and sees the patient again weeks or months later, usually without a score in between. Most patients will not reach remission on their first antidepressant, many stop taking it early, and a treatment that is not working can go on for months.
Measuring symptoms with validated scales at every step, and acting on the scores, is called measurement-based care. In randomised trials it raised remission and adherence. It is rarely done, mostly because there is no time to collect and review the scores. Chairstead collects them on WhatsApp and reviews them against your clinic’s rules.
- 74.7%of people in Latin America with a moderate to severe mental disorder receive no treatment.Kohn and colleagues, Rev Panam Salud Publica, 2018
- ~30%of patients reached remission on their first antidepressant in STAR*D, the largest effectiveness trial of depression treatment.Trivedi and colleagues, Am J Psychiatry, 2006
- 42%stopped their antidepressant within the first 30 days, and 54% of Hispanic patients did.Olfson and colleagues, Am J Psychiatry, 2006 (United States)
- 18%of psychiatrists in the United States routinely use rating scales to monitor improvement.Fortney and colleagues, Psychiatric Services, 2017
How it works
Four weeks of one treatment, from the visit where the drug starts to the next one. Patients join with a QR code that opens WhatsApp: nothing to install.
Week 0Visit
Joins with a QR codePHQ-9 at the visit: 18
Opens episode 1Baseline: 18
Starts sertraline 50 mgSigns the visit note
Week 12-minute check-in
PHQ-2: 5Took it 7 of 7 days. No side effects
No rule triggered
Nothing to do
Week 2Check-in
PHQ-9: 187 of 7 days. Nausea, burden 2 of 6
No rule triggeredNausea recorded
Nothing to do
Week 3Check-in
PHQ-2: 56 of 7 days. Voice note:
Can I take it at night?
No rule triggeredQuestion kept for the visit
Nothing to do
Week 4Next visit
PHQ-9: 177 of 7 days. No side effects
Below 20% at week 4One-page brief, with the rule and the data
Reads one page, raises the doseEpisode 2 starts from its own baseline
Any weekRisk language, or PHQ-9 item 9 above 0, goes at once to a named clinician. It never waits for a visit.
Validated scales, fixed answers
PHQ-2 every week, the full PHQ-9 every two weeks, days the drug was taken and side effects on the FIBSER scale. No AI asks or rewords a question.
Your clinic’s rules
Each week the episode is checked against a short list of transparent rules. Raising the dose opens a new episode, measured from its own baseline.
One page, then a person decides
Built from templates, never written by AI, so every line can be checked in the record. Chairstead never says whether a drug works.
Transparent rules, set by your clinic
Each episode is checked against the same short list of rules. The defaults follow published definitions; your clinical team can change every threshold.
No early improvement
PHQ-9 down less than 20% from baseline at about 4 weeks.
CANMAT 2023: without a 20% improvement by 4 weeks, response or remission at 8 to 12 weeks is unlikely.
Response and remission
Response: PHQ-9 down 50% or more. Remission: PHQ-9 under 5.
The definitions used in collaborative care programs and in CANMAT.
Worsening
PHQ-9 up 5 points or more from baseline or from the previous score.
Five points is the minimal clinically important change on the PHQ-9 (Löwe and colleagues, 2004).
Relevant side effect
FIBSER burden of 4 or more on a 0 to 6 scale.
FIBSER was built for STAR*D to rate side effects of antidepressants.
Stopped treatment
3 days or fewer two weeks in a row, or the patient says they stopped.
In one large US study, 42% of patients stopped their antidepressant within 30 days.
Risk
PHQ-9 item 9 above 0, or risk language in a message or voice note.
Never waits for a visit: a red alert goes at once to a named clinician.
Every alert names the rule that triggered it and shows the data behind it, and ends with the same words: review using clinical judgement. Chairstead never says that a drug works or does not, and never suggests a dose.
On track
The bot
Sends the weekly check-in and keeps the record. No team time.
Needs attention
Nurses, social workers, community health workers
Reach out when a patient stops the drug or reports a troublesome side effect, within their competence: ask why, record it, never suggest a dose change, and refer to the psychiatrist in one click.
Medication
Psychiatrist
Reads one page per patient before the visit and decides: keep, raise, lower or switch.
Urgent
Designated psychologist
Any sign of risk goes straight to the designated psychologist, or to the one on call after hours. Red alerts are never left with staff without clinical training.
The pilot will measure whether this saves the team time, in minutes per patient per month. Until then, it is a hypothesis.
Safety and privacy
- Risk
- Every message is checked for crisis language with fixed rules before any AI model runs, and the check still works if the model fails. A match, or a PHQ-9 item 9 above 0, creates a red alert for a named clinician with a response time set by the clinic. The patient receives one pre-approved message with local emergency numbers; the bot does not handle crises.
- Clinical decisions
- Chairstead measures; it does not diagnose, prescribe or decide whether a treatment works. Scales are asked with their validated wording and fixed answers. AI is used only for what has no format: transcribing and summarising voice notes. Referrals, dose changes and closing an alert are always done by a person.
- Privacy
- Patients consent to each type of data separately and are told who will read it. Scores and summaries are never sent back over WhatsApp. Voice notes are deleted after transcription. Access depends on role, and every record opened is logged.
Urgent
PHQ-9 item 9 answered “several days”: alert to Carlos, psychologist. The patient sees emergency lines; the care manager is copied.
Response due in 30 minEvidence
The evidence below is for the method Chairstead puts into practice, not for Chairstead itself, which has not yet been tested with patients.
-
In a randomised trial with blinded raters, 73.8% of patients reached remission with measurement-based care, against 28.8% with standard care, and they responded sooner (5.6 against 11.6 weeks).
Guo T, Xiang YT, Xiao L, et al. Measurement-based care versus standard care for major depression: a randomized controlled trial with blind raters. American Journal of Psychiatry, 2015.
-
In Pakistan, measuring symptoms and side effects (with FIBSER) at fixed weeks and adjusting the dose to the scores raised remission at 12 weeks from 51.5% to 71.9%, and halved the median time to remission.
Husain MI, et al. Measurement-based care to enhance antidepressant treatment outcomes in major depressive disorder: a randomized clinical trial. JAMA Network Open, 2025.
-
Across seven randomised trials and 2,019 participants, measurement-based care was associated with more remission and better medication adherence.
Zhu M, et al. The efficacy of measurement-based care for depressive disorders: systematic review and meta-analysis of randomized controlled trials. Journal of Clinical Psychiatry, 2021.
-
Without at least 20% improvement by 4 weeks of treatment, response or remission at 8 to 12 weeks is unlikely; guidelines recommend validated scales at regular intervals.
Lam RW, Kennedy SH, et al. Canadian Network for Mood and Anxiety Treatments (CANMAT) 2023 update on clinical guidelines for management of major depressive disorder in adults. Canadian Journal of Psychiatry, 2024.
-
Collaborative care, of which repeated measurement is a core part, improved depression outcomes compared with usual care across 79 randomised trials and 24,308 patients.
Archer J, Bower P, Gilbody S, et al. Collaborative care for depression and anxiety problems. Cochrane Database of Systematic Reviews, 2012.
-
WHO recommends that trained non-specialist health workers deliver mental health care under specialist supervision, particularly where specialists are scarce.
World Health Organization. mhGAP Intervention Guide, version 2.0. 2016.
The check-in uses the PHQ-9, validated in Chilean primary care (Baader and colleagues, 2012), and its first two questions as the PHQ-2. FIBSER has no validated Spanish version yet: we use a provisional translation with synthetic patients only, and are seeking the author’s permission before any pilot.
See it with one psychiatrist’s patients.
A 20-minute walkthrough with synthetic patients, in English or Spanish. We are looking for clinics and public mental health services in Latin America for pilots in 2026–27.
Book a 20-minute walkthrough