Les Roberts – “Part a: Triangulation – a scientist’s best friend when data quality is not an option.”

Les Roberts – Port Loko, Sierra Leone – April 18, 2015

I had a wonderful experience yesterday about my own inability to be analytical. We were repeating a sensitivity assessment that we had done in November to see where things stand and to document any improvement in surveillance over time. The process goes like this: in each village we ask the chief about the details of any deaths this year, articulated as since Christmas, and if there have been any Ebola cases in his or the neighboring village this year. After starting with the chief, we always try and triangulate the findings with some other source like counting recent graves or speaking to a local clinic nurse. In the end, we see the fraction of credible reports that had been detected by the surveillance system… and while not perfect, it gives us a reasonable insight into the system sensitivity. We went to a village… maybe with around 300 people… where the chief reported five survivors and about 17 deaths this year, all but one from Ebola. This was really uncool in several ways. This would be more cases than this Chiefdom (their equivalent of a county) had recorded for this entire year. If this were true, this would be a huge undetected flare-up and cause us to doubt the entire knowledge about the state of the outbreak in this District. According to the chief, six of the cases had been in March and one was just last week. Even before analyzing the data fully, I was sensing that the surveillance system was not performing well but this was a monumental oversight. Making it worse, most of the cases had gone to a local government clinic and died there making the under-reporting more inconceivable. On the other hand, the chief and the villagers around him seemed unusually unsure about the dates of the deaths and we had seen elsewhere that people often report deaths from 2014 as being in 2015, so I was not completely distraught yet.

We did our usual procedure; at the end of the interview with the chief we asked if we could go see the graveyard where he has told us all of the safe burials occurred. In this rather remote village there was not one graveyard but the chief graciously gave us a tour and showed us 14 graves, some of which could have been four or five months old.  It has not rained since October and lack of vegetation proves little.  But, a couple were very fresh, like only a couple weeks old and probably another seven sure looked like they were from this year. Given that most of the 20 deaths had supposedly happened at the clinic and the patients were buried there, this sure seemed to verify the chief’s report. Making it more disturbing and plausible, the six deaths reported for March had all not gone to the clinic and had died at home.  People grew to think if you go to the clinic you will die, so late in the outbreak they tended to stay home. I was bummed. We thought we were at ten cases a week nationwide. If we could randomly pick 12 villages in Port Loko and just by chance find a village like this, how many hundred others must be out there under the radar?

The surveillance system has so many shortcomings that this was completely possible. Given how major a surveillance failure this was, we decided to go to the clinic where most of the deaths had taken place.

On the drive there, I was thinking of how to break this news to my boss Dr. Yoti, who has been here for months and months working unbelievably hard and barely seeing his family in Uganda… and the NERC (National Ebola Response Centre) in Freetown… oh my gosh will they be apoplectic! I started to write my e-mail to Dr. Yoti in my head. Should I come back with a team the next day and interview the families and get all the details before bringing this to my superiors? With so many graves and so many deaths reported by the chief, something had to be going on.

When we arrived at the Cadick Maternal and Child Health Post we were lucky enough to encounter both the nurse and her assistant. They were really impressive, and more importantly, had kept great records.  They confirmed every name and every death that the chief had attributed to this clinic, but showed how almost all had been in November and October, back when we knew there were widespread undetected cases. Whew! The one death the chief said happened at the clinic last week did indeed, and while it was suspected EVD, the lab results were not yet known. Thus, even with many fresh graves, there apparently was not a huge outbreak that had gone undetected in recent weeks.

A team of local staff went back to the village today and now the chief says only two deaths happened this year. Based on the graves I saw I do not believe that either. I suspect that this chief wanted support and aid for his village where ten percent have died since October and they have been given little beyond the blessing of quarantine. In this case, I suspect they had a bad outbreak in 2014 and a few cases spilled into early 2015 but he was allowing the families to bury their own so the last cases went unreported… but this was not evidence of a recent (post February) massive outbreak going undetected.

Oh my gosh! Thank goodness for triangulation and the ability to put some check on crude and deceptive data. Thank goodness I don’t tweet (… which if we are honest, sounds rather rude in polite company).



les april18

A pretty recent grave! That is weeks not months.

les april18 2

A grave probably from recent months but not recent weeks.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s