This manuscripts presents an analysis of surveillance data on the incidence of severe acute respiratory infection (SARI) due to influenza in New Zealand. The analysis compares pre-pandemic trends from 2012-2019 to the resurgence that occurred in 2022 following a two year period in which community transmission of influenza was absent due to the measures introduced to control Covid-19. Results are stratified by age and ethnicity and comparisons are made on monthly and annual incidence rates.

Overall the manuscript is a very interesting and important analysis. Given the high data quality with a long-term consistent surveillance baseline and the disappearance of influenza during the pandemic, the results are likely to be of broad interest internationally. They also have clear public health implications within New Zealand due to the striking disparities among ethnicity groups which appear to have been exacerbated following the pandemic.

I mostly only have minor comments and suggestions (see below), however one thing that struck me was the seemingly different patterns to the SARI influenza positive data published by ESR on their weekly dashboard at https://www.esr.cri.nz/digital-library/respiratory-illness-dashboard/ . For example:
• The dashboard appears to show a cumulative average (winter) incidence for 2015-19 of approximately 37 per 100,000, i.e. 0.37 per 1,000 in the authors units. That’s only about 10% of the rate shown in Table 1 for 2012-19 of 3.26 per 1,000. Similarly, 2022 looks to be approximately 0.36 per 1,000 on the dashboard vs 2.93 per 1,000 in Table 1
• Although the annual incidence for 2022 is very close the pre-pandemic average, the temporal pattern appears quite different with a much higher/sharper peak on the dashboard. Yet this is not seen in the authors results (Figure 1) where, although the peak occurred earlier than in pre-pandemic years, it appeared comparable in height for most age/ethnicity groups. (Possibly sharp but asynchronous peaks in the single-year data are disguised by pooling into a multi-year average, but I doubt this fully accounts for the difference.)
I am not suggesting that there is any problem with the authors’ analysis – possibly the data are measuring a different outcome, or the rate calculations use a different denominator e.g. Auckland versus national, or possibly I have misinterpreted it. But given they are both influenza positive SARI data from sentinel Auckland hospitals, it may be of interest to the general reader if the authors could comment on the apparent differences.

Minor comments
1. P3 line 50 – I assume this means that people in quarantine facilities who had respiratory symptoms but were negative for Covid were PCR tested for influenza? (At present the sentence is slightly ambiguous as to what they were tested for, and all people in quarantine were PCR tested for Covid regardless of symptoms, so a minor clarification would be helpful)
2. P5 line 31 – as above were national pop denominators used, or denominators specific to the Auckland population covered by the participating hospitals?
3. P5 line 58 – “no cases in weeks 42-34” which year does this refer to?
4. P6 line 5 – I am confused here because my understanding was that Byar approximation provides a more accurate alternative to normal approximation when counts are relatively low. So why was Byar used when counts were high (annual incidence) and normal approximation when counts were low (monthly incidence)?
5. P6 line 29 – can you clarify whether you used prioritised or total ethnicity data?
6. P8 line 3 – the first part of this sentence is slightly ambiguous about which groups are being compared. Perhaps a clearer wording would be “Among younger (15-49) and older (>65) groups, annual incidence in 2022 was higher in Māori and Pacific people than it was in NMNP”
7. P9 line 3 – this is a very interesting point. Are the authors saying that infants in 2022 still had normal levels of maternally acquired immunity derived from maternal influenza infections that occurred prior to 2020 ?
8. P9 line 33 – as currently written this sentence implies that NMNP adults had a higher prevalence of comorbidities. I don’t think this is correct. Should it say something like “Compared to NMNP over 65s, Māori over 65 years had 10-15% lower influenza vaccination coverage between 2020 and 2021 and a higher prevalence of comorbidities”?
9. P9 line 40 – not clear what “3-fold” is referring to. According to Table 2, the 2022 incidence in Māori and Pacific over 65s was about 10-fold and 18-fold higher respectively than NMNP over 65s.
10. P9 line 45 – but this hypothesis (household crowding) presumably would not explain the observation that the IRRs for Māori and Pacific relative to NMNP were much higher in 2022 than pre-pandemic? Unless the authors are suggesting that household crowding has markedly worsened in this time (in which case they should make that clear)
11. P9 line 58 – I suggested avoiding the term “alert levels” which had a specific meaning in NZ which (i) will not be clear for international readers and (ii) was discontinued in late 2021. I think the authors mean more generally that public health measures and social distancing behaviour were still at a high level.
12. P10 line 15 – what is confounding depends on what causal effect you are trying to estimate. It could be argued that the variables being implied here as confounders (I presume vaccination coverage and comorbidity prevalence) are actually on the causal pathway from the exposure (ethnicity) to the outcome, and household crowding is not really any different from a causal inference point of view. Do you mean that the observed increase is unlikely to be accounted for solely by vaccination and comorbidity, and therefore other factors such as household crowding may be relevant?
13. P10 line 24 – there was some (albeit very low) community transmission of SARS-CoV-2 in 2020 and 2021.
14. P10 l28 - I think this means that patterns of SARI-influenza were similar to pre-pandemic levels in children of all ethnicities, but the way it’s currently written could be interpreted to mean that patterns were similar across ethnicity groups. It would be helpful to clarify.
15. Tables 1-2 – how was significantly different to 2022 tested? Was it just whether the 95% CIs were non-overlapping (if so this will tend to bias results somewhat as the 95% CIs for the two rates can be overlapping even in cases where the 95% CI for the difference between the rates excludes zero)

Michael Plank.