Page 50 - Timeliness of Infectious Disease Notification & Response Systems - Corien Swaan

48 Chapter 2
creased by one day (statistically significant only for shigellosis). Apparently, a one day decrease in the Pd is not significant enough to result in a difference in the means of the total time delay. Nonetheless, the median values and per- centages of infections reported within the incubation period do show a clear improvement by direct reporting.
Interestingly, the use of e-mail by reporting physicians or laboratories does not seem to accelerate the process, largely due to security problems.
Some countries have experienced improvement of timeliness through di- rect and automated electronic laboratory reporting (ELR) [20-24]. Such systems allow physicians to report over the internet (using web forms), and data can even be imported from the laboratory computer systems, eliminating manu- al data entry. Reporting methods and procedures should be optimised in the Netherlands, for example, by developing more secure e-mail approaches, ad- justed web forms for reporting, or automated ELR.
In this study we chose to compare median incubation periods with median notification times as a measure of timeliness. However, this comparison has two major shortcomings. One is that incubation time measures the time between infection and appearance of symptoms, but not the time to infectiousness and further transmission. With diseases in which onward transmission occurs before symptoms appear, the latent period is a better measure of timeliness. Therefore we introduced the period Ic for HAV infection and measles as measure of time- liness instead of the incubation period. It follows that the optimal way to define timeliness would be to base it on the generation interval, i.e. the time between infection of the index case and infection of its secondary case. However, as the generation interval is difficult to observe the serial interval is usually used, i.e. the expected time between symptom onset in the index case and symptom on- set in its secondary case [11,17].
The second problem in measuring timeliness is that intervals for notifica- tion and latent periods are not fixed but distributed, possibly with a large vari- ance.In Figure 3, it is readily noticeable that for shigellosis, a disease with a very short incubation time (low variance), the distribution of the reporting time lies far beyond this incubation interval and has a large variance. In contrast, for HAV infection the curves are more alike, showing a large proportion of cases within the incubation period.
A comprehensive analysis requires taking into account the distribution of notification time intervals and comparing it to the distribution of generation times. We are currently working on an analysis of timeliness based on the distri- butional properties of the relevant time intervals.