Docs · dashboard

The roster and concern detection

Your patient list, how it is triaged, what the at-risk flag means, and every concern tier explained in plain clinical language.

The roster is your home base. It lists every patient in your institution and tells you, at a glance, who needs your attention today. Open a patient from here to reach their full workspace.

The roster table showing columns for Member, Status, Consent, Last active, Streak, Unread, and At-risk, with at-risk patients sorted to the top
The roster, sorted so the people who need you are already at the top.

How the list is ordered

You do not sort this list yourself. It arrives in triage order, so the top rows are always the ones worth reading first:

At-risk patients first, most urgent first
Anyone flagged at risk floats to the top, ranked by concern tier: Critical, then At risk, then Watch.
Then most unread messages
Among the rest, patients with unread messages to you come next.
Then most recently active
Everyone else falls in by how recently they used the app.

The search box filters by name or email. A patient who deleted their own account shows a greyed Account deleted row for the 30-day retention window; the name stops linking anywhere, because their detail access is gone.

What "at risk" means

The at-risk badge on the roster is a rollup of the same signals you will see, spelled out, inside the patient. The single most important reason shows under the badge so you know why before you even open them. A sustained mood decline is one of these rollup signals now: a patient whose recent mood drifts below their own baseline surfaces as a medium mood trending down reason on the roster, not only on their member page, so you can catch it without opening them. The badge is tiered to match the concern behind it: a red Critical for the acute lane, a red At risk for a high concern, or an amber Watch for a medium one.

At risk is a prompt, not a diagnosis

The flag says "look here today." It never replaces your clinical judgment, and it is never shown to the patient. Everything behind it is transparent once you open the patient.

Concerns, tier by tier

Inside a patient, the concern panel sits at the very top of every tab. Each concern shows a colored severity band, a tier badge (Acute, High, Watch, or Info), whether it is a one-off event or a sustained pattern, and a plain-language "Flagged because" line so you can see the evidence, not just a score. Clicking a concern jumps you to the tab that holds the underlying data.

The concern panel showing several stacked concern cards, each with a colored left border, a severity badge, a headline, an event-or-pattern tag, and a 'Flagged because' explanation line with an Acknowledge button
Every concern explains itself. No opaque risk number.
TierConcernFires whenReads as
AcuteCrisis content in chatAI moderation flagged one or more messagesEvent, same-day eyes
AcuteSuicide screen elevatedA recent C-SSRS scored 4 or higherEvent, review today
AcuteHigh-risk clinical noteYou marked a recent note high-riskEvent
HighWeight down 5%+Rapid loss across recent synced weigh-insPattern
HighUrge to purge after mealsThe urge to purge was flagged at 2+ meals in 7 daysEvent
HighRestrict-then-binge patternA binge urge follows restriction (a skip, a restrict urge, or a very low day) the same or prior day, 3+ times in 7 daysPattern
WatchRestrict-then-binge patternThe same restrict-then-binge coupling seen 2 times in 7 daysPattern
HighPossible binge episodes3+ meals in 7 days far larger than the patient's own usual, each carrying a binge urgePattern
WatchPossible binge episodes2 such loss-of-control episodes in 7 daysPattern
HighUrge to binge after mealsThe urge to binge was flagged at 5+ meals in 7 daysEvent
WatchUrge to binge after mealsThe urge to binge was flagged at 3 to 4 meals in 7 daysEvent
HighMeals not eaten60%+ of answered meals marked not eatenPattern
WatchMeals not eaten40 to 59% of answered meals marked not eatenPattern
WatchUrges reported after mealsAny urge flagged at 3+ meals in 7 daysEvent
WatchHard mealsThe patient tapped a struggle feeling at one or more mealsPattern
WatchMood trending downRecent mood dips below the patient's own prior baselinePattern
WatchEngagement downCheck-ins fell 40%+ versus their own prior ratePattern
WatchModerate-risk noteYou marked a recent note moderate-riskEvent
HighVery high water intake5+ days at or above roughly double their water goalPattern
WatchVery high water intake3 to 4 such daysPattern
InfoLow water intake5+ logged days at or under half their water goalPattern
InfoMeal gapThe longest stretch without a logged meal in range hit 3+ daysPattern

A few things worth knowing about how these read clinically:

  • Weight, skip-rate, and mood are baseline-relative and windowed. A loss or a skip streak from months ago will not keep firing today. Concerns look at recent data; the per-tab charts still show full history.
  • High water outranks low water, and the app never mentions either. Drinking to excess is a compensatory behavior, so it is ranked as one; under-drinking is informational. The patient's app stays silent about both, because flagging high intake in-app teaches people to log less rather than drink less. See Water.
  • Binge-eating patterns require loss of control, never size alone. The binge concerns key off the self-reported urge to binge (the app's proxy for loss of control), not calorie amounts. A large meal on its own is never flagged, because a big meal can be normal, a celebration, or recovery-appropriate eating. The restrict-then-binge concern reflects the dietary-restraint model (Fairburn CBT-E; restriction raises binge odds the same and next day), and "possible binge episodes" needs both an unusually large amount, relative to that patient's own logs, and the binge urge. This lane is behavioral and non-evaluative by design; it never scores calories or weight.
  • The composite amplifier. When two or more of the pattern signals above are happening at once, Pippa raises one extra High concern: "concerning patterns at once." Co-occurring signals predict relapse far better than any single weak one, so they get combined and lifted.
  • Red is reserved. Only the acute lane (crisis chat, positive suicide screen, high-risk note) gets red. This keeps red meaningful and guards against alarm fatigue.

Acknowledging and hiding a concern

Each open concern has an Acknowledge button. Acknowledging records your name and the time, then moves the concern into a quiet "Acknowledged" line at the bottom of the panel. It does not delete anything; it says "I have seen and handled this."

A new event re-alerts

Acknowledgements are keyed to the specific occurrence. If a new crisis message, a new severe measure, or a fresh weight drop into a worse tier comes in later, it raises a new concern rather than staying hidden under an old ack. A pattern drifting within the same tier (say 38% to 39% skips) will not churn your acknowledgement.

If a concern is not clinically useful for a given patient, hiding it removes the row from your view without recording a formal acknowledgement.

Sharing your screen?

When patient view is on, the roster replaces itself with a "Patient list hidden" card and every concern disappears, so navigating mid-session never exposes another patient or an alert.

A single concern card with the Acknowledge button focused, and below it a quiet 'Acknowledged' summary line showing a previously handled concern with its date
Acknowledging keeps an audit trail without cluttering the live panel.
Was this page helpful?