Skip to content
Coordinates of ThoughtHome of Philosophy
Join
Question · Open

At my hospital “no evidence of harm” keeps being heard as “evidence of no harm”. When is absence of evidence evidence?

Omar Ragab Fellow
· 4 replies
Medicine & healthThomas Bayes#absence-of-evidence#argument-from-ignorance#bayesian-reasoning#medical-reasoning
Follow

Its coordinates: 6.13 Philosophy of Medicine

In meetings at our hospital, I hear the same move again and again. Someone asks whether a new procedure, a cheaper supplier's product, or a shortcut in cleaning has any risks. Someone else answers: "There's no evidence of harm." And the meeting moves on as if that were a reassurance.

I am an administrator, not a doctor, but I have learned the old saying that absence of evidence is not evidence of absence. The textbooks call the opposite move the appeal to ignorance: something is true because it hasn't been proven false.

But I also notice that the saying can't be entirely right. If we have used a product on ten thousand patients and watched carefully, and nobody was harmed, surely the absence of harm is evidence of safety. And if nobody has looked at all, the same words mean almost nothing.

So my questions:

  1. When does "no evidence of harm" become real evidence that there is no harm?
  2. Is there a simple way to say this to a committee that doesn't sound like an attack on whoever said it?

In kalam I have read arguments that the absence of proof for something is grounds for denying it. I suspect it is the same question in a turban.

LikeAnswer
SaveFollow4 answers

Join the conversation: members reply here, and joining is free. Reading is always free.

Join free to reply Sign in

4 answers

  1. Wanjiru Njenga

    Fellow

    The Bayesian answer is short: absence of evidence is evidence of absence to the degree that you would have expected to see evidence if the thing were there. Ten thousand patients carefully watched: you would very likely have seen harm if it were common, so not seeing it is strong evidence. Nobody looking: you would see nothing either way, so seeing nothing tells you nothing.

    For the committee, one question does the work: "If there were harm, how likely is it we would know by now?"

  2. Tsion Bekele

    Fellow

    Add a second question for the committee: "Who would have reported it?" At our hospital, side effects in patients who don't speak Amharic are reported far less, because nobody asks them in their language. So "no evidence of harm" may mean "no evidence of harm in the people we talk to". The silence has a shape, and you need to know it before you can read it.

  3. Omar Ragab

    Fellow

    Wanjiru, I will use your sentence word for word at the next meeting. Tsion, "the silence has a shape" is the part nobody in our meetings thinks about, and it's the one that would actually change what we do, starting with who collects the reports.

  4. Callum MacLeod

    Reader

    Search and rescue runs on Wanjiru's sentence, and we even put a number on it. When a lifeboat or a helicopter sweeps an area and finds nothing, the planners don't write "nothing there". They record the probability of detection for that sweep: given the sea state, the light, the height and the size of what we are looking for, how likely were we to see it if it was there? The chance that the dinghy is in that area goes down by that much, not to zero, and the next search goes where the remaining chance is highest. Reasoning of this kind helped direct the 2011 search that found the wreck of Air France 447, after earlier searches had come back empty.

    So for your committee, Omar, one question: "If there were harm, how likely is it that we would have seen it?" Nobody can answer that without saying how hard they looked.