The prep bag is by the door. This is the important fact of the room at 10:13 PM.
The zipper is closed. She checked it at six — CAS-card (physical laminate, subscriber record embedded in the chip, necessary because the imaging intake desk has gone to non-digital verification for first-party documents since the portability registry dispute last winter), water bottle (empty, to be filled in the morning from the B-wing fountain or not at all), the confirmation printout from the UPMC Scheduling agent in a pocket folder. She printed the confirmation because the portal's terms of service specify that printed confirmation is accepted where digital access may be limited. She does not know if access will be limited. She printed it because the cost of printing one page is zero and the cost of not having it is not.
She drew the line at six and she has held it. The bag is complete. Checking it again would be doing the correct procedure with the wrong intention.
The CAS-monitor runs its 9:30 calibration pulse from the kitchen windowsill, blue-green light against the glass, THRESHOLD-STABLE. She knows the numbers behind the status field: 84 IU/mL, plus or minus four depending on hydration and dose timing, holding that range for ninety days. THRESHOLD-STABLE is not a medical opinion. It is a machine's report that nothing has moved outside the corridor the attending set in July. Tomorrow morning, at Building C, the MRI will look at things the CAS-monitor cannot. The CAS-monitor watches surface chemistry. The MRI watches structure. They are not the same instrument and she has learned not to read one's silence as the other's answer.
She has two questions for the readout.
The trajectory question: is the seven-point IU/mL drop from the August baseline meaningful or within variance? She can hold this one. It is a real question with a real answer that the imaging data will bear on directly.
The baseline-gap question: how much of the August reading was affected by the three-day dehydration event at the end of that month? She can hold this one too. It requires calibration data the attending has and she does not.
The third question has no form yet. She has been preventing it from forming all day.
At 11:30 AM she was making tea. The household agent had surfaced the pre-imaging protocol checklist at the two-day mark yesterday — CAS-card charged, morning dose timed correctly, prep bag assembled — and today it surfaced the one-day items: Attending readout typically scheduled within 72 hours. Clear schedule where possible.
Standard text. She has received it before, for the two previous scans. She read it and set the phone down and waited for the kettle.
The third question came close while the water heated. Not the question itself — its shape, the weight of it, the territory it covered. The 72-hour window. The attending's posture when she has something to say. She made herself stay with the kettle. She poured the tea. She drank it standing at the counter, looking at the CAS-monitor in the window, which was cycling through its late-morning background mode, a quiet amber tick every ninety seconds. She waited until the shape went back under. It did.
At 1:45 PM she was not doing anything in particular, which is probably why the third question found more room. She was at the window looking at the building across the way — the Zabriski-adjacent building with the two UPMC relay nodes on the south face, identifiable by their amber bezel housings and the provider code stenciled below them. She was not thinking about anything. The shape came up from wherever it lives and she could feel its outline clearly for about eleven seconds before she recognized what was happening.
She put her hand on the CAS-monitor's charging surface and held it. The device ran a 1:47 PM ambient pulse, the standard mid-afternoon check against the rolling baseline window. THRESHOLD-STABLE. She counted four more pulses — the amber tick, ambient, nominal, nominal — before the shape went back under.
She is not proud of either of these moments and she is not ashamed. They are simply what happened.
The UPMC Intake agent sent the pre-imaging packet at 3:07 PM. The interface has changed since her last scan; the new subscriber portal uses a cascade form rather than a paged questionnaire, which means each question appears after the previous one is confirmed rather than displaying all sections at once. For returning CAS-card holders the expedited flow removed twelve questions from the standard battery. She answered the remaining forty-eight.
Question twelve is the one she had been expecting. It has appeared in some version on every intake form she can remember.
How would you characterize your psychological relationship to the upcoming diagnostic?
Option A: Prepared and calm. Option B: Anxious, with manageable symptoms. Option C: Anxious, with symptoms requiring clinical attention. Option D: Uncertain of my current state. Option E: Defer to clinical review.
She hovered over Option D for eleven seconds. She knows because she times things; she cannot stop herself and she stopped fighting it. Option D was not right. She was not uncertain of her state. She was certain of her state and uncertain of what the correct verb was for carrying it — whether carrying it was the same as managing it, whether managing was different from suppressing, whether the question's framework had a slot for the thing she was doing. It did not. The cascade form has five options.
She selected Option E.
She learned six months ago that Option E has history. She had been searching for something else in the patient-experience literature and found a behavioral systems paper from the UPMC AI Services integration team, a dry institutional document about intake completion rates and abandonment points in the first deployment cycle of the cascade form. In the first deployment, a significant fraction of patients were abandoning the form at Question 12. Not because they were distressed. Because they were doing what she does: holding something accurately and not being able to put it in any of the available slots. The paper noted this and the system team added Option E: Defer to clinical review.
The system learned from them and built the option they would have needed. She found this comforting in a way she has not fully mapped. The form knows her kind. Not her specifically — her category, her behavior pattern, the shape of what she does at Question 12. Someone else's hesitation became her exit.
The household agent sends the 10 PM quiet-mode confirmation at 10:16. She has it set to do this; it was one of the configurations she made in July, after the August dehydration event, when she was waking at 3 AM to check the CAS-monitor manually and realized that manual checking at 3 AM was not giving her anything the system was not already giving her — except wakefulness, which was the problem. The household agent sends a confirmation; she acknowledges it; she knows the overnight monitoring profile is active; she does not go to the kitchen to verify. That is the purpose.
Overnight monitoring active. CAS-monitor THRESHOLD-STABLE as of 21:30. Next scheduled pulse: 06:00.
She acknowledges it. She sets the phone face-down.
At 10:23 PM the CAS-monitor sends an unscheduled notification.
Not THRESHOLD-CONCERN. The color is the same blue-green and the status field still reads within range. But unscheduled, and the supplement text is different from the standard ambient pulse: Evening ambient variance detected — 4.2 IU/mL above 14-day rolling evening baseline. Within 90-day window.
She picks up the phone.
The CAS-monitor's 90-day window is the wider corridor; the 14-day rolling average is the tighter one, calibrated to her seasonal baseline rather than the full quarterly record. A 4.2 point variance above the 14-day evening average is within the 90-day boundary. The notification is informational. Green flag indicator, not amber, not red.
She sits on the edge of the bed and reads the notification three times.
The question she has been preventing from forming all day is now very close.
She puts the phone down. She picks it up. She opens the monitoring history in the CAS portal app and looks at the past two weeks of evening readings. The 14-day evening average is 83.1. Tonight's 10:23 PM reading is 87.3. She scrolls back: last Monday was 84.1. The Monday before was 84.8. Three of the past fourteen days show evening readings above 85.
This is information. She does not know if it is the information she is looking for, because she does not know yet what she is looking for. That is the problem with the third question: it is shaped like a question about this, about the variance and the trending and what it means relative to the structure the MRI will show tomorrow. But she cannot form it tonight because she does not have the structural data, and a question formed without the data it needs is not a question yet — it is the shape of a question, a demand the mind makes before it knows what answer would satisfy it.
She can feel the shape clearly now. She does not want to push it back under again. She also does not want to form it without its basis.
She opens the Adamski notepad on the nightstand — she moved it there without deciding to, sometime in September, and has not used it yet — and writes: Evening variance. 14-day rolling. Last three weeks? She caps the pen.
That is not the third question. It is a note to bring to the readout. A datum she will hand to the attending rather than a question she is carrying alone. There is a difference between information and what you build from it. She has been carrying the information correctly all day without knowing that was what she was doing.
The CAS-monitor runs its midnight ambient pulse from the kitchen window. THRESHOLD-STABLE.
The monitoring system holds information, she thinks. It does not hold me.
In July this meant: let the device do its job and go back to sleep. Tonight it means something with more layers — something about the difference between information and its implications, about how much of the weight she has been carrying is the information itself versus what she has been building on top of it without the structural data to support it. The system can watch. The system cannot interpret. The interpretation belongs to tomorrow, when the attending sits down across from her and pulls up the comparison scan and she is holding two questions and a note on a notepad about evening variance.
Alarm: 8:30 AM. UPMC Scheduling agent confirmation: 7:00 AM. Building C intake desk: 9:45 AM. MRI: 10:00 AM.
The prep bag is by the door, zipper closed.
The notepad on the nightstand has three words on it that were not there this morning.
She turns off the light. In 72 hours she will know what question she was almost asking.