Strongest predictor of SIDS was mom’s education level, followed by whether or not she was smoking; stomach/back was third. It’s the strongest one that can (easily) be changed, but something is very fishy about the data.

Proponents point out the inverse correlation between sleeping on back and SIDS, starting at the time the recommendations were made, as proof that this is a causal relationship. However, inclusion of time prior to recommendation shows that SIDS is inverse correlated to time; and sleeping on back is positively correlated with time only after recommendations were made.

Also, it’s possible data in the last 10 years paints a different picture - but at the time, the case for causal “back to sleep” was very flimsy.

Note the decline in SIDS rate, which starts in 1988 (see e.g. https://pediatrics.aappublications.org/content/105/3/650 ), before the recommendation (1992) and "back to sleep" campaign (1994) and though it does track the decline of belly sleeping position, I found in the past graphs of belly sleeping from before 1986 in which it was essentially uncorrelated (or even inverse correlated, if you go back enough - but I don't trust that data).

But what's more alarming, is that although there is a continuing decline in SIDS, if you look at the combined SUID (Sudden Unexpected Infant Death) rate, it is essentially flat since 1996 (following a decline from 1990 as visible in the CDC graph, and which goes much farther IIRC). As a data scientist, this hints either (a) change of classification system that was not properly noted, or (b) that there's an underlying reason which causes infant death, which may manifest in different ways depending e.g. on whether the infant sleeps on their belly or their back.

As a new parent, this was terrifying (is anything really ever safe?) but also reassuring (the expert advice might technically be true, but seems to be useless).

I've found this over and over in many other cases - e.g., statins seem to reduce the change of a heart attack, but do nothing for (or even slightly increase) all cause mortality. Also financial advisers give advice which is technically correct but often useless.

Experts rarely qualify the limits of their advice, and that goes double when it is delivered 2nd hand (by nurses or even doctors who never bothered to look at the data but just parrot the official recommendations).

As a data scientist, you should really consider more than these two causes. For example, perhaps some other cause of infant deaths was rising during that same period. Or, perhaps there is a change in classification, AND there has been a saturation point reached in terms of how the existing public health initiatives can affect population behaviors. You really just cannot reason from coarse, population level death rates like this to a specific condition or cause.

Also: https://www.cdc.gov/sids/data.htm

The graph on that site very clearly shows that most of offsetting increase is caused by strangulation and suffocation, not an actual unexplained cause. It appears to be mostly a change in classification.

Finally, I really want to point out that we should not expect that a public information campaign would decrease SIDS rates to zero. After all, we can see in this thread that there are still plenty of parents who put their babies to sleep on their tummies. As far as most people know, the data could show that SIDS is 100% caused by prone sleeping and people would still doubt the evidence and put their child to sleep prone. SIDS rates would still not drop to zero, and that would not be evidence against the hypothetical cause.

And I started with saying that my conclusion was that it is likely safer for babies to sleep on their back from the SIDD/SUID perspective, but the evidence didn’t seem as conclusive or overwhelming as I was led to believe.

The hospital staff made it sound like letting my baby sleep on their tummy is signing their death warrant.

Do note, also, that prevalence of back sleeping went from less than 30% to more than 80% over the period graphed. One should not expect compliance from such a campaign in general, but (as far as medical compliance goes) a very good one was achieved. New parents are perhaps the most compliant group.

I am unlikely to dig as deep into the data again, but my well supported (I believe) conclusion at the time was that, all-cause-infant-mortality-wise, the recommendation may be somewhat justified, but the strength, might and terror in which it is given is definitely not.