The case for 14-day mood check-ins between therapy sessions
Between-visit drift is the quiet failure mode of outpatient behavioral health. Here is why a structured 14-day reflection arc closes the gap without becoming a second EHR.
There is a quiet failure mode in outpatient behavioral health that nobody likes to name: between the last visit and the next one, the clinician's working model of the patient slowly goes out of date. Two weeks later, you ask "how have things been since we last spoke?" and the answer is a compressed, edited, often optimistic summary. You do your best work in that hour, but the hour is built on a smaller piece of evidence than it should be.
Mood-tracking apps have been trying to fix this for over a decade. Most of them fail in one of three ways. They become a second EHR the clinician has to log into. They generate so much noise that signal disappears. Or they cross the line into telling the patient what is wrong, which is exactly the line a decision-support tool should never cross.
A 14-day reflection arc, scoped carefully, can do better than that.
Why fourteen days
Two weeks lines up with the cadence most outpatient psychiatric and primary-care behavioral-health visits already use. It is short enough that a patient will commit to a daily reflection without it becoming a chore. It is long enough to capture more than a single bad weekend. And it gives PHQ-9 and GAD-7 enough room to sit at three meaningful checkpoints (baseline, mid-arc, and pre-visit) rather than being administered once at the door, where the score is biased by the act of entering a clinical setting.
The arc also gives the patient a finish line. "Reflect every day" with no defined end is a permission-less commitment that quietly degrades into nothing. "Reflect every day for two weeks until your next appointment" is a concrete ask, with a payoff the patient can see.
What the patient should actually do
The temptation is to ask for a lot. Sleep, food, mood scores, energy, social contact, panic episodes, suicidal ideation, substance use. The temptation is the trap. Patients who get a Likert form on day one will stop opening the app by day four.
The version that works is closer to a journal than a data entry surface. A short prompt: "Tell me about one moment from today or yesterday that's been on your mind." The patient writes (two sentences, two paragraphs, doesn't matter) and the app extracts signals from that text rather than asking for them directly. Anxiety, anhedonia, rumination, sleep disruption, suicidal ideation: all of these can be classified out of a single open prompt with reasonable accuracy.
The exception is the validated instruments. PHQ-9 and GAD-7 must be administered with their original wording: every word, every answer option, in the original order. Reword them and you have invented a new instrument with no validation behind it. So on days 1, 7, and 14, the patient sees the actual scales, exactly as the literature wrote them.
What the clinician should actually see
Not a dashboard of scores. Not a feed of every reflection. The hour before the next visit is the wrong time to read fourteen days of journal entries.
What the clinician needs is an evidence map: a summary that points at the visit, with which signals appeared and how often, which scales moved and by how much, which moments are worth opening verbatim. Source-backed: every signal cites the patient turn that produced it, so the clinician can verify the classification before relying on it. Auditable: every action (patient submission, classifier output, scale capture, clinician note) sits in an append-only log with a hash chain, so disputes about what was shown when can be settled with evidence rather than memory.
This is the part the existing mood-tracking apps fail at the hardest. They show the patient a chart of their PHQ-9 score over time, which is exactly the wrong audience for that chart. The PHQ-9 trend belongs in the clinician's review. The patient should see their reflections, not a scoreboard.
What the system must never do
Three lines that decision-support has to stay behind:
- Never diagnose. Even when a classifier is reasonably sure something is happening, the output to the clinician is "signal detected with these citations", never "patient appears to have generalized anxiety disorder." The diagnostic act stays with the licensed clinician.
- Never medicate. No surfacing of medication options, no "patients with similar profiles often respond well to X." Treatment recommendation requires evaluation, history, formulation. A 14-day check-in app has none of those.
- Never replace the crisis line. If a patient indicates active suicidal ideation, whether PHQ-9 item 9, classified intent in a reflection, or anything else, the screen escalates immediately with 988 and 911 visible as plain readable text. The app is not monitored in real time. It cannot be the first line of crisis response, and the patient must know that before they ever sign in.
The economics, briefly
A clinician who walks into a follow-up visit already knowing which patterns moved gets that hour back. The patient doesn't spend the first ten minutes summarizing two weeks under prompt. The clinician doesn't spend the next ten asking the questions the summary skipped. The visit starts at the actual work.
We have not run a randomized trial that proves this saves time. The framing matters anyway: the goal is not faster visits, it is better-evidenced ones. The patient who shows up with fourteen days of context behind them is a different patient than the one who walks in cold, and that is true whether or not the visit ends sooner.
What we are not promising
Nyra is decision support. It is not medically approved. It is not an emergency service. It does not replace a clinician's judgment, and it explicitly blocks any claim, anywhere in the product, that it does. We will not ship a feature that quietly drifts into clinical recommendation, and we will not light up integrations with EHRs until the compliance work behind that integration has been done in the open.
The 14-day check-in is the smallest version of this that can be useful without crossing those lines. That is the version we are building.
Where to go next
If you are a clinician thinking about how this would land in your practice, book a thirty-minute walkthrough. It is a live tour of the patient flow and the clinician review surface, with time at the end for the questions this post does not answer.
If you are a patient curious about Nyra, the app is invite-only through your clinic. Ask your clinic whether they are taking part, or read more about what the patient experience actually looks like.