The half that belongs in a clinic conversation rather than in a forum: aerophagia is one of the classic reasons a pressure prescription gets revisited. Comfort features that ease the pressure on exhalation exist partly for this, and so do other modes — but which of those applies is a judgement made from your data and your history by the person who prescribed it.

What you can bring is precision. Note which nights were bad, what the pressure was doing on those nights compared with good ones, whether the bad nights were also the high-leak nights, and whether it correlates with position or with the evening. "It is worse on the nights the machine spends longest at the top of its range" is a genuinely useful sentence.

What not to do is lower your own pressure to stop the bloating and leave your apnea untreated in exchange. That trade is a real one and it should be made by somebody who can see both sides of it.

Nothing on SomniTalk is medical advice. Talk to your own clinician before changing therapy.

One data note, because it decides more of this choice than the spec sheet does.

Machines differ in what they record and how you get it out. Some write a full night of detailed data to a card that you own; some report a reduced summary to an app and nothing else; some do both but only if the card is actually in the slot. If you ever want to answer a question more specific than "was my AHI under 5", that difference matters more than the model number.

So the practical question to ask before buying anything: does this machine take a card, and does it write detailed data to it? If the answer is no, every future conversation about your own therapy is limited to what a summary screen chooses to show you.

Nothing on SomniTalk is medical advice. Talk to your own clinician before changing therapy.

Why the distinction is worth the effort, for anyone reading their own charts for the first time.

Two events can be the same length and the same drop in flow and mean completely different things. One is a blocked pipe: the effort is there, the airway is not open, and pressure is the treatment. The other is the instruction to breathe not arriving: the airway is open, there is no effort to see, and more pressure is not automatically the answer — sometimes it is part of the cause.

Practical points when you look at your own data:

• Read the event mix, not just the total. The same AHI made of different events is a different night, and a shift in the mix over weeks is a genuine finding.
• A handful of clear-airway events is ordinary, particularly as you fall asleep and after an arousal. It is runs and trends that carry information.
• What your machine labels is a best guess from flow. It is a good guess, and it is not a diagnosis; only a study measures effort directly.

Which is why the useful output of reading your own charts is a description — this pattern, this often, since this change — handed to the person who can act on it.

Nothing on SomniTalk is medical advice. Talk to your own clinician before changing therapy.

This is one of the most alarming things a chart can show a newcomer, largely because of the name, so it is worth describing what it actually looks like in your own data.

The pattern people arrive with: obstructive events dropped away nicely once therapy started, and clear-airway events climbed to take their place — sometimes over the first days, sometimes over the first weeks. The total AHI can even look worse than it did before, which is what usually prompts the message.

What the chart shows: pauses with the airway open, so no effort against a blockage — the flow simply stops for a while and resumes. Frequently in runs, and often clustered around the transitions into and out of sleep rather than spread evenly through the night. That clustering is worth noticing, because events at sleep onset are extremely common and mean something different from events distributed across a whole night.

In a large share of cases this settles on its own over the first weeks to months as breathing regulation adapts to the new pressure. That is the usual course, not a promise about you.

The practical difference between the two routes, since this is the article people land on when they ask what the cloud actually gets.

The card holds the detailed night: flow, pressure over time, individual events with their timestamps, leak as a trace rather than a number. It is what makes questions like "when in the night does this happen" answerable.

The cloud summary holds the headline: hours, an event index, a leak figure, a pressure figure. Enough to confirm you used it and roughly how it went; not enough to work out why a night was bad.

Which is why "my app says my AHI was 4.2" and "I looked at my data" are different sentences. If the only record you keep is the summary, the detailed night is gone once the card is overwritten — cards do get reused, and the oldest detail is what goes first.

One thing the data is good for here: it separates the two problems for you.

A night that starts late puts the trouble at the front. A night that ends abruptly at the same hour puts it at the back — and if usage stops at 3 a.m. with a mask-off marker, that is not insomnia at all.

That distinction is worth bringing to an appointment, because "I sleep badly" and "my therapy ends at 3 a.m. four nights a week" get very different responses.

One practical note, since compliance is the part that makes people anxious.

Your own record of your usage is worth keeping regardless of what any portal shows. Systems lose nights, cards get read late, and cellular reporting has gaps. Having your own history means a disputed fortnight is a five-minute conversation instead of an argument you cannot win.

And the thing worth saying plainly: hitting an hours target is not the same as therapy working. Plenty of people clear a compliance bar and still feel terrible, and plenty of people miss it during a rough month while the therapy is doing exactly what it should.

How to tell whether what you tried actually worked, rather than relying on how the morning felt.

Mouth leak has a characteristic shape: fine early, then a step up that holds as the jaw relaxes. If you try a strap or a different mask, compare a week before with a week after and look for that step disappearing — not for a lower average, which moves for a dozen unrelated reasons.

The part that deserves its own post, because it inverts the intuition:

A large leak can make your event count look better than it is. When the leak is big enough, the machine cannot reliably detect events, so fewer get scored. The low number is an artefact, not an improvement. This is why "my best night ever" and "my worst leak ever" showing up together is a warning, not a win.

Two practical notes on leak numbers:

• The brands do not report the same thing. Some report only the unintentional excess leak; others report total leak including the air the mask is designed to vent. A higher raw number on one brand is not automatically a worse seal, and the two are not directly comparable.
• Short spikes are normal — rolling over, repositioning. It is the sustained periods that corrupt a night's data.

And the thing not to do about it: raising pressure to chase a leak generally makes the leak worse, not better. Seals are fixed mechanically — size, fit, tube routing — and pressure belongs to your prescriber.

Adding what those comparisons look like when people send them back, because the pattern is consistent.

The good month and the bad month are rarely far apart on AHI — a tenth or two, well inside the night-to-night noise. What separates them is one of the other lines: half an hour less on the mask, a leak figure that has crept up by a quarter without tripping anything, or a pressure the machine now reaches most nights and used to touch twice a week.

The seasonal note is real, too. A run of these arrives every spring about a winter that went wrong.

A thread for the habit almost everyone develops and nobody admits to: checking the number before checking whether you feel rested.

Some grounding, so the number stops running the morning:

• Under about 5 is the widely used benchmark for effective therapy, and it is the range clinical guidance treats as a good titration outcome. Targets are individual — yours is set by the person treating you.
• Zero is not the goal and is not realistic. Some residual events are normal on effective therapy.
• There is no recognised standard that makes "under 1" a target. Some clinicians aim lower when it is comfortably achievable; that is a preference, not a threshold.
• A single night is noise. A spike on one night surrounded by good nights is almost always a story about position, alcohol, congestion or a shifted mask.

Which means 4.9 and 5.1 are the same night with different rounding, and treating one as a pass and the other as a fail is the fastest route to being miserable about a therapy that is working.

And the data half of point 4, since it is the one this board can help with most.

Night-to-night variation is normal and large. Sleep position, alcohol, a cold coming on, how much of the night was spent in REM sleep, a mask that shifted at 3 a.m. — all of them move the numbers, none of them mean your therapy stopped working. The published guidance is to read a trend over roughly two to four weeks, not any single reading.

Two habits worth forming in week one:

• Read the leak figure next to the event count, never on its own. A large leak stops the machine from scoring reliably, which can make a bad night look like a good one. A great number on a high-leak night is one of the least trustworthy things on the screen.
• Write down what was different on a night that looks odd. Half the mysteries solve themselves that way.

The shelf article for exactly this: Why Does My AHI Change Night to Night? Normal Variation Explained.

The third version of the tired question, and the one where the data actually earns its keep. Anonymised:

"Therapy was brilliant for about a year. Some time over the winter it stopped being brilliant. My AHI still looks roughly the same, I have not changed anything, and I am back to falling asleep on the sofa. Nothing on the screen tells me what happened."

Something on the screen usually does tell you what happened — but not on last night's screen. This is a comparison question, answered by putting two months side by side. And since "I have not changed anything" is in nearly every one of these messages, it goes to the desk that deals with what changes without anybody deciding it.

One thing worth saying to anybody weighing this up who is new to CPAP data: you do not need the deep-analysis engines to answer your first three months of questions.

Usage, AHI, leak and pressure — the numbers that a clinic actually reviews — are all on the cheaper tier. The engines earn their keep for the specific case of "my numbers look fine and I still feel terrible", which is a real and common case, but it is not where anybody starts.

Anonymised, and this one genuinely does get reported as a bug:

"I cancelled an import because I had uploaded to the wrong profile. It said cancellation started, and then the button sat there greyed out saying Cancelling for several minutes. I gave up and uploaded again. Now I am not sure what state anything is in."

From the data side, two things worth knowing before you compare your leak figure with anybody else's.

The brands do not report the same thing. Some report only the unintentional excess; others report total leak including the air the mask is designed to vent. A bigger raw number on one brand is not automatically a worse seal, and comparing across brands is a trap.

Shape matters more than the average. Short spikes when you turn over are normal. A step up that lasts an hour corrupts a night: while it happens the machine cannot score reliably, so a heavy-leak night can produce a flatteringly low event count.

A framing that helps, since this is really a question about what a number over 90 days is for.

A quarter-long average is not a better version of last night. It answers a different question — "is the direction of travel right?" — and it answers it best when you look at it rarely. Reading a 90-day average every morning is a way to feel bad about noise: the number moves for reasons that have nothing to do with your therapy, and the wide scope exists precisely so a single bad night stops mattering.

If a trend genuinely is moving the wrong way over weeks, that is exactly the thing to take to the person who can act on it, with the printable report from the Reports page in your hand.

Nothing on SomniTalk is medical advice. Talk to your own clinician before changing therapy.

Ours from here. What is worth looking at when the headline numbers are good, roughly in order of how often it turns out to be the answer.

1. Flow limitation. Breaths that are restricted but never collapse fully. They do not score as apneas or hypopneas, they can carpet a whole night, and they are the commonest explanation for "great numbers, terrible sleep".
2. RERAs — arousals caused by respiratory effort. Not every machine flags them, and none of them count towards AHI. A night with a low AHI and a pile of arousals is still a fragmented night.
3. UARS. Restricted breathing and repeated arousals without the events a scoring rule recognises. Real, under-diagnosed, and the reason some people feel unwell on numbers that look perfect.
4. Arousals from the therapy itself — pressure that is uncomfortable at some point in the night, aerophagia and the bloating with it, a mask that wakes you when you turn over. Worth separating from the apnea, because the fix is different.
5. Leak that is intermittent rather than high. A short run of large leak in the middle of every night can wreck the sleep and barely move the nightly average.
6. Everything that is not sleep apnea. Iron and ferritin, thyroid, medication side effects, depression, restless legs, alcohol timing, plain chronic short sleep. Several of those are ordinary blood tests, and that part is a conversation with a doctor rather than with a chart.

The practical move is to bring something specific rather than argue about AHI: a few weeks of data, the pattern rather than one night, and a plain sentence about what the tiredness looks like in your day. "My AHI is 0.8 and I fall asleep at my desk every afternoon" is much harder to wave away.

One more sighting of the same family, worth knowing before you report it: a view that still shows data you just deleted or cancelled.

Deletions and cancellations finish in the background, and the app refreshes itself when the cleanup completes. If a chart or calendar looks stale in the meantime, press F5 — browsers cache these pages hard. If it is still there after a refresh and a minute of patience, that is a real bug and worth a report in Bug Reports.

And the standing rule for everything on this board: numbers on a screen are a description of one night, not a verdict on your therapy.

Nothing on SomniTalk is medical advice. Talk to your own clinician before changing therapy.

Anonymised, and a fair complaint about our own naming:

"There are three engines with three names and I cannot tell what the difference is supposed to be. Do I run all of them? Do they disagree with each other?"