robertlathamesq.org

I wrote last month that if there wasn’t a big jump in the number of kids being removed by October (the typical high-point in Fall removals) then I was going to call the COVID prophecies bunk. And here we are with the October data and…nothing. The October removal numbers came in lower than expected.

Below are the numbers for January 2014 to October 2020, taken right from the DCF dashboards. The grey line is a trend line that calculates the median value for +/- 6 months. And, seriously, if you didn’t know there was a global pandemic that had locked down entire countries, closed schools, and possibly changed the course of history, you certainly wouldn’t have guessed it based on these charts (and I even highlighted COVID for you). The child welfare system experienced some shocks at the beginning, but for the most part it adapted and carried on.

Below is the exact same set of numbers with the normal seasonal variations taken out. This lets you see what is unusual. You can see that abuse calls in April 2020 were unseasonably low, but they got back up there. If you look at Hurricane Irma in 2017, you’ll see that the whole system took a pause — calls, removals, and discharges were all lower that month. But the pandemic seems to have only temporarily reduced calls, delayed some exits to September 2020, and pushed removals into an abnormal pattern that is reflected by that saw-tooth line starting around September 2019. What COVID didn’t do: impact the out-of-home care population hardly at all.

It looks like discharges had a rebound in September, clearing out cases that had stalled due to lack of court and closed services. You can see the bump in all three types below: reunifications (17%), adoptions (34%), and guardianship (27%). The adoption number should be put into context, though — the adoption trend line is dropping, suggesting things are slowing down, but only time will tell if that trend continues.

If I were to write a short factual statement on how COVID affected Florida’s statewide numbers, it would be this:

Florida’s been in a very gradual contraction period for a few years, due in large part to a reduction in the number of kids entering the system. The trend was 1,428 kids removed per month in February 2016 and 1,128 in February 2020. When COVID hit, removals got thrown off their normal schedules, but the trend line remained the same. Discharges had been very stable before COVID and were slightly down going into 2020. COVID dropped discharges a lot more, from about 1,100 per month to 960 per month. We saw a huge spike in discharges in September, but we need to do more to get the trendline back up.

Not everywhere in Florida had the same experience

Florida’s system is a patchwork of community based care agencies, and they rarely move in lockstep. Some agencies did in fact experience dips and surges not seen in the statewide numbers. I made a charts for each of the 19 CBCs. And, I’ve included a short factual statement you can make about each of them based on the numbers. The big takeaway is that across the state COVID had way more impact on discharges than removals, but mostly it had little effect at all on the out-of-home care population. (A more sophisticated analysis would look at the demographics and reasons kids came into care during COVID and test for shifts there — this is not that.)

If you want to go through the data in more detail (down to the county level), you can do so at this dashboard I made.

One important note: COVID significantly affected the quality of life of kids and families in the system by denying in-person visits, reducing court access, closing schools, and a million other ways. The DCF dashboard numbers don’t measure those things. These charts are only about how COVID affected the core system metrics of removals, discharges, and overall population.

Another important note: I’m not including the hotline calls for two reasons. First, DCF does not report hotline calls by CBC, so the info is not readily available. Second, the number of hotline calls have very low (almost no) correlation to actual removals.

I started this post a few days ago and hoped it would be my last COVID post. But with the numbers going up and states entering back into lockdowns I’m worried I’ll be staring at these numbers for a long time to come. Be safe.

I haven’t written about the COVID numbers for a few months because there hasn’t been much new to say. We’ve needed more time to tell if the initial shocks in April were a blip or a new normal. We now have 6 months of data to compare and the answer looks pretty clear to me.

It also looks pretty clear to DCF, which tweeted out this thank you to teachers for calling in more cases on kids.

That’s a very odd thing to thank people for, along the lines of “thanks for calling 911!” or “thanks for reporting so many forest fires that we investigated and discovered were not actually a fire at all!”

Before we get to the numbers, a quick reminder that foster care numbers are time-series data, along the lines of the stock market, ice cream sales, and average temperatures. Here are three important components of time-series data that this post will reference:

1. The seasonal ups and downs are the changes that happen, on average, every year at around the same time. For example, an ice cream stand may do 250% more business every summer, and the firecracker industry may have two or three spikes per year. When you’re interested in what’s normal, the seasonality is interesting. When you want to measure what is abnormal (like a pandemic) you remove the seasonal variation from the data to get a smoother graph and a clearer picture.
2. The long-term trend is also measured and removed. If ice cream sales have been going up 10% a year for the last decade, that is interesting but should be factored out when determining whether you had a good month. In the child welfare graphs below I’ve used a 12-month median to determine the trend line. A smaller window would result in a jumpier line, and a bigger window would give a smoother line.
3. If you take out the seasonal and long-term trends, you’re left with the remainder amount. This shows how far the value is outside of what you expected. If you’re asking whether something was an anomaly, you could look here to see if anything stands out. The graphs below look for spikes that are four or more standard deviations from the mean. That means they are very, very unlikely to be normal variations.

So now, with six months under our belts (masks?), let’s see what the numbers show. The full dashboard down to the county level is here. This post is a summary of the statewide numbers.

First, something that the pandemic impacted a lot: institutional abuse reports

If you believe in a direct correlation between abuse calls and safety, then the lockdown orders undeniably made one set of kids safer: the 500-600 kids per month who are allegedly abused by their schools, daycares, summer camps, churches, and other institutional caretakers. This isn’t a situation where the numbers dipped because of a lack of “eyes on the child.” The “eyes” are the abusers in these cases.

Below you can see that the number of institutional abuse reports was down 58% in April 2020, 44% in May, and then started to rebound. The trend line (in blue) had been pretty steady since 2014, and may have even started dipping a little in 2019. The pandemic significantly changed the trend — and I suspect the number won’t completely rebound because fewer people are sending their kids to institutional care settings.

The reduction in calls on children in their homes wasn’t nearly as drastic

Looking at the institutional abuse calls puts the in-home abuse calls into better perspective. Yes, they went down in April (24%) and May (13%) but then they headed back up. The trend line wasn’t much bothered by these dips. I’d call this a V-shaped recovery. The next six months will tell whether that slight increase in September is just a rebound or a new trend. The pandemic dip was a blip.

A trend that wasn’t affected at all: removals

The number of removals per month has trended down since 2016. During that period the largest single month decrease was September 2017 during Hurricane Irma (down 19%). April 2020 came close (down 15%). By summer, though, removals were back above the trend line. And most importantly, the trend line itself didn’t budge.

This is the time to mention that the child welfare system is very bad at its core job of identifying and getting help for kids who need it. The system relies on citizen reporters, as opposed to, say, random home checks of everyone with kids or, say, having such amazing services that people voluntarily call for help instead of having to be sued by the state to accept it (imagine!). These reporters notoriously over-report cases in some communities and under-report them in others. Even the cases that are reported typically don’t result in a removal (only 5-6 removals per 100 alleged victims), and somewhere between 40-50% of investigations result in no indicators of child maltreatment all. Investigating those cases is a waste of time, results in needless trauma to the family, and takes away resources and attention from kids who really need help.

So how did calls to the abuse hotline go down but removals not budge? One possible explanation is that the calls that got skipped would have largely fallen into the “no findings” group. Another explanation could be that DCF weakened its risk thresholds and removed kids who normally they would not have. Yet a third possibility is that the lower investigation caseloads resulted in deeper investigations and kids being removed who would have been overlooked before.

It could be some combination of all three.

There’s some evidence that DCF’s risk threshold changed. For example, removals in April 2020 decreased most for kids aged 5-12 and 13-17. The same level of decrease didn’t happen for newborns or toddlers. These removals were based on calls made in February and March, before the pandemic , so you can’t blame the callers on this one. The investigators’ decision making seems to have change. Teens had a very sharp rebound in June that wasn’t seen in the other age groups.

Below you can also see that neglect-type maltreatments (e.g., inadequate supervision and housing) saw greater decreases — relative to their normal ups and downs — than abuse-type harms. Some maltreatment removals rebounded over the summer (e.g., sexual and physical abuse), and some saw very little change over time at all (e.g., drug abuse and domestic violence). Not shown here, domestic violence removals went up a little in the Northeast Region and stayed flat everywhere else.

I should note that removals typically rise 7% in October and then are down through the end of the year. If the predicted wave of removals doesn’t appear by October, I don’t think it’s ever happening.

Discharges are more complicated

The discharge rate has been very flat for a while. It arguably started trending downward in mid-2019 and continued trending downward through the pandemic. You can see some green spikes that look like corrections but the general trend is downward.

Breaking that into exit types, reunifications trended downward and hit their lowest point in May when most courts were shut down (down 20%). They were then down all summer. There appears to be a U-shaped rebound, but we won’t know for sure until we get more information.

Guardianships took a huge hit in March when courts first closed down, but have more or less corrected. Guardianships have been on a downward slide for a while (which I think is very bad policy), and the pandemic didn’t do anything to change that.

Adoptions took a big enough dive to affect their trend line, going from about 321 per month to 243 per month. We need more time to see whether the correction is U-shaped or whether we get a new normal L.

Well-being measures mostly improved

If you thought the pandemic would lead to worse well-being measures in foster care, you would have been wrong. Most of the measures have actually improved, with the slight exception of the number of kids getting timely dental care.

• The percent of kids abused during in-home services is the second–lowest ever reported = 4.55%.
• The rate of abuse of kids in out-of-home care is the lowest it’s ever been = 6.3 per 100,000 bed days.
• The percent of kids abused again within 6 months of closing services is the near the lowest it’s ever been = 3.49%
• The percent of alleged victims not seen within 24 hours by an investigator is the lowest it’s ever been = 6.76%.
• The percent of investigators with more than 20 open investigations bounced back a little but is still historically low = 1.93%
• The percent of kids placed outside of their removal circuit is pretty flat = 17.04%.
• The percent of kids receiving timely dental services bounced up slightly but is still pretty low = 80.75%.
• The percent of kids receiving timely medical services is a little elevated but not by much = 95.31%.
• The percent of kids with kinship care as their first placement has gone up a lot to 62%.

And this next one is my favorite, so I’m showing the graph. The statewide number of placement moves per 1,000 bed days in out-of-home care was the lowest it’s ever been (3.53). Black children had the most stability they’ve seen in the last 13 years (3.9) and the pandemic erased an 8-year trend of increasing instability for Black children in care. It’s not surprising that making caregivers bear more costs from placement instability (i.e., risks to their own health) and reducing placement departments’ willingness to accept placement change requests had a net positive effect for kids.

So that’s it. The pandemic didn’t crash the system. It made some of it better. One more month like this and I think we can call it stabilized.

Florida DCF’s numbers are out for June and, in terms of investigations and removals, they were spectacularly normal. The predicted wave of removals isn’t here yet, if it’s ever coming at all. To the contrary, the system is struggling to discharge kids, as reunification and adoptions have both significantly decreased. We also have our first look at medical care, dental care, placement stability, sibling separation, and other measures that DCF only reports quarterly. Go ahead and form your own theory about what happened on those before you read on. You might be surprised.

Here are the details.

There were 10,054 initial in-home abuse calls accepted in June 2020. That is only 230 calls below June 2019, making it a fairly normal summer. You can see in the chart above that June calls have been dropping since at least 2016. The rate of decrease actually slowed this month, coming in at 11% above the seasonal trendline. That bump could be due to delayed calls from April and May or the drumbeat of media stories about how people need to report. The short version: people can’t say that calls are down this month, but I’m not quite ready to say they’re rebounding either.

Looking at the pipeline effects, with fewer calls in April and May, there were fewer investigations to close in June — though some investigations may have also been closed faster due to reduced workload. The number of verifications also went down while the number of removals rose slightly. None of this is particularly unusual, except it happened to occur during a pandemic.

And that’s the takeaway thus far: we haven’t seen the drastic increase in removals that people whose jobs depend on caseloads keep predicting. The chart below shows that there has been a gradual but steady decrease in removals since around 2016. June was about 2% up, which is right with the trend.

Now that we have more data for context, it appears that April 2020 was our lowest month for removals, as everyone was getting adjusted to the lockdown. That dip of 15% in April was still less than Hurricane Irma in November 2017, which shut down four of the five regions and reduced removals by 19%. There was a slight rebound in 2017, but it came two months later and quickly leveled out.

Meanwhile, we are seeing lower numbers of kids exiting care across the board. Total exits were about 18% below the trendline. Reunifications were 14% below, and guardianships were pretty even. Adoptions in June, typically one of the biggest adoption months right before the end of the fiscal year, were down a whopping 42% from the expected value. I’ve got a few theories on why. First, trials have been postponed, which could reduce the number of kids free for adoption. Countering that, though, appeals usually take 4-8 months, so we should still be clearing out adoptions from December 2019 to maybe March 2020. Second, the summer adoption campaigns that normally result in lots of adoptions happening all at once probably got paused this year. All those adoptions may wind up spread over the summer months more evenly. The lower rate of adoptions for June isn’t a headline — yet.

New this month, we have data from DCF reports that are run quarterly. Take a look at the DCF chart below. The pandemic seems to have solved placement instability. We don’t yet know why that is, but I suspect foster parents were less likely to ask for a kid to be removed, and agencies were less likely to act on requests that were made. Whatever the reason, the decrease in placement changes happened statewide. The Southern Region hit the 4.12 moves per 1,000 bed days standard for the first time in five years. It will be interesting to see if we can hold these numbers down.

The lockdown appears to have significantly disrupted dental care. Note that the measure covers the entire quarter and looks back seven months, so some of these kids may have gone to the dentist prior to the lockdown.

The lockdown only slightly reduced medical care. Maybe more kids wound up going to the doctor with suspicious symptoms, or maybe the expansion of telehealth made it possible to see a doctor without going in. The window here is 12 months, so maybe we’ll see a dip next quarter if the lockdown continues.

As far as the other reports, there was no noticeable change in the percent of kids placed in their removal circuit, seen by case managers every 30 days, sibling groups placed together, or just about any of the other quarterly reported measures. You can review DCF’s dashboards here.

You can play with the dashboards I created for these posts here and view the data by region, CBC, circuit, or even county. Until next month, be safe.

The May numbers are out and they begin to show how the pandemic is affecting things later in the pipeline from investigation to removal. Remember that investigations take about 60 days from intake. Intakes were down 6% in March and 16% in April from their expected numbers based on historical trends. Closures on investigations in May, however, dropped nearly 34% from the expected trend. That suggests investigations may be taking longer to close. Meanwhile, verifications were down 22% from trend, and removals were only 7% below trend.

Having removals come in 7% under the trendline is very normal. It happened in November 2019 and didn’t make headlines. This suggests that the reduction in intakes was more heavily weighted toward low-risk cases that would not have resulted in a removal anyway. DCF only verifies about 15% of calls and only removes about 4-6% children under investigation, so a reduction in frivolous cases is a win for everyone. We need to wait another month to see how the remaining March-April intakes come out.

On the back-end, exits from care are still very low. They were 40% lower than last year and 20% below the trend line. Reunifications and adoptions were both down. Courts should probably set a few rocket docket days to hear the cases that are pending an uncontested reunification, adoption, and guardianship.

Looking ahead, intakes in May dropped about 33% from 2019 levels, or 9.9% below trend. That puts us back to 2010-level numbers, and it’s still too early to say whether that is an anomaly or a new normal.

You can explore the numbers all the way down to the county level using the dashboard.

Here are the statewide details for this month.

Intakes were down about 9.9%. It’s very hard to put that decrease in context because the intake numbers started behaving oddly in mid-2019. The “turn” may have predated the virus, but we’ll only know if this was a dip or a new normal with more time. The numbers for April have been adjusted up to -16.3% (from -26.5% last month) based on the new data. That’s still hurricane levels low.

Investigation closures were down 34%. We are starting to see the pipeline effects of the lockdown. Intakes were low in March, so closures — which take up to 60 days — are predictably low in May. Based on the numbers of intakes in April, closures should be down even more in June.

Verifications were down 22%. Verifications have been steadily decreasing for years. The fact that verifications didn’t drop as much as closures suggests (but definitely doesn’t prove) that the lockdown resulted in a larger percentage of lower risk cases from being called in. It doesn’t fully prove that because we can’t say how the pandemic affected DCF’s perception of what should be verified.

Removals were only down 7%. That’s well within normal ranges for removals. This is the strongest indication yet that the pandemic resulted in fewer low-risk cases being reported — with the same caveat that we don’t know how DCF’s risk analysis changed. Removals per 100 victims went up a little, again suggesting that more low-risk cases were removed from the investigation pool.

There’s been a lot of discussion about homebound children and sexual abuse cases, and sure enough removals were up for sexual abuse (+23%), inadequate housing (+14%), and inadequate supervision (+10%). That could mean more egregious sexual abuse cases or less risk-tolerance from DCF. Removals came in lower than expected for physical abuse (-11%), drug abuse (-7%), and domestic violence (-3%). All of these, however, were within normal ranges, meaning we saw similar ups and downs throughout the last 16 years. The removal of Black children was down 7.7% in May and 22% in April. That April number was very low.

Discharges from care were down 20%. That’s very low. Courts are starting to get back up and running, so this number should go back up over time as cases that were ready to close get put on the docket. Here are the numbers by permanency type: reunifications were down 13%, adoptions were down 13%, and guardianships were flat.

That’s it for now. The full dashboard lets you see the numbers all the way down to the county level.

We are finally starting to see what impact a nationwide quarantine will have on the foster care system. DCF handled 18,894 intakes in April, which was the lowest number in at least a decade. You could call that 40% lower than last year, or you could say that it’s 26% lower than expected, given that the numbers have been on a downward trend anyway. Either way, as you can see in the full dashboard, we’re beyond hurricane numbers, which typically give a month of lower intakes and then a return to normal. We are instead moving into something unprecedented: what happens if the foster care system doesn’t get new kids?

Ins and outs

For context to the numbers below, I want to give an analogy. The foster care system is a pipeline. Kids come in through the front and leave out the back. Governmental policies, more than anything else, determine the rates at which kids enter and exit. Over the decades, various federal and state administrations have prioritized removals, adoptions, reunifications, in-home services, licensed placements, family placements, and a slew of other policies that had the primary effect of changing the rate of kids coming in and the rate of kids going out.

The middle of the system is like a firehose. If you push water into the hose faster than it can exit, then your hose will expand and eventually burst. If you don’t keep enough pressure in the hose, the water will sit and become stagnant. Pressure comes from pumps and valves and other mechanisms that keep things moving.

Putting that water hose in child welfare terms, the policies around permanency timelines, judicial review, and funding limitations are designed to keep kids moving through the system. The policies around placements are meant to provide enough flexibility that the system can expand and contract without suffering serious shocks.

The most significant expansions and contractions in systems have largely been a result of intentional governmental policies. Since the acculturation of Americans to mandatory child abuse reporting from the 1960s through the 1990s, there has been a steady stream of reports, but not always a consistent level of governmental response. People have built careers, contracts, and budgets around handling a certain number of cases or children per month. Foster parents make work decisions based on the board rate, and programs hire staff on the reasonable assumption that every child who leaves will be replaced by a new one.

A lot of discussions have centered on the question of whether kids will be left in unsafe homes due to the pandemic. I think that’s a valid question. We don’t know whether those 26% of calls would have resulted in removals or not. We don’t even have a good baseline, because we don’t know how the pandemic will change the rates and nature of child maltreatment. We have reason to believe financial insecurity raises the risk of abuse and neglect, but we also know that income inequality and racial disparities impact the numbers by targeting certain families for scrutiny. We have no idea what happens when everyone goes through social and financial trauma at the same time.

The pandemic also raises a second question of what happens to the system if people don’t call cases in and that steady stream of kids suddenly dries up. Will investigators start removing children on facts that would have otherwise gotten a pass? Will the system hold onto existing children longer to minimize the time that beds stay empty? Will programs close or change their focus to take in different kids? Will foster parents not take one child because they had built their home’s budget around two or they now have their own families to look after? Will children who had been bounced around before suddenly find stability when there is no other child waiting to take their place.?

We have never in modern history had a prolonged supply-side shock to the child welfare system. How it responds will tell us a lot about how it works.

Here are the numbers for Florida.

Intakes were down 26.5% from the expected rate. You can see that the March numbers have been adjusted up to -4% because they don’t look so low compared to the new trend. The numbers were down in every region of the state.

Verifications were normal, maybe even a little up (+1%). They’ve been going down for a very long time. That dip in December 2019 predates any serious talk of pandemics. I have no explanation for that. Verifications follow this pattern everywhere except the Southern and Northeast Regions, where they were flat.

Removals were down 19%. There’s usually a two-month lag between an intake and a removal. These investigations would have happened during the pandemic, but the calls likely came in right before it. We are still seeing hurricane levels here. Removals for physical abuse (-30%), inadequate housing (-11%), inadequate supervision (-18%), and sexual abuse (-14%) were all significantly down. Drug abuse (-4%) and domestic violence (-6%) removals were low but normal.

Removals per 100 intakes were normal (-2%). Again, these removals are based on intakes that came in right before the lockdown. We will be watching this number over the next months to see what happens.

Exits were low-normal (-5%). This will also be interesting to watch.

Reunifications were normal (-1%). But, you can see that drop-off getting ready to happen. If that line jumps back up next month, we’re continuing our normal downward trend. If it falls off even more, we have something new happening.

Guardianships may be back to normal (-0.4%). You can see that guardianships took a huge hit in March, but hopped back up in April. Again, that is either part of the old decrease or it’s a new normal. Only time will tell.

Adoptions recovered (+3%). I suspected that judges would handle the adoptions that got postponed, and that seems to be what happened. Those cases were already ready, so it’s easy to just set the one final hearing. We do not yet know how the pandemic will affect the permanency timelines of cases that come in now.

The full dashboard is available here. You can check the numbers for your region, CBC, circuit, or county. You can also see the breakdown of how removals are different for different maltreatment types. I hope you’re doing well in these strange times.