In this epidemic with a novel virus, it was of great importance to rapidly collect as much accurate information as possible in order to characterize the public health threat and support the health authorities in its management. This national double-level hospital-based surveillance was implemented very rapidly after the first COVID-19 patients were hospitalized and revealed to be crucial. The two approaches are complementary and serve different needs, for different end-users.
First, SC survey data allows a daily follow-up of Belgian hospital capacity. This close follow-up of the daily capacity enables rapid interventions regarding the potential saturation of local hospital networks. The HTSC Committee, together with the federal hygiene inspectors, regulates transportation of patients based on the individual situation within the hospitals. The close follow-up also allows to regulate the restart of elective activities within the hospitals. The occupancy rate of ICU beds is considered as the most important indicator during the different phases of the lockdown de-escalation plan. The ICU bed capacity for COVID-19 patients is regulated by formalizing an ICU bed surge capacity. Further, the data on ICU, IV and ECMO are used for logistic purposes, e.g. supply management by the Federal Agency for Medicines and Health Products. For the purpose of hospital capacity monitoring, it is very important that patients suspected of having COVID-19, in addition to those confirmed, are also registered, since they also occupy beds in the isolation units. Despite the fact that the SC survey was originally intended for operational purposes, the collected data allows the monitoring of epidemiological trends in hospitals as well. The surveillance of COVID-19 mortality takes advantage of the mandatory nature of the SC survey for the reporting of hospital deaths due to COVID-19. The daily reporting by all Belgian hospitals, ensures exhaustiveness of in-hospital deaths due to COVID-19.
The individual patient data obtained through the Clinical survey allows to study the different in-hospital care trajectories and lengths of stay, as well as insights into risk groups and patient outcomes. Specific research questions could be addressed such as defining priority groups for vaccination, assessing the effectiveness of treatment regimens (e.g. hydroxychloroquine), characterizing certain patient groups (e.g. pediatrics), etc. Furthermore, supplementary ad hoc analyses are performed to respond to questions raised by health authorities such as proportions, characteristics and outcomes of patients in ICU and ECMO, proportion of non-Belgian residents among hospitalized patients, proportions and outcomes of hospitalized healthcare workers and residents in nursing homes.
Between March 14th (date of the first epidemiological bulletin) and June 30th, 94 daily epidemiological bulletins were published in which daily trends of key indicators (e.g. number of hospitalizations, number of admissions, and number of discharges) obtained through the SC survey were presented. Moreover, during the same period, additional information on hospitalized patients’ characteristics and risk factors for hospitalization and death obtained through the Clinical survey were presented in 12 weekly bulletins. These two different bulletins aimed to inform both authorities and the general public about the evolution of the epidemic and are made available on the dedicated COVID-19 Sciensano web page (https://covid-19.sciensano.be/fr/covid-19-situation-epidemiologique). During this period, there were more than 600 000 downloads of the daily bulletin. Since the SC survey underwent several changes over time, the presentation of data in the daily bulletin was adapted accordingly, which was sometimes a challenge, as these changes were not always easily understood by the general public. In order to tackle this, a FAQ on the surveillance systems and the way of reporting its data is also made available. An additional thematic report on the highlights of this double surveillance in hospitals was published recently [17]. Supplementary aggregated information at national and provincial level is available on the EpiStat website [18]. Further, the data from both surveys are used for external scientific efforts such as modelling different short and long term scenarios in the context of the exit strategy.
Next to serving needs at the national, regional, or provincial level, the collected data over time proved to be useful at hospital level as well. Individual hospitals consult the statistics to better assess the epidemiological situation in their province and use them for their own planning. Findings related to clinical and care aspects are provided to the hospital practitioners to support them in the management of the COVID-19 patients. Exports of individual hospital data from the SC survey is often requested for internal evaluations. Specific reports for hospitals on the Clinical survey results were distributed every 2 to 3 weeks from the beginning of April, with the aim of informing caregivers about the clinical characteristics and hospital evolution of COVID-19 patients an supporting them in their management. From the 29th of April, the data recorded by each hospital is transferred back to them after data cleaning and management, via the eHealthbox, to be used for hospitals’ own analyses and statistics. This return of information to the data providers partly compensates for the heavy registration burden.
These data are also intended to be used for scientific publications by multidisciplinary teams in national and international projects. A Belgian collaborative group on hospital COVID-19 surveillance, including representatives of the most actively reporting hospitals in the Clinical survey and scientists from Sciensano, was created with the aim to establish a forum to discuss about scientific aspects of the surveillance and to exchange views on topics of interest that could be further analyzed.
This surveillance has both strengths and limitations. A major strength is the rapidity of implementation of both surveys. Another major advantage is the exhaustiveness of the SC survey thanks to the mandatory participation of the general hospitals as supported by a legal framework. This allows a complete and timely overview on the situation in the hospitals throughout the country. Although not mandatory, the Clinical survey is strongly recommended and this resulted in a high participation rate thanks to the engagement of the hospitals. Nevertheless, we observed missing admission or discharge report forms for a sizeable proportion of patients in the Clinical survey, but the absence of difference for most baseline characteristics for subgroups with missing data, comforted us in the reliability and generalizability of the findings. The comparison of the mortality data reported in the Clinical and the SC surveys confirmed representativeness of the Clinical survey for baseline demographic characteristics of the deceased patients. But a very slightly higher mortality was estimated in the Clinical survey, particularly among those with only discharge information, maybe due to a tendency to report more those deceased given the well-established practice of mandatory reporting in hospitals.
A specific limitation of the SC survey is that there is no clear view on the number of CT-scan-confirmed patients, as they are registered together with the possible cases. Only at a later stage, a distinction was made concerning the new admissions. Another limitation of the SC survey is that the daily number of new lab-confirmed COVID-19 hospitalizations are underestimated. The exceeding coverage of the Clinic survey with regard to the SC survey in week 12 and 13 indicates misclassification of lab-confirmed cases in the SC survey at the start of the epidemic in Belgium. Indeed, test results are not always available at the moment of reporting. COVID-19 patients newly admitted to the hospitals but with pending test results are reported in the category ‘CT-confirmed/suspected’ of the survey. When the test result turns out positive, these patients are reclassified and included in the prevalence number of lab-confirmed COVID-19 patients in the hospital, leading to a discrepancy between the incidence and prevalence figures.
A limitation of both surveillances is the continuous changing of variables which impacted the profile of reported patients over time. Important information was not captured from the start, such as the number of COVID-19 patients coming from nursing homes, distinguishing between real COVID-19 cases and COVID-19 positive patients diagnosed through screening, ethnicity, etc. However, the flexibility by which the data collection was adapted to incorporate such variables and promptly react to evolving aspects of the epidemic, modifications of the case definition, and changes in testing strategies is a clear strength of both surveys. Similarly, additional variables were added to enable the investigation of specific research questions, such as the addition of questions more specific for ICU and pediatric patients. We are conscious of the fact that the reporting is a supplementary burden for hospitals already under pressure during the COVID-19 crisis. Particular attention was devoted to ensure a trade-off between information gain (adding variables to the survey) and assuring compliance. The changes in case definition were particularly challenging for data providers to accurately classify their patients. With a similar objective of facilitating access to reporting, an access code was not set up at the initiation of the surveys but later on when it appeared necessary to ensure the integrity of the encoded responses and when the reporting habit was already well established. Clear communication on the changes and support was available throughout the process and widely used to help data providers report in a correct way. All analyses and interpretations of results have taken into account the timing of survey modifications.