Thread by @GidMK

I've been tangentially watching this critique by @ikashnitsky of another worrying paper in JAMA Psych, so I thought I'd briefly discuss my own thoughts on the paper in a bit of twitter peer-review 1/n

https://twitter.com/ikashnitsky/status/1340474056005603331

2/n Paper is here, it looks at suicide rates in Maryland during 2020, and finds that there are racial disparities in these rates. In particular, there were fewer White individuals dying of suicide during lockdown, but more Black people died this way jamanetwork.com/journals/jamap…

3/n @ikashnitsky and one of the authors have had a fantastic back-and-forth, and I strongly recommend you read the threads here:

https://twitter.com/TheNeurocyclist/status/1340060128029061123?s=20

4/n Now, I am not a suicidologist, and while I have my own experiences of mental health issues it is not my primary area of focus

5/n I'm also a white Jewish guy from London, so at best my insight into racial issues is minimal aside from an academic understanding

6/n That being said, I think it's worthwhile to actually look at what the study found in terms of numbers

The authors essentially looked at the mean rate of suicides in three 2-month time periods (Jan/Feb, Mar/Apr, May/Jun) for different races

7/n Statistically, the choice to use the crude number of daily suicides is very strange. Suicides don't happen every day in this dataset (there are too few), and crude numbers are usually not very useful without a denominator

8/n That being said, this is what the authors found. Suicides decreased by ~20% in Black people in periods 1 and 3 (Jan/Feb+May/Jun) but doubled in period 2 (Mar/Apr), coinciding with the lockdown and COVID-19

9/n In contrast, suicides in the White population increased in period 1, and dropped significantly in periods 2+3 by 25 and 30% respectively

10/n Now, the authors have not standardized these rates to population, but that is fairly easy to do. The population figures for Maryland are here: datausa.io/profile/geo/ma…

11/n So, standardized to population we get these revised numbers

Still interesting, but immediately a bit less enormous in terms of the change

12/n It's interesting at this point to note that the overall suicide rate in Whites in Maryland appears to be 2-3x higher than that of Blacks, which raises the question of age-adjustment and whether it would impact these results

13/n That being said, it is immediately pretty obvious that the overall rates have not changed at all in Maryland over the first 6 months of the year for Black people. The yearly avg noise is bigger than the margin between 2019 and 2020

14/n Conversely, there is clearly a strong decreasing trend for the suicide rate in White people in Maryland, which has gone down by quite a bit every year for the last four

15/n So overall in this time period Black suicides haven't changed, and White suicides have gone down quite a lot per 100k person-years

This doesn't really seem to support the authors conclusion that COVID-19 has caused divergences in the suicide rate by race

16/n If we look at the individual time-periods, the data gets a lot more noisy, and it's very hard to discern an impact. In period 3, for instance, both Black AND White suicides appear to be trending down

17/n In period 2, this is reversed, with White suicides still trending down but Black suicides trending up

18/n Indeed, the exact thing that the authors have attributed to COVID-19 - a doubling in the Black suicide rate - happened in 2017/18 in period 2

19/n Ultimately, the numbers just aren't there to take home much meaning from this in my opinion. Overall, there have been no changes in suicide rates by race in Maryland in the first half of 2020 - the figures show that quite clearly

20/n This is not to say that there should not be targeted efforts to help Black people with mental health, which is of course a priority that cannot be ignored. Public health equity is a really important consideration

21/n And it's also not to say that a study with larger numbers may not be able to show such a difference. It is certainly possible that there is a meaningful change here that is simply not possible to see with so few events

22/n That being said, numerically I think it's important to note that the doubling of suicides in the Black community during Mar/Apr in Maryland was entirely offset by a 20% reduction in suicides in the preceding and following months

23/n There may be a statistical test that shows significance of a sort here, but conceptually I don't think there is much meaning to be garnered from numbers like this