Table 2 Needs of large-scale automated healthcare facility surveillance and potential of federated systems
From: Federated systems for automated infection surveillance: a perspective
Current barriers precluding large-scale automated surveillance | Needs for large-scale automated infection surveillance in healthcare facilities | Potential benefits of federated systems for surveillance | Potential downsides of federated systems for surveillance | Prerequisites for federated systems for surveillance |
|---|---|---|---|---|
Case-definitions are not always suitable for fully AS | Changes in definitions, centrally implemented semi-AS, or more complex case detection methods | Enables automated case detection in healthcare facilities by (complex( algorithms, supporting a broader range of definitions that may better align usefulness for local practitioners | Adaptability of algorithms to local situation more difficult to realize | Transparency and explainability of algorithms for case detection; sound methods for algorithm development and validation |
Differences in registration procedures of routine care data limits comparability of case definitions across centers and over time | Uniformity when reusing EHR data (FAIR) ensures consistency of case-definition | Improved comparability of surveillance results due to harmonization of source data and improved insight in coding practices | Loss of information with data transformation; Knowledge and IT capacity for EHR data harmonization | Stakeholder agreement on a minimal data set, data model and terminology; Willingness and capacity for FAIR data curation; Robust measures for quality control |
Limited flexibility in information needs or granularity of existing national registries currently demands additional infrastructures or additional manual data collection (SARI) | Less resource intensive AS systems that are agile to meet current and possibly changing information needs | Sustainable systems that meet information needs for various purposes and topical issues by performance of various analyses on harmonized data sources at different time points, and different level of details | Methodology may not be accepted if data ownership and good governance are well arranged | Regulatory framework for data access, handling and publication of results; Transparency of methods |
Limited data linkage, due to limitations in interoperability or privacy restrictions | Authorized access to more detailed machine-readable data, allowing for less resource-intensive and more widely implemented surveillance systems | Potentially increased coverage rate of healthcare facilities (more complete information of catchment area); Privacy by design tailored to local situation; Methodology analyzing free text may be supported | Methodology may not be accepted unless validated and well understood; Lack of access to individual patient-level data on a central level; Data on rare events or combination of traits can make identifying individuals a risk, new information security issues or risk of hacking may not be excluded | Regulatory framework for data access, handling and publication of results; transparency in technical applications (freely accessible source code and pseudo code); Technical solution needs to fit the local needs and (regulatory) requirements; Thorough method validation |
Insufficient capacity of IPC and IT in healthcare facilities for development, implementation, and maintenance of AS systems | Reduced workload for development, implementation, and maintenance of AS systems; Interoperable machine readable EHR data allowing exchange of algorithms and scripts | Surveillance systems programmed and maintained at one location; Easily scalable solution with interoperable, machine-readable data | Software ETL processes necessary every healthcare facility; accountability more complicated; IT capacity at the expense of IP capacity if the importance and shift in IP tasks is not recognized | Technical applications need to be transparent (open source); Governance framework addressing accountability, enhanced collaboration |
Insufficiently broad and thorough knowledge on AS in all participating healthcare facilities hinders development or procurement of methodologically sound and sustainable AS systems | Central support in the development of methodologically sound and sustainable AS systems; easier applicable, not requiring complete knowledge from AS by a large multidisciplinary team within the healthcare facilities | Programme of requirements of a sustainable AS system can be determined collectively within federated community; healthcare facilities are not dependent on individual agreements with vendors | Knowledge in all participating healthcare facilities is required to understand, accept and validate outcomes of federated AS. Potential loss of view on local events without manual assessment | Digital literacy and minimal training in data intelligence for understanding of (maintenance of) algorithms and interpretation of outcomes, for decision making; supervision and oversight of AS methodology by qualified professionals |
External validation of locally implemented AS systems difficult (HAI) | Source codes and data handling procedures transparent and easy to review | Surveillance algorithms programmed, validated and maintained at one location | Continuous quality monitoring, maintenance and version control to be performed by coordinating center | Framework for validation; transparency in data processing through whole pipeline from curation to publication of results for quality assurance, and avoidance of gaming in surveillance |