Elopement & Alerts · Investigation
An autistic child was missing for two days before an alert went out
Parker Wells was 11, autistic and non-speaking. He left a Calgary day home late on a Thursday morning in July 2026 and was recorded on CCTV kilometres away roughly an hour later. The emergency alert reached phones about fifty hours after he was last seen. His case never met Alberta’s AMBER Alert criteria — because no one had taken him. Half of autistic children leave safe places. Water is what kills them. And the alert systems built to find missing children were designed around a crime that, in these cases, has not happened.
Direct answer
Government of Alberta · Verified 2026-07-31
Editor’s note: End The Wait Ontario covers Ontario. Parker Wells died in Calgary. We are writing about it because elopement is not a provincial risk and the alert rules are: Canada’s emergency alerting runs on shared federal infrastructure, but the criteria for using it are set province by province. Ontario wrote its own vulnerable-person alert into a bill, took it to committee, and let it die in 2025. What failed in Alberta is the same gap still open here.
The evidence base
49%
of autistic children had attempted to elope after age 4
Anderson et al., Pediatrics, 2012 · Verified 2026-07-31
~160×
the general-population drowning death rate for autistic children
Guan & Li, Am. J. Public Health, 2017 · Verified 2026-07-31
73.9%
of reviewed fatal drownings of autistic children began as wandering
Injury Epidemiology, 2017 (23 US cases, 2000–2017) · Verified 2026-07-31
74%
of elopement incidents began at a home — the child’s own or someone else’s
Anderson et al., Pediatrics, 2012 · Verified 2026-07-31
What happened in Calgary
The facts below are drawn from Calgary Police Service statements and contemporaneous reporting. They are set out plainly because the argument that follows depends on the clock, and the clock is not in dispute.
Thu 16 Jul, 11:08–11:41 a.m.
Parker Wells, 11, autistic and non-speaking, is last seen at his day home on Connaught Drive N.W. in Calgary.
CP24 / CTV News, July 28, 2026 · Verified 2026-07-31
Thu 16 Jul, around noon
Articles of clothing believed to be Parker’s are found on Northmount Drive N.W.
CP24 / CTV News, July 28, 2026 · Verified 2026-07-31
Thu 16 Jul, 12:52 p.m.
CCTV captures Parker on Travis Crescent N.E. — across the city from where he started, roughly an hour after he left.
CP24 / CTV News, July 28, 2026 · Verified 2026-07-31
Sat 18 Jul, 1:35 p.m.
An Alberta Emergency Alert is issued — about fifty hours after Parker was last seen. The case does not meet Alberta’s legislated AMBER Alert criteria, so the province grants a one-off exemption. The alert describes him as 4 feet 1 inch tall.
Government of Alberta, July 18, 2026 · Verified 2026-07-31
Wed 22 Jul, 10:55 a.m.
The alert is updated and re-issued province-wide. The description is now 4 feet 11 inches — ten inches taller than the first broadcast.
Government of Alberta, July 18, 2026 · Verified 2026-07-31
Tue 28 Jul
After more than 8,600 km searched on foot, a member of the public reports evidence matching the clothing Parker was last known to be wearing.
CP24 / CTV News, July 28, 2026 · Verified 2026-07-31
Wed 29 Jul
A body is recovered from a small pipe on the east side of Deerfoot Trail. Police believe Parker entered a culvert and travelled through a series of pipes.
CBC News, July 29, 2026 · Verified 2026-07-31
Thu 30 Jul
The Office of the Chief Medical Examiner formally identifies Parker. Police say the death is not believed to be criminal.
CBC News, July 29, 2026 · Verified 2026-07-31
Read the timeline again and notice where the alert falls. Parker was already on camera on the far side of a major roadway roughly one hour after he was last seen. The broadcast that asked an entire city to look for him arrived roughly forty-seven hours after that.
Nothing here establishes that an earlier alert would have changed the outcome. No one can honestly claim that, and this page does not. What the timeline establishes is narrower and harder to argue with: the tool that mobilises a city was withheld for two days, not because anyone judged Parker to be safe, but because the legislation defining when it may be used was written around abduction — and abduction was never what happened to him.
The alert gap: built for a crime that did not occur
AMBER Alerts are abduction tools. Every Canadian province operates its own criteria, and across them the common requirement is a reasonable belief that a child has been taken. The system is extraordinarily good at what it was designed for. It was simply never designed for this.
Calgary police were explicit about the mechanics. The case did not meet Alberta’s legislated AMBER Alert criteria; the Government of Alberta granted an exemption permitting the Calgary Police Service to use the Alberta Emergency Alert system given the circumstances of the investigation. The infrastructure was there the whole time. The wireless broadcast network, the provincial alerting authority, the cell towers, the emergency tone on every phone in the city — all of it existed on the Thursday morning Parker walked out.
The Ontario Autism Coalition put it in one line in its statement on Parker’s disappearance: “The gap is not technology. The gap is policy.” An exemption granted case-by-case, at provincial discretion, after someone senior enough decides the circumstances are unique, is not a safety system. It is a favour. And favours take time that a child in a storm drain does not have.
What improvising a category costs
One detail belongs in the record, carefully. The alert issued at 1:35 p.m. on 18 July described Parker as 4 feet 1 inch tall. The updated alert issued at 10:55 a.m. on 22 July described him as 4 feet 11 inches. Both went out province-wide, and both sit on the public Alberta Emergency Alert record.
This is not a scandal, and it is not a criticism of the officers involved. Those messages were assembled at speed, under an exemption that had to be requested before anything could be sent at all, by people trying to save a child’s life. That is precisely the point. A ten-inch discrepancy in a description broadcast to millions is what standardised, pre-written vulnerable-person alert templates exist to prevent. When the category itself has to be invented in an afternoon, its contents get invented too. A province that had built this tool in advance would have had a form to fill in.
Elopement is not misbehaviour. It is a predictable feature of autism
The largest study of the behaviour remains Connie Anderson and colleagues’ 2012 paper in Pediatrics, drawn from 1,218 families in the Interactive Autism Network registry. Its headline finding is the one every first responder should know: 49% of autistic children had attempted to elope at least once after age 4 — roughly four times the rate reported for their non-autistic siblings.
The consequences in that same sample were not hypothetical. Of the children who eloped, 53% went missing long enough to cause serious concern. Among those, 65% of incidents involved a close call with traffic and 24% a close call with drowning. Nearly two-thirds of families — 62% — said fear of wandering had stopped them from attending activities outside the home.
A second figure belongs beside that one, and it is lower. Kiely and colleagues, writing in PLOS ONE in 2016, used the CDC’s nationally representative Pathways survey of 4,032 families and found that roughly a quarter of children with developmental disabilities had eloped in the preceding twelve months, rising to about 35% among children with a dual diagnosis of developmental disability and autism. The two studies are often presented as contradicting each other. They do not. Anderson asked whether a child had ever eloped since age four; Pathways asked about the past twelve months. A shorter recall window returns a smaller number. Both describe a behaviour common enough that a public system should expect it rather than be surprised by it.
| Study | Measure | Finding | Source and date |
|---|---|---|---|
| Anderson et al. | Anderson asked whether a child had ever eloped since age four | 49% of autistic children had attempted to elope at least once after age 4 | Pediatrics (American Academy of Pediatrics), October 2012 · Verified 2026-07-31 |
| Kiely et al. | Kiely et al. asked about the preceding twelve months | Roughly a quarter of children with developmental disabilities had eloped in that window, rising to about 35% for children with a dual diagnosis of developmental disability and autism. | PLOS ONE, 11(2), e0148337, February 2016 · Verified 2026-07-31 |
Where it happens — and why blaming the day home is wrong
The reflex after a death like Parker’s is to ask who took their eyes off him. The research answers that question in a way that should retire it. In the Anderson sample, the most common place a child eloped from was not a park, a shop or an outing. It was a home — the child’s own or someone else’s — reported in 74% of cases. Shops accounted for 40% and classrooms or schools 29%.
Elopement happens most often in exactly the settings we treat as safe, under exactly the supervision we consider adequate, because the behaviour takes seconds and gives no warning. That finding is not a defence of any particular provider, and nothing here is a finding about the day home Parker left; the investigation concluded his death was not criminal. It is a statement about where the remedy lives. If the most common site of elopement is a home, then vigilance alone was never going to close the gap, and a response system that only activates when someone is at fault will keep arriving late.
Why a child leaves
Behaviour-analytic and clinical literature converges on two broad patterns, and they call for opposite responses:
- Goal-directed elopement. The child is moving toward something: water, a train line, traffic lights, a fire station, a favourite shop, a relative’s house, a specific sensory experience. Fixated interests are powerful predictors of destination, which is precisely why they belong in the alert itself.
- Escape-driven elopement. The child is moving away from something intolerable: noise, fluorescent light, crowding, transitions, a demand they cannot meet. A child leaving under this pressure will often keep moving until the input stops, and will seek somewhere enclosed, dark and quiet — which is exactly what a culvert is.
Neither pattern is defiance, and framing it as a discipline problem actively costs lives, because it delays the call. A caregiver who has been told their child is “running off to get attention” hesitates before dialling 911. A caregiver who understands that half of autistic children do this, and that water is the first place to look, does not.
Why elopement kills: water, silence, and enclosed spaces
The mortality signal is unusually high for a paediatric safety issue. Guan and Li, analysing US death-certificate data in the American Journal of Public Health in 2017, found that people with autism died from injury at far higher rates than the general population, and that the excess was concentrated in drowning — with autistic children at roughly 160 times the general-population drowning death rate. A companion review of 23 fatal drownings of autistic children reported in US media between 2000 and 2017 found the victims averaged 7.7 years old, that ponds accounted for 52.2% of locations, and that wandering was the activity that led to the drowning in 73.9% of cases.
Three factors compound, and each one inverts ordinary search intuition:
- Water is an attractor, not a deterrent. For many autistic children, moving water is among the most regulating sensory experiences available. The feature an ordinary risk assessment treats as a hazard to be avoided is the feature that draws the child in.
- Calling the child’s name may not work — and may backfire. A non-speaking child, or one who does not orient to their name, will not respond to searchers. Some will deliberately stay hidden from strangers, from shouting, from flashlights and sirens. Search teams trained on this use recorded family voices, familiar music and preferred objects rather than volume. Calgary’s searchers played Parker’s favourite Disney songs.
- Enclosed spaces read as safety. Drainage systems, culverts, pipes, crawl spaces and cupboards offer dark, quiet, tightly bounded environments. Search literature on lost-person behaviour has long noted that linear features — streams, ditches, drainage channels — funnel and hold subjects. Police believe Parker entered a culvert and travelled through a series of small pipes beneath the city.
Why the first hour is the entire argument
Search and rescue works on a simple, brutal geometry. The area that must be searched grows with roughly the square of the time elapsed, because the missing person can travel outward in any direction.
Double the delay and you roughly quadruple the ground.
This is why SAR doctrine speaks of golden hours, and why the National Center for Missing & Exploited Children and paediatric-search guidance both stress that the earliest hours carry most of the survival probability.
Parker demonstrates the geometry exactly. He was last seen shortly after 11 a.m. and was on camera at 12:52 p.m. in a different quadrant of the city. Within one hour, a containment ring drawn around his day home was already obsolete. Within two days — the interval before the alert — the searchable area had expanded past what hundreds of officers and volunteers could cover, and they went on to search more than 8,600 kilometres on foot.
An emergency alert is not primarily a request for information. It is the fastest available way to convert an entire population into observers while the containment ring is still small. Issued at hour one, it reaches the driver who is about to pass the child, the resident who is about to see him near a culvert, the shift worker walking to a bus stop. Issued at hour fifty, it reaches people who are being asked to remember something they saw two days ago.
Ontario has already had this case, and shelved the answer
This is not a Calgary problem, and Ontario cannot treat it as somebody else’s tragedy. On 12 June 2022, Draven Graham — an autistic boy who did not answer to his name and had a sensory aversion to touch — wandered from his home in Lindsay. His body was recovered from the Scugog River the following day.
What followed was the ordinary Ontario sequence. A petition. A vigil. A private member’s bill: Bill 74, brought by Hamilton Mountain MPP Monique Taylor, which would have amended the Missing Persons Act, 2018 to create a Vulnerable Persons Alert — an AMBER-style broadcast for missing people who cannot keep themselves safe, whether or not a crime is suspected. It was shaped by Draven’s case and by that of Shirley Love, a senior who died after leaving her home underdressed in winter. Families testified. The bill went to committee. It was shelved when the 2025 election was called, and it has not returned.
Four years after Draven, Ontario still has no vulnerable-person alert. When Parker Wells went missing in July 2026, the Ontario Autism Coalition — whose advocacy on this began in 2022 — called on Senior Officials Responsible for Emergency Management, the federal-provincial-territorial body governing Canada’s alerting system, to build a disability-informed alert for vulnerable missing persons.
That gap sits inside a wider pattern this site documents daily: 71,263 children are registered in the Ontario Autism Program without a funding agreement. The same province that cannot fund the therapy also has not built the alert. Both are choices about how much a delay is allowed to cost.
What a working alert actually looks like
“Expand AMBER” is not a specification, and vagueness is what lets a proposal die in committee. Below is a concrete one, drawn from what the Parker Wells and Draven Graham cases exposed, from the elopement research, and from what search teams say they need on arrival.
01
Trigger on vulnerability, not on a suspect
The alert fires when a person who cannot reliably seek help, give their name, or recognise danger is missing — regardless of whether anyone took them. Abduction is one reason a child is at risk. It is not the only one, and for autistic children it is the rarest one.
02
Minutes, not days — a hard clock
A defined maximum interval between the report and the broadcast, measured and published. Parker’s alert came about fifty hours after he was last seen. By the time it reached phones, he had already been recorded on CCTV kilometres away, an hour after he left.
03
Police-initiated, not committee-approved
The officer holding the file issues it against published criteria. Alberta had to grant a special exemption before Calgary police could use the emergency alert system at all. A system that needs a cabinet-level favour to work is not a system.
04
Carry the search-relevant facts, not just a description
Height and jacket colour do not find a non-speaking child. What finds him is: he will not answer to his name; he may hide from you; check water, culverts, drainage and storm infrastructure first; do not shout, do not chase, do not approach with sirens.
05
Geofence tightly, then widen
Cell-broadcast alerts can be bounded geographically. A tight initial radius around the point last seen, widening on a published schedule as the plausible travel distance grows, targets the people who can actually look — and is the direct answer to alert fatigue.
06
Opt-in profiles held before the emergency
A voluntary, consent-based record — sensory triggers, communication method, fixations, whether the child is drawn to water, a current photo — attached to the address and released to searchers the moment a file opens. Assembling this during the search costs the hours that matter most.
07
One standard across every province and territory
A child does not become less findable at a provincial boundary. The Ontario Autism Coalition has asked Senior Officials Responsible for Emergency Management (SOREM) — the federal-provincial-territorial body that governs the alert infrastructure — to build a single disability-informed standard.
08
Publish the outcomes
Every issuance logged: time missing before broadcast, time to location, outcome. Without published data, no one can tell whether the system works, and the next review starts from anecdote again.
The objections, answered
“People will mute alerts. Alert fatigue is real.”
It is real, and it is an argument for design, not for absence. Fatigue comes from broadcasts that are untargeted, unactionable and repetitive. A vulnerable-person alert is the opposite on all three counts: geofenced to a defined radius around the point last seen, carrying a specific instruction (check your pool, your ditch, your storm drain, your shed), and rare — vulnerable-person disappearances that reach emergency threshold are uncommon events. The honest version of this objection is that we have decided some silence is worth more than some children, and that should be said out loud if it is going to drive the policy.
“Broadcasting a disability is a privacy harm.”
It is a genuine cost, and it is why the profile registry must be opt-in, consent-based and revocable, and why disclosure at alert time should be limited to what materially changes the search. But note what the current arrangement does instead: it withholds the alert entirely, and the child’s name, photograph, diagnosis and death are published anyway — in the news coverage that follows. The choice is not between privacy and disclosure. It is between disclosure that might help and disclosure that comes with an obituary.
“Most of these children are found quickly.”
True, and it is not a defence. Most fires are extinguished without a fatality; we still fund fire departments against the ones that are not. The relevant number is not how often elopement ends safely but what happens when it does not — and there the evidence is unambiguous: drowning, at roughly 160 times the general-population rate, overwhelmingly beginning as wandering.
“Police already have discretion.”
Discretion is what produced a two-day delay. Alberta had to grant a formal exemption before Calgary police could use the emergency alert system for a non-abduction case. Discretion exercised upward through a provincial approval chain is not speed; it is a queue. Published criteria that an incident commander can apply at hour one are.
What Ontario should do
Reintroduce and pass a Vulnerable Persons Alert
Bill 74’s substance already exists and has already been through committee. Amend the Missing Persons Act, 2018 and be done with it.
Set a published time standard
A maximum interval from report to broadcast, reported publicly each year alongside outcomes.
Fund an opt-in vulnerable-person profile registry
Held by police services, with consent, revocation and audit built in from the start.
Make autism-specific search doctrine mandatory training
For first-response and SAR units: water first, do not shout, do not chase, expect hiding.
Take it to SOREM
Ontario should carry a national disability-informed alert standard to the federal-provincial-territorial table, so that the answer does not depend on which province a child goes missing in.
What families, schools and day programs can do now
None of this replaces the policy fix, and none of it should be read as putting the burden back on parents. It is what is available while Ontario decides.
- Call 911 immediately
- Say “autistic” and “non-speaking” if they apply. There is no waiting period. Ask directly whether an emergency alert can be issued, and ask them to record the answer.
- Send someone to water in the first minute
- Pool, pond, creek, ditch, storm drain, culvert — before searching anywhere else.
- Search the home again, including enclosed spaces
- Under beds, inside cupboards, behind and inside furniture, in vehicles.
- Prepare a one-page profile before you need it
- Current photo, height, communication method, sensory triggers, fixations, whether the child is drawn to water, what sounds or objects will draw them out, and what will make them run. Give copies to the school and day program.
- Ask every setting who supervises your child what their elopement protocol is
- Specifically, how many minutes pass before they call 911, and who makes that call.
- Consider swim and water-safety instruction
- Adapted for autistic learners, and locking or alarm hardware on exit doors, as layered measures rather than a single point of failure.
More detail on each of these: our Ontario elopement-prevention guide for home measures, and school and day-program elopement safety for what to put in a safety plan.
The gap is not technology
Every piece of machinery needed to put Parker Wells’ face on every phone in Calgary existed on the morning he walked out of his day home. What did not exist was a rule saying that a child who cannot ask for help is reason enough to use it.
Ontario had that rule drafted, debated and sitting in committee. It let it lapse. Writing it back into law is a smaller task than the search that follows when we do not.
Questions people are asking
What is elopement in autism?
Elopement — also called wandering or bolting — is when an autistic person leaves a supervised or safe setting without their caregiver’s knowledge. It is not disobedience or a discipline problem. It is usually either goal-directed (moving toward a place, object or sensory experience the person wants, such as water, trains or traffic) or escape-driven (moving away from noise, crowding, lights or demands that have become intolerable). In the largest study of it, published in Pediatrics in 2012, 49% of parents of autistic children reported at least one elopement attempt after age 4.
Why is elopement so dangerous for autistic children?
Because of where autistic children go and how they respond once they are there. Drowning is the leading cause of death associated with autism elopement: Guan and Li, writing in the American Journal of Public Health in 2017, found autistic children were roughly 160 times more likely to die by drowning than children in the general population. In a review of 23 fatal drownings of children with autism reported in US media between 2000 and 2017, wandering was the activity that led to the drowning in 73.9% of cases. A non-speaking child will often not answer searchers calling their name, and may actively hide from the people looking for them.
Was the day home at fault for letting Parker leave?
Nothing in the public record supports that conclusion, and Calgary police said the death was not believed to be criminal. The research points the other way as well: in the largest study of autism elopement, the most common place a child eloped from was a home — the child’s own or someone else’s — in 74% of cases, ahead of shops at 40% and schools at 29%. Elopement most often happens in the settings adults consider safest, because the behaviour takes seconds and gives no warning. That is the argument for a systemic alert response rather than a search for someone to blame: if vigilance alone could close this gap, it would already have closed it.
How common is autism elopement — 49% or 27%?
Both figures are real and they measure different things. Anderson et al. (Pediatrics, 2012) surveyed 1,218 families and asked whether a child had ever eloped since age four; 49% had. Kiely et al. (PLOS ONE, 2016) used the CDC’s nationally representative Pathways survey of 4,032 families and asked about the preceding twelve months; roughly a quarter of children with developmental disabilities had eloped in that window, rising to about 35% for children with a dual diagnosis of developmental disability and autism. A shorter recall window returns a smaller number. Neither study supports treating elopement as rare.
Why was no AMBER Alert issued for Parker Wells right away?
Because AMBER Alerts are built for abductions. Alberta’s legislated criteria require, among other things, reason to believe a child has been abducted. Calgary police did not believe Parker had been taken, so his case did not meet the threshold. The Government of Alberta granted a specific exemption allowing Calgary Police Service to use the Alberta Emergency Alert system given the circumstances — and that alert went out roughly two days after he was last seen.
What is an "indigo alert"?
Indigo alert is the name proposed by Canadian families and advocates for an AMBER-style emergency alert covering missing neurodivergent and vulnerable people, where abduction is not suspected. The name comes from the deep purple associated with autism awareness. It is a proposal, not an existing program: as of July 2026 no Canadian province operates a dedicated vulnerable-person alert of this kind.
Does Ontario have a vulnerable person alert?
No. Bill 74, a private member’s bill from Hamilton Mountain MPP Monique Taylor, would have amended the Missing Persons Act, 2018 to create a Vulnerable Persons Alert. It was brought forward after the deaths of Draven Graham, an autistic boy who drowned in the Scugog River in June 2022 after wandering from his home in Lindsay, and Shirley Love, a senior who died after leaving her home in winter that same year. The bill reached committee and was shelved when the 2025 provincial election was called. Ontario families have been asking for this tool since 2022.
What should I do in the first ten minutes if an autistic child is missing?
Call 911 immediately and say the words "autistic" and "non-speaking" if they apply — do not wait, and do not let anyone tell you to wait 24 hours. Send someone to the nearest water first: pond, creek, pool, ditch, storm drain, culvert. Search your own home again, including under beds, inside cupboards and behind furniture, because hiding in enclosed spaces is common. Tell responders the child may not answer to their name, may run from searchers, and what sounds or interests will draw them out.
The gap is not technology
Evidence for this guide
Parker Wells’ name and the circumstances of his death are used here as reported publicly by Calgary Police Service and Canadian news media. No photograph of Parker is published on this page: the images in circulation were released by his family for the search, and End The Wait Ontario holds no permission to republish them.