Demographic change in Germany and Europe is leading to a steadily growing number of people in need of care, especially older people with multiple health problems [1]. A growing problem in this age group is falls, which place considerable physical and psychological strain on both those affected and the care systems and increase the need for treatment [2,3]. The continuous scientific development of nursing processes is essential for improving the quality of care, promoting evidence-based knowledge, and further developing nursing practices [4,5].

The application of artificial intelligence (AI) provides new possibilities for nursing research by enabling the efficient analysis of large amounts of care data, recognizing patterns, and obtaining scientifically useful insights [6]. This strengthened nursing research contributes significantly to improving and facilitating nursing care, as scientific findings can be used to develop evidence-based care strategies [7]. In turn, data from nursing care form the basis for AI-supported analyses (Figure 1) [8]. The cyclical process between (AI-based) research, health care and nursing care, and AI innovations, with data serving as the central interface, is shown in Figure 1. Using the example of fall prevention, AI innovations can enhance research opportunities, such as studies on falls in long-term care (LTC). The findings from this research can then improve the quality of nursing care and help prevent falls. In nursing care, data are continuously generated, for example, when falls are documented, which feeds back into research and supports further AI-driven innovations in fall prevention. The challenge is to bring the data collected in the health care system into a uniform format that is accessible for research.

The secondary use of existing documentation data offers an alternative to traditional prospective studies for research purposes. However, accessing and harmonizing these data are challenging due to heterogeneous systems and varying documentation practices, which often result in unstructured or nonstandardized data.

Existing solutions, such as internationally recognized nursing terminologies or so-called standardized nursing languages, aim to standardize the documentation of nursing data by specifying certain terms, topics, and terminologies [9]. However, the use of standardized nursing terminology and specialist languages is not common across Europe. The use of structured nursing documentation varies widely across countries, and studies by Thoroddsen et al [10] and Dos Santos et al [11] suggest that nursing diagnoses and interventions are used most frequently.

The general use of standardized nursing terminology is still lacking, which makes access to nursing data more difficult [10]. A cross-sectional study from 2020 in the Netherlands revealed that approximately half of the 667 nurses and assistants surveyed would use standardized terminology in electronic documentation [12]. The reasons for this include burdensome documentation (use barrier) and a still predominant use of medical terminology (terminology barrier), although medical terminology is insufficient for describing nursing processes, such as assessing unsteadiness or unsuitable footwear. Other studies also show that complex documentation is inefficient in care practice, consumes resources necessary to perform care measures, and leads to a considerable documentation burden for nursing staff [11,13-16].

In the case of Germany, LTC facilities took a step away from standardization toward more patient-oriented documentation in 2015 with the introduction of a law to reduce bureaucracy in care documentation, which led to an increased use of unstructured free-text entries [17,18]. The absence of legal provisions governing the use of standardized terminology constitutes a structural barrier to the implementation of such concepts [19]. Furthermore, there is currently a paucity of terminology that has been specifically developed for German LTC, with the literature consisting of translations of foreign approaches. Nursing care systems are characterized by a high degree of heterogeneity in different countries, which reinforces the use barrier [20].

This paper addresses a research gap in German LTC caused by the lack of integration and routine use of nursing terminology. This situation gives rise to a paucity of structured and standardized data, while existing approaches to data use remain ineffective, resulting in the untapped potential of valuable data for research purposes remaining unused. Our approach aims to standardize the format of research data rather than the methods of data collection, allowing nursing staff to continue working as usual, while their collected data are converted for research purposes (Figure 1).

The following research questions are addressed:

The conceptual and technical development of an NMDS according to the definition by Werley et al [21] for the context of LTC is presented via the “fall prevention” use case, and its potential is discussed in terms of the challenges mentioned. In our use case, an analysis is conducted of data on risk factors, interventions, and outcomes of falls with the aim of preventing such incidents. In addition, a possible practical application in the context of the AI-supported analysis of nursing data is presented.

This paper presents a new approach to enable the secondary use of nursing data without additional effort in nursing practice. To this end, the following objectives are pursued: (1) outline the necessary steps for the development of an NMDS based on the “fall prevention” use case, guided by the literature and expert input; (2) describe the technical approach to systematically record fall risk factors, including the collection and prioritization of key factors; and (3) provide a perspective on an NMDS for data collection and harmonization, demonstrating its applicability for research and AI-based analysis.

Some related work must be accounted for when addressing the research objectives. The German research project PFLIP (Pflege-Kerndatensatz und Intersektorales Pflegedaten-Repository [Nursing Minimum Data Set and Intersectoral Nursing Data Repository]), on which the following results rely, aims to standardize previously collected nursing data to make the data available for research and AI applications and to offer the possibility of harmonizing data across sectors [22]. This project is based on the German Medical Informatics Initiative (MII), which has developed a medical core dataset on which the NMDS presented in this paper is based [22,23].

With respect to the scientific foundations of fall risk assessment, the “fall prevention expert standard” guidelines served as a foundation for the development of a use case. In Germany, the nursing guidelines that define best practices and quality benchmarks to ensure high standards of nursing care are called “expert standards” [2].

With respect to standardization and interoperability, the National Association of Statutory Health Insurance Physicians in Germany publishes relevant health care transition documents for interoperable data exchange in a telematic infrastructure. The work of this organization is used as a basis for the development of conceptual and technical use cases [24].

To enable standardized fall data to be used on a broader basis, a suitable data format must be chosen. The Health Level Seven International (HL7) Fast Healthcare Interoperability Resources (FHIR) standard is an internationally recognized framework for the exchange of health information, and its technical terminology is used by health care providers, software developers, and governments worldwide to improve interoperability between different systems and facilitate access to health data. The active community of FHIR includes professionals from various health care fields, who work together to further develop and implement the standard to optimize global health care.

This paper examines the development of an NMDS for LTCs using the example of the “fall prevention” use case. Building on initiatives such as the MII, German expert standards, and international standards such as HL7 FHIR, the NMDS aims to harmonize nursing data, integrate it across sectors, and make it usable for both practice and research, assuming that such an approach is feasible, relevant to practice, and sustainable to implement.

The methodology for achieving our research goals consists of an approach of conceptual standardization, which we combine with and translate into technical standardization. An overview of the procedure is presented in Figure 2.

In the next step, the interim results were translated into the technical terminology of the HL7 FHIR standard and made available for use in digital applications and AI-based systems.

This study did not require approval by an institutional review board, as it did not involve interventional or experimental research with human participants. The study aimed to collect opinions and assessments from nurses as experts regarding fictitious fall risks derived from the literature. All participants received written information about the purpose, scope, and procedures of the questionnaire. Participation was entirely voluntary, and informed consent was obtained prior to data collection. Participants were informed that their data would be processed anonymously. The study was conducted in accordance with the principles of the Declaration of Helsinki, the guidelines of the Council for International Organizations of Medical Sciences, the World Health Organization, the European General Data Protection Regulation, the German Federal Data Protection Act, and the Data Protection Act of the State of North Rhine-Westphalia (Landesdatenschutzgesetz NRW). As all survey data were fully anonymized and no information was collected that would allow the identification of individual participants, the study does not involve the processing of personal data as defined in Article 4(1) of the European General Data Protection Regulation. All data were stored on protected servers. All participating institutions approved the study as part of a data protection agreement, with the involvement of data protection experts.

The “fall prevention” use case was conceptually developed on the basis of an NMDS based on the MII core data (Multimedia Appendix 2).

The development of a German NMDS use case for fall prevention in LTC is based on a literature analysis of international guidelines and German practice documents, resulting in fall risk factors, interventions, and outcomes, and was developed in collaboration with 12 experts from various nursing fields. The relevant minimum items were prioritized by experts, reflected upon in a quantitative survey with 158 participants, and finally transferred into an FHIR-based IG.

The developed use case can benefit nursing research by making standardized, documented data available for secondary use and enabling AI-supported analyses. Since the dataset is based on routine documentation in nursing facilities and standardizes already documented data, it has the potential to support problems that have arisen thus far in efforts to standardize care documentation, such as a lack of manageability and the effort of documentation by nursing staff [13-16].

In contrast to broader international NMDS initiatives, the approach presented here takes a more detailed approach to a specific use case, resulting in practical, actionable information, better reflecting clinical variability, and creating a solid foundation for digital applications and AI.

In addition, the NMDS has been linked to the technical standardization HL7 FHIR from the outset, ensuring interoperability and direct technical implementability and guaranteeing connectivity to modern digital systems.

However, it should be noted that the dataset does not automatically ensure that documented data are also transferred to standards (such as FHIR). Data that are documented in other facilities and care documentation systems are not always automatically available in FHIR and may be in different formats. To counteract this limitation, a system solution, such as a data repository, is needed to automatically standardize routine data. The standardization of content and technology also facilitates the exchange of data between sectors and professions. In addition, the structural orientation toward an existing medical core dataset represents a step toward harmonizing nursing and medical data. This can support the need to give nursing terminology equivalent weight in documentation as medical terminology and integrate them instead of overwriting them [9,23].

Unlike previous approaches in LTC, which often failed to be implemented, the NMDS developed here is based exclusively on existing routine documentation to promote acceptance and long-term use in everyday nursing care without increasing the workload [34]. In accordance with extant recommendations in the literature, the development of NMDS should be undertaken in a multidisciplinary manner and integrated into (digital) documentation systems, with a focus on nursing content as opposed to medical content; thus, it should be oriented toward nursing standards, which we followed and built a base for [34,35].

The co-design approach chosen to develop the use case indicates a way of standardizing nursing data while maintaining a high degree of user centricity. This gives this NMDS an advantage over previous NMDSs and assessment instruments. In the future, a person-centered approach can be ensured by broadening the perspective to include affected persons, that is, residents of LTC facilities. The results in the NMDS table (Multimedia Appendix 5) illustrate the differences between risk factors that are perceived as relevant by practitioners and the scientific literature and those that are documented in a German geriatric care facility; this can indicate which important fall risk factors have not yet been mapped in routine documentation, such as “environmental factors” in this case. This information highlights the need for further research.

Although the results were collected nationally, findings from other European countries demonstrate difficulties in introducing standardized care documentation systems and standardization. This approach can serve as a model for the development of NMDSs for other countries [13-15].

There lies a challenge in bringing the data collected in the health care sector into a uniform format that is accessible for research. Once the data are accessible for research, AI innovations can benefit from a much larger data pool that is also very close to the point of care; this significantly improves the results of AI applications. The FHIR translation makes the data usable for AI applications that can assess and predict falls. The IG thus serves as a valuable resource for health care providers and researchers alike and provides a structured and standardized approach to capturing and analyzing fall risk factors. It is hoped that the future application of this guide will contribute to the development of more effective and personalized fall prevention strategies and ultimately improve patient outcomes. The developed IG also provides a methodological basis for expanding the minimum dataset to include other nursing topics in addition to falls as well as other care sectors and is internationally applicable.

An AI-based nursing decision support system (NDSS) for elderly people to predict falls in nursing homes can be used as an example of an application of the content and technology developed for our NMDS.

Developed in a user-centered design process based on International Organization for Standardization (ISO) 9241-210-2019, the NDSS covers routinely documented aspects of falls in German nursing homes and reflects the most relevant fall-related factors from the literature and practice [36]. Nursing staff with different levels of education and specialized training, administrative employees from an outpatient care facility and day care, and people who worked in the field of nursing science were asked about the practical suitability of the proposed NDSS in 4 workshops. Based on the workshop results, the NDSS was improved iteratively. The decision support system is designed to predict the risk of falling using a 5-step scale and focuses on the interpretability of AI so that predictions are comprehensible and transparent for users. Therefore, the NDSS can provide information about recognized risk factors that are responsible for corresponding fall risk. The proposed decision support system also suggests prophylaxis to reduce the risk of falling and show the progression of this risk over several months.

The results have several limitations. For reasons of feasibility, we focused on one use case. In this use case, we initially focused on a limited number of risk factors that were technically implemented. In the future, we plan to expand NMDS thematically. It must also be considered that there are different contexts, causes, and types of falls. Falls are presented in a highly standardized manner in this paper. In addition, to date, only aspects of inpatient LTC have been examined, but we plan to extend the findings to different settings.

It can be assumed that both the participants in the co-design workshops and the participants in the quantitative survey showed a certain level of interest in and affinity for the topic. As participation was voluntary, it is possible that less-interested participants did not take part due to participation bias.

Another potential bias in the quantitative study arises from the fact that participant evaluations were used as a criterion for item selection. Since documentation routines, levels of knowledge, and familiarity with existing evidence vary in everyday clinical practice, there is a risk that empirically relevant content will be underestimated or overlooked; this can lead to an overemphasis on subjective assessments, while evidence-based aspects remain potentially underrepresented. Moreover, we deliberately choose a user-centered approach to differentiate ourselves from previous solutions and ensure greater practical relevance. To reduce that bias, a broad survey covering a range of backgrounds and nursing staff was conducted. In the future, the perspectives of residents within the designated dataset should be incorporated. Furthermore, the participants’ ratings were based on international nursing standards and fall prevention guidelines and were not indiscriminately determined by the participants.

Nevertheless, it is believed that this approach is effective in enabling a user-centered methodology to be adopted, thus counteracting the lack of use and implementation of NMDS and nursing terminology and standards, as outlined in the problem statement in the introduction. It is important to validate the results in future studies.

The results are oriented toward national standards, practical procedures, and documentation in German LTC facilities. However, the procedure and methodology can be applied internationally.

The “fall prevention” use case developed for a German NMDS in LTC shows how nursing data can be recorded in a standardized and user-centric way. The use of previously documented routine data reduces the documentation burden for nursing staff and facilitates cross-sector data exchange. This concept creates a valuable basis for nursing research by providing high-quality, structured data for secondary use. It also offers the potential for international application and can serve as a model for the development of nursing datasets in other subject areas, particularly through FHIR translation for use in AI-based applications.

The IG and the profiles it contains are to be submitted for an HL7 FHIR balloting. This structured mechanism is designed to solicit feedback from members of the HL7 community to ensure quality and relevance. The suggestions for changes and improvements collected in this process can be evaluated during further efforts and, if approved, can be integrated into the IG to promote the joint creation and further development of the NMDS.

In conclusion, our approach shows that co-design makes user-centered and applicable NMDS development feasible. Moreover, FHIR enables interoperability. This work represents a preliminary step toward harmonizing medical and nursing care in Germany by aligning the data format with MII structures, with future efforts needed to fully integrate the nursing sector. This paper enables us to adapt the approach internationally for AI-ready nursing data.