From Investigation to Reunification: Rethinking What Child Welfare Data Is Actually For

In 2019, child protection workers across Canada opened roughly 299,171 maltreatment investigations. Fewer than 4% of the substantiated ones involved physical harm to a child. Nine in ten investigations were about chronic family need: poverty, unstable housing, a caregiver stretched past their limit.

That gap should stop us. The system pours enormous effort into investigating families, but the thing it's actually trying to do, in the vast majority of cases, is keep those families safely together. So here's the question worth sitting with: if reunification is the goal, why is almost every case management system built to document the investigation instead?

The mission is relational. The data model is transactional.

Every provincial child welfare statute says roughly the same thing. Protect children, and preserve families wherever it's safe to do so. The Truth and Reconciliation Commission put it even more plainly in its first Calls to Action: keep families together, and resource communities to do it.

But the operational backbone of the sector, the case management system, was built for a different job. Most systems were designed to capture what regulators ask for: incidents logged, risk assessed, standards met, timelines hit. They're good at producing a compliance record. They're far less good at showing whether a family got stronger.

Many child welfare information systems were built primarily to support reporting and compliance, and they capture data well while doing little to support decisions in the moment. The result is a mismatch between what the system says it's for and what its data is actually shaped to measure.

We count investigations. We barely count reunifications.

Here's a telling contrast. In the United States, federal foster care data treats reunification as a headline number. A Child Trends analysis of that data shows nearly half of children leaving foster care each year are reunited with family, making it the single most common way children exit care.

Canada has no equivalent national figure. Ontario reports that about 68% of children who came into care in 2019-20 left within a year, usually reunited with family. But that's one province's snapshot, and researchers note there's almost no Canadian study examining what actually predicts a successful reunification.

This is the outcomes-versus-outputs problem in its sharpest form. Outputs get tracked exhaustively because the systems were built to report them. Outcomes, families safely back together, kinship ties preserved, stability that holds a year later, barely register.

The most striking illustration comes from Cindy Blackstock's early research, which found that support to families often falls away right after reunification, arguably the moment it matters most. Worse, the data couldn't confirm it, because there was no systematic way for workers to record what happened post-reunification. The system literally couldn't see the part of the story it exists to produce.

Most cases aren't about danger. They're about poverty.

If you only read the intake fields, child welfare looks like a system responding to abuse. The evidence says otherwise.

In the CIS-2019 data, 90% of investigations focused on chronic needs rather than urgent protection, and more than 98% of those chronic-needs cases involved no physical harm at all. Neglect, not abuse, is the single biggest reason children enter care. In British Columbia, neglect accounts for roughly 73% of court-ordered admissions.

And neglect is often a stand-in for poverty. The Ontario Association of Children's Aid Societies says it directly in its own guidance: poverty can be mistaken for neglect, which draws extra scrutiny onto low-income families. The Ontario Human Rights Commission's review, Under Suspicion, found risk criteria that quietly read poverty in racialized and Indigenous families as parental failure.

When your data is organized around risk flags and incidents, a family in poverty looks like a case to investigate. When it's organized around strengths and supports, the same family looks like one to help. Same family. Different lens. The system usually installs the first one by default.

The law already moved. The data hasn't caught up.

Nowhere is the case for a reunification-first design more urgent than for Indigenous families. In 2021, Indigenous children were 7.7% of kids under 15 but 53.8% of children in foster care, roughly 14 times more likely to be in care than non-Indigenous children.

The legal response is unambiguous about direction. Bill C-92, the federal act affirming Indigenous jurisdiction over child and family services, came into force in 2020 and requires the system to consider whether supports could keep a family together before a child is removed. In February 2024, the Supreme Court unanimously upheld the Act as constitutional in full.

So the mandate is now explicit: prevention, family, community, connection. But the systems most agencies run on still encode an older logic, one built to open, assess, and close investigations. Policy has turned toward reunification. The data model is still facing the other way.

What data built for reunification would actually capture

A system designed around the real goal would track different things. Not instead of safety information, but alongside it.

It would record family strengths and protective factors, not just risks. It would map kinship networks and note which relatives have been found and engaged. It would follow reunification as a live process with milestones, and it would keep watching after a child goes home, so a family that's struggling shows up before a crisis does.

This isn't abstract. A generation of practice models already works this way: Signs of Safety, Family Group Conferencing, family finding, and kinship care all put family relationships at the centre. The trouble is that these models are hard to evaluate, precisely because conventional systems don't capture the variables they're built to change.

We have direct Canadian evidence of the disconnect. When Trent University's Research for Social Change Lab studied Ontario's mandated data system, front-line workers described it as focused on compliance with standards rather than on service quality or outcomes, and said it made equity-focused work difficult to measure. That's the whole thesis in a single case study: the data backbone counts compliance, not family progress.

A caution: data can support families, or it can surveil them

Critics have argued that because the model treats past system involvement as a proxy for harm, it can bake in and amplify the very biases the system already carries, aiming more scrutiny at poor and racialized families. Other jurisdictions pulled back after similar tools underperformed or drew backlash.

The lesson for case management design is sharp. More data is not automatically more help. Data built to flag and predict tends to expand surveillance. Data built to support families has to measure something else entirely: connection, capacity, and progress toward home.

What this looks like when it works

The encouraging part is that Canadian communities are already building the reunification-first model, most of them under Bill C-92.

Cowessess First Nation in Saskatchewan passed its own child and family law in 2020 and created the Chief Red Bear Children's Lodge, framed by then-Chief Cadmus Delorme around prevention rather than intervention. By early 2022, the community reported no children in provincial ministry care on its reserve lands, with genealogy work actively reconnecting kids to relatives.

In Manitoba, Peguis First Nation became the first community in the province to take over child welfare under the Act, and its leadership has described transferring roughly 100 members back to the nation's care from a starting point of more than 250. Different governance model, same direction of travel: fewer children in care, more families supported.

None of these communities got there by collecting more risk data. They got there by reorganizing the work, and the information that supports it, around keeping families whole.