Integrating narrative medicine into laboratory quality improvement: a sustainable model for reducing coagulation specimen rejections

Abstract

Background

Pre-analytical errors, particularly specimen rejections due to insufficient volume, represent a persistent challenge in coagulation testing, especially in oncology settings where patients often have compromised vascular access. Conventional quality improvement (QI) approaches frequently focus narrowly on procedural compliance and may inadvertently reinforce a punitive, blame-oriented culture between laboratory and nursing teams, thereby undermining systemic problem-solving. This study evaluated whether integrating narrative medicine principles into QI could reduce rejection rates while fostering a just culture grounded in empathy, shared accountability, and patient-centered learning.

Methods

A longitudinal QI study using iterative Plan-Do-Study-Act (PDSA) cycles was conducted at a large tertiary oncology hospital (Jan 2019–Nov 2025). Root-cause analysis identified a “culture of blame” as core drivers. A narrative medicine curriculum utilizing “parallel charts” to foster empathy and mutual understanding was implemented. Monthly coagulation specimen rejection rates were analyzed across three phases using non-parametric statistics and staff behavioral changes were monitored.

Results

Rejection rates decreased by 98.6%, from a median of 1.862% (baseline) to 0.026% (optimization phase; p < 0.001). This reduction remained stable during a major staff rotation, indicating the resilience of this system. Qualitative feedback indicated that narrative reflection rebuilt trust and created a “just culture” focused on patient safety rather than individual blame.

Conclusions

Embedding narrative medicine in laboratory QI is a rigorous, evidence-informed strategy. By shifting the focus from technical compliance to human-centered understanding, healthcare systems may better bridge the gap between the laboratory and the clinic, and potentially achieve breakthroughs in patient safety.

Supplementary Information

The online version contains supplementary material available at 10.1186/s43058-026-00937-8.

Introduction

The quality of clinical laboratory specimens is essential for diagnostic accuracy [1]. Pre-analytical errors account for up to 94.6% of all testing errors [2, 3], with insufficient volume being a leading cause of spurious abnormalities in coagulation and biochemical assays [4, 5]. This challenge is particularly acute in coagulation testing, which requires strict adherence to blood collection protocols (90%–110% fill volume) to ensure a precise anticoagulant-to-blood ratio of 1:9 [6, 7]. In practice, however, this standard often conflicts with clinical reality, particularly in oncology, where patients frequently have poor venous access and fragile health [8]. This creates a persistent gap between rigid technical requirements and clinical feasibility.

Prior to 2019, our institution addressed this issue with periodic, generalized training sessions based on standard phlebotomy manuals to address this issue. However, improvements were minimal and unsustainable. This suggested the root cause extended beyond knowledge gaps. The nature of the problem became clear during a single incident.

In early 2019, a nurse arrived at the laboratory, visibly frustrated, holding a coagulation tube. “Why was this rejected?” she asked, pointing to the sample. “There’s clearly enough blood.” From the laboratory technologist’s perspective, the volume was clearly insufficient, risking a falsely elevated activated partial thromboplastin time (APTT) result. What followed was not just an explanation of the 1:9 ratio, but a moment of shared understanding. The nurse explained the clinical reality: “Our patients are often very ill. We are sometimes hesitant to draw too much blood, and their veins are so poor that a successful draw is difficult.” This interaction revealed that the issue was not merely technical ignorance, but a communication breakdown and a lack of shared context between the bedside and the bench. It was a clash of two narratives: the laboratory’s story of precision and the nurse’s story of patient vulnerability.

That moment was our “aha!” —it signaled that sustainable improvement required shifting from a blame culture, where rejections are seen as individual failures, to a just culture, where they become opportunities for system learning [9, 10]. We hypothesized that achieving this shift necessitated moving beyond punitive enforcement to mutual understanding, retaining technical rigor while honoring clinical reality. Narrative medicine offered the framework for this transformation. By cultivating narrative competence: the ability to absorb, interpret, and be moved by stories, clinicians and laboratory staff can bridge the gap between evidence and experience [11–15]. Though widely applied in patient care and nursing [16], its use in resolving cross-departmental, system-level QI challenges, especially in the pre-analytical phase, remains unexplored.

Therefore, this study evaluated a narrative medicine-based intervention to bridge the communication gap between laboratory and clinical staff, reduce coagulation specimen rejection rates, and establish a sustainable model for pre-analytical quality improvement (QI).

Methods

Setting and context

This QI project was conducted from January 2019 to November 2025 at a large tertiary oncology hospital in Hunan, China. It was a cross-departmental initiative co-led by the Department of Clinical Laboratory and the Nursing Office, with formal engagement from the Department of Nursing, and the Department of Materials Management. The scope was limited to the pre-analytical phase of coagulation testing for adult oncology patients across all inpatient and outpatient units. This study was reported in accordance with the SQUIRE 2.0 (Standards for Quality Improvement Reporting Excellence) guidelines [17].

Prior to the intervention, the institutional context reflected a fragmented accountability structure. Specimen rejections were treated as individual failures rather than system errors. The reporting process was unidirectional: the laboratory issued “Rejected” notices via the Laboratory Information System (LIS) without explanatory rationale or feedback channels. This structure reinforced a blame culture, in which nurses perceived rejections as punitive, while laboratory staff perceived clinical units as dismissive of analytical validity. This adversarial baseline was the explicit target of cultural transformation.

Measurable aims and outcomes

The primary aim was to reduce the monthly rejection rate of coagulation specimens due to insufficient volume. The specific, measurable, achievable, relevant, and time-bound (SMART) goal was: To decrease the average monthly rejection rate from the 2019 baseline of 1.88% to below 0.5% within one year, and to sustain it at this level for at least three years. The primary outcome was the monthly coagulation specimen rejection rate, calculated as (Number of Rejected Specimens/Total Number of Specimens) * 100%. Data were retrospectively extracted from the LIS; all specimens were from routine clinical testing (no additional collection).

Rationale for the intervention: from blame to just culture

After root-cause analysis confirmed communication breakdown as the root cause of specimen rejections (Introduction), the QI team systematically evaluated intervention options. Prior didactic training on phlebotomy technique yielded only transient behavioral improvements—confirming that knowledge transfer alone was insufficient. We concluded that sustainable change required cultural intervention.

Narrative medicine was selected as the theoretical framework, grounded in three evidence-based principles: attentive listening, recognition of the other’s narrative (whether patient, nurse, or technologist), and reflection on one’s own assumptions. We posited that narrative competence, specifically the ability to understand the colleague’s perspective, was the critical mediator required to shift from a blame culture to a just culture.

The intervention: a narrative medicine-based, four-step QI cycle

We designed and implemented a structured intervention integrating the Attention-Understanding-Reflection-Response framework of narrative medicine [18] —a model endorsed in the Chinese Expert Consensus on Narrative Medicine in Clinical Practice—with the Plan-Do-Study-Act (PDSA) cycle [19].

Plan (attention & understanding)

Narrative Diagnosis: We framed the index incident (the “aha!” moment) as a systems failure. The core tool was parallel charts: reflective, first-person narratives written in non-technical language to explore the human dimensions of a clinical event [20, 21]. All charts were strictly anonymized (no patient identifiers, staff names, or unit specifics) and used solely for internal QI training—ensuring psychological safety and eliminating legal or professional liability. An anonymized example from a laboratory technologist:

“A nurse stood before me, eyes weary, holding the tube. She wasn’t just questioning a number; she was defending her patient’s comfort and her own competence. In that tube, I didn’t just see insufficient volume. I saw fragile veins, a clinical dilemma between pain and protocol, and a chasm between our worlds. My role was not just to reject, but to bridge.”

This reflective writing served two critical functions: First, it fostered empathy by forcing laboratory staff to articulate the nurse’s moral distress, and vice versa; Second, it revealed a hidden system flaw: the standard 3 mL collection tube was a significant burden on frail oncology patients, a fact obscured by rigid protocol adherence. While routine parallel charting for every rejection was impractical in a high-volume setting, this initial exercise transformed the incident into a shared educational text, defining the problem as a communication gap, not non-compliance. It subsequently helped us diagnose the root causes across four key relationships (Table 1).

Table 1.

Narrative diagnosis of a specimen rejection incident using the “Four relationships”

Relationship Type	Example of Recorded Content	Guideline Reference [18]
Physician-Nurse Relationship	Nurses’ stress due to performance evaluation linked to blood collection failure rates	Recommendation 11
Physician–Patient Relationship	Pain and distress experienced by cancer patients during repeated venipuncture due to poor vascular conditions	Recommendation 5
Physician-Technician Relationship	Laboratory staff’s limited awareness of instrument minimum volume requirements and rigid rejection criteria	Recommendation 14
Physician-Society Relationship	Rigid medical supply procurement processes affecting tube availability and standardization	Recommendation 12

Do (reflection & response)

Communication & training (April 2019)

A 90-min joint training session was held for all laboratory staff and head nurses, co-facilitated by the Nursing Office and Medical Affairs Department. The session included a presentation on collection standards and sampling precautions, followed by an interactive Q&A to address practical concerns. It comprised three components: (i) A narrative presentation: the index incident and parallel chart were read aloud to establish shared emotional and intellectual understanding (Attention and Understanding). (ii) Protocol redesign: Rejection notices were revised to explain the “why” (e.g., “Volume is insufficient, risking a false result. We recognize your patient’s venous fragility—how can we support optimal collection?”), replacing abrupt “Rejected” labels; (iii) Technical anchoring: A concise, evidence-based explanation of the 1:9 ratio was provided to maintain technical rigor alongside empathy. Each rewritten rejection notice became a micro-practice of narrative “Response,” sustaining the framework without requiring continuous formal writing of parallel charts.

System-level optimization (planned)

Narrative diagnosis identified the 3 mL tube as a key system barrier. A proposal to procure 2 mL low-volume tubes was submitted. However, institutional procurement constraints prevented implementation during the study period (2019–2025).

Study & act (sustaining the improvement)

Continuous Monitoring (April 2019 onwards): Monthly rejection rates were extracted and aggregated from the LIS by the principal investigator, enabling rapid detection of deviations.

Targeted Re-education (July 2021): A large-scale staff rotation triggered a transient spike in rejection rate. New technologists—though trained in 2019—had lost familiarity with narrative protocols while rotating through non-coagulation units. This prompted a targeted refresher, emphasizing explanatory language in rejections. It marked a critical evolution: from static one-time training to dynamic, context-responsive reinforcement. This adaptation successfully restored performance. It validated that the narrative culture was durable but required intentional maintenance during transitions.

Data analysis

Statistical analysis

Normality was assessed using the Shapiro–Wilk test for each phase. Homogeneity of variance was evaluated using Levene’s test for pairwise comparisons. Given non-normal distributions and heteroscedasticity across phases, non-parametric tests were used throughout. Pairwise comparisons between phases were conducted using the Mann–Whitney U tests. A Bonferroni correction was applied to adjust for multiple comparisons, setting the significance threshold at α = 0.017(0.05/3). Differences in median monthly rejection rates across the three phases were assessed using the Kruskal–Wallis test. Effect sizes for Mann–Whitney U tests were reported as rank-biserial correlation coefficients (r), interpreted as: small (r = 0.1), medium (r = 0.3), large (r = 0.5) [22]. Given the small baseline sample (n = 3), 95% confidence intervals (CIs) for median rejection rates were estimated using bootstrap resampling (10,000 iterations). All analyses were performed in Python 3.10.17 (SciPy 1.14.0, NumPy 1.26.1, Pandas 2.2.2). p-values < 0.05 were considered statistically significant (two-tailed), with Bonferroni-adjusted thresholds applied for post hoc tests. Effect sizes and CIs were reported to ensure clinical interpretability beyond statistical significance.

Data visualization

Time series plots of monthly rejection rates were generated in R 4.0.3 (2020-10-10) to visualize temporal trends. Phase boundaries were clearly marked using different background colors and vertical reference lines. Key events (intervention initiation, staff rotation periods) were annotated with descriptive labels to contextualize temporal changes in performance.

Results

A three-phase analysis of sustainable quality improvement

This retrospective time series analysis examined coagulation specimen rejection rates from January 2019 to November 2025 (83 consecutive months). The data reveal a distinct three-phase trajectory following implementation of the narrative medicine intervention (Fig. 1).

Fig. 1.

Three-phase trajectory of coagulation specimen rejection rates from January 2019 to November 2025. The chart illustrates the evolution of rejection rates following a narrative medicine intervention. Phase 1 (Baseline, January–March 2019) shows a high and unstable rejection rate with a median of 1.862%. Phase 2 (Post-Intervention Stabilization, April 2019–December 2021) indicates a significant reduction to a median of 0.108%, with a minor fluctuation observed during a large-scale staff rotation in July 2021 (marked by an orange arrow). Phase 3 (Long-Term Optimization, January 2022–November 2025) reflects further continuous improvement, achieving a median rejection rate of 0.026%. The vertical dashed lines denote the boundaries between phases, and the red arrow marks the start of the intervention in April 2019

A total of 1,106,996 specimens were analyzed across the three phases. Descriptive statistics, including specimen counts, rejection counts, and monthly rejection rates per phase, are summarized in Table 2. As shown in Table 3, pairwise Mann–Whitney U tests confirmed statistically significant reductions across all phase comparisons.

Table 2.

Descriptive statistics of coagulation specimen rejection rates across three phases

Phase	Duration	n (months)	Total Rejections	Total Specimens	Mean Monthly Rate (%)	Range (%)	Median (%)	IQR (%)
Phase 1 (Baseline)	Jan-Mar 2019	3	418	22,124	1.881%	[1.784, 1.996]	1.862	0.106
Phase 2 (Post-Intervention)	Apr 2019-Dec 2021	33	497	359,101	0.140%	[0.016, 0.487]	0.108	0.093
Phase 3 (Long-term Optimization)	Jan 2022-Nov 2025	47	310	725,771	0.046%	[0.006, 0.174]	0.026	0.058

IQR Interquartile Range

The Mean Monthly Rate was calculated as the average of the monthly rejection rates within each phase. Range indicates the minimum and maximum monthly rejection rates observed in each phase

Table 3.

Pairwise comparisons of rejection rates between phases

Phase Comparison	Mann–Whitney U	Original p	Adjusted p	Median Diff (95% CI)	Relative Reduction (%)	Effect Size r	Effect Interpretation
Phase 1 vs 2	U = 99.0	0.00028	0.00084	1.754% [1.640, 1.895]	94.2	0.472	Moderate
Phase 2 vs 3	U = 1346.0	< 0.001	< 0.003	0.082% [0.059, 0.129]	75.9	0.623	Large
Phase 1 vs 3	U = 141.0	0.000102	0.000306	1.835% [1.750, 1.975]	98.6	0.407	Moderate

CI Confidence Interval

Adjusted p values were calculated using the Bonferroni correction for multiple comparisons (significance threshold α = 0.017). Effect size r represents the rank-biserial correlation, interpreted according to Cohen’s conventions. Relative Reduction was calculated based on the median rejection rates of the compared phases. The 95% CI for the median difference was estimated using bootstrap resampling (10,000 iterations)

Phase 1: baseline period (Jan-Mar 2019)

Prior to intervention, the system exhibited a high and unstable rejection rate. The median monthly rejection rate was 1.862%, with a peak of 173 rejected specimens in March 2019, indicating an urgent need for systemic improvement.

Phase 2: post-intervention stabilization phase (Apr 2019—Dec 2021)

Immediately following intervention initiation, the median rejection rate declined to 0.108%—a 94.2% reduction from baseline (U test, P < 0.001, effect size r = 0.472). This 33-month period demonstrated system robustness. Notably, a large-scale staff rotation occurred in the laboratory in mid-2021. New technologists, unfamiliar with the established just culture and narrative communication protocols, temporarily reverted to prior adversarial practices, triggering a transient spike in rejections. Although formal interrupted time-series analysis was not performed, visual inspection of the control chart identified this deviation as a signal requiring action. A rapid, targeted re-education session was implemented, after which the rejection rate promptly returned to its low baseline. To determine whether this spike reflected a sustained shift, we compared rejection rates before and after rotation: no significant difference was found (p = 0.467). Thus, the entire period was retained as a single, cohesive Phase 2—evidence of system resilience.

Phase 3: long-term optimization phase (Jan 2022—Nov 2025)

Building on Phase 2 stability, the system entered a period of sustained excellence. The median rejection rate further declined to 0.026%—a 75.9% improvement over Phase 2 (P < 0.001) and a 98.6% reduction from baseline. This continued improvement suggests progressive internalization of the just culture and the effectiveness of iterative PDSA cycles—including ongoing dashboard monitoring and protocol reinforcement during staff transitions. These mechanisms collectively sustained adherence to empathetic communication standards.

Critically, this improvement occurred alongside a substantial increase in testing volume: median monthly specimen volume rose from 7,374 (Phase 1) to 15,442 (Phase 3)—a > twofold increase reflecting growing clinical demand. Despite a 3.3-fold rise in total specimens from Phase 2 to Phase 3, the absolute number of rejected specimens decreased from 497 to 310. Monthly rejections fell from a peak of 173 (March 2019) to a median of < 1 in Phase 3—demonstrating both scalability and reliability of the intervention.

Discussion

This study indicates that integrating narrative medicine with iterative PDSA cycles can drive dramatic and sustained reductions in coagulation specimen rejection rates. The three-phase analysis trajectory—94.2% reduction in Phase 2 and a further 75.9% improvement in Phase 3, culminating in a 98.6% overall reduction from baseline—reveals a critical principle: sustainable QI requires both human-centered intervention and continuous system optimization. The core strength of our intervention lay in transforming rigid technical standards into a shared narrative, bridging the longstanding gap between laboratory precision and clinical complexity.

Shifting from blame to a just culture: a communication transformation

Our findings indicate that pre-analytical errors often reflect a systemic failure—a blame culture that frames rejections as individual lapses rather than safety signals. By implementing parallel charts, we facilitated mutual Attention and Understanding [18, 23, 24]. Laboratory staff gained direct insight into the clinical constraints nurses face when drawing blood from frail oncology patients. Nurses, in turn, received empathetic, evidence-based explanations for volume-critical protocols.

This process fostered consensus and became the catalyst for collective Reflection on the root causes and proactive Response to solutions [18, 25].Its most tangible outcome was the transformation of rejection notices from impersonal, technical alerts into contextualized communication. For example, a directive “Rejected” was replaced with an explanation of the “why” and an offer of assistance. This shift aligns with evidence that narrative medicine strengthens trust and enables collaborative problem-solving. The sustained reduction in rejections and observed change in communication behavior reflect a transition from adversarial to collaborative interprofessional dynamics. These findings are consistent with the establishment of a just culture, one in which errors trigger system learning rather than individual accountability [26]. We emphasize, however, that this cultural shift is inferred from behavioral data and was not directly measured with validated tools.

The “data-storytelling” dual-driver model: a novel framework for QI

Building on the Attention–Understanding–Reflection–Response framework of narrative medicine and the PDSA cycle, our project led to a conceptual model for QI: the data-storytelling dual-driver model (Fig. 2). This model integrates two interdependent drivers. Data-driven elements were operationalized as: (1) the objective metric of monthly specimen rejection rates; (2) adherence to standardized protocols (1:9 citrate-to-blood ratio); (3) continuous monitoring via a monthly dashboard; and (4) technical validation through evidence-based education. Conversely, story-driven elements were implemented as: (1) parallel charts for narrative diagnosis; (2) empathic understanding of clinical constraints; (3) structured reflection on system-level flaws rather than individual error; and (4) shifting to rational communication, exemplified by rewriting rejection notices to explain the “why” and propose collaborative solutions.

Fig. 2.

A Conceptual Model of the “Data-Storytelling” Dual-Driver Framework for Sustained Quality Improvement. This model synthesizes the core mechanism of the intervention. The Data-Driven gear (blue) represents the fundamental, metrics-based components of quality management (e.g., standardized protocols, continuous monitoring). The Storytelling-Driven gear (orange) represents the narrative medicine components (e.g., parallel charts, contextual reflection). Their interlocking operation within the PDCA (Plan-Do-Study-Act) cycle illustrates how the systematic integration of objective data with empathic, narrative-based understanding was theorized to foster a just culture and achieve the sustained quality improvement outcomes observed in this study

The uniqueness of this model lies in its intentional integration of technical rigor and human context. Traditional QI often overemphasizes data-driven enforcement, which may fail when protocols conflict with patient needs. Purely humanistic approaches, in turn, risk lacking technical rigor. This model demonstrates how these two drivers can be synergistic rather than mutually exclusive. For example, we upheld the 1:9 ratio with full technical rigor, and we used clinical narratives, such as “fear of repeated venipuncture” and “poor venous access”, to identify system inefficiencies. This revealed 3 mL collection tube as a key burden. We prioritized it alongside quality targets. In the era of precision medicine, this model offers a blueprint for balancing technical reliability with person-centeredness, suggesting that “rigor” and “empathy” can be mutually reinforcing.

Sustaining excellence: the role of system resilience

The sustainability of our intervention, particularly its resilience during staff rotation, offers critical insights for QI projects in high-turnover settings. The transient rejection spike in July 2021 and its rapid reversal through targeted training reflect a positive shift in team culture and mindset. Yet this shift requires continuous reinforcement. Although we did not use validated psychometric tools to measure safety culture directly, the sustained behavioral outcomes support a transition toward a Just Culture. Low rejection rates persisted for over four years, including during multiple staff rotations. This reflects more than temporary compliance—it reflects a shift in underlying norms. Informal staff discussions corroborate this.

As one senior nurse stated, “We don’t just get a rejection slip anymore. We understand the ‘why’ behind it, which makes us feel like partners with the lab, not opponents”. This perceptual shift reflects enhanced shared purpose and proactive engagement. The narrative approach transformed quality compliance from a mandated task into a shared professional value. However, the staff rotation served as a system stress test. The initial reversion to prior practices indicates the just culture had not yet become institutionalized. Rapid recovery after targeted training highlights the importance of training protocols that go beyond technical skills to include narrative competence training.

This longitudinal study (2019–2025) occurred amid multiple external changes: staffing fluctuations, evolving patient case mix, LIS upgrades, accreditation cycles, and the COVID-19 pandemic. We did not formally measure all these variables. But three features of our data strengthen causal inference. First, coagulation testing volume increased substantially—yet rejection rates fell by 98.6%. Second, the temporal pattern was highly specific: a sharp drop followed the April 2019 intervention; low rates were sustained throughout the pandemic; and a transient spike occurred only during the mid-2021 staff rotation—and resolved immediately after targeted training. This pattern aligns with a targeted intervention and resilient culture—not with slow, unmeasured secular trends. Nonetheless, unmeasured confounders remain a recognized limitation of longitudinal QI studies.

Limitations and future directions

The primary strength of this study is its long-term follow-up, providing robust evidence for intervention sustainability. However, four limitations warrant acknowledgment. First, the three-month baseline reflects the reactive nature of the intervention, triggered by a single clinical incident, constraining statistical power for trend analysis. Second, we observed behavioral changes but lacked validated tools, such as the Safety Attitudes Questionnaire, to quantify cultural shift objectively. Third, as a single-center study in an oncology hospital, generalizability to other settings is limited. Fourth, while temporal association is strong, unmeasured secular trends cannot be fully excluded. Future research should adopt interrupted time-series designs and validated psychometric instruments to strengthen causal inference.

We recommend embedding narrative competence training into formal curricula for laboratory and nursing professionals. Bridging the bench-bedside gap requires bidirectional understanding: laboratory staff must appreciate clinical constraints; nurses must understand the scientific rationale for technical standards. This interprofessional education addresses a critical gap in medical training, building shared language and mutual respect early in practice. The model is transferable to any setting with interprofessional communication friction.

To standardize communication and reduce turnover-related variability, health information technology offers promise. Examples include AI-powered analysis of rejection reasons or empathetic, context-aware prompts embedded in the LIS [27]. Finally, system-level optimization identified through our narrative diagnosis, such as procuring low-volume tubes, represents the next actionable step to reduce patient burden.

Conclusion

This study indicates that integrating narrative medicine into laboratory quality management effectively reduces pre-analytical errors. It achieved a 98.6% reduction in coagulation specimen rejections, a change sustained for over four years. The intervention succeeded by replacing punitive enforcement with a just culture grounded in empathetic communication. This reflects a novel data-storytelling dual-driver model, which aligns technical data-driven management with human-centered narratives. The model is transferable across healthcare settings. It shows that robust solutions may be found at the intersection of data and humanity.

Abbreviations

AI

Artificial Intelligence

APTT

Activated Partial Thromboplastin Time

CI

Confidence Interval

IQR

Interquartile Range

LIS

Laboratory Information System

PDSA

Plan-Do-Study-Act

QI

Quality Improvement

SQUIRE

Standards for QUality Improvement Reporting Excellence

SMART

Specific, Measurable, Achievable, Relevant, and Time-bound

Authors’ contributions

YMH: conceptualization, writing – original draft, reviewing & editing, visualization, funding acquisition. ZXC and QLL: Data querying, organization, and data analysis. HYD and YHH: conceptualization, reviewing & editing, supervision.

Funding

This work was supported by the Hunan Provincial Natural Science Foundation of China (2025JJ80848 and 2025JJ80831).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information. Rejection counts and specimen counts are provided in Supplementary Table 1.

Declarations

Ethics approval and consent to participate

This study was reviewed and approved by the Medical Ethics Review Committee of Hunan Cancer Hospital (Approval No. [40]). As this was involved the retrospective analysis of de-identified, aggregated data from the Laboratory Information System, the requirement for written informed consent was waived by the committee. The study followed the principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.