Power calculations are seldom used in the multiple regression context, particularly with panel data and population-level data. These tend to be rather made with trial-based data to estimate appropriate sample sizes. Many would argue that post hoc power calculations are misleading and irrelevant. ^{Reference Levine and Ensom1–Reference Fogel3} Nevertheless, a post hoc power calculation based on the ordinary least squares model which uses the total number of valid cases used in the analysis, the total number of predictors in the model, the model R-squared, and the assumed P-value (set at 0.05), suggests that for all models the power is 1.00. By convention, this value should be greater than or equal to 0.80.

More importantly though, the benefit of the difference-in-difference methodology is that it provides for more precise estimates than the previous analysis and also allows for the simultaneous inclusion of covariates such as the team fidelity criteria (e.g. crisis resolution and home treatment teams (CRHTTs) offering a 24-hour service) as well as overall time trends. There are fundamental differences between the two types of analyses with the difference-in-difference methodology being a far more potent and robust policy evaluation tool.

We agree that future studies should ideally look at analysing admissions (and potentially other factors) at CRHTT level. We explored the possibility of doing this by contacting several teams to ask about their geographical boundaries, but found, surprisingly, that many teams were in fact unable to clearly delineate their geographical ‘patch’ and that even if they could define their current boundaries, these had often changed over time, making an analysis of long-term trends with difference-in-difference methodology unfeasible. Moreover, a large-scale national longitudinal study would require data from before the policy change (circa 1998) to effectively assess the policy impact, for which routine administrative data is more suited than data from individual electronic records systems, which have huge variation in detail, quality and method of collection.