On 2021-02-17 17:12:29, user Tim Pollington wrote:
Dear Epke and colleagues,
I would like to share some comments following reading your (v. relevant) paper on impact of COVID on VL in India at the country level. This is the second time I've commented on a preprint like this on medrxiv, and shared an 'open review' so I hope you receive it in the spirit it was intended. As I'm interested in doing similar studies your manuscript was relevant to me. And since I am funded by BMGF I thought it would be a waste of my funded time if I do not share these thoughts with you too, especially since you're at the preprint/pre-accepted stage.
I thought the paper could benefit from an additional author who has field experience of the IRS/ACD activities occurring there to back-up your assumption that "no IRS and ACD take place and that only passive case detection" during an interruption.
Given that the role of Asx in infecting others is still debated (some say recent xeno shows near zero contribution while ours last year did fit estimates consistently when relative Asx infectiousness of 0,1 or 2% were used), your use of the models E1 & E0 is a smart move to err on the cautious side.
Model structure and quantification section<br /> Thanks for much for following best practice and using PRIME-NTD. It is the first time I have seen it and I definitely plan to use it in my next modelling publication and also when initially planning a model re engagement with policymakers.
Given that the model runs for 30 years has population growth been taken into account?
Impact assessment section<br /> Although adding incidence rates in the same period is acceptable, as events share the same 'person time at risk' denominator (and if the events are mutually exclusive), I'm not sure if epidemiologically it's a correct calculation to sum up rates over the 30 years since the population will be changing in this time and thus the denominators are changing. Perhaps one can convert it into absolute cases in each year and then sum those up?
Discussion section - First paragraph<br /> It may help the reader if more emphasis was made on how a 1-year impacted delay by describing how it is amplified. ie How just one year interruption causes growth which needs to be curtailed before it turns over and falls, and the excess cases this generates. This concept of amplification could be strengthened.
Second paragraph<br /> "80% of [VL-endemic???] sub-districts..." Did this cover just Bihar or all 4/5? endemic states.
Third paragraph<br /> I think mortality rates are really relevant but can understand your caution re scant data. I think it's so important now considering the 1%CFR 2021-2030 target. Could this independent review help provide some rough estimates from pages 12-15 & 40? <br /> Even rough estimates from your model on excess VL cases and when they would likely be seen in the coming years, could be a useful starting place for resource planning of drugs.
I think a caveat needs to be noted that this analysis is country-level whereas the threshold targets are at the block-level, to avoid the reader making an ecological fallacy.
I hope that helps and also encourage you to comment on my work if I get to that stage!
All the best, Tim.