Some of these have become more evident as a result of recent crises in different parts of the world, and the papers in this volume are particularly valuable in drawing our attention to the issues involved.

Any discussion on the use of health statistics, for example, must consider the perennial question of denominator data and the difficulties that are almost inevitably encountered in obtaining these in the context of complex emergencies and natural disasters. Knowing what the risks are and who is really “at risk” are questions that have long plagued the assessment of complex emergencies and natural disasters and their health impact.

There are no easy solutions to this, especially where national health and other statistics prior to and certainly during crises have been poorly kept, tampered with, or damaged, or are simply not accessible. The papers discussed here reflect these difficulties and rightly raise them as limiting factors in any discussion of mortality in the context of emergencies and disasters. More attention clearly needs to be given to accessing denominator data from countries where the perceived likelihood of complex emergencies and natural disasters is high. It is also important that more attention is given to defining the risk factors and events that are likely to be of concern in crisis situations so that planning for data needs can occur early on. Burkholder et al., for example, refers to “war-related deaths” and highlights the question of what are war-related deaths. Are they the deaths that result from war injuries? Are deaths associated with “unintentional” injuries and exposure that occur during flight from persecution and accidents en route to safe havens to be included? And what about deaths that occur as a result of lack of access to health care services and medication? Shall we include the deaths among the elderly and others who “give up” the will to live in these situations? And what of the suicides that so often occur but also go unmeasured? If the latter are to be included, and surely the case must be made for this, then the concept of war-related deaths takes on a whole new scope and importance.

In the same way, it is important that we try to distinguish between mortality in the location of the crisis (e.g., primarily intentional and unintentional injuries and homicide), mortality during uprooting and forced movement (e.g., primarily “natural” causes, unintentional and intentional injuries), and mortality during resettlement (e.g., primarily “natural” causes and unintentional injuries).

Robinson et al.'s data suggest that in the case of North Korea the movement of people was not associated with mortality either as a prompting event or as an associated impact. In the case of Kosovo, on the other hand, at least two and possibly three very different pictures emerged. Mortality, or the fear of it, was a major prompting event, as it often is in situations of ethnic cleansing. There is also evidence that mortality peaked