CA Prisons as COVID-19 Incubators: Data Analysis

Building on the work we did in the last couple of weeks based on the CDCR COVID-19 tool, my partner Chad Goerzen spent a few days and nights synthesizing the numbers from the tool with the numbers for the surrounding counties from the L.A. Times tool. We think these plots tell stories about the interplay between prison and community infections, but the stories are incomplete because testing (and retesting) is so lacking, so take them with a grain of salt.

Up on top you see our plots for Lassen county prisons and for Lassen county population. Please keep in mind that the L.A. Times ticker does not include prison populations (though it does include residential homes, as per their data page, and we don’t know whether it includes county jails.) As you’ll see, the Lassen plot confirms that the prison and outside community outbreaks happened in tandem, and that we cannot rule out a causal story that explains the spike in Lassen County as an outcome of the botched transfer from Quentin to Lassen. The spike makes more sense if you notice that our Y axis is exponential, not linear.

The Marin county plot tells a very similar story. Notice that the outbreak in Quentin slightly preceded the sharp spike in county cases, confirming the theory floated in the Chron a few days ago that attributes the Bay Area spikes in great part to the Quentin outbreak.

In other counties, it’s more plausible that the prison outbreaks occur either as a consequence of a community contact or some CDCR snafu, against a backdrop of a county that sees exponential increase in new cases (seen in this graph as linear). A classic exam is Kern County, which we looked at a few days ago. Kern is a relatively open county with low levels of shelter-in-place compliance, and it’s not surprising we’re seeing contagion in all three prisons, which is more consistent with a story of surrounding county chaos than a particular transfer to a particular prison.

Seeing a similar pattern in Kings county. You can see a discrete outbreak in the local prison, against a backdrop of rising cases in the neighboring county. We see the county spikes closely following prison spikes (or vice versa; we’re not sure whether the 6-day testing lag in prisons is the same in the counties), but it’s hard to tell a causal story.

Same thing in San Luis Obispo. We’re seeing the outbreak at CMC against the backdrop of community infections, and it could be attributed to a community contact or to CDCR mismanagement:

At Imperial county, we see parallel outbreaks in prison and county, both of which follow the same pattern over time.

Same deal (with worse numbers) in San Bernardino, where the virus continues unabated in both county and prison:

At L.A. County, which has the worst numbers in the state, there was early on a serious outbreak at the local prison, which has now abated, but their jail is facing some serious problems.

Now take a look at Riverside county, where outbreaks at local prisons are staggered (and all in different stages of abatement.) The county numbers continue to grow–started rising exponentially in mid-May after rising at the same rate since late March — and it’s hard to tell a story about community contacts.

Given the variety of patterns and the low quality of the data, it’s hard to tell a consistent story about this, except the Marin and Lassen stories, which are the most obvious. The only takeaway, which I think is not an unimportant one, is the two-way permeability of prison and county populations. This should provide a rather solid answer to whoever in your life is telling you “but I don’t care about ‘those people'” when you sound the alarm about prison outbreaks.

UCSF Town Hall Report on Quentin COVID-19 Crisis from Amend

My UCSF colleagues Brie Williams and David Sears, among others, are at the helm of Amend, an organization seeking to transform the toxic correctional culture inside U.S. prisons and jails to reduce its debilitating health effects. They partner with correctional institutions to provide a multi-year immersive program drawing on public health-oriented correctional practices from Norway and elsewhere to inspire changes in correctional cultures and create environments that can improve the health of people living and working in American correctional facilities.

Recently, Drs. Williams and Sears gave a talk at the UCSF Town Hall. You can hear and see their findings here (from minute 19:00 to 34:00.) The team visited San Quentin on June 13 and were horrified by what they found:

The AMEND team made a series of recommendations. I recommend reading their entire report, which details possible isolation sections within the prison, as well as the importance of creating a true sense of partnership with the prison population instead of frightening them even more.

Among the AMEND recommendations was the urgent need to prioritize tests coming from San Quentin, which now take an astounding 5-6 days to come back positive or negative. They also noticed a disturbing neglect in staffing shifts, where staff was not “cohorted” with the same people, but mixed around to mill with new people every day. This was the situation when they visited:

Now, of course, things are more dire; we already have five confirmed deaths. But, and this is important, the picture we are getting is partial and misleading, because testing is so lacking and inconsistent. This gives you a comparison of cumulative testing and the testing positive rate (TPR). You’ll notice that the testing has slowed down, and there’s very little in the way of repeat testing.

If anyone reading this is in a position to help AMEND, either by offering your medical skills or in another way, here’s the contact information:

What Can We Learn from Prisons where COVID-19 Has Abated?

We got more much-needed media attention yesterday to the crisis at San Quentin and elsewhere; here’s my interview at KALW, and here’s a fantastic episode of Fifth and Mission. Also, the Quentin outbreak is now considered one of the “three big reasons” for the outbreak in the Bay Area–as per the graphs in my post a few days ago.

Let’s take a moment to look at two other prisons this morning: Avenal and Chuckawalla. Both prisons belong to the first group in my prison typology from a few days ago: places where there was serious outbreak that seems now to be petering out. The Avenal data is in the image above; the Chuckawalla data is in the image below.

The two prisons have numerous features in common. First, they are both overcrowded below, but near, the limit set in Plata. Avenal has 4,158 prisoners in a facility built for 2,920, housed at 124.4% of design capacity, and Chuckawalla has 2261 people in a facility built to house 1738, at 130.1% of design capacity. The course of the pandemic in both prisons has been remarkably similar: an alarming rise in cases, to the tune of hundreds of cases, which then gradually slowed down. In both prisons, 99% of the population has been tested, so the numbers tell a fairly complete story. A few people died of the virus (three at Avenal, two at Chuckawalla.) A few people were released (30 at Avenal, 17 at Chuckawalla.) The vast majority of cases (871 at Avenal, 710 at Chuckawalla) resolved with the person still in custody. Overall, about 1,000 people in each prison tested positive, and the contagion seems to be abating.

Even though I’m not an epidemiologist, it seems to me that studying what happened at Avenal and Chuckawalla has immense epidemiological importance. The most important question is: How did the contagion abate? This is where I enter the realm of speculation. One possibility might be that, at some point, even without prison intervention, the virus simply reaches saturation, the population develops herd immunity, and infection rate gradually slows down. Another possibility might be that the populations at these prisons are younger and healthier, and they recover more quickly. We know that a quarter of California prisoners are aged 50 and up, but they might not be evenly distributed throughout all facilities; San Quentin, for example, has a higher concentration of older prisoners. Yet another possibility might be that the few releases they did were targeted toward key transmitters, though I doubt there’s that level of epidemiological knowledge within CDCR at this point. If the answer is mostly the former–natural abatement–then the follow-up question might be: what is the risk of a second outbreak if there’s a new botched transfer into the prison, or a staff member contracts the virus outside and brings it in? Does the herd immunity hinder a second outbreak?

The answers to these questions are important because they can shed light on other epidemiological questions we are facing. The topic de jour in my social media circles today seems to be the reopening of schools–if, when, and how. It strikes me that, given the real possibility of outbreaks in schools (albeit minimal, because kids do not seem to be carriers or transmitters to the same degree), the experience of Avenal and Chuckawalla can provide the worst-case scenario of contagion and give us a sense of what to expect–as well as how to prevent it. This is relevant to other indoor spaces in which social distancing may be a challenge: workplaces, movie theaters, etc. If epidemiologists want to provide knowledgeable advice, they might want to learn from the experience of these prisons–what can be expected when the virus runs its course and, if any interventions were used, which of them was fruitful.

Rising COVID-19 Rates in Kern County Prisons Likely Reflect County Statistics

Media attention has thankfully shifted to COVID-19 outbreaks in prisons, focusing, understandably, on the horrific crisis unfolding in San Quentin. Yesterday’s KTVU story (below) and another one at The Appeal are a step in the right direction (also, today at 5pm KALW will broadcast an interview in which I explain some dimensions of the problem.)

I think it’s important, though, to perceive what is happening not just at the individual prison level, but on a systemic level, and through the lens of organic connections between prisons and the surrounding community. Which brings us to the site of some recent rises in infections: prisons in Kern County.

A few important things to know about Kern County: Just by looking at the CDCR tracking tool, you’d think that there are four prisons there – California Correctional Institution (CCI), Kern Valley State Prison (KVSP), North Kern State Prison (NKSP), and Wasco State Prison (WSP). But there’s a fifth one, California City Correctional Facility (CAC). The story of CAC explains a lot about the dynamics of California corrections. It was originally built in 1998 on speculation by Correctional Corporation of America (CCA), now rebranded in its gentler, kinder image as CoreCivic. By contrast to CCA’s wild success nationwide, it was unable to open a private prison on California soil because our powerful prison guards’ union, the CCPOA, resisted. Think about it as a Terminator-vs.-Godzilla epic fight: as Josh Page explains in his book The Toughest Beat, the union was so powerful that it beat the private contractors. The facility lay empty until 2006, when it was used as an ICE detention center (this was part of the “portfolio investment diversification” I talk about in Cheap on Crime.) But in 2013, as part of the state’s difficulty complying with the Plata population reduction mandates, they leased CAC from CCA, and it remains a privately-owned, state-run facility–confirming the speculative strategy of CCA, encapsulated in “if you build it, they will come.” I don’t know why the CDCR tracking tool does not provide information about CAC, and if you do, please email me–if there are people incarcerated there under CDCR management, their health is as important as that of people in state-owned facilities.

Let’s talk about what we do know. California Correctional Institution (CCI) started seeing cases on June 16. There are now 109 cases (there were 110; one person was released) and they have tested 41% of their population of 3,655 prisoners. Their infection rate is, therefore 7% of their tested population, and with an overcrowding of 131.3% as of last count, this could become a more serious problem.

Kern Valley State Prison (KVSP) has no cases at all (here’s hoping that, barring staff carriers or more botched transfers, it will stay that way), but North Kern State Prison (NKSP) has a few. They had one isolated case in late March, which resolved itself in April, and on June 3 they had one case. They currently have five. They have tested 18% of their population, so there are probably more, and they are at 107.3% capacity.

The situation at Wasco State Prison (WSC) seems more recent. They currently have 24 cases; the first five were diagnosed June 1st. Again, they have only tested 13.5% of their population, so it’s hard to say how things might evolve. They are at 109.7% capacity.

I’ve looked at the corresponding numbers in Kern County, and there doesn’t seem to be a corresponding spike. In fact, contagion Kern County is an ongoing disaster regardless of what happens at the prisons–their infection rates has been rising, unabated, since March. You can see the overall county picture in yellow in the graph below; the county numbers are dwarfing the prison numbers, even though the latter, in themselves, seem significant.

As I’ve explained before, it is impossible to tell airtight causal stories based on these graphs without careful contact tracing. Nonetheless, it seems like what is happening in prisons there is a consequence of excessive reopening countywide: their restaurants and bars are open for indoor dining, as are their gyms, salons, and tribal casinos, to name just a few. The L.A. Times page for Kern County offers another dimension to the story: a tragic focal point of infections there is their nursing homes which, like prisons, are vulnerable to contagion once the virus is introduced from outside. It seems more probable, then, that infection in Kern County prisons is attributable to staff who live, shop, dine, or gamble in the county. The imperative seems to be to avoid transferring anyone into Kern Valley and to release everyone over 50 or otherwise immunocompromised/vulnerable.