This study was conducted at Zhejiang Provincial Hospital of Chinese Medicine, a general tertiary hospital located in Hangzhou, Zhejiang Province, China. Applying FMEA to assess risks in preantibiotic therapy microbiological specimen submission, formulate targeted interventions, and evaluate their effectiveness. A complete checklist is available in Checklist 1. The implementation steps are as follows (Figure 1):

A multidisciplinary expert team was established to ensure comprehensive identification of workflow risks across clinical, laboratory, managerial, and informatics domains. The team consisted of 9 members, including clinical physicians, Disease Prevention and Control Experts, infection prevention and control (IPC) professionals, microbiology laboratory staff, senior nurses, and health informatics engineers. Team members were selected based on predefined criteria: (1) at least 5 years of professional experience in their respective fields; (2) direct involvement in antimicrobial stewardship, specimen management, or infection surveillance activities; (3) familiarity with routine workflows and known bottlenecks within the hospital; and (4) willingness to participate in structured meetings, independent FMEA scoring, and consensus discussions. This interdisciplinary composition ensured that all key steps of the specimen submission process, from clinical decision-making to specimen collection, laboratory processing, and information system operation, were adequately represented, providing a balanced and comprehensive perspective for risk analysis.

We first conducted a systematic review of the literature to identify common influencing factors in the microbiological specimen submission workflow (Multimedia Appendix 1). Commonly reported factors affecting microbiological specimen submission were extracted and categorized into four predefined domains: (1) health information systems factors, (2) personnel-related factors, (3) administrative and management factors, and (4) external support factors. This framework was used to guide subsequent discussions. Building on this framework, the multidisciplinary expert panel participated in 2 rounds of structured brainstorming sessions. During each session, experts were instructed to identify specific, concrete scenarios within each domain that could plausibly lead to specimen submission failure or delay. To enhance consistency, failure modes were defined as observable deviations from the expected workflow (eg, missed test ordering, delayed collection, incomplete data capture), rather than abstract causes or outcomes. All proposed failure modes were collated, merged, and refined by the research team. Redundant items were consolidated, and ambiguous descriptions were clarified through iterative discussion with panel members. Items were retained only if consensus was reached that the scenario was (1) operationally distinct, (2) locally relevant to routine practice in the study hospital, and (3) theoretically supported by either literature evidence or expert experience. A final set of 30 locally relevant failure modes was generated. These items formed the basis for subsequent FMEA scoring and prioritization.

Before formal scoring, all team members received standardized training on the FMEA scoring criteria provided by the research team. Each member independently evaluated the Severity (S), Occurrence (O), and Detection (D) of each potential failure mode. After all 9 members completed their independent assessments, a consensus meeting was convened to discuss discrepancies and clarify scoring decisions. Final S, O, and D scores for each failure mode were determined through group consensus. The RPN was calculated using the formula: RPN=S×O×D.

The scoring range for S, O, and D was 1‐10. Conventionally, corrective actions are mandated when the RPN exceeds 125. To address limitations of the RPN method, we introduced the Action Priority (AP) index. The determination of AP followed the standard methodology outlined in the Automotive Industry Action Group–Verband der Automobilindustrie FMEA guidelines. Unlike the RPN, the AP level is not calculated through a mathematical formula; instead, it is assigned by consulting the guideline’s AP lookup table, which specifies the corresponding action level based on different combinations of Severity (S), Occurrence (O), and Detection (D). As shown in Multimedia Appendix 2, AP levels are categorized as high, medium, or low. High AP indicates that corrective actions should be taken with priority, medium AP suggests that action is recommended but can be scheduled, and low AP indicates that existing controls are acceptable and require only routine monitoring. High-priority failure modes identified by the FMEA (based on both RPN and AP indices).

For each mode, the multidisciplinary team collaboratively designed corresponding digital health solutions. All digital interventions underwent iterative refinement through pilot testing in selected departments. Feedback from clinicians, nurses, infection control practitioners, and IT personnel was collected to optimize alert thresholds, interface design, workflow integration points, and data visualization formats.

The final intervention package consisted of three interrelated components:

After developing the intervention measures, we proceeded with the implementation phase following the standard quality improvement process. The preparatory stages of FMEA, including workflow mapping, identification and scoring of failure modes, prioritization, and intervention design, were completed between January 1 and March 31, 2024. Beginning on April 1, 2024, the FMEA-informed interventions were formally implemented. Physicians, nurses, and support staff received training on the updated specimen submission workflow and information system functions through the online learning platform. The effectiveness evaluation covered the period from January to December 2024. Improvements were assessed by comparing pre- and postintervention microbiological specimen submission rates.

Three core outcome indices were evaluated:

Specimen submission data from January 1, 2024, to December 31, 2024, was extracted from the Hospital Infection Surveillance System (Xinglin System). The dataset included antibiotic order time, drug category (restricted-use, special-use), and timing of administration relative to specimen submission. To ensure data quality and consistency, only complete records with clearly documented timestamps for both specimen submission and antibiotic prescribing were included in the analysis. Duplicate entries, canceled orders, and specimens rejected by the laboratory due to quality issues were excluded.

Statistical analyses were performed using SPSS (version 27.0; IBM Corp). Categorical variables were reported as counts (percentages). The Mann-Kendall trend test was used to examine the trend of submission rate over time. When the z score was positive, it meant that the composition ratio of each indicator showed an upward trend. When the z score was negative, it meant that the composition ratio of each indicator presented a downward trend. Statistical significance was set at 2-tailed P<.05.

This study was approved by the Ethics Committee of the First Affiliated Hospital of Zhejiang Chinese Medical University (2025-KLS-038-01). As this study used retrospective data, the ethics committee waived the requirement for participants to provide written informed consent. The research data does not involve individual patient information. No compensation was provided to participants as this study was a secondary analysis of preexisting data with no new patient recruitment or intervention. The research was conducted in accordance with the Declaration of Helsinki.

The top five failure modes limiting preantibiotic specimen submission were (1) PDA barcode scanning failures, (2) inadequate clinical decision support alerts, (3) insufficient clinician awareness of testing protocols, (4) suboptimal oversight mechanisms, and (5) patient-related health constraints (Table 1).

For the potentially high-risk failure modes, recommended interventions were identified with a view to reducing their occurrence or improving their detection (Table 2). An LMS-based online training was implemented. All personnel involved in specimen collection or antibiotic order entry were required to complete the module and pass a competency assessment. A score of at least 80% was considered a passing result. The LMS automatically recorded training completion and assessment outcomes; individuals who did not complete the module or did not achieve a passing score were required to undergo retraining before being allowed to independently perform the related tasks. A real-time CDSS was embedded into the antibiotic prescribing process.

When clinicians initiated orders for therapeutic antibiotics, particularly restricted-use or special-use agents, the CDSS automatically evaluated whether an appropriate microbiological specimen had been ordered. If noncompliance was detected, the system generated real-time alerts prompting clinicians to initiate specimen submission. This workflow enabled prospective intervention at the point of care, reducing missed submissions caused by time pressure or oversight. Supportive process-enhancement measures were integrated to reinforce execution and oversight. Enhanced PDA-based barcode scanning strengthened specimen traceability during collection and transport, reducing data loss related to scanning failures or rapid operations. In parallel, automated monitoring dashboards aggregated specimen submission compliance data by department and antibiotic category. Regular feedback reports were provided to clinical units and infection control teams, supporting targeted supervision and continuous workflow optimization.

Following the implementation of FMEA management in April 2024, significant improvements were observed in the overall specimen submission rate (Figure 2), the submission rate for restricted antibiotics (Figure 3), and the submission rate for special-use antibiotics (Figure 4). The Mann-Kendall trend test revealed that all 3 categories of specimen submission rates exhibited a statistically significant upward trend (P<.001).

Microbiological specimen submission plays a pivotal role in hospital infection control and promoting rational antibiotic use, representing a complex systemic endeavor [18,19]. This study implemented the internationally recognized risk assessment tool—FMEA, to conduct a multidimensional evaluation of the entire etiological specimen submission workflow, spanning information systems, health care personnel, hospital administration, and external support. Through expert panel discussions, the research identified multiple failure modes and prioritized targeted intervention strategies, providing scientifically grounded decision-making support for improving the quality of specimen submission practices. This study provides a replicable pathway and methodological framework for improving microbiological specimen submission workflows, and may indirectly support hospital infection prevention efforts by enhancing antimicrobial stewardship and strengthening infection surveillance.

The advancement of intelligent health care information systems has emerged as a critical driver for improving specimen submission rates [20]. FMEA-based investigation identified PDA barcode scanning failures and insufficient CDSS alerts as high-priority intervention targets. Mechanistically, real-time alert mechanisms within information systems can trigger specimen collection reminders before antimicrobial administration, thereby minimizing workflow delays caused by human oversight and ensuring precise temporal control of submissions [21]. Concurrently, integrated multisource data platforms, by consolidating heterogeneous data from EHRs, laboratory results, and other sources, enable dynamic decision-making models that assist clinicians in accurately evaluating the necessity of testing indications and the specificity of required assays [22]. These findings reveal a dual-driven mechanism of information system optimization: process monitoring reduces operational deviations, while data integration enhances diagnostic decision-making efficacy.

This study also revealed that inadequate clinician awareness of specimen submission constitutes a critical bottleneck in improving submission quality. To address this, the research team implemented a dual-track capacity-building framework, “theoretical training+practical supervision,” alongside intelligent system upgrades: standardized operational training was used to reinforce protocol adherence, while interdepartmental supervision mechanisms established closed-loop quality improvement cycles. Postintervention evaluation demonstrated statistically significant improvements in specimen submission rates (P<.001). Although the observational study design precludes causal inferences regarding direct relationships between interventions and outcome improvements, the synergistic decline across multidimensional surveillance metrics indirectly suggests potential systemic spillover effects of FMEA-based management in hospital infection prevention and antimicrobial stewardship.

Despite achieving improvements in nosocomial infection specimen submission processes and promoting rational antimicrobial use through hospital-wide informatization, several areas for development remain. Externally, budget constraints limit access to advanced yet costly testing modalities, hindering comprehensive testing. Variations in patient case mix create department-specific testing demands that cannot be uniformly addressed. Additionally, diagnosis-related group reimbursement constraints introduce cost-benefit trade-offs that indirectly influence testing decisions. Therefore, hospital management must continue refining collaborative mechanisms across clinical, technical, and administrative departments. Strategic initiatives should include optimizing test menu configurations through cost-effectiveness analysis, developing disease-specific testing guidelines, and leveraging informatization for continuous monitoring to enhance staff efficiency. These measures will ultimately strengthen AMR control, promote rational antibiotic use, improve surveillance capabilities for emerging pathogens, and drive health care quality toward higher precision and patient safety.

This study has several limitations that warrant consideration. First, some of the interventions relied on a relatively well-developed hospital information infrastructure, such as artificial intelligence-driven CDSS alerts. Institutions with limited digital capacity may need to adopt manual checklists or simplified reminder mechanisms as alternatives. Second, the study was conducted in a tertiary TCM hospital. Although the hospital includes Western medicine departments such as the intensive care unit, respiratory medicine, and emergency medicine, with antimicrobial use patterns comparable to general hospitals, the patient population and service model of TCM hospitals may differ from other health care settings, which may limit generalizability. Nevertheless, the methodological framework of FMEA is highly transferable, and individual institutions may adapt specific interventions according to their local resources and workflows. Additionally, this was a single-center, before and after study without a control group, which imposes inherent limitations on causal inference. Improvements in specimen submission rates may not be attributable solely to the FMEA-based interventions; they may also reflect the hospital’s broader emphasis on antimicrobial stewardship, underlying secular trends in workflow optimization, or concurrent quality improvement initiatives or policy changes. Future research should validate the effectiveness of these interventions using multicenter designs with appropriate control groups.

This study validated FMEA as a risk management tool for optimizing preantimicrobial specimen submission workflows. By addressing critical failures like barcode errors and inadequate decision support, FMEA-based interventions significantly improved specimen submission rates for restricted/special-use antibiotics, aligning with national quality standards. Challenges, including cost constraints and patient barriers, necessitate policy-level solutions. This model offers a scalable framework for antimicrobial stewardship, with future research needed to evaluate long-term outcomes and advanced diagnostic integration.