In The Netherlands, medical physicians and microbiological laboratories are required, by law, to notify the Gemeentelijke Geneeskundige Dienst (GGD, municipal health services) of patients diagnosed with notifiable infectious diseases. The GGD are the regional authorities responsible for receiving preliminary notifications so that immediate control measures can be initiated. The GGD are required by law to send summaries of these reports as soon as possible to the Chief Medical Officer (CMO) at the Inspectorate of Healthcare (IGZ). There is voluntary reporting of surveillance data to the National Institute for Public Health and the Environment (RIVM). Before 2002, reporting from the GGD was paper-based and involved two different processes for reporting to IGZ and RIVM.

The internet-based reporting system OSIRIS, developed in the RIVM, was introduced in the Netherlands in 2002. Therefore, at regional level, as a result of this web-based system, mandatory and voluntary reporting (to IGZ and RIVM) merged into a single process. By December 2002 all 38 GGD in the Netherlands used the internet as the sole means of notifying infectious diseases to the CMO at IGZ and the RIVM. Physicians and laboratory staff continued to use paper, fax and e-mail to send their notifications to the GGD [FIGURE 1].

Authorised users at the GGD, IGZ and RIVM have password-protected access to the system. OSIRIS makes preliminary reports available to both the IGZ and RIVM for early warning of significant adverse events. However, the GGD can continually update information until the report is finalised.

Methods
We compared diseases reported by the conventional paper-based system for 2001 with diseases reported by OSIRIS for 2003. The study was confined to diseases with a minimum of 100 cases reported for each study year (tuberculosis notifications were excluded from the analysis, as the data collection logistics for this disease are substantially different from other notifiable conditions).

To determine the timeliness of the surveillance systems, three separate time points were defined. T1 was defined as the first day of illness as entered into common fields in both the conventional reporting system and OSIRIS. T2 was defined as the date that illness was reported to the GGD. T3 was defined as the date that illness was first reported to the IGZ/RIVM. Two distinct types of delays were compared in both systems [FIGURE 2]. Total delay was defined as the time lapsed between the onset of symptoms and reporting of illness at a national level: T3- T1. Central delay was defined as the difference between T3 and T2 and represented how much sooner or later the electronic system identified notifiable diseases than the paper-based system. If a date required for calculation of a specific delay was missing only that specific delay (and not the total case) was excluded from analysis. To increase the validity of our results we corrected the data, where appropriate, for digit attraction. The presence of digit attraction was confirmed by analysing illness onset/notifications by frequency table of calendar date of onset (i.e.1-31). Records with a calendar date of onset/notification that occurred more frequently than the expected average were excluded from further analysis.

Median delays were calculated and expressed with an interquartile range. Median delays between both systems were compared using the Wilcoxon Rank Sum-Test. Also, electronic reports were compared with those received through the conventional reporting system for completeness of specific data fields. For our study, completeness was defined as the proportion of selected data fields completed in each surveillance system. This analysis was confined to five selected conditions: legionellosis, bacillary dysentery, hepatitis A, pertussis and malaria. These diseases were selected for data quality evaluation as they represented different categories of notifiable diseases in the Netherlands: vaccine preventable diseases, enteric infection, respiratory infection, laboratory-notified infection and travel-associated infection. The Fisher exact test and the Mantel-Haenszel test with Yates correction was used to determine the significance of two proportions. Data was analysed using Epi Info ™ version 6.04c, SAS version 8.2 and MS Excel 97 ®.

Results
Nine diseases with more than 100 cases reported in 2001 and 2003 were included in the study: bacillary dysentery, hepatitis A, hepatitis B, hepatitis C, legionellosis, malaria, meningococcal disease, pertussis and foodborne outbreaks.

Digit attraction was only evident for first day of illness (T1). Thus, we corrected total delay, for digit attraction (T3-T1). We excluded all cases with illness date of onset on 1,5,10,15,20,25 and 30 as these dates were more frequently recorded than expected if illness onset was equally likely on all days. Correction for digit attraction resulted in a decrease in the estimated total delay (T3-T1) for all person-based infections in 2001 and 2003. (There was no correction for digit attraction for hepatitis B and hepatitis C as less than one in five patients with these illnesses had a recorded date illness onset).

Between 2001 and 2003 the central delay for all nine diseases was significantly reduced
[FIGURE 3]. Overall, the central delay was reduced from a median value of 10 days (interquartile range 4) in 2001 to 1 day (interquartile range 1) in 2003. Except for malaria, the total delay (T3-T1 ) was also significantly reduced for diseases studied

Electronic reports contained more complete information on variables common to both conventional and electronic reporting formats. In 26 of 36 data fields studied, those completed electronically contained significantly more information (p<0.05). Overall, in 2003, 91.3% of examined data fields were complete in comparison with 82.3% in 2001 [TABLE 2].

Discussion
To our knowledge, this is the first report comparing electronic and conventional reporting of infectious disease surveillance data on a national basis. Electronic reports were received at the national level significantly quicker than conventional reports for the nine diseases studied. This improved timeliness was not detrimental to data quality as electronic reports also contained more complete information than conventional reports. Similar results have previously been reported for electronically notifiable disease reporting from clinical laboratories [3,4].

The improved timeliness was almost exclusively due to the reduction in reporting delay between the GGD and the national authorities. This reduced reporting delay can be attributed to OSIRIS as there was no other major change in work practices at GGD level that would have resulted in a reduced local reporting delay (T2-T1). In fact, using this system lead to an estimated 50% reductions in administrative workload in relation to reporting infectious diseases at GGD level [5]. Correction for digit attraction resulted in a reduction in the estimated total delay for bacillary dysentery, hepatitis A, legionellosis, malaria, meningococcal disease and pertussis in both study periods. This suggests that some patients tend to overestimate the time period during which they are ill by ‘rounding-up’ to the nearest convenient date. While correcting for this phenomenon is impractical in routine practice, time intervals should be measured in a consistent way to allow comparison between different outbreak detection reports and surveillance systems [6].

The noted improvement in data quality is also important as this availability of more complete information should enable national authorities to respond in a more timely and appropriate manner. While we only selected 7-8 data fields per disease as indicators of data quality the general superiority of electronic reports suggests that improved completeness is also likely in unexamined data fields.

A potential concern in comparisons such as this is variation in coding between the fields in the electronic and paper-based systems. However, in this study as we only selected variables that were equivalent on the hardcopy and the electronic surveillance forms, direct comparability was ensured. Also, before the introduction of the electronic system staff training, technical assistance was provided at local level to ensure any data entry and coding problems were identified and managed appropriately [5]. Another potential concern is that the relative benefits of electronic reporting in this study could be secondary to deterioration in the conventional system. As the transition from conventional to electronic reporting occurred mid-year in 2002 and we selected only years when one system functioned at GGD level, a decline in the conventional working process could not explain the improved reporting times in 2003. In addition, the consistency of our results for all nine conditions suggests that the improved reporting times are real.

OSIRIS has achieved its objectives. Data received at national level is more timely and of better quality than with conventional reporting. However, the primary purpose of surveillance is not merely speedy and complete transmission of data. Technologically innovative reporting systems, as OSIRIS, also need to be consistent with the purpose of disease reporting, that is, of translating information into action [1,7]. Thus, it must be a two-way communication process of information exchange between public health agencies and the clinical community. Even in this technologically advanced age, observations made by astute clinicians still remain important, in timely reporting of certain notifiable diseases [8]. In these instances, electronic surveillance systems help us verify suspicions of outbreaks as was recently observed in the Netherlands when action was taken as a result of the observed increased notifications of hepatitis A cases. This action was due to a combination of clinical observation and national notification by OSIRIS [9,10].

This study documented improved timeliness and completeness of national infectious disease surveillance data that has occurred as a result of the use of electronic communication.