key: cord-346908-21hahh03 authors: Fan, Shihe; Blair, Corinne; Brown, Angela; Gabos, Stephan; Honish, Lance; Hughes, Trina; Jaipaul, Joy; Johnson, Marcia; Lo, Eric; Lubchenko, Anna; Mashinter, Laura; Meurer, David P.; Nardelli, Vanessa; Predy, Gerry; Shewchuk, Liz; Sosin, Daniel; Wicentowich, Bryan; Talbot, James title: A Multi-function Public Health Surveillance System and the Lessons Learned in Its Development: The Alberta Real Time Syndromic Surveillance Net date: 2010-11-01 journal: Canadian Journal of Public Health DOI: 10.1007/bf03403963 sha: doc_id: 346908 cord_uid: 21hahh03 Objective: We describe a centralized automated multi-function detection and reporting system for public health surveillance–the Alberta Real Time Syndromic Surveillance Net (ARTSSN). This improves upon traditional paper-based systems which are often fragmented, limited by incomplete data collection and inadequate analytical capacity, and incapable of providing timely information for public health action. Methods: ARTSSN concurrently analyzes multiple electronic data sources in real time to describe results in tables, charts and maps. Detected anomalies are immediately disseminated via alerts to decision-makers for action. Results: ARTSSN provides richly integrated information on a variety of health conditions for early detection of and prompt action on abnormal events such as clusters, outbreaks and trends. Examples of such health conditions include chronic and communicable disease, injury and environmentmediated adverse incidents. Discussion: Key advantages of ARTSSN over traditional paper-based methods are its timeliness, comprehensiveness and automation. Public health surveillance of communicable disease, injury, environmental hazard exposure and chronic disease now occurs in a single system in real time year round. Examples are given to demonstrate the public health value of this system, particularly during Pandemic (H1N1) 2009. Zone. The Alberta Real Time Syndromic Surveillance Net (ARTSSN) was developed with three goals: 1) improve upon traditional paperbased, fragmented public health surveillance using a centralized automated system; 2) enhance routine public health surveillance through effective use of existing Provincial Electronic Health Record data for earlier detection of cases, clusters, outbreaks and trends of communicable disease, injury, and environmental hazard exposure; and 3) track the effectiveness of public health interventions. This paper describes ARTSSN and the lessons learned in its development as a real-world application of syndromic surveillance. ARTSSN is a modular system 14 consisting of multiple real time Health Level 7 (HL7) data feeds from operational systems and selective data from the Provincial Electronic Health Record repository Alberta Netcare, a supporting data repository, a provincial Enterprise Master Patient Index, a terminology translation service and a user interface. In its first phase of development, ARTSSN obtained a privacy impact assessment approval for four streams of electronic data for residents of the Edmonton Zone: 1) telephone calls to Health Link Alberta; 2) emergency visits; 3) laboratory tests; and 4) school absenteeism (Table 1) . Health Link Alberta (HL) provides a 24/7/365 province-wide service for health advice and information. Registered nurses use a decision support tool of protocols to assess calls; each protocol consists of detailed algorithms for triaging symptom-based chief complaints (CC) and providing advice and information for managing symptoms and directing patients to appropriate health care. At the time of writing, HL had 210 protocols; ARTSSN used 133. Six hospitals and three urgent care centres provide emergency visits data through an Emergency Department (ED) Information System (EDIS) 15 or an electronic triage information system (eTRIAGE). 16 Online data forms Syndromic, Varying with specific forms As required Diagnostic * Average time from a nurse answering a call at Health Link Alberta to the time when the call record is seen by an ARTSSN user is 27.9 (SD = 7.0) minutes (30 tests on 3 different days). Note that the nurse may spend a variable amount of time answering the call and ARTSSN receives the record only after the call is closed. The data transmission time is therefore substantially less than indicated. † Average time from a visit being triaged at an emergency department to the time when the record is seen by an ARTSSN user is 15.7 (SD = 4.6) minutes (30 tests on 3 different days). ‡ Not tested. The proprietary data repository is a scalable Oracle™ database. Data elements (see Table 1 ) are first extracted from the feed databases, transformed, cleansed, checked for completeness and reorganized. Personal identifiers are replaced with meaningless pseudo-keys. The repository also receives data manually entered into electronic forms through an online survey module of the user interface. Population data for the Edmonton Zone, as derived from the provincial health insurance registry and broken down by gender, age group and geographic area, are contained in the repository for rate calculation. The terminology translation service uses look-up tables for syntactic and semantic translation. It unifies terms within and between data sources and translates CCs (ED data) and protocols (HL data) to a standard set of syndromes and "health concerns", and translates school absenteeism to a standard set of reasons for absence. Each health concern consists of similar syndromes (physiologically or anatomically) and is named accordingly, e.g., all cardiovascular syndromes are grouped under the cardiovascular heading. The user interface was developed by Voxiva (Washington, DC). It is intranet-based and secured behind a firewall for authorized uses only, and has submodules for 1) data analysis and visualization, 2) automated notification, and 3) system administration. The data analysis and visualization submodule can present results in tables, charts or maps by person, place, time, CC, discharge disposition, laboratory test and result, or disease agent either by count or by rate. These products can then be saved for future reference, exported for reporting, or placed on online dashboards for real-time surveillance. Line listings of data can also be exported for off-line analyses or scientific research. The automated notification submodule continuously compares incoming data to pre-set thresholds for the diseases or syndromes under surveillance by person, place and time according to algorithms configured within a series of notification templates. The algorithms are based either on rate or count, depending on the diseases or syndromes. If a threshold is surpassed, prompt notification is automatically dispatched to designated personnel via means specified in their notification subscription. For instance, 43 alerts of school absenteeism due to illness ≥10% were sent to the ARTSSN epidemiologist on November 5, 2009 during Wave 2 of pH1N1 2009. This automated notification improves traditional surveillance markedly since a very large number of conditions can now be effectively surveyed by many combinations of person, place and time. This degree of examination would be extremely laborious and expensive using traditional methods. The system administration module controls, among other functions, access to the system, data sources, and functions of the system. Users are first grouped into a role matrix according to their responsibility and institutional affiliation. A profile of privileges is built for each role; users sharing a common role are further controlled by manipulating their privileges, such as being limited to only a single data source or reduced system functions. A novel feature of ARTSSN is the patient re-identification mechanism. Under the Public Health Act in Alberta, notifiable diseases must be reported to the local public health department for investigation. Consequently, a drill-down function was built in the system for authorized users such as MOH, communicable disease control nurses or environmental health officers, which allows them to reidentify individuals using the Enterprise Master Patient Index via the pseudo-key. ARTSSN is multi-functional and has many potential uses in public health surveillance. Figure 1 shows its use in pandemic surveillance, 17 using ED visitation data during pH1N1 2009 to inform the timely opening of a mass influenza assessment centre and then evaluating its impact on relieving pressure in the EDs. Figure 2 demonstrates use for injury prevention. In January and February of 2009, the local public health department used media assistance to advise the public of treacherous road conditions and precautions required when driving or going outdoors. Injury prevention is one of the five key areas recommended for improving the health of Canadian children. 18 From 2005 to 2009, approximately one third of ED visits by young people were injury-related (Table 2) . ARTSSN followed four design principles: automated, real-time, routinely useful and locally useful. Automation minimizes interruptions to real-time surveillance even during the most resource-challenging conditions like pH1N1 2009 (Figure 1 ). Each ED visit or HL call about influenza-like illness (ILI) or laboratoryconfirmed pH1N1 case is automatically acquired by the data repository and updated on online dashboards for timely understanding of the pandemic progressing in the community. The information is then relayed to Emergency Operation Centres to inform their pandemic responses. Real-time information also enables ED managers to promptly re-allocate patient care and infection prevention and control resources based on ED patient flow, punctually intervene in screening ILI patients and significantly improve ILI data quality. The pH1N1 2009 experience demonstrates that ARTSSN is truly useful to both the data user and provider. 19 Real-time surveillance reduces recall bias and improves interpretation of statistical associations since immediate temporal, spatial and population connections are possible as events are unfolding ( Figure 2 ). Real-time surveillance in ARTSSN gives public health practitioners a new tool to survey injury, environmental hazard exposure, chronic and communicable disease comprehensively for improved and targeted health protection and promotion. Being routinely useful enables ARTSSN to gain wide user support; 31 users (groups), including data providers, MOH, communicable disease and environmental health epidemiologists in the Edmonton Zone, as well as all MOH in the province, now have real-time information for decision-making. Although biological, 6,7 nuclear or radioactive terrorist attacks are all potential threats to public health, such attacks are rare compared to the frequency of infectious disease outbreaks, unintentional injuries or severe weather-related events. A system solely dedicated to small probability events may be unsustainable over time at the local level. Being locally useful 20 means that ARTSSN carefully balances the needs of both the data provider and data user. Strategies included rapid feedback about the quality, meaning and limitation of the data; help in defining the requirements and functions of ARTSSN; and joint efforts in the development of surveillance products. Evaluation was done in parallel with system development by independent evaluators for simplicity, acceptability, utility and other aspects following established frameworks for evaluation of public health surveillance systems. 21, 22 Issues or concerns raised were assessed and resolved promptly. Consequently, the user interface is more intuitive; new users can become proficient with two to three hours of instruction. ARTSSN was developed through rapid successions of data collection, analysis, interpretation and dissemination in iterative "plan, do, check, and act" cycles, avoiding past practice of sequentially optimizing each individual step before moving to the next. For example, HL data were first extracted into the repository; line listings, graphs, tables and maps were then developed using other software before the user interface was ready. Feedback about these prototypes was incorporated into the next development cycle. These rapid, iterative cycles uncovered deficiencies earlier, gained user acceptance faster, and obtained support more broadly, allowing ARTSSN to be implemented more Relationship between daily emergency department visits for injuries (a) and daily maximum/minimum temperatures (b) from January 1 to February 12, 2009 During the specified data period, unusual warming periods (daily maximum temperatures above zero) intermingled with rain and light-snow events. Each warming episode melted the packed snow surface in daytime and refroze it at night, which caused icy road conditions and increased emergency visits for injuries. One such warming period (between 1/15 and 1/22) is highlighted in the figure. Daily injury visits a b Rain Snow swiftly (<2 years from launching to operation) than some legacy surveillance systems. The data repository is critical to ARTSSN. It facilitates data-flow from the source to the user interface without interfering with the stability and functionality of the operational databases. Data pseudonymization simplifies privacy impact assessment approval for the system and ethics approval for research use of data. This central repository approach differs from other distributed systems where the data reside with their originator. [23] [24] [25] A key advantage to ARTSSN is that the database is a centralized model within a single health organization, making the repository development and maintenance easier. This centralization significantly improves public health surveillance through the multi-functions of ARTSSN by avoiding fragmentation that has plagued traditional surveillance systems. The Pandemic (H1N1) 2009 showed that automated data acquisition in ARTSSN could be disrupted if the source capacity was exceeded 19 (see Figure 1 ). For instance, the repository was unable to receive call data when Health Link Alberta was forced to streamline call recording using paper instead of its time-intensive electronic system. Moreover, unlike other systems, 14, 26, 27 ARTSSN lacks statistical modeling capability. Thresholds for the syndromes and diseases under surveillance require individual programming based on historic data. Difficulties remain in simultaneously deciphering aberration signals from multiple data streams and in determining the syntactic and semantic relationship in free-texted medical records in some new databases as ARTSSN is expanding from a regional to a provincial system. These issues are being addressed by working with data providers and by collaborating with university researchers on natural language processing and statistical analysis. A retrospective study on clinical features of and treatment methods for 77 severe cases of SARS Severe acute respiratory syndrome epidemic in Asia Identification of severe acute respiratory syndrome in Canada Emergence of a novel swine-origin influenza A (H1N1) virus in humans Transcript of statement by Margaret Chan, Director-General of the World Health Organization. Geneva: WHO Secondary aerosolization of viable Bacillus anthracis spores in a contaminated US senate Office Forensic application of microbiological culture analysis to identify mail intentionally contaminated with Bacillus anthracis spores Syndromic surveillance: The case for skillful investment Implementing syndromic surveillance: A practical guide informed by the early experience Syndromic surveillance systems A massive outbreak in Milwaukee of cryptosporidium infection transmitted through the public water supply A fatal waterborne disease epidemic in Walkerton, Ontario: Comparison with other waterborne outbreaks in the developed world Syndromic surveillance of gastrointestinal illness using pharmacy over-the-counter sales AEGIS: A robust and scalable real-time public health surveillance system Grafstein E for the Canadian Emergency Department Information System (CEDIS) working group. A Consensus-based process to define standard national data elements for a Canadian emergency department information system Emergency triage: Comparing a novel computer triage program with standard triage Are we ready for pandemic influenza? Researching for the top: A report by the advisor on healthy children and youth. Ottawa, ON: Health Canada Operation of a real-time syndromic surveillance system during Pandemic (H1N1) 2009. CCDR (accepted) Signals come and go: Syndromic surveillance and styles of biosecurity Public Health Agency of Canada. Framework and Tools for Evaluating Health Surveillance Systems Framework for Evaluating Public Health Surveillance Systems for Early Detection of Outbreaks Information system architectures for syndromic surveillance Distributed data processing for public health surveillance Propagation of program control: A tool for distributed disease surveillance A systems overview of the electronic surveillance system for the early notification of community-based epidemics (ESSENCE II) Technical description of RODS: A real-time public health surveillance system Objectif : Nous décrivons l'Alberta Real Time Syndromic Surveillance Net (ARTSSN), un système multifonction, centralisé et automatisé de détection et de production de rapports de surveillance de la santé publique. Il s'agit d'une amélioration par rapport aux anciens systèmes sur papier, souvent fragmentaires, limités par la collecte de données incomplètes, par des capacités d'analyse insuffisantes et par l'impossibilité de fournir de l'information en temps utile pour les interventions de santé publique.Méthode : L'ARTSSN analyse simultanément plusieurs sources de données électroniques en temps réel et en présente les résultats sous forme de tableaux, de diagrammes et de cartes. Les anomalies décelées sont immédiatement communiquées aux décideurs pour qu'ils puissent agir.