You open your heart-rate feed and find a wall of tachycardia episodes. The alerts cluster around grocery runs, hot showers, dog walks, and the few minutes after you climb stairs. Then a genuine 145 bpm episode arrives at 3 a.m., but it's buried beneath everything your watch and app have flagged since Monday.
That pattern is familiar to people living with POTS or dysautonomia. The detection system isn't necessarily broken. It's responding to increased heart rate, but it may not know whether the rise came from standing, exercise, heat, motion, anxiety, or a clinically important event. False positive reduction isn't about making the graph look cleaner. It's about restoring trust in alerts and giving your clinician an episode log they can truly interpret.
The practical fix is a workflow, not a single sensitivity dial. Start with context-aware workout exclusion, then tune the detection criterion, attach symptoms and triggers, improve watch placement, and export reports with the relevant settings preserved.
Why False Positive Reduction Matters for Tachycardia Alerts
A heart-rate alert can be physiologically accurate and still be clinically unhelpful. If your pulse rises during a shower, a brisk walk, or a startled reflex, a threshold-only detector may record the event exactly as designed. The problem is that the alert lacks the context needed to distinguish a predictable exertion response from a sustained orthostatic episode.
For someone with POTS, that distinction affects daily decisions. After enough irrelevant notifications, you may stop checking every alert. That habit creates a dangerous failure mode, not because every alert represents an emergency, but because the one event worth reviewing can disappear in a feed dominated by ordinary activity.
Practical rule: Don't judge an alert by the number alone. Judge the rise, its duration, your position, your symptoms, and what you were doing immediately beforehand.
Noise damages the record
A crowded episode feed also weakens clinical communication. A cardiologist or primary care clinician has limited appointment time, and a long list of unclassified spikes forces them to spend the visit separating exercise, motion, recovery, and possible orthostatic events. The result is less attention on the episodes that include dizziness, palpitations, breathlessness, chest discomfort, or near-fainting.
Research in high-stakes screening shows why this principle matters beyond heart-rate monitoring. A 2021 Nature study of hybrid breast-cancer screening models reported a 37.3% average reduction in radiologists' false positives, lowering the false-positive rate to 12.0% while maintaining the same number or fewer false negatives. The lesson is broader than breast imaging: a useful system must suppress avoidable alarms without weakening detection of meaningful events.
The goal is actionable data
False-positive reduction should therefore improve three things at once:
- Attention: You should know which notifications deserve a look.
- Interpretation: Each episode should retain enough context to explain why it happened.
- Clinical review: Your exported report should show the criterion, timing, symptoms, and surrounding circumstances.
Start with the largest predictable source of noise, workouts and recovery. Then adjust the criterion only after you've seen how much context-aware exclusion removes from the feed. That order matters. If you change the threshold first, you won't know whether the new setting solved the actual problem or just hid useful episodes.
Let Cardiogram Handle Workout and Recovery Exclusion First
The highest-value cleanup step is to exclude exercise before changing your tachycardia criterion. Cardiogram can use Apple Workout data and heart-rate session information from Apple Health to identify periods when a high pulse is expected. When Workout Mode is enabled, tachycardia alerts that begin during a logged workout can be suppressed, along with a configurable recovery period afterward.
That recovery window is essential. Heart rate often remains high after exercise, exactly when a threshold-only system is most likely to create another benign alert. A cooldown of twenty to sixty minutes can remove that predictable post-workout cluster. The right duration depends on how long your heart rate stays high, so choose a window that reflects your usual recovery rather than treating it as a universal medical rule.
Open the app and go to Settings → Notifications → Workout Mode. Turn on workout exclusion, select the recovery period, and then leave the rest of your detection settings unchanged for a full week. You need a clean observation period before deciding whether the criterion itself needs adjustment.
Log activities the system can't see
Automatic exclusion only works when the activity is recorded. Log Apple Watch workouts, even if you're mainly interested in POTS episodes. A genuine arrhythmia or abnormal episode that begins during a logged workout may also be excluded, so Workout Mode is a contextual filter, not a diagnostic safeguard.
For yoga, slow cycling, stretching, household exertion, or other activity that isn't captured as a formal workout, add a manual activity note. Don't skip the workout record to force an alert. That approach creates a noisier dataset and makes later clinical interpretation less reliable.
A POTS-focused heart-rate monitoring guide can help you think through how activity, posture, and recovery fit into your broader tracking routine. Keep the workflow simple: record the activity, allow the recovery exclusion to run, and observe the feed before touching the threshold.
The same principle appears in patient-monitor alarm research. An alarm-filtering study found that adaptive time-delay methods reduced clinically irrelevant false alarms across alarm types, with the strongest results coming from context-aware suppression rather than tightening thresholds. Your first move should be the same: remove predictable context, then tune sensitivity.
Tune Detection Criteria to Match Your Real Episodes
Once workout and recovery noise is under control, adjust the two settings that shape the episode feed: magnitude and sustained duration. Magnitude asks how large the heart-rate rise must be. Duration asks how long the high rate must continue before the app records it as an episode.
Cardiogram's automatic detection is built around a 30+ bpm rise within five minutes, with sustained capture that records the event rather than treating every momentary spike as a complete episode. That structure fits an orthostatic pattern, but it won't suit everyone equally. A person with a resting baseline near 90 bpm may generate more everyday alerts than someone whose baseline is lower, especially if the person regularly climbs stairs, handles pets, cares for children, or moves quickly between tasks.
Change one dial at a time
Open Episode Settings and decide what problem you're solving.
If you're missing episodes that clearly follow standing, consider a lower bpm threshold only after confirming that the events aren't being excluded as workouts or distorted by motion. If your feed is full of short spikes that disappear quickly, increase the sustained-duration requirement. A window of thirty or sixty seconds can filter startled reflexes and brief movement artifacts that resolve before a sustained orthostatic pattern develops.
Use these profiles as starting points, not medical prescriptions:
| Profile | bpm Threshold | Sustained Duration | Expected Daily Episodes |
|---|---|---|---|
| Conservative POTS profile | 30 bpm | 60 seconds | Lower noise, with short events filtered |
| Dysautonomia with anxiety | 40 bpm | 30 seconds | Fewer reflex-related spikes |
| High-baseline profile | Above 110 bpm | 30 seconds | Focus on sustained high-rate periods |
The heart-rate threshold guide can provide additional context for choosing a criterion that matches your monitoring goal.
Avoid the noise combination
The worst combination is a low bpm threshold paired with a short duration. That setup catches nearly every abrupt transition, including standing quickly, reaching, talking while moving, or reacting to stress. It may feel sensitive, but a feed that requires constant manual sorting isn't giving you better monitoring.
A better process is deliberately boring:
- Keep workout and recovery exclusion active.
- Choose one duration or magnitude change.
- Run that setting for a week.
- Compare episode count, symptoms, posture, and activity.
- Keep the change only if it improves interpretation without hiding events you care about.
False-positive reduction is successful when the remaining episodes are easier to explain, not fewer.
Log Symptoms and Context to Separate Noise From Signal
An alert without context is just a timestamp and a number. Log what you felt and what was happening at the moment, even when the episode seemed harmless. A high heart rate during a shower with no symptoms is different from the same rate after standing, accompanied by dizziness and palpitations.
Use a small set of repeatable fields. You don't need to write a long diary entry for every notification. Consistency matters more than literary detail.
| Episode field | Examples to record |
|---|---|
| Symptoms | Dizziness, palpitations, fatigue, brain fog, chest discomfort, unusual breathlessness, fainting |
| Body position | Supine, seated, standing |
| Activity | Shower, grocery shopping, dog walk, stairs, yoga, prolonged sitting |
| Hydration and salt | Recent fluid intake, salt intake, skipped fluids |
| Recovery factors | Sleep quality, illness, heat exposure |
| Timing factors | Caffeine, medication timing, skipped meal, emotional stressor |
Record position and timing
Body position often gives an episode more meaning than the peak value alone. Note whether you were lying down, seated, or standing, and roughly how long you'd been upright. A rapid rise shortly after standing may deserve closer review than the same reading during a planned workout or recovery period.
Record likely contributors without treating them as proof. Poor sleep, low fluid intake, heat, caffeine, medication timing, and a skipped meal can all help explain patterns across several days. They don't automatically make an episode false, and they shouldn't be used to dismiss symptoms that concern you.
A benign-feeling episode can still be useful data. Label the context, not the result.
Review your entries weekly. Look for recurring combinations, such as symptoms appearing after poor sleep and limited hydration, or alerts clustering after showers. Keep the original readings and timestamps intact. The purpose of logging is to make the data more interpretable, not to edit the record until it tells a cleaner story.
If a symptom is severe, new, or associated with fainting, chest pain, serious breathlessness, or other urgent concerns, seek appropriate medical care rather than relying on an app alert or its absence.
Fix Device Placement and Wear Habits at the Source
Cardiogram can interpret the heart-rate data it receives. It can't reliably distinguish a true tachycardia episode from motion artifact when the Apple Watch sensor is sliding, partially lifted, dirty, or pressed against an unstable surface. False-positive reduction starts with a stable signal.
Wear the watch snugly against clean, dry skin. The band shouldn't hurt, but it should prevent the watch from shifting during standing tasks and exercise. Keep the sensor back fully covered, avoid placing the watch directly over a prominent wrist bone, and don't let it slide toward your hand.
During a questionable episode, stop moving. Rest your wrist on a stable surface and keep the watch still for several minutes before comparing the reading with how you feel. If the rate changes sharply after you reposition the watch, treat that as a signal-quality clue rather than immediately changing the detection criterion.
Common sources of spurious readings
- Loose fit: The sensor loses consistent skin contact as your wrist moves.
- Excessive movement: Typing, clapping, gripping, or contact with a vibrating surface can disrupt optical readings.
- Moisture or buildup: Sweat, water, lotion, and residue can interfere with the sensor seal.
- Poor placement: A watch that presses on bone or sits too close to the hand may move more than expected.
- Post-exercise checking: Immediate hand movement and recovery can create a confusing combination of real elevation and noisy sampling.
Tighten the band slightly for exercise or tasks that involve repeated wrist movement, then return it to a comfortable fit afterward. Tightness can improve contact, but it isn't a replacement for correct placement or stillness.
Verify before you suppress
When an episode looks wrong, take a second measurement after cleaning and repositioning the watch. Compare the reading with your symptoms and activity. Don't delete or mentally discard a reading because you dislike it. Mark it as questionable and preserve the conditions under which it occurred.
The Apple Watch heart-rate tracking guidance can support better wear habits. The practical standard is straightforward: improve the physical signal first, then decide whether the software criterion needs changing.
Read Weekly Trends and Export Clean Reports
A single alert can mislead you. A weekly pattern usually tells you more. Use summaries and heatmaps to see whether elevations cluster at a particular time, after a repeated activity, on exercise days, or alongside logged symptoms and triggers.
Start by sorting episodes into practical categories. Narrow spikes that repeatedly follow the same activity may point to an exclusion or fit problem. Sustained elevations that occur after standing and coincide with dizziness or palpitations deserve closer attention. Neither pattern proves a diagnosis, but each gives your clinician a better question to investigate.
Use the feed as a pattern-finding tool
Compare:
- Time of day and day of week
- Exercise and recovery periods
- Standing or prolonged sitting
- Sleep quality and hydration notes
- Medication timing
- Symptom presence and severity
Don't delete legitimate episodes because they make the report look messy. Instead, preserve the raw event and add an interpretation such as “after shower,” “sensor questionable,” or “standing with dizziness.” A clinician can assess the original record more effectively when your notes show both the event and your uncertainty.
A critical review of remote-monitoring alarm reduction methods warns that lower alert counts don't automatically prove safer or more useful care. The important question is whether the remaining alerts are actionable and whether true events remain visible. That's why weekly review should include symptoms and downstream usefulness, not just episode totals.
Export for clinical review
Before sharing a PDF report, check the selected date range, timestamps, time zone, episode data, and detection criterion. The report should make clear which rule was active when the episodes were recorded.
Add a short summary with:
- The symptoms that matter most
- The activities or positions linked to episodes
- Any major change in sleep, hydration, illness, or medication timing
- Which events look questionable and why
Send the original report rather than a collection of screenshots. A report preserves context and gives your clinician something they can annotate or compare at a later appointment. Keep the raw details alongside your summary, especially when you've changed a criterion or enabled a new exclusion.
Build a Repeatable Routine That Keeps Alerts Honest
False-positive reduction works best as a routine you can follow without making every alert a new investigation. The app should handle predictable filtering, while you provide the context it can't infer safely from heart rate alone.
Daily
Open the episode feed once in the morning. Tag overnight events, record sleep quality and morning medication timing, and check that the watch sits securely before you start moving through the day.
During the day, respond to a live prompt when it appears, but don't keep refreshing the feed. Constant checking encourages overreaction to isolated readings and makes ordinary fluctuations feel more significant than they are.
Weekly
Review the weekly summary and heatmap. Compare the timing of high-rate periods with symptoms, posture, activity, hydration, and sleep. If the pattern still looks noisy, change one criterion only and observe the result before making another adjustment.
Monthly
Export a clean PDF before a medical appointment and archive it somewhere you control. Confirm that the report identifies the date range and detection criterion, then redact your name, address, or other information that isn't clinically necessary before sharing it.
Keep privacy in the workflow. Cardiogram's stated approach is on-device analysis with read-only Apple Health access and iCloud-based syncing, so your monitoring data remains within your device and account ecosystem unless you choose to export or share it. Whatever system you use, don't alter readings to make the report look cleaner, and don't treat fewer alerts as success unless the remaining episodes are more actionable.
The routine is simple: filter known exercise context, tune one criterion, log symptoms, stabilize the sensor, review patterns, and export the original record. Leave the settings alone when the data is already interpretable. Constant optimization can create more confusion than the original noise.
Cardiogram offers automatic tachycardia episode detection, workout and recovery exclusion, symptom and trigger logging, weekly trends, and clinician-ready PDF reports built around your Apple Watch data. Visit Cardiogram to turn a flooded heart-rate feed into a structured record you can review and discuss with your clinician.
