Connected Health and Wearables · Study deck

Elderly Care IoT: Monitoring, Economics, and Autonomy

A fall-alert service can meet its response budget, but daily care depends on slower patterns as well as sudden events.

Ada is your guide for this deck.

caseselderly

Ada, the module guide, in a scene from this chapter.

After studying this chapter

Learning objectives

You will be able to:

Explain: A phone or a dedicated reader picks up the signal and shows a plain status with a room number attached, which is what turns a reading into an assignable task.
Explain: The proposed monitoring pattern runs passively in the background: motion sensors record changes in movement, audio sensors track changes in speech, and a camera measures how long a task takes.
Explain: A moving average can still suit small random jitter on a slow signal; the median is the stronger first defense when a sensor occasionally produces a lone extreme reading.
interpret behaviour baselines and prevention economics

Major section

Elderly Monitoring Architecture

Above them the same chain is drawn concretely.

A thin disposable tag sits on the outside of a garment, so nothing invasive touches the resident.
A phone or a dedicated reader picks up the signal and shows a plain status with a room number attached, which is what turns a reading into an assignable task.

Elderly care monitoring workflow with a disposable moisture sensor tag, wireless transmission to a mobile reader, an app alert, and a caregiver response.

Major section

Elderly Monitoring Architecture (continued)

The claim here is dignity as much as data.

Every other sensor in this section repeats those five steps, and each one adds alerts a caregiver must judge.
Clinical value: detecting unusual inactivity and behavior changes.
Data collected: room temperature, water-use patterns, and light usage.

Major section

Sensor Patterns to Clinical Insight

That filter rejects one impossible spike without spreading it into nearby estimates.

A moving average can still suit small random jitter on a slow signal; the median is the stronger first defense when a sensor occasionally produces a lone extreme reading.
Reduced kitchen activity: Sensor pattern shows 50% fewer fridge opens and 70% less cooking time.

The same three sensor samples produce different outputs: averaging spreads an isolated spike, while the median rejects it.

Major section

Passive Progression Monitoring Between Assessments

A Parkinson's disease example in the source material makes the difference between an event alert and a progression record explicit.

Symptoms can fluctuate continuously, so a monthly clinical check may not represent what happens between visits.
The proposed monitoring pattern runs passively in the background: motion sensors record changes in movement, audio sensors track changes in speech, and a camera measures how long a task takes.
For an operating record, keep the three evidence streams distinct.

Major section

Compelling Statistics

Predictive Analytics: Gait analysis can predict fall risk days in advance.

Independence: Seniors can age in place rather than moving to care facilities.
Individual or family: Value received is peace of mind, slowing decline, and aging in place.
Health insurance: Value received is fewer hospitalizations and lower long-term care cost.

Major section

Privacy and Autonomy Considerations

Elderly IoT systems must balance safety monitoring with dignity and independence.

Strip or blur identity before data leaves the home, and keep it encrypted at rest and in flight.
The last panel closes the loop with a deletion right that must reach every downstream service.

Privacy flow for wearable data showing local identity stripping or blurring before cloud sharing, with controls for scope, revocable consent, encryption, and downstream deletion.

Deck summary

Key takeaways

Above them the same chain is drawn concretely.

The claim here is dignity as much as data.
That filter rejects one impossible spike without spreading it into nearby estimates.
A Parkinson's disease example in the source material makes the difference between an event alert and a progression record explicit.
Predictive Analytics: Gait analysis can predict fall risk days in advance.

Retrieval practice

Recall check

Ada says: answer from memory, then check your reasoning.

Q1A fall detection wearable uses a 3-axis accelerometer, gyroscope, and heart rate monitor. During testing, the system correctly detects 95 out of 100 actual falls (sensitivity = 95%). However, it also triggers 50 false alarms out of 1,000 non-fall events (specificity = 95%). If the device is deployed across 500 elderly patients who collectively experience an average of 2 real falls per day and 10,000 non-fall events per day, approximately how many total alerts (true + false) will caregivers receive daily?

AAbout two daily alerts, because the true falls represent all alerts that reach the care team.

BAbout 500 daily alerts, most caused by false positives in routine activity.

CAbout 50 daily alerts, undercounting the test's false-positive volume.

DAbout 100 daily alerts, counting fewer routine events than the test implies.

Show answer

Answer: B see answers page

Q2An elderly-care team needs remote anomaly alerts and daily digests from motion sensors in 12 rooms of a patient’s home. Under the applicable privacy and minimum-necessary rules, which processing architecture best fits that care purpose?

AStream every raw room trace to cloud, because the digest requirement is satisfied by storing the underlying traces.

BEncrypt and upload all raw data, because encryption makes unrestricted collection private.

CProcess locally; send anomaly alerts and daily summaries to cloud only.

DProcess only locally, preventing the required remote alerts and digests.

Show answer

Answer: C Edge processing converts raw motion events into clinical-level summaries (activity score, sleep quality, routine adherence) without exposing intimate behavioral details.

Q3Under an illustrative scenario where each listed recurring service is furnished and billable in all 12 months (99453: $19 one-time setup, 99454: $62/month device, 99457: $51/month communication, 99458: $41/month additional), what is the first-year revenue per patient, and how does it compare with $40,000?

A$1,867 in year one per patient; a $40,000 event is about 21 patient-years.

B$744 in year one, omitting listed recurring services from the scenario.

C$2,076 in year one, treating listed charges as a different monthly bundle.

DNo covered revenue, because remote monitoring services fall outside reimbursement categories.

Show answer

Answer: A The recurring monthly components total $62 + $51 + $41 = $154.

Q4What gives a fall-alert user a cancellation window while preserving retry and escalation state?

AWearable and Ambient Fall Evidence

BCancel Window and Reliable Alert Path

CCaregiver Escalation and Response

Show answer

Answer: B The middle path offers a clear self-cancel window, preserves acknowledgement and retry state, and escalates only when evidence remains credible and unresolved.

Q5Complete the elderly fall-alert escalation gate:

Aif event['confidence'] < 0.85: return False

Bif event['confidence'] > 0.85: return False

Cif event['seconds_since_detection'] < 0: return True

Dif event['caregiver_acknowledged']: return True

Show answer

Answer: A Fall-alert escalation should respect wearer self-cancel, device state, posture evidence, confidence, and the 30-second cancel window before notifying caregivers.

Q6An older adult starts rising much later than their usual pattern. What should a monitoring service do?

ADiagnose a fall from late rising alone without reviewing other causes.

BIgnore the change because a separate fall sensor did not fire.

COffer a wellness check and review possible causes with the care team.

DCall emergency services after every late morning without triage.

Show answer

Answer: C A change from the person’s own routine can prompt a human check, not a diagnosis.

Print reference

Answers 1 of 2

Answer key.

B · True positives: 95% of 2 falls = 1.9 alerts. False positives: 5% of 10,000 non-fall events = 500 false alerts. Total = ~502 alerts/day. This is why alert fatigue is the #1 operational challenge in deployed fall detection systems. Even 95% specificity produces overwhelming false alarm volumes at scale.
C · Edge processing converts raw motion events into clinical-level summaries (activity score, sleep quality, routine adherence) without exposing intimate behavioral details.
A · The recurring monthly components total $62 + $51 + $41 = $154.

Print reference

Answers 2 of 2

Answer key.

B · The middle path offers a clear self-cancel window, preserves acknowledgement and retry state, and escalates only when evidence remains credible and unresolved.
A · Fall-alert escalation should respect wearer self-cancel, device state, posture evidence, confidence, and the 30-second cancel window before notifying caregivers.
C · A change from the person’s own routine can prompt a human check, not a diagnosis.