AI-assisted Fall Prevention Through Evidence (SAFE)

Safe AI-assisted Fall Prevention Through Evidence

The goal of this multi-method study is to investigate how AI-assisted fall-prevention are implemented in routine hospital care what their effects are. The main questions it aims to answer are how these AI systems influence patient safety outcomes, how they affect healthcare professionals work and healthcare resource use, and what factors support or hinder their sustainable integration into hospital environments.

Study Overview

• Status: Recruiting
• Conditions: Fall
• Intervention / Treatment: Other: Not applicable- observational study

Detailed Description

Artificial intelligence (AI) offers new opportunities to strengthen patient safety, particularly in preventing in-hospital falls through real-time, sensor-based monitoring and alerts. As hospitals across Europe begin adopting these proactive fall-prevention technologies, evidence on their routine implementation and impact remains limited. The Safe AI assisted Fall Prevention through Evidence (SAFE) project aims to address this gap by examining the large-scale introduction of an AI-assisted fall prevention system in hospitals within the Västra Götaland Region (VGR), Sweden. Conducted between 2026 and 2028, the multicentre, multimethod project involves collaboration between Halmstad University and VGR hospitals, encompassing up to 2,400 patient beds. Using a multi-method design including surveys, interviews, observations, and a retrospective study, the project will follow the implementation process and evaluate effects on patient safety, healthcare workflows, and resource use multiple sites. Additionally, two learning labs will engage patients, relatives, and healthcare professionals to co-develop strategies that support sustainable system integration. The project will generate evidence-based insights and practical guidance for implementing AI-assisted fall prevention, with relevance for healthcare professionals, patients, hospital managers, and policymakers. While centred on VGR, the findings will offer valuable lessons for future initiatives in Sweden and internationally, contributing to the broader evidence base needed for responsible and scalable use of AI in healthcare fall prevention.

• Study Type: Observational
• Enrollment (Estimated): 23425

Contacts and Locations

• Study Contact: Name: Project leader; Phone Number: +46706924613; Email: elin.siira@hh.se

Study Locations

• Sweden
  • Gothenburg, Sweden
    • Recruiting
    • Sahlgrenska Hospital
    • Contact: Project leader; Phone Number: +46706924613; Email: elin.siira@hh.se

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

Key actors in the implementation, managers, staff, patients and patient representatives

Description

Individual interviews with key actors in the implementation

Inclusion criteria:

1. Be employed at one of the participating hospitals
2. Hold a role as a key stakeholder in the implementation work
3. Have experience with the implementation of the AI-assisted fall prevention
4. Have the ability to understand and communicate in Swedish

Exclusion criteria:

1. Have insufficient proficiency in Swedish to participate in an interview or observation and to understand the purpose and content of the study

Individual interviews with managers

Inclusion criteria:

1. Be employed as a manager at one of the participating hospitals
2. Have experience with the implementation or use of the AI-assisted fall prevention
3. Have the ability to understand and communicate in Swedish

Exclusion criteria:

1. Have insufficient proficiency in Swedish to participate in an interview or observation and to understand the purpose and content of the study

Individual interviews with staff

Inclusion criteria:

1. Be employed as staff on a ward at one of the participating hospitals where the AI-assisted fall prevention has been decided to be implemented
2. Have experience with the implementation or use of the AI-assisted fall prevention
3. Have the ability to understand and communicate in Swedish

Exclusion criteria:

1. Have insufficient proficiency in Swedish to participate in an interview or observation and to understand the purpose and content of the study

Observations of staff work

Inclusion criteria:

1. Be employed as staff on a ward at one of the participating hospitals where the AI-assisted fall prevention has been decided to be implemented
2. Have experience with the implementation or use of the AI-assisted fall prevention
3. Have the ability to understand and communicate in Swedish

Exclusion criteria:

1. Have insufficient proficiency in Swedish to participate in an interview or observation and to understand the purpose and content of the study

Individual interviews with patients and family members:

Inclusion criteria (for family members, only inclusion criterion 1b applies):

1. Either (a) have been a patient on a ward at one of the participating hospitals where the AI-assisted fall prevention has been implemented, or (b) be a family member involved in the care of a patient who meets the inclusion criteria but lacks the ability to provide informed consent. The family member must be able to understand and communicate in Swedish
2. Have experience with the AI-assisted fall prevention as part of their care
3. Have the ability to understand and communicate in Swedish
4. Be 18 years of age or older
5. Have the ability to provide informed consent. If there is any uncertainty regarding a patient's ability to provide informed consent, the research team will refrain from conducting the interview.

Exclusion criteria:

1. Have insufficient proficiency in Swedish to participate in an interview or observation and to understand the purpose and content of the study
2. Lack the ability to provide informed consent

Web-based surveys with staff:

Inclusion criteria:

1. Be employed as staff on a ward at one of the participating hospitals where the AI-assisted fall prevention is decided to be implemented.

Exclusion criteria:

1. Not being employed on a ward at one of the participating hospitals where the AI-assisted fall prevention is decided to be implemented
2. Lack the ability to access information and respond to the survey in Swedish

Retrospective medical record data:

Inclusion criteria:

1. Patients who have been cared for on a ward at one of the participating hospitals where the AI-assisted fall prevention has been implemented, either (a) up to 24 months after implementation or (b) up to 12 months before implementation.

Exclusion criteria:

1. Patients who have only received care outside the defined time period, meaning not within 24 months after or 12 months before the implementation of the AI-assisted fall prevention

Learning labs:

Inclusion criteria:

1. (a) Be employed at one of the participating hospitals and hold a role as a key stakeholder in the implementation work, or be staff or a manager with experience of the implementation or use of the AI-assisted fall prevention; or (b1) have been a patient cared for on a ward at one of the participating hospitals where the AI-assisted fall prevention has been implemented and have experience with the AI-assisted fall prevention as part of their care; or (b2) be a family member of a patient with such experience; or (c) be a patient representative for a patient group, patient organization, or user organization where falls are an identified issue
2. Have the ability to understand and communicate in Swedish

Exclusion criteria:

1. Have insufficient proficiency in Swedish to participate in an interview or observation and to understand the purpose and content of the study
2. Lack the ability to provide informed consent

Study Plan

How is the study designed?

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Time Frame
Fall rate	From earliest January 2025 to latest December 2028

Other Outcome Measures

Outcome Measure	Measure Description	Time Frame
NoMAD instrument	The NoMAD (Normalization Measure Development) instrument is a 23-item survey used to assess how complex interventions become implemented, embedded, and integrated into everyday healthcare practice. Grounded in Normalization Process Theory (NPT), it focuses on the collective work staff perform to make a new method part of routine care. Twenty of the items map onto the four core NPT constructs: coherence, which concerns how staff understand and make sense of the new practice; cognitive participation, which captures the work required to engage and involve people; collective action, which reflects the practical efforts needed to enact the intervention; and reflexive monitoring, which assesses how staff appraise its effects once it is in use. Three additional general items measure how "normal" the practice currently feels to users. NoMAD has demonstrated strong validity and reliability.	Baseline, 12 months follow up, 24 months follow up
COPSOQ III	The Copenhagen Psychosocial Questionnaire (COPSOQ) III is an instrument designed to measure, assess, and manage the psychosocial work environment. The Swedish standard version of the tool evaluates 33 different dimensions of work life through 76 items, covering categories such as quantitative and emotional demands, social support, leadership quality, and health outcomes like stress and burnout. Validated at both individual and workplace levels, the instrument demonstrates strong reliability and construct validity, making it a robust tool for identifying psychosocial risks and evaluating organizational interventions. A key feature of its application in Sweden is the establishment of population-based benchmarks, which provide reference values that allow organizations to interpret their specific results in relation to the national average.	Baseline, 12 months follow up, 24 months follow up
BAT	The Burnout Assessment Tool (BAT) is a validated, theory-based instrument for measuring work-related burnout. It assesses four core dimensions; exhaustion, mental distance, cognitive impairment, and emotional impairment, using 23 items supported by Rasch analysis, allowing the subscales to form one reliable overall score. The BAT performs consistently across demographic groups and enables conversion of ordinal scores into interval-level metrics, making it a precise tool for identifying burnout risk in both clinical and organizational settings.	Baseline, 12 months follow up, 24 months follow up
Resource use		January 2026- December 2028

Collaborators and Investigators

• Sponsor: Halmstad University
• Investigators: Principal Investigator: Elin Siira, PhD, Halmstad University

Study record dates

Study Major Dates

• Study Start (Actual): January 1, 2026
• Primary Completion (Estimated): December 1, 2028
• Study Completion (Estimated): December 1, 2028

Study Registration Dates

• First Submitted: March 13, 2026
• First Submitted That Met QC Criteria: March 25, 2026
• First Posted (Actual): March 31, 2026

Study Record Updates

• Last Update Posted (Actual): March 31, 2026
• Last Update Submitted That Met QC Criteria: March 25, 2026
• Last Verified: March 1, 2026

More Information

Terms related to this study

• Keywords: effects; artificial intelligence; implementation; patient safety; hospital; work; fall prevention; resource use
• Other Study ID Numbers: 2025-03701-01

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No