Confirming the outbreak requires that we have reliable data about the number of cases that we would expect in the same area and within the same time period in which the cases were reported. So can we define what we mean with «more than expected»?

We can't.

Sometimes we even do not have reliable information about the background rate of cases. And even with reliable expected numbers, how much higher does it make an outbreak? One standard error? Two standard errors? Partly it will depend on the level of background incidence. In the end, it remains a subjective choice.

Most of the time it will take judgement in addition to data to confirm an outbreak. A single case of a communicable disease long absent in population can therefore be an outbreak. (Eg : one case of rabies in a country with the 'rabies free status', one indigenous case of polio in the EU). For some diseases it may also depend on the season: 1000 cases per week in the summer may be consider 'an outbreak', while the same number in winter may be considered 'normal'.

Information sources that can help to determine the expected number of cases, such as surveillance systems, hospital registries or surveys.

OK, we have more cases than expected. Now what?

Any increase could mean several things:

1. An increase in population at risk (=denominator), for example due to an influx of migrants. If the number of cases doubles, because the population has doubled, then we would have expected that, and in the light of our definition, it would not be an outbreak (however, it could still require public health attention).
2. Random fluctuations. If in a city of 1 million inhabitants a rare disease occurs in only 1 patient per year, and suddenly we have 2 cases in a year, then this could still be within the variance of the expected. Again here. it helps to have the denominator
3. Registration artefact: it could be a classification error, or a better diagnostics was used, or a screening programme introduced. Any kind of change to 'the system' that suddenly detects more cases, while the 'true' number of cases in the population has not changed
4. A true rise in cases (a 'real outbreak').

When we have no denominator information whatsoever, then we still want to try to rule out causes 1,2 and 3 so that we are more confident that it is 'more cases than expected'

Explanation 1 is difficult to assess without reliable information about the denominator. However, we may still have some information. If we know that migration did not take place, and that birth rate and death rate have not changed significantly, then we can assume that the population has been stable. So without knowing the exact rate, we can consider population changes as an unlikely cause of the increase

Explanation 2 is easier to test, even when we only have absolute case counts, and no rates. Obviously, we first need to rule out differences in the size of the population at risk (see above). Then it helps to have historical data. By observing the fluctuations in the past, we can calculate the expected (usually the mean of the observed from the past) and the variance. This will help us determine how many standard deviations the observed value if from what we expected. To be pragmatic, if the observed value is more than 2 standard deviations from the expected, we consider this explanation 2 unlikely.

Explanation 3 is also independent of population size: we need to now the diagnostic and reporting system very well.

So even in absence of accurate denominator information, we should be able to make a reliable assessment if an increase in cases is due to 'a real outbreak' or due to one of the other three explanations.

In retrospect it is always easy to recognise an outbreak, or an increasing trend. Yet when you are confronted for the first time with an increase, it may be surprising how much doubt on the interpretation is around. Morabia tells us: [Around 1900, lung cancer was extremely rare. Its incidence seem to grow at a fast pace, but evidence did not convince everyone. It was argued that a better diagnosis and aging population could explain the trends. An editorial in the British Medical Journal in 1942 stated:

"It is doubtful whether the higher incidence of cancer of the long observed in recent years is real or only apparent."

It took the same journal 10 years to first comment that "few trends are more dramatic than the rise during the last 30 years in the notified death rates from cancer of the lung." In retrospect we all have 20/20 vision. The question is rather: how much of an open mind do we all have interpreting the present?

References:

• Alfredo Morabia (editor). A history of epidemiologic methods and concepts. Birkhauser Verlag, 2004. ISBN 3-7643-6618-7