In this work, we present SmICS as a digital supporting tool for infection control tasks such as monitoring of bacteria, contact networks and epidemiological analysis. Nosocomial outbreaks caused by different bacterial and viral pathogens, e.g., MDRBs²⁸ or the severe acute respiratory syndrome coronavirus 2, are one of the major challenges in IPC²⁹. Depending on the kind of pathogen and source, transmission dynamics vary greatly, necessitating a broad and deep IPC expertise in timely recognition to implement the best suitable intervention measures aiming at reducing the spread and controlling the outbreak. Smart software solutions which support IPCTs in their daily work are very promising and can help them to anticipate and better understand potential transmission chains in the light of massive amounts of available data³⁰. Tasks such as contact tracing and sophisticated epidemiological analysis of MDRBs at the ward level are particularly time-consuming when using the traditional manual approach. Especially against the background of the shortage of skilled labor and the difficult recruitment of new IPCT members, digital supporting systems that release time capacities of IPCTs needed for verification of transmission events, identifying potential sources, choosing the most appropriate intervention strategies and finally evaluating their success are of special value. This is particularly true in times of pandemic, which dramatically highlights the increasing workload pressures in the healthcare sector³¹. Our digital solution to this challenge, SmICS, provides multiple functions for efficient support in routine infection control tasks. For this purpose, microbiological data and movement data are combined as essential data sources and automatically analyzed in a suitable manner. Using historical, real-world data sets, the functionality of SmICS was successfully evaluated on the basis of three practical IPC tasks. Our analysis showed that (depending on the site and the tools available locally from routine care), by using SmICS, routine infection control tasks can be carried out efficiently, resulting in an average time saving of 39.2 min (95%CI [24.0-54.4]) for all defined tasks and individual savings of up to 68.5 min (95%CI [30.5-106.5]) for one specific task. All participants were able to complete the tasks without errors.

Furthermore, SmICS is fully openEHR-based and provides open interfaces to allow secondary use of data, enhancement of algorithms, sharing of results and reuse of the application across institutions. The use of openEHR templates, reviewed and accepted by domain experts, helps to provide information about which data is required for the flawless use of the SmICS, e.g., by using cardinalities on an archetype and element level. On top of that, openEHR models permit automated constraint and plausibility checks during the storage of data, thereby ensuring syntactic and semantic correctness to a higher extent. Nonetheless, the quality and completeness of data captured in routine clinical settings can vary considerably, influenced by differences in documentation practices, data entry errors, and incomplete records. There still is the risk that such variability may affect the reliability, validity, and generalizability of real-world data-based tools. Since we fully rely on openEHR-based data, no such technical variability has been seen in our SmICS evaluation. To evaluate data quality and variability effects, an enhanced analysis of, on the one hand, the extraction, transformation and loading processes in the medical data integration centers (MeDICs) and, on the other, the original documentation processes are potential mitigation strategies. Furthermore, given that our SmICS is conceptualized as a fully routine data-based approach, it is important to note that utilization is constrained to the available data. For instance, data pertaining to healthcare workers may be of significant interest in the context of infection control. However, the collection of the requisite data with the necessary level of detail for such purposes (e.g., by means of tracking devices) is subject to strict legal regulations, thus, such data is not widely available in routine data sets. In addition, more extensive forms of contact tracing—such as those conducted during the COVID-19 pandemic, which also included non-hospital settings (e.g., nursing homes)—are currently not implemented in SmICS. If data on patient presence in functional areas are available in the primary data systems (e.g., as part of the movement data), they could, in principle, be integrated into SmICS.

The interoperable and openEHR-based design of SmICS has enabled us to roll out the application in a timely manner to various German university hospitals with different primary source system infrastructure landscapes. Our interoperable concept makes future cross-location analysis and sharing concepts possible, which will become particularly important in an increasingly interconnected healthcare system that calls for quick collection and analysis of data, particularly in times of crisis.

Alleviating the burden of daily IPC workload by means of innovative digital solutions has been the subject of research before. In 2024, Arzilli et al. published a scoping review of articles from 2018 to 2023, presenting an overview of new practical technologies to support traditional infection control and surveillance work in real settings. The use of different health informatics technologies, natural language processing, digital health/e-health/m-health applications, mobile computing, and the reuse of electronic health record data has been recently investigated, with machine learning-based approaches emerging as the predominant area of research³². Related, the authors also acknowledge the need for both in-depth expertise and large volumes of sensitive data to be available in order to be able to train and apply these machine learning-based approaches in real settings. Moreover, the majority of research work focuses on the timely detection of specific infections or general surveillance statistics³². Digital, low-barrier, routine data-based approaches for infection control support in multidrug-resistant bacteria as in the sense of digital contact networks or tracing, as anticipated by SmICS, are scarce. One related example, however, focusing specifically on the retrospective contact network analysis of Vancomycin-resistant Enterococcus transmissions, has been researched by Neumann et al. in 2020³³. They point out the advantages of infectious-contact networks and the potentials of also considering negatively tested patients to reveal intra-hospital pathways of spreading and to support infection prevention planning³³. Although delivering a visual representation for analysis, their objective has not been to implement a practical application, as was achieved with SmICS. However, it should be noted that Neumann et al. further incorporated genomic analyzes, which could prove to be a valuable prospective feature to be integrated into SmICS.

Obviously, in the Corona pandemic, related approaches focusing virologic pathogens have emerged. For example, in the Hospital Clínic de Barcelona, a surveillance system called CoSy-19, was developed to interrupt intra-hospital transmission of infections and enabling timely implementation of interventions, by using contract tracing and epidemic curves, amongst other³⁴. Another line of research of significant importance within the domain of infection control pertains to the detection of clusters and outbreaks, a function not yet integrated into SmICS. In 2009, the Brigham & Woman’s Hospital in Boston, Massachusetts, conducted a retrospective study on microbiology data from 2002 to 2006 with a tool called WHONET-SaTScan. This is an automated cluster detection tool based on two freely available software packages: SaTScan, a software developed for geographical disease surveillance, and the WHONET/BacLink, a software for descriptive analysis of microbiology data¹³. Another exemplary cluster alert system, named CLAR, was implemented in the Charité, a tertiary care hospital in Germany. This system uses six algorithms for detecting potential outbreaks on a ward. In 2019, a prospective study was conducted, in which the system generated automated alerts, 35% of which were classified as relevant alerts requiring further investigation³⁰.

In contrary to the presented related work, the idea of SmICS is to have a software that serves as an IPC program fully operating on an interoperability standard. To the authors’ knowledge, utilization of clinical routine data in a digital application through an interoperable, standardized approach with interactive visualization interfaces that is not restricted to a specific pathogen is a distinctive feature of the present study.

SmICS was fully implemented and evaluated at only three of the participating sites. The technical implementation proved to be challenging because there were difficulties in integrating microbiological findings and movement data into the local MeDICs due to strict data protection issues or limited access to proprietary primary clinical systems. An identified issue with SmICS is its technical performance, resulting in long waiting times for the users and, thus, dissatisfaction of the users with the system’s usability. These waiting times depended on the amount of data which were stored within the MeDIC data repository and how many resources were allocated to the data repository, in the form of RAM and CPU, and if the data repository was set up directly or by using a dockerized version. Due to the different infrastructural strategies of each participating MeDIC, waiting times were not considered in the evaluation, as they were not due to the waiting times of the core software, but to delays caused by the underlying MeDIC data repository. Prior to the start of the study, it was not recognized that waiting times would exert such a significant effect across all sites. Consequently, the study protocol did not include the measurement of delays, thus constituting a limitation in study design. Meanwhile, the MeDIC infrastructures underwent changes, so it is not possible to reproduce the waiting times. Basically, it is important to note that there are simply a lot of data points to process, resulting in long query response times. These loading problems have also been reported in other related work (such as the WHONET-SaTScan). Of course, data integration optimizations or the inclusion of further indexing structures are currently underway to solve this major issue. For example, the new version of EHRbase, which is an open-source database implementation of openEHR, now allows indexing at the level of each data element path, which can result in improved query performance³⁵. Another solution could be to overcome live querying by implementing a user-modifiable daily morning batch load, as reported in Stachel et al.¹⁴.

These waiting times also seem to be the main reason for the System Usability Score of 51.6 (“OK’ according to Bangor et al.²⁷). 14 out of 15 participants scored the SUS lower than 71.4 points (“good” according to Bangor et al.²⁷). It is evident that SmICS, in light of the technical issues that have been identified, did not attain a satisfactory level of usability. The performance issues were also mentioned in the answers to the open-ended questions. Additionally, in contrast to the quantitative time efficiency results, subjective usability-based feeling of time-efficiency was rather low: the second statement of the cSUS (“Effective support for this software is hard to access in a clinically-appropriate timescale”) resulted in an average of 3.5 (5 referring to “Agree strongly”, meaning not time efficient). However, it has to be mentioned that contact tracing with extensive movement data and huge amounts of lab data in hospitals with ~60,000 inpatients per year is a data-intensive, demanding task, and—considering the limited human IPCT resources—may be even impossible without computational support.

The technical platform of SmICS offers the possibility of developing into an advanced, interoperable, open-source infection control assistance suite. Important functions such as contact chains and the tracing of potential transmission events—to be verified or falsified by the IPC team later on—and the performance of smaller routine epidemiological tasks (for instance creating a ward-based epidemiological curve for a specific pathogen based on raw data not verified to belong to the event yet) are integrated. Another important task that IPCTs are involved in on an ongoing basis for a significant proportion of their working time is surveillance for certain bacterial and viral species, sometimes based on resistance patterns. This is a legal requirement for hospitals in many countries, including Germany (German Infection Protection Act). These requirements also cover a broad range of different, complex ways to define healthcare-associated infections (HCAI), with the standard parameters being infection rates and incidence densities. One entity of HCAI with a high impact for patients, healthcare workers, hospitals and the community are bloodstream infections. In this context, venous catheter (line) associated bloodstream infections (CLABSI) serve as a worldwide accepted benchmark parameter. In addition to this well-established, but laborious to define, parameter, there has been increasing discussion among IPC experts as to whether hospital-onset bacteremia (HOB) is also a suitable measure for surveillance^36,37. In a subsequent medical informatics, multi-site project (RISK PRINCIPE: risk prediction for IPC)³⁸, we will address automated, digital surveillance and prediction of HOB as another perspective of an infection control support system.

Furthermore, in the future it is planned to integrate a new outbreak detection algorithm from our group, which is on par with or better than used state-of the art approaches³⁹. Based on weekly case counts from microbiological data, the algorithm generates alarms to inform the user about potential outbreaks for pathogens and wards which are being investigated. The algorithm models the observed number of cases using a hidden Markov model combined with a statistical background model. As the architectural setup in Fig. 1 hints, the outbreak detection algorithm is already planned to be incorporated into the SmICS. An integration of this component into SmICS is waiting for the final phase of the validation. As a next step in translation into routine medical practice, certification as a medical device needs to be assessed. In this way, SmICS’s potential to significantly accelerate the completion of routine tasks that must be performed regularly comes into effect. The IPCT’s capacity that is freed up can then be reinvested into advanced tasks, such as in-depth clarifications, analyzes, and practical actions.

In this article, we present the design, implementation and evaluation of an open-source, openEHR-based application for supporting automated monitoring of bacteria and transmission events and associated epidemiological tasks (called SmICS). Despite the demonstrated feasibility and benefits of developing a standard-based, openEHR application for the IPC domain, broader adoption of open standards and interfaces by software manufacturers is rare. This reluctance is often driven by business strategies favoring proprietary solutions, as well as skepticism regarding the maturity, functionality, and practical value of open data models and associated technologies. Through our work, we aim to showcase the potential and added value of open, interoperable systems for IPC while also openly acknowledging current challenges, such as infrastructural variability impacting usability. Importantly, we also contribute to the broader community by making our newly created, standardized openEHR data models freely available through the Clinical Knowledge Manager, fostering transparency, reuse, and further development within the community. In our work, initial evaluation in an efficiency and usability study based on real-life examples demonstrated that, by using SmICS, some of the most time-consuming basic IPC tasks can be successfully managed in a timely and efficient manner, but SmICS still needs major improvement in terms of usability. Query performance due to varying site-specific infrastructural implementations of the underlying data repository was identified as a remaining issue. When solved, the potential time savings with SmICS will allow IPCTs to use their capacity to focus on verification of outbreaks, advanced analysis and expert-driven intervention and outbreak control tasks in future scenarios. The openEHR-based and open design of SmICS allows roll out, reusability and cross-institutional sharing as well as community-driven enhancements.