Set Up Heat Safety Automations for Elderly, Infants, and Pets

During extreme heat, the useful question is not whether the forecast looks ugly. It is whether the room where the older adult is sitting, the baby is sleeping, or the dog is crated has crossed a line that should make something happen.

That is where smart home devices for extreme heat safety are most valuable: not as a clever dashboard, but as a room-level early warning layer. A temperature and humidity sensor watches the actual space. Presence detection keeps the automation from cooling an empty room or missing an occupied one. A smart plug, smart thermostat, or IR controller can start a fan or AC when it is safe to do so. Alerts escalate from yellow to red so the person receiving them knows whether to check in soon or act now.

The strongest evidence for this kind of setup comes from older-adult heat monitoring, not from infant nurseries or pet rooms. In the Ethos proof-of-concept study, 78 older adults in Queensland, Australia used a digital heat early warning system during the 2023-24 summer; participants acted on 90% of yellow alerts and 94% of red alerts, heat preparedness increased 3.4x, and the system scored 78 on the System Usability Scale.[1] That matters because an alert that nobody understands, trusts, or acts on is just another notification.

The same study also keeps the recipe honest. It was built for adults 65 and older, in one subtropical Australian setting, with a modest sample. It does not establish baby-safe or pet-safe temperature thresholds. For infants and pets, the automation pattern can still be useful, but the thresholds need to be set conservatively from general pediatric or veterinary guidance, household knowledge, and professional advice when risk is high.

The Heat Safety Flow

Build the automation around states, not gadgets. The exact platform can be Home Assistant, Apple Home, Google Home, Alexa, or another hub. The logic is the part that should survive a device swap.

The fan cutoff is not optional decoration. WHO guidance says fans can help below 40°C / 104°F but can heat the body above that threshold, so a smart fan rule needs a hard stop rather than a simple “hotter room equals more fan” curve.[2]

Why the Room Matters More Than the Weather App

Outdoor warnings are still useful, but they are too blunt for vulnerable-room automation. A shaded downstairs bedroom, a west-facing nursery, and a pet room with the door closed can behave like three different houses. The automation should care about the room where the person or animal actually is.

That focus is especially important for older adults. The Ethos literature review notes that older adults spend up to 90% of their time indoors and that most heat fatalities occur in the home.[1] WHO also reports that heat-related mortality among people older than 65 increased by about 85% between 2000-2004 and 2017-2021, with an estimated 489,000 heat-related deaths annually worldwide.[2] For a household automation recipe, those facts point to one practical decision: measure the occupied room, not just the ZIP code.

If you already use NestGrid-style presence automations for elder care, the same basic idea applies here: occupancy has to be confirmed before the system decides what state the room is in. A motion sensor may miss someone asleep in a chair. A door sensor may know the nursery is closed but not whether the baby is in the crib. mmWave or bed/chair presence sensors can help in some rooms, and the elder-care presence approach in smart home elderly fall detection is relevant because heat safety depends on knowing whether someone is actually there.

Devices to Use, Without Turning This Into a Shopping Project

You need four device roles. Brand matters less than whether each role is reliable in the vulnerable room.

• Temperature and humidity sensor: place it near the occupant’s real breathing zone, away from direct sun, AC vents, radiators, humidifiers, and electronics that run hot.
• Presence signal: use room motion, mmWave, bed or chair occupancy, crib routine, door state, camera-based pet detection, or a manual “occupied” toggle when automation-grade sensing is not trustworthy.
• Cooling control: use a smart thermostat, IR controller for a mini-split or portable AC, smart plug for approved fans, or a smart relay only when the appliance is safe to automate.
• Alert channel: send yellow and red alerts to the person who can act, not just to the person who owns the hub.

Do not put a fan, portable AC, or smart plug into an automation until the appliance itself is safe for unattended use. A plug can remove power; it cannot make a cheap fan mechanically safe, keep a cord away from a crib, or guarantee that a portable AC drains correctly.

Copyable Platform-Agnostic Logic

Use variables instead of hard-coded universal temperatures. The Ethos study supports the older-adult alert model, but it does not provide one universal yellow or red temperature to copy into every home. For infants and pets, the numbers should be chosen even more carefully because this evidence comes from general care guidance, not a smart-home trial.

INPUTS
  room_temperature
  room_humidity
  room_occupied
  occupant_type = elderly | infant | pet
  cooling_available = ac | fan | both | none
  fan_safe_for_room = true | false
  caregiver_primary
  caregiver_backup

THRESHOLDS
  yellow_temp = occupant_specific_yellow_temperature
  red_temp = occupant_specific_red_temperature
  red_duration = occupant_specific_time_limit
  fan_cutoff = 104°F / 40°C

NORMAL MONITORING
  IF room_occupied is false:
    keep monitoring
    do not start cooling unless this room must be pre-cooled by schedule

YELLOW STATE
  IF room_occupied is true
  AND room_temperature >= yellow_temp
  AND room_temperature < red_temp:
    send yellow alert to caregiver_primary
    include room name, current temperature, humidity, and occupant type
    recommend immediate low-risk actions
    IF AC is available:
      start AC or lower setpoint within the household's safe range
    ELSE IF fan is available
      AND fan_safe_for_room is true
      AND room_temperature < fan_cutoff:
        start fan

RED STATE
  IF room_occupied is true
  AND (room_temperature >= red_temp OR yellow state lasts longer than red_duration):
    send red alert to caregiver_primary
    send red alert to caregiver_backup
    IF AC is available:
      start AC or set stronger cooling mode
    IF fan is running
      AND room_temperature >= fan_cutoff:
        turn fan off
    require human check-in or confirmed relocation

FAN CUTOFF
  IF room_temperature >= fan_cutoff:
    prevent fan-only cooling
    turn off automated fan rules
    send alert that fan-only cooling is unsafe at this room temperature

In Home Assistant, this becomes helpers, template sensors, automations, and notify actions. In Apple Home, Google Home, or Alexa, it becomes routines with temperature triggers, presence conditions where available, smart plug or thermostat actions, and notification steps. If your platform cannot express “occupied and hot for more than a set duration,” keep the alerting logic in the platform that can, then let the voice assistant handle the visible notification or appliance action.

For more appliance-specific examples, NestGrid’s smart fan automation recipe and Home Assistant heat-wave YAML recipes are better places to copy platform syntax. This recipe is the safety logic those automations should respect.

Elderly Rooms: Use the Evidence-Backed Pattern

For older adults, the Ethos trial gives this recipe its best footing. Participants were not just shown data; they received yellow and red warnings and acted on them at high rates.[1] The most-used cooling strategies were drinking more water, turning on a fan, using a hand bath, turning on AC, and removing clothing.[1] Those are exactly the kinds of actions an alert should suggest because they are familiar, concrete, and possible without turning the home into a laboratory.

Privacy deserves a small but real note here. In the Ethos design, no participant data was sent via Wi-Fi; the transmitted sensor metadata was limited to sensor ID, room name, temperature, and humidity over WPA2/HTTPS.[1] A home recipe does not need to copy that architecture exactly, but it should copy the restraint. For heat alerts, the caregiver usually needs room, temperature, humidity, alert level, and acknowledgement status. They do not need a constant feed of private activity unless the care plan explicitly requires it.

If you are configuring heat alerts for someone living separately, pair this recipe with a remote-caregiver path such as heat alerts for elderly caregivers or the troubleshooting playbook for elderly heat dome automations. The alert is only useful if it reaches someone who can intervene.

Infant Rooms: Treat Thresholds as Conservative Household Rules

The infant version should borrow the room-monitoring pattern, not the older-adult evidence. The Ethos study does not tell us what nursery threshold is safe for a baby under 12 months. Set yellow and red values from pediatric guidance you already follow, your clinician’s advice when relevant, and the specific room’s behavior during heat events.

A nursery automation should be more cautious about unattended appliance control than an adult bedroom automation. A fan may be useful in some rooms, but the automation still has to respect cord placement, tip-over risk, airflow direction, noise, and whether the device is approved for use around a crib. If there is any doubt, make the yellow state alert-only and reserve automated cooling for a central AC, mini-split, or other appliance already considered safe for that nursery.

For a practical setup, place one sensor near the crib’s general sleep zone but out of reach, then compare it against the thermostat reading during a hot afternoon and an overnight period. Nurseries often surprise people because the room can lag behind the hallway thermostat or heat up after the rest of the house feels comfortable.

Pet Rooms: Occupancy and Appliance Safety Matter More Than Elegance

Pet automations need their own rules because a pet-safe automation is not just a comfort automation with a lower threshold. A dog in a crate, a cat in a sun-facing room, or an older pet confined while recovering from surgery cannot open a door, move to the basement, refill water, or tell you the room is wrong.

Again, the Ethos study does not establish pet thresholds. Treat pet yellow and red values as veterinarian-informed household rules, and set them more conservatively for short-nosed breeds, older animals, overweight pets, animals with heart or respiratory disease, very young animals, and pets confined to crates or small rooms.

The common failure is assuming that a fan automation is harmless because the pet likes the fan on normal days. During extreme heat, the automation needs to know whether the fan is helping, whether AC is available, and whether the room has crossed the fan cutoff. If the house has no safe automated cooling, the red state should be an urgent human-response alert, not a loop that keeps toggling a plug.

Set the Yellow and Red Thresholds

Because there are no universal numeric room thresholds for each occupant type, the safest copyable pattern is to define the thresholds deliberately instead of pretending one chart fits every home.

Do not make yellow and red too close together if that causes alert fatigue. Do not make them so far apart that yellow becomes decorative. A useful yellow alert gives a tired caregiver time to act before the room becomes urgent. A useful red alert is rare enough to be taken seriously and specific enough to say what happens next.

Humidity belongs in the notification even when your first version triggers only on temperature. It changes how the room feels and helps the caregiver judge whether a fan, AC, dehumidification, relocation, or direct check is the next sensible move. If your platform supports combined heat rules, test them carefully; if it does not, keep humidity visible in the alert rather than hiding it in a dashboard nobody opens at 2 a.m.

Escalation Rules That Someone Can Actually Follow

The automation should never end with “notification sent” if nobody is responsible for the next move. Yellow can go to the on-duty caregiver. Red should go to the on-duty caregiver and a backup. If the vulnerable person lives alone, the backup should be someone local enough to check the room or arrange help.

ALERT MESSAGE TEMPLATE

YELLOW:
  Heat safety yellow alert: {room_name} is {temperature} with {humidity} humidity.
  Occupant: {occupant_type}.
  Suggested next step: check room, start safe cooling, offer water if appropriate, or relocate if needed.

RED:
  Heat safety red alert: {room_name} is {temperature} with {humidity} humidity.
  Occupant: {occupant_type}. Occupancy is confirmed.
  Action required: direct check or confirmed relocation. AC command sent: {yes/no}. Fan disabled by cutoff: {yes/no}.

For older adults, include a gentle acknowledgement path when possible: “I am cooling the room,” “I am drinking water,” or “Please call me.” In Ethos, older adults acted on alerts at high rates, which is one reason this kind of system feels practical rather than speculative.[1] For infants and pets, acknowledgement comes from the caregiver, not the occupant.

Test It Before the Heat Event

Testing is where many smart-home safety automations either become useful or become theater. Run the recipe with temporary low thresholds on a normal day so you can see every state without waiting for dangerous heat.

• Trigger yellow and confirm the right person receives the alert with room name, temperature, humidity, and occupant type.
• Trigger red and confirm the backup contact receives it without depending on the primary caregiver’s phone being awake.
• Verify that AC starts when commanded and that the room temperature actually begins to fall.
• Verify that fan-only cooling is blocked at 104°F / 40°C and above.
• Turn off Wi-Fi or unplug the hub briefly and check what failure alert, if any, reaches the caregiver.
• Replace any “assumed occupied” rule that creates false confidence with a manual toggle or better sensor.

If you already use a broader heat playbook such as extreme heat safety automations or extreme heat watch vs warning automation, keep this vulnerable-room recipe separate. Weather-triggered pre-cooling is useful, but the room-level yellow/red system should still be able to fire when the forecast missed the problem or one room behaves badly.

What to Build Today

Pick the vulnerable rooms first: the recliner or bedroom used by an older adult, the nursery, the pet room, the crate area, or the room that always runs hotter than the thermostat claims. Put the sensor where the occupant spends time. Add an occupancy signal you trust. Define yellow and red thresholds separately for elderly, infant, and pet spaces. Add the fan cutoff. Send alerts to people who can act. Then test the whole path while the stakes are low.

A smart home is not a medical device and it cannot replace caregiver checks. It can, however, stop treating heat as a whole-house average and start watching the rooms where vulnerable people and animals actually are.

References

1. A digital heat early warning system for older adults, npj Digital Medicine, 2025.
2. Climate change: Heat and health, World Health Organization.