[farah (OP)]

when ever you see a graph where there is an effect of something on the deaths of people e.g. disease, accident etc there is always number of deaths given per 100 000 why?

[Chemistry4me]

Eg, 50 000 out of the 100 000 deaths were due to automobile accidents
2 out of 100000 deaths were due to excess posting on internet forums etc...
It could be out of 1000 or any number really.

[RD]

It could be out of 1000 or any number really.

If the "per" number were too low then it could not be used to describe rare events using natural numbers:

e.g. some rare illnesses are quantified in X per million, ( 1 < X < 10 ). 

[Chemistry4me]

Aye aye sir, point taken.

[glovesforfoxes]

I think it is so that many different countries', cultures, places etc stats can be compared easily. Say Ireland has 2 deaths for alcohol poisoning a year but England has 50 a year - you might think that Ireland's death rate is lower, but you can't really tell unless it's expressed as deaths by alcohol poisoning a year per 1000 people. This is because of the different overall populations Ireland and England have - Ireland's population is much smaller, so it will have fewer deaths to alcohol poising anyway. Expressing it as per 1000 people removes this problem, and allows accurate comparisons of X rate per Y.

As an example, Ireland's death rate for fatal familial insomnia, a rare genetic disease that causes sleeplessness and eventually death, might be 0.1/1,000,000 a year, and England's might be 0.05/1,000,000 a year. Now because Ireland's pop is around 6 million, this means there will be 0.1*6 deaths a year - 0.6 deaths a year, so a little more than 1 every 2 years. England's population is about 51 million, so 0.05*51=2.55 deaths a year, so 5 deaths a year in total. If I just said Ireland has only 1 death every two years for FFI, & England has 5 every 2 years, you might automatically think it was more prevalent - but it is not. The rate is actually lower in England, even though there are more deaths by it. A graph would be fairly useless if it just gave you number of deaths over time if you want to compare the two places, so the rates are calculated and used in the graph instead.

[farah (OP)]

I am sorry but I completely disagree,
every time the number of deaths in a graph is taken per 100 000.
so my question still remains the same []

[glovesforfoxes]

For more common things, perhaps road accident deaths, yes, sure, which might be 500/100,000 people. It would be more appropriate to use a million in the case of fatal familial insomnial, since the death rate is so low from it, otherwise you end up with a number like 0.005/100,000, which is not as intuitively understood. Graph's purposes, in the media at least, are to tell us something fast. If you have to do some maths to work out what they mean, they have not done their job properly. This does not detract from the accuracy of what I said previously - the same principle applies to deaths expressed as 1,000, 100,000 or 1,000,000.

[Bored chemist]

Google shown me that there are 431,000 references to "deaths per million" and 310,000 references to "deaths per thousand" as opposed to 2930 for "deaths per ten thousand" and 18000 for deaths per hundred thousand.
With numbers rather than words, there are more referencess to per 1,000 than for per 100,000 so it seems that you have been looking at an odd set of graphs.
Here,
http://en.wikipedia.org/wiki/List_of_countries_by_death_rate
for example the rates are given per thousand.
Anyway the answer is that deaths per 100,000 generally gives a reasonable number without messing about with decimals or lots of zeros. Also, as glovesforfoxes pointed out, unless you have a rate rather than an absolute figure you can get misled by population sizes.