We are of course pleased by Dr Kapoor's interest in our paper on overcrowding in hospital wards and physical assaults on staff. ^{Reference Virtanen, Vahtera, Batty, Tuisku, Pentti and Oksanen1} The impact of overcrowding is a serious, albeit understudied, problem in healthcare research. In addition to the potentially increasing risk of violence perpetrated by patients, overcrowding has been shown to be associated with work overload in hospital staff and an increase in their risk of mental health problems. ^{Reference Kivimäki, Vahtera, Kawachi, Ferrie, Oksanen and Joensuu2–Reference Virtanen, Batty, Pentti, Vahtera, Oksanen and Tuisku4}

Dr Kapoor suggests that we were in error in reporting in the text that men were more likely than women to be working in high-occupancy wards. This is a misunderstanding. In Table 1, the proportion of men was indeed higher in overcrowded wards. More specifically, 264 of all 343 men in the study (77%) worked in wards with excess bed occupancy; 193 men (56%) worked in wards with the highest overcrowding. The number of women in overcrowded wards was 506, that is, 67% of all 755 women; 317 women (42%) worked in wards with the highest overcrowding. Conversely, 79 men (23%) and 249 (33%) women worked in wards with no overcrowding.

We agree with Dr Kapoor's view that simply by satisfying one of the Bradford Hill criteria of causation (in this case, temporality) does not provide sufficient evidence of a causal link between exposure and outcome. There is currently no consensus on the number of criteria required for determining whether an observed association is causal. ^{Reference Bradford Hill5} Dr Kapoor also referred to another of Bradford Hill's criteria – consideration of alternate explanations for a given association. Interpretation of findings from observational studies are inevitably constrained by concerns over confounding; that is, the role of unmeasured or poorly measured covariates. As we were careful to do in the paper, Dr Kapoor also describes some examples of such confounding factors.

We agree that the Overt Aggression Scale could provide interesting comparison to our findings. However, this scale (or its newer revised version) does not specifically measure physical assaults on staff, which was our study question, but instead a large spectrum of aggressive behaviours ranging from unspecified verbal aggression (loud noises, shouting) to physical attacks, which are not defined specifically as attacks on staff. ^{Reference Mattes6} However, owing to the extra resources needed and their time-consuming nature, such detailed instruments are most suitable for smaller-scale studies. In a large study involving 1098 staff drawn from 90 bed-wards, use of those instruments would not have been feasible.

Finally, just as any discussion section based on analyses of observational data inevitably touches on the problem of confounding, though similarly trite, it is also true to state, as Dr Kapoor indicates, that additional studies are now required to replicate and extend our findings before we can conclude with certainty that overcrowding increases physical assaults on staff.