This was a very straightforward study examining whether a simple text message could increase flu vaccination uptake. The trial was well designed and the write-up is crystal clear. The results show a small increase in uptake, but given the low costs this is a potentially useful intervention. The per protocol analysis showed a larger increase than the intention-to-treat, which suggests that the intervention might be more effective in more ideal circumstances.

I have only minor comments: - The practices involved already used texting, so some set up costs have already been spent and the intervention may be slightly more expensive in non-texting practices. - It's a shame that the ethical complexities of studying pregnant women have excluded them from this study. - I assume the 5% significance level for the power calculation was two-sided? - page 11, line 14, it might be worth inserting 'observed' for 'intra-cluster correlation' as the previous page used an assumed one for the power calculation (and it's the closest I've ever seen between an a priori 0.024 and observed 0.029 ICC). - page 15, line 26, the important thing for significance testing is not whether the confidence intervals overlap, but whether the mean of one estimate is contained within the other interval. Given that, it does look like mornings are worse, although a formal interaction test would be preferable. - page 24, not every author appears in the contributions - Figure 2 does not add much information as there's no investigation of any geographical bias (plus the data are also in Table 1)