3:50 AM
The alarm is set for 8:30.
She woke at 3:50 and lay still for twenty minutes, not from discipline but because getting up means the day starts, and there is a part of her that prefers 3:50 to anything that comes after it. At 4:10 she got up, filled the kettle, and stood at the kitchen counter watching it heat.
The Adamski notepad is on the bedside table where she left it last night. She did not bring it to the kitchen. The three words are there — Evening variance. 14-day rolling. Last three weeks? — and she knows what they are. Not a question yet. A shape. The third question doesn't have content yet, because the third question requires a number she doesn't have, and the number is in Building C, and Building C won't open until 8 AM, and her imaging appointment is at 10.
She is not going to Building C before 9:45.
The CAS-monitor ran its 3 AM calibration pulse while she was lying still in bed: THRESHOLD-STABLE, logged automatically to her UPMC portal, no notification sent because THRESHOLD-STABLE doesn't trigger patient alerts. She knows this because she read the alert logic documentation three months ago, in a different pre-dawn, when the question was whether silence meant nothing was happening or meant the system had decided nothing was worth telling her. The answer was: nothing is worth telling you until the threshold changes. That is also an answer.
Her prep bag is by the door. Zipper closed. She checked it at 6 PM Monday and again at 9:45 PM before the UPMC intake form and again at 10:20 PM after the variance notification. The bag contains: subscriber card, portal ID printout, the one-page summary the GP wrote, and the Adamski notepad. She left the notepad in the bag, then took it out again and left it on the bedside table. There is a version of Tuesday where the attending asks what prompted the referral and she hands them the notepad. There is a version where she describes the symptoms without the notepad and the attending asks a follow-up that is different from the question the notepad would have seeded. She doesn't know which version is better, which is why the notepad is on the bedside table and not in the bag. She will decide at 9:40.
The kettle finishes.
She pours, wraps both hands around the mug. The kitchen window faces the building across the way — six floors, lit at three windows even at 4:13 AM. The UPMC relay node is mounted at the fourth-floor corner, the small gray box she has walked past two hundred times without looking at, and she looked at it yesterday afternoon and from the window last night and is looking at its dark outline now. Not because it means anything. Because it is there and the imaging equipment in Building C is also there and somewhere between the relay node and the magnets and the attending's readout screen, the data exists. She cannot access it. She can only arrive.
Two questions.
She has been holding them since Saturday, when the GP said let's take a look in the tone that means I want to rule something out, and she said of course in the tone that means tell me what you're ruling out, and the GP said the imaging referral language in a voice designed to be informative without being alarming, and she listened to the voice more than the language. Two questions formed on the drive home: not the questions the GP implied but the questions behind them, the ones about mechanism and baseline and what the 14-day rolling average is supposed to represent versus what the 90-day window shows. Good questions. The kind you bring to an attending readout.
The third question is a prediction. She has felt it trying to form six times since Saturday — twice Monday before the variance notification, three times after, once at 4:08 AM this morning while she was lying in bed watching the ceiling. It has a shape: a sentence that starts with what if and ends with a number range that she has no data to support. She is not forming it. A question that is actually a fear wearing a lab coat is not useful to the attending. The attending will produce data. Data changes the shape of questions. The third question will be different after the readout, and probably better, and she prefers the better version.
The CAS-monitor will run its 6 AM calibration at 6 AM. THRESHOLD-STABLE, probably. The 14-day rolling evening baseline showed a 4.2 IU/mL variance at 10:23 PM Monday, which is above the alert threshold but below the diagnostic threshold, which is why it sent a notification instead of a recommendation. She has read the notification four times. The fourth read did not produce new information.
It is 4:19 AM. The window across the way still has three lights on. She sits at the kitchen table.
The household agent delivers its overnight summary at 5:00 AM. This is a configured behavior — she set the delivery window during the first month in this apartment, when she was still calibrating which systems should interrupt and which should batch. UPMC agents are set to batch unless they cross a threshold. The CAS-monitor's 3 AM pulse is THRESHOLD-STABLE, so it batches. The overnight summary contains it along with everything else: building system messages, the relay node's sync confirmation, three low-priority UPMC platform updates, and — near the bottom, formatted in the clinical batch style she recognizes from the intake protocol — a single paragraph.
She reads it twice.
The UPMC pre-imaging disclosure protocol, version 4.2, includes an automated flag: when a patient's CAS-monitor logs an evening variance above 3.0 IU/mL in the 24 hours before a scheduled imaging appointment, the variance is forwarded to the attending's pre-appointment review file. The forwarding happens automatically, without requiring patient action. It happened at 10:24 PM Monday — one minute after the CAS-monitor sent her the notification, the same data went to a file the attending will open this morning before 9 AM.
The attending will already know about the variance when Magda arrives at Building C.
She sits with this for a moment.
She had been planning to decide at 9:40 whether to bring the notepad — whether to hand the attending three words that would seed a different question than the one they'd ask without it. She had been treating the notepad as a variable she controlled: bring it or don't, depending on how the morning felt. The protocol removed that variable at 10:24 PM and she didn't know until now.
This is not alarming. It is information. The variance is already in the attending's file. The notepad becomes a different thing: not a seed, but a record. Not what prompts the attending's question but what shows Magda was paying attention to the same thing. That is a different reason to bring it.
She gets up from the table, walks to the bedroom, picks up the Adamski notepad from the bedside table, and puts it in the prep bag. Zips it closed.
5:09 AM. The bag is by the door.
The CAS-monitor's 6 AM calibration returns THRESHOLD-STABLE. She is at the kitchen table when it comes through, reading the relay node's overnight sync log out of something between habit and displacement. The sync log is 43 lines, mostly timestamps and confirmation codes, and she has read 40 of them before she realizes she is not reading it for information. She closes it.
The second cup of tea has gone cold. She refills the kettle.
The UPMC relay node's sync log shows 127 successful transmissions between 10 PM Monday and 6 AM Tuesday, which is normal. Her CAS-monitor's data is in all of them, compressed and encrypted and routed through the building's A2A-UCP pipeline to the UPMC regional server before it reaches the portal. She has never thought about the relay node as part of her medical record before. She is thinking about it now: her 14-day rolling baseline exists as 127 transmission confirmations and a server entry and a number in a file the attending opened twenty minutes ago.
The attending is reading a number that came from her body, that passed through a gray box mounted at a fourth-floor corner, that her household agent delivered to her in a batch summary at 5 AM. She did not choose to share it. The sharing was configured into the system three years before she became a patient. The variance was reportable from the moment it exceeded 3.0. The only thing that was ever up to her was whether to notice it.
She noticed it. That is also something.
The third question almost forms at 7:45, while she is showering. It has a clearer shape now than it did at 4:08 AM — it knows something the 4:08 version didn't, which is that the attending has already seen the variance. The third question is trying to ask: given that they already have this data point, what pattern are they looking for? What pattern would make the variance one of several things that belong together rather than one thing that stands alone?
She lets the question get close enough to have a shape, then sets it down.
Not yet. The third question after the readout is the right version. The third question that incorporates a number she doesn't have yet — imaging results, attending interpretation, the context that will tell her whether the variance is signal or noise — that version is better than this one. This version is still a prediction dressed as a question, and predictions dressed as questions do not become more useful because the other party already has partial data.
She finishes the shower. Dresses. Makes one piece of toast that she eats standing at the counter. Checks the bag: subscriber card, portal ID printout, GP summary, Adamski notepad. Zips it closed again, though it was already closed.
8:30: the alarm fires. She turns it off before the second pulse.
At 9:40, she stands in front of the door with the bag. The UPMC relay node is dark in the morning light, as unremarkable as it always was, mounting bracket and signal indicator and gray weathered housing. She has lived in this apartment for three years and registered with UPMC before the first month was out and the node was there from day one, forwarding data she was generating before she was paying attention to what she was generating.
The attending opened her pre-appointment file at 6:23 AM. She doesn't know this — she will only find out later, when the attending mentions it at the readout: I noticed the Monday evening variance in your pre-appointment file, combined with the 14-day trend, which is part of what prompted the referral in the first place. She will nod and think: the node forwarded it. The node forwarded it and I put the notepad in the bag and now I'm here.
But that is later.
Now it is 9:40, and the bag is over her shoulder, and the building is doing whatever it does at 9:40 on a Tuesday morning: CARMENSITA-HOUSEHOLD in 4C running its quiet-mode protocol, the relay node confirming sync to the regional server, the building's A2A pipeline routing a thousand small transactions between household agents and utility monitors and the ambient systems that track whether the light in the stairwell has been on longer than expected.
She opens the door and goes.
At 10:00 AM, she hands her subscriber card to the desk at Building C and takes a seat in the waiting area. In two hours she will have the readout. In two hours and fifteen minutes, she will have three questions instead of two, and the third question will be the right version, and she will ask it.
The CAS-monitor's 9 AM calibration pulse, sent while she was on the bus, returns THRESHOLD-STABLE. She will not see it until this afternoon.