Thank you for the opportunity to review stage 2 of this interesting manuscript.

The authors did adhere to the procedures described in Stage 1. Their introduction, rationale and stated hypotheses are the same as in Stage 1. 

The data seem to be able to address the study aims. 

There is an additional analysis not mentioned in the Stage 1 protocol, i.e., comparing overlapping HDA- and SUD-based PSMU cases with non-overlapping SUD-based cases, but this analysis is justified. 

The conclusion that HDA, and especially the HDA2 scoring method, may represent a valid conceptual alternative to the GD-based scoring is supported by the data. 

In the discussion section, the relation of these findings to the ICD-11 criteria for gaming disorder could also be mentioned. I know this was already mentioned in the introduction, but it could be elaborated in the discussion as well. Which unique element(s) or advantages over ICD-11 does HDA bring to the table?

Finally, I agree with what the authors said in the Limitations section: the current analysis used existing data and not spefically tailored items for social-media-induced dysfunction and harm. Besides stating this in the Limitations section, I believe it should be mentioned at the beginning of the Future directions section as well and the section could be organized with this in mind. This section esentially discusses ways to better tailor HDA to PSMU (and to behavioral addiction in general) but this is not stated at the beginning, which is why it reads somewhat confusing. Especially since this is then stated for the first time in the Limitations section that follows. Also, the Future directions section could address PSMU more specifically. 

​​​

During Stage 1, I was a reviewer for one round and it turned out that one of the authors was serving as a reviewer for a paper where I am an author. Therefore, I dropped from the process, as cross-reviewing can easily generate COIs. As I received a review invitation for this Stage 2 version, I decided to accept the invitation with the following caveats

- I'm curious about the results so I'll read them anyway, thus I can share some notes while doing that

- These notes should not be considered as a "review" but extra feedback, and proper reviewers should be invited to carefully review the full Stage 2

Here are my notes:

- Based on one initial read, the' reporting seemed solid. In particular, reporting language was approriate as it did not push confirmation or oversell the results; it all seemed very honest and clear.

- In one small part, I'd consider toning down a bit: "These findings suggest that the HDA approach, as it is designed to do, eliminates false positive diagnoses..." --> I would say "might eliminate" to be safe, as there's indeed no clinical evidence (which is correctly mentioned later). 

- For the most central results regarding overlap, I would help the reader by visualising them at least via one Venn diagram 

- This is the most important point I have: when the paper reports small, medium, and large effects, I would really try to transform those effects into some raw, real-world effects. There are many ways to do this, but any will be better than rules of thumb thresholds. The discussion could be a place to address this in more detail, eg. by comparing obtained effect sizes to those found in similar designs/tests in other mental health research fields. The PCI RR author guidelines have some tentative guidance/links on this topic.

- At the very end, the paper concludes by addressing design features. Although I fully agree these are very central to the field, the section appears entirely separate from the HDA topic. If you wish to keep that part, I'd suggest trying to build more explicit bridges (or even hypotheses) between design features and HDA.

Veli-Matti Karhulahti