Thread by @tomtom_m

A new paper from @mrcZA and @uct_news published today by the #SAMJ, looking at unnatural deaths, alcohol restrictions, and curfew.
A (longish) thread, explaining what we did - and did not - find. Paper, open access, available here: samj.org.za/index.php/samj…

2/We use data for 68 weeks, from early 2020 to mid-April 2021 (the paper was submitted in early May).We then categorise each week by the nature of alcohol restrictions, and duration of curfew, in force based on the changing sequence of lockdown regulations. >fig1<

3/The expected deaths are derived from the collaboration's mortality surveillance system. We then proceed to model the excess unnatural deaths each week, to remove effects associated with (e.g.) month ends, with covariates of alcohol restriction and categorical curfew duration.

4/The model fits the observed series of unnatural excess deaths very well - explaining nearly 3/4 of the variation >Fig2<. We then derive estimates of the overall effect for all combinations of alcohol restriction and curfew duration >Fig3<.

5/Or, in tabular form >Table 1<. Four of the seven combinations of restriction and curfew had a statistically significant reducing effect on unnatural deaths.

6/ 3 of those are associated with full restrictions on alcohol, with the effect reducing with reduced duration of curfew. The fourth seems anomalous: Partial restrictions on alcohol, with no curfew.

7/Reductions were large: ranging from nearly 50% (517 unnatural deaths averted/week, during Level 5), to 26% (292 deaths/week, with full restrictions & 4-7h curfew) to 13% (the 'anomaly' mentioned).

8/All other combinations had no statistically significant effect, and - in the case of no restriction with a 4-7h curfew saw slightly elevated (not significantly) unnatural deaths.

9/The 'anomaly' is not actually that much of an anomaly, reflecting the period in June-July 2020 immediately after hard lockdown when alcohol restrictions applied to BOTH on- and off-site consumption.

10/Our conclusions. a) full bans on alcohol reduce unnatural deaths. b) partial bans are largely ineffective at reducing unnatural deaths. c) it is alcohol, more than curfew, that drives the pattern of unnatural deaths.

11/ d) The greatest observed effect is likely the upper bound of such restrictions: the effect of such may diminish over time, as (inter alia) new restrictions now tend to be telegraphed in advance.

12/ e) Deaths represent the extreme outcome from trauma and injury. During hard lockdown, unnatural deaths fell by 50%. Other data suggests that trauma and accident hospitalisations fell by around 65%. Our paper does NOT look at hospitalisations.

13/ f) Bans on alcohol sales are effective in the short-term, but blunt, and do not offer a long-term solution. These are complex policy issues.
Other solutions should be considered, for example limits on volumes, unit pricing, constraints on marketing etc.

14/ An appendix shows that using different categorisations of curfew duration, or a different measure of mobility based on Google residential mobility data, does not materially change the results. samrc.ac.za/bod/UnnaturalD…