Page 206 - Timeliness of Infectious Disease Notification & Response Systems - Corien Swaan

204 Chapter 9
timeliness of notification systems. Most authors use medians, either with or without interquartile ranges, but many studies provide means as shown in sup- portive information 1 table (Chapter 3). As delays are commonly not equally distributed but skewed to the right, the average tends to be higher than the median, overestimating delays and underestimating timeliness. The number of notification systems fulfilling the different timeframes will therefore in reality be higher.
Secondly, using only medians to determine timeliness, as done by us in Chapter 4 and several studies in Chapter 2, results in an overestimation of time- liness; with only half of the notifications being timely, the timeframe is achieved. Applying a threshold of 80% for sufficient timeliness for medians of groups of diseases or reporting units, this theoretically can be achieved by only 40% timely notifications. In order to measure timeliness according to a timeframe, percent- age of notifications in time should be used. Furthermore, Yoo et al. recommend to measure and examine the complete distribution of the respective delays in a notification chain, as the entire shape of the distribution may provide important information on differences of delays between diseases (10). We found that the proportion of infections caused by the index case and secondary cases at time of notification depends mainly on medians and not on the standard deviations of the delay distribution (Chapter 4). However, disease specific comparison of the proportion of patient, doctor and laboratory delay in these distributions can generate hypotheses for causes, and recommendations for improvement. We recommend using medians in measuring notification delays, and cumulative percentages for measuring reaching thresholds for timeliness.
We developed a disease specific outbreak control timeframe for delay from disease onset until notification at the MHS (Chapter 4), for 6 person-to-person transmissible diseases. Although it addresses disease specific characteristics as speed of developing new cases and reproduction number, it also has limitations. The model assumes that control measures by the MHS are immediately imple- mented at the moment the index case is notified, which is difficult to achieve. This implies that even tighter outbreak control timeframes are necessary for timely measures to prevent outbreaks. While the model provides timeframes for outbreak control, nowadays MHS are expected to prevent any further re- lated cases. Notifications within the outbreak control timeframe might provide the MHS more time to prepare larger interventions as hepatitis A vaccination campaigns at schools, direct response to prevent any further cases once a case is notified is still necessary.