Les Roberts – Freetown, Sierra Leone – October 17th, 2014
Day 13: Bias, bias everywhere, but not a chance to think.
Great news. I was meant to have dinner with the two top guys here from CDC but they had to cancel to have a conference call with Tom Frieden. What a great director, what an impossible position. But, that can be a discussion for someone else’s blog. What this really means is that I have a couple free hours for the first time in days and I can write to you.
I had a great chat recently with Craig Spencer (PFMH ’2013, one of Rachel Moresky’s international emergency medicine fellows) who is with MSF in Guinea. He said that a quick scan of their data suggests that the longer between the onset of symptoms and when one shows up in the clinic, the better the chances of survival. This is of course, the opposite of the message we want to get out. This is clearly a bias issue, probably survivor bias.
CDC thinks that the median time from the onset of symptoms to death in 6 days. In Sierra Leone, we think that on average people have symptoms for seven days before coming into a health care facility. Thus, in terms of Craig’s data, with no treatment at all, if someone is on day two of symptoms, there is probably a 70ish percent chance they will die. If they show up on day ten, there is probably only a 20ish percent chance they will die. Thus, of course the longer a patient has been symptomatic when they show up at your clinic, the better the chance they will survive. But, as I said before, the primary health benefit of a patient going for treatment is not that they will get medical care and survive. The primary benefit is that they will not infect a slew of other people as they become viremic and shed via every pore and liquid in their body. I can think of no other disease with the possible exception of rabies where the number one health reason for medical care is not related to the patient. Thus, if the message got out, “the later you come for treatment the more likely you are to survive,” this would completely undermine the control efforts.
Almost everyone understands that this survivor bias is going on. There is one hospital here in town that is very proud and loud about how they have a 70+% survival rate. It does not matter which one. They are an Ebola treatment facility and patients have to be laboratory confirmed positive to get referred to this hospital. Now, it takes three to four days in Freetown to get lab results back (and a LOT of people are working on this issue). Thus, given seven days from symptoms to clinic presentation plus the lab delay, patients have been symptomatic 10-11 days without dying upon arrival at this facility… Of course, they have a 70+% survival rate. I care much more about the fact that this facility has handled a couple hundred patients and none of their clinicians have been infected. To me, this is far more impressive and more important in terms of stopping the outbreak.
There are other big biases operating as well. Before coming, I heard repeatedly (not just from the talking heads on the television but in my WHO briefing) that this outbreak seems to have a very low case fatality rate compared to past Ebola outbreaks, perhaps 46-48%. I strongly suspect and the evidence is building that this is just an issue of low sensitivity surveillance; and the burial surveillance is worse than the case surveillance. (That is what the CDC meeting tonight was about. I think between CDC and WHO, we will have some quantitative insight into this before long.) Likewise, I keep hearing that in Sierra Leone, this is largely an urban outbreak. While Ebola has never been in cities before and this is huge and noteworthy unto itself, I strongly suspect that this is more an issue that surveillance is almost complete (especially for burials) in the urban centers compared to the rural areas. Now it is Government policy that bodies be swabbed for testing before burials and if a body is Ebola positive, it gets entered as a case and a death, improving the surveillance sensitivity of both surveillance measures. Burial teams are providing nearly complete coverage in Freetown and are not available in many rural areas.
I would love to spend my time assessing these things but the fact is that if the surveillance system sensitivity is 30% or the urban fraction is exaggerated by a factor of two, the strategy for controlling the outbreak is the same. In Bo and Kenema, centers of the crisis last summer, the outbreak is on the decline. They now have enough beds for referrals. They have good contact tracing (maybe 90%????) and follow-up. The message has been sent out – come for treatment early. Most bodies get buried effectively by trained teams. Where it is not the case, groups spontaneously arose and developed their own no-touch approaches which use ropes with slip-knots to move corpses. We need to get that going everywhere as we have those efforts going well only in the Eastern areas. (Also, lab testing makes the clinics function better by getting the negatives out quickly, freeing beds, protecting the non-Ebola cases and allowing contact tracing to happen…) Time spent on improving data quality is potentially time away from getting facilities up and running. So on we run.
I hate bias! But I am just going to have to ignore its brazen manifestations for a while.