Investigating precise control pathway for haemostatic clip usage in laparoscopic cholecystectomy based on patient clinical variations: an exploratory retrospective observational study

Abstract

Objectives

To explore a precise control pathway based on patient clinical variations for haemostatic clip usage in laparoscopic cholecystectomy (LC) through on-site research data from a teaching hospital. Specifically, the study aimed to: (1) to calculate the optimal haemostatic clip consumption in LC based on diagnostic grouping and surgical combinations, and (2) to provide evidence for cost containment of high-value medical consumables used in LC.

Design

Retrospective observational study.

Setting

Hospital in southwest China.

Participants

The inclusion criteria were set as inpatients whose medical records included International Classification of Diseases codes 1–3 with code 51.2300 (corresponding to LC surgery) and who were discharged on medical advice. A dataset containing 1001 patients without any haematological diagnoses was collected.

Primary and secondary outcome measures

Two ordered multinomial logit models were established to identify factors affecting haemostatic clip use in LC. Two-step clustering was used to form subgroups. The premodel included preoperative variables (acute cholecystitis and scarring/fibrotic atrophy), while the full model added intraoperative variables (anatomical variation and severe adhesions). Both models met application prerequisites.

Results

Key factors influencing haemostatic clip usage were identified, including acute cholecystitis, scarring/fibrotic atrophy, anatomical variation and severe adhesions. Consumption references for haemostatic clips were established for LC patients with good prognoses on discharge. The ordered multinomial logit model revealed that acute cholecystitis decreased the odds of using four or more clips (p<0.001), while anatomical variation, scarring/fibrotic atrophy and severe adhesions increased the odds (p<0.001 for all). The model suggested using no more than three clips in 17.30% of cases, precisely four clips in 81.72% of cases and five or more clips in 0.98% of cases, depending on the co-occurring factors. Model-predicted clip usage was consistent with actual usage (consistency=0.56).

Conclusions

This study provides a framework for evidence-based consumable management in LC, demonstrating the potential for extrapolation to other consumables and surgical combinations. Continuous monitoring and timely adjustment will be imperative as medical products and practices evolve.

Strengths and limitations of this study.

• Two-step clustering method employed to form subgroups for analysis.

• Ordered multinomial logit models used for multicategorical dependent variable analysis.

• Two-stage modelling approach separated preoperative and intraoperative variables.

• Single-centre data over 18 months may limit generalisability of findings.

• Retrospective design precluded verification of some potential influencing factors.

Introduction

Efficient monitoring and management of medical consumables are crucial for strengthening autonomous cost control in hospitals.^{1 2} The official definition categorises medical consumables as high value, taking into account their impact on health, safety standards, clinical usage, prices and the financial burden on the public.³ These have been the focus of attention from national-level initiatives.⁴ The intricacy of high-value medical consumables, encompassing their various types and sizes, leads to losses and overflow of inventory. The lack of consensus among healthcare practitioners on medical consumable usage results in discretionary behaviours, incurring in inappropriate expenses for hospitals, significant wastage of insurance funds and preventable harm to patients’ interests. Value-based healthcare aims to align the interests of healthcare payers, hospitals and patients; however, inadequate refinement in consumables management hinders this goal. Proposals include leveraging quantitative approaches to regulate the physicians’ behaviour, assist patients in reducing surgical expenses and inform insurance payment standards.

Laparoscopic cholecystectomy (LC) has become the standard procedure for treating gallbladder diseases, offering several advantages.5,7 Previously contraindicated cases now fall within the scope of surgical indications.^{8 9} LC is widely performed in hospitals in China^{5 10} and is considered a representative of minimally invasive surgery that uses laparoscopy.^{11 12} Although LC uses many consumables, there is limited research on their fine-grained management and proper use. Haemostatic clips, often managed as high-value consumables, are used at an average of 4.9 per LC at the study hospital. By calculating the optimal clip consumption based on diagnostic grouping and surgical combinations, the authors aimed to provide evidence for cost containment of medical consumables. Previous research has focused on challenging factors in LC, cost variations or preventive strategies for clip migration, yet limited attention has been given to optimising haemostatic clip usage in LC.

The primary objectives of this study were twofold: (1) to calculate the optimal haemostatic clip consumption in LC based on diagnostic grouping and surgical combinations, and (2) to provide evidence for cost containment of high-value medical consumables used in LC. By establishing precise consumables management, this study aimed to empower hospitals and contribute to an efficient healthcare system. The findings could help regulate physician behaviour, assist patients in reducing expenses and inform insurance payment standards, ultimately aligning the interests of the three parties involved.

In summary, using LC as an example, this study sought to explore strategies for precise control and cost containment of high-value medical consumables. In the context of public hospital reform and payment system reform,² evidence-based research on the appropriate use of consumables, such as haemostatic clips, is crucial for promoting effective utilisation of healthcare resources and improving overall healthcare system efficiency.

Materials and methods

Study setting, participants and variables

Data were collected from a hospital in southwest China between January 2021 and June 2022. The inclusion criteria were set as inpatients whose medical records included International Classification of Diseases (ICD) codes 1–3 with code 51.2300, corresponding to LC surgery. Only medical records with a discharge mode of ‘discharge on medical advice’ were retained to distinguish patients who were transferred to higher level hospitals, transferred to lower level medical institutions or deceased.

The study variables included age, sex, ABO blood type, Rh blood type, department, primary surgeon, ICD codes 1–7, length of stay (LOS), out-of-pocket (OOP) expenses and disposable medical material fee (DMMF). Seven factors affecting common intraoperative and postoperative comorbidities were also considered: the presence of acute cholecystitis,¹⁰ scarring/fibrotic atrophy,¹³ liver cirrhosis,^{14 15} portal hypertension,¹⁵ anatomical variations,¹⁶ severe adhesions^{10 17} and associated injuries.^{14 18} Admission and discharge diagnoses were checked for blood disorders.

The authors had no access to information that could identify individual participants during or after data collection.

Data sources and measurements

Patient information regarding physiology, pathology, surgical procedures and costs was extracted from the hospital and other internal information systems.

The names of consumables were normalised, including ligature clips (item code: 28834; specifications: 544230, 544240, 544250; manufacturer: Telifor Medical, USA) and ligature vascular clips (specifications: WD-JZ S, WD-JZ M, WD-JZ L; manufacturer: Zhejiang Weidu Medical Devices, China), which were collectively referred to as haemostatic clips, as they simply differed in purchase lots. Laparoscopic puncture devices (specifications: type A, type B, type D; manufacturer: Jiangsu New Zhiyuan Medical Technology, China), trocar puncturer (specifications: WD-CCQ O05, WD-CCQ O10, WD-CCQ O12; manufacturer: Zhejiang Weidu Medical Equipment, China) and laparoscopic punctures 42288 (specifications: type A; manufacturer: Jiangsu New Zhiyuan Medical Technology) were collectively termed as trocars. Medical laryngeal mask airways, breathing circuits (sets), specimen retrieval devices, stone retrieval baskets, high-pressure angiographic syringes and accessories and bacterial filter dosage fluctuations exhibited minimal fluctuations; thus, the focus was on haemostatic clips and trocars. Four-hole, three-hole, two-hole and single-hole LC methods^{19 20} are widely used, while hidden scarring²⁰ and pneumoperitoneum-free²¹ methods are emerging. The corresponding trocar usages of LC surgery were taken as 4, 3 and 2 in all cases.^{9 22}

During the modelling process, data generation timing was categorised as preoperatively or intraoperatively available, depending on whether the influencing factor was known before the surgery. Various data generation methods, such as blood tests, imaging examinations or intraoperative judgements by physicians, were used to identify influencing factors. Data sources included surgery, discharge or medical records, from which the occurrence of influencing factors could be determined. The text contents of the operation and discharge records were converted to structured variables using specific rules. If the diagnosis was acute cholecystitis or the diagnosis content included ‘gallbladder (triangle) congestion’, acute cholecystitis was assigned a value of 1. If the diagnosis was XX cirrhosis or the diagnostic content included ‘varices of the abdominal wall’, ‘tortuous veins of the abdominal wall’ or ‘portal hypertension’, the presence of liver cirrhosis and portal hypertension was assigned a value of 1. If the diagnosis contained ‘encapsulated adhesions’, ‘wrapped gallbladder’ or ‘partial fusion with the wall of the gallbladder without obvious boundaries’, the presence of severe adhesions was assigned a value of 1. If the diagnostic content contained ‘XX (organ, blood vessel) injury’, the presence of associated injuries was assigned a value of 1.

Statistical methods and study size

The dataset was divided into subgroups using a two-step clustering method with ICD codes 1–7 and the DMMF as clustering variables. This method first preclusters the cases into small subclusters using a sequential clustering approach, and then clusters these subclusters into the desired number of clusters using the agglomerative hierarchical clustering method. The log-likelihood distance measure and the Bayesian information criterion were used to automatically determine the optimal number of clusters.²³ Subgroups were merged, and records with missing values were excluded.

Each surgery and operation (SO) corresponds uniquely to an ICD code, and the medical records contain information on ICD codes 1–10. Specifically, SO 1–7 and DMMFs were used as no information was available for ICD codes 8–10 in all cases. Multiple comparisons were conducted using the Scheffé and Student-Newman-Keuls methods.²⁴

Given the multicategorical nature of the dependent variable (haemostatic clip usage levels) and the binary nature of the independent variables (presence or absence of clinical factors), ordered multinomial logit models instead of linear regression were employed.²⁵ This model estimates the cumulative odds of being at or below a particular category of the dependent variable, considering the ordinal nature of the categories. Three levels of haemostatic clip usage were measured: 5 or more, precisely 4 (the mode of haemostatic clip usage among the included cases) and no more than 3. These levels were coded as y=1 (high consumption), y=2 (moderate consumption) and y=3 (low consumption), respectively. A two-stage modelling approach was used to separately consider preoperative and intraoperative variables. The premodel included only preoperative variables (presence of acute cholecystitis and scarring/fibrotic atrophy), while the full model added intraoperative variables (anatomical variation and severe adhesions) to the premodel. This staged approach allows for the development of clip usage guidance based on the timing of information availability. Collinearity was checked with tolerance above 0.1 and variance inflation factor above 10. The model equation is as follows:

$${\displaystyle \mathit{l}\mathit{o}\mathit{g}\mathit{i}\mathit{t}\left({\mathit{P}}_1\right)={\mathit{\alpha }}_1+{\mathit{\beta }}_1\ast {\mathit{x}}_1+{\mathit{\beta }}_2\ast {\mathit{x}}_2+\cdots +{\mathit{\beta }}_{\mathit{n}}\ast {\mathit{x}}_{\mathit{n}}}$$

,$${\displaystyle \mathit{l}\mathit{o}\mathit{g}\mathit{i}\mathit{t}\left({\mathit{P}}_1+{\mathit{P}}_2\right)={\mathit{\alpha }}_2+{\mathit{\beta }}_1\ast {\mathit{x}}_1+{\mathit{\beta }}_2\ast {\mathit{x}}_2+\cdots +{\mathit{\beta }}_{\mathit{n}}\ast {\mathit{x}}_{\mathit{n}}}$$

where $x_n$ denotes the nth independent variable, ${\alpha }_n$ is the constant term and ${\beta }_n$ is the coefficient of the independent variable.

The first regression equation, P1=P (y=1)÷P (y=2 or 3), represents the ratio of the probability of clip usage exceeding 4 to the probability of usage not exceeding 4; P2=P (y=1 or 2)÷P (y=3) represents the ratio of the probability of clip usage exceeding 3 to the probability of usage not exceeding 3.

Following the 1 in 10 rule,^{26 27} the number of independent variables was restricted to ensure at least 10 events per variable (EPV) should exceed 10, although some suggest an EPV of 5–9.^{28 29}

Two sensitivity analyses were conducted: one included the sex variable (coded as 1 for female and 0 for male), and the other included the aged 65 variable (coded as 1 for age ≥65 years and 0 for age <65 years).

Table connection was completed using SAS V.9.4 (SAS Institute), while statistical tests and modelling were performed using SPSS V.26.0 (IBM). Figures were created using Microsoft PowerPoint (Microsoft, Redmond, Washington, USA), and tables were generated using Microsoft Excel (Microsoft).

Patient and public involvement

As this was a retrospective observational study, patients or the public were not involved in research.

Results

Participants

A dataset containing 1001 patients without any haematological diagnoses was collected. After clustering, three categories were formed: category 1 (114 cases), category 2 (680 cases) and category 3 (180 cases). The average silhouette value of 0.7, which is greater than 0.5, indicated an acceptable clustering effect. A total of 59 combinations of SOs were performed. The combinations in category 1 were irregular, with SO1 including seven procedures, such as LC and endoscopic retrograde cholangiopancreatography; SO2 including 15 procedures, such as laparoscopic adhesiolysis and laparoscopic incision of renal cysts; and SO3 including 36 procedures, such as unilateral inguinal hernia repair and fascial repair. Few patients underwent SO4. Categories 2 and 3 exhibited distinct features: all cases in category 2 were combinations of LC and laparoscopic adhesiolysis, all cases in category 3 had LC as SO1 and over 90% had either no SO2 or laparoscopic adhesiolysis.

Based on the clustering variable importance ranking, SO3 was the key factor distinguishing category 1 from the other two, while SO2 was key for distinguishing category 2 from category 3. The auxiliary factor, SO5, ranked second in importance. SO1 and SO7 were the least important factors; the former resulted from the inclusion criteria, making LC the most dominant SO1, and the latter due to the limited number of cases involving the seventh surgical operation. Categories 2 and 3 were merged, missing values were dealt with (figure 1) and the final sample size was 815.

Figure 1. Data cleaning process.

Descriptive data

The mean differences in DMMFs among clustering groups were compared (table 1). Additionally, a population characteristics analysis of the uncleaned dataset was conducted (table 2). Exploratory analysis of the relationship between relevant clinical characteristics and haemostatic clip usage was performed on the cleaned dataset (table 3). In the uncleaned dataset, no significant differences were observed regarding sex and age. Statistically significant differences were found in ABO blood type, LOS, OOP expenses, department, primary surgeon and ICD codes 1–3. Variations in ICD codes 1–3 indicated differences in the types of surgical operations performed on patients undergoing LC, which might contribute to the observed differences in LOS and OOP expenses. Differences in the primary surgeon and their respective departments suggested potential clinical characteristic disparities among LC patients treated by different departments.

Table 1. Multiple comparison results for disposable medical material fees.

	Cluster	n	1	2
SNK*†	3	180	1432.3001
	2	680	1473.4575
	1	141		1655.5772
	Significance		0.143	1.000
Scheffe*†	3	180	1432.3001
	2	680	1473.4575
	1	141		1655.5772
	Significance		0.342	1.000

Means for groups in homogeneous subsets are displayed.

^*Subset for alpha=0.05.

^†The group sizes are unequal. The harmonic mean of the group sizes is used (harmonic mean sample size=212.490).

SNK, Student-Newman-Keuls

Table 2. Clinical characteristics by clustering groups.

Variables	Cluster			F/χ2	P value
	1	2	3
Age	51.55±14.42	49.28±13.48	50.14±13.91	1.696	0.184
LOS	8.32±3.81	5.27±2.61	6.51±2.92	70.979	<0.001
OOP expenses	5836.21±3518.97	4530.32±2234.54	4404.12±2733.52	16.781	<0.001
Sex
Male	44 (31.20)	225 (33.10)	71 (39.40)	3.121	0.210
Female	97 (68.80)	455 (66.90)	109 (60.60)
Department
Hepatopancreatobiliary	98 (69.50)	563 (82.80)	153 (85.00)		<0.001
General	32 (22.70)	65 (9.60)	11 (6.10)
Others	11 (7.80)	52 (7.60)	16 (8.90)
Primary surgeon*
CSW	2 (1.40)	24 (3.50)	10 (5.60)	44.210	<0.001
FMS	4 (2.80)	15 (2.20)	7 (3.90)
FY	1 (0.70)	17 (2.50)	2 (1.10)
JSH	21 (14.90)	159 (23.40)	18 (10.00)
LYC	38 (27.00)	155 (22.80)	38 (21.10)
XDP	35 (24.80)	122 (17.90)	60 (33.30)
XF	26 (18.40)	147 (21.60)	34 (18.90)
Others	14 (9.90)	41 (6.00)	11 (6.10)
Blood type†
A	45 (31.90)	234 (34.40)	48 (32.70)	17.544	0.007
B	38 (27.00)	153 (22.50)	46 (23.70)
O	40 (28.40)	214 (31.50)	46 (30.00)
AB	12 (8.50)	56 (8.20)	11 (7.90)
Unknown	6 (4.30)	23 (3.40)	29 (5.80)
ICD code 1‡
51.2300	133 (94.30)	680 (100.00)	180 (100.00)	49.187	<0.001
Others	8 (5.70)	0 (0.00)	0 (0.00)
ICD code 2§
54.5100×005	7 (5.00)	0 (0.00)	37 (20.60)	923.152	<0.001
54.5101	94 (66.70)	679 (99.90)	0 (0.00)
54.5903	21 (14.90)	0 (0.00)	61 (33.90)
Others	18 (12.80)	1 (0.10)	10 (5.60)
Null	1 (0.70)	0 (0.00)	72 (40.00)
ICD code 3¶
54.5100×005	1 (0.70)	0 (0.00)	0 (0.00)	919.317	<0.001
54.5101	8 (5.70)	0 (0.00)	0 (0.00)
54.5903	5 (3.50)	0 (0.00)	0 (0.00)
83.8901	73 (51.80)	0 (0.00)	0 (0.00)
Others	44 (31.20)	0 (0.00)	0 (0.00)
Null	10 (7.10)	680 (100.00)	180 (100.00)

^*Primary surgeons who performed fewer than 20 cases of laparoscopic cholecystectomy (LC) within the one-and-a-half-year study period are combined and referred to as ‘others’ for better display. CSW, FMS, FY, JSH, LYC, XDP and XF were all initials of the names of the surgeons.

^†There was no Rh-negative blood in all cases, so the blood type here refers specifically to the ABO classification.

^‡The ICD code 51.2300 corresponds to LC. Similarly, ICD codes with infrequent occurrence were combined and referred to as ‘others’.

^§The ICD code 54.5100×005 corresponds to laparoscopic abdominal adhesion lysis, 54.5101 corresponds to laparoscopic adhesion lysis of the intestine and 54.5903 corresponds to intestinal adhesion lysis. The term ‘null’ indicates that the field is empty or has no information in the medical record. Likewise, ICD codes with infrequent occurrence were combined and referred to as ‘others’.

^¶The ICD code 83.8901 corresponds to fascioplasty, and the remaining processing for this variable is the same as for ICD code 1 and ICD code 2.

ICDInternational Classification of DiseasesLOSlength of stayOOPout-of-pocket

Table 3. Results of the univariate analysis.

Variables	Sum of squares	df	Mean square	F	Significance
Sex
Between groups	177.234	65	2.727	1.114	0.259
Within groups	1833.925	749	2.448
Total	2011.158	814
Age
Between groups	177.234	65	2.727	1.114	0.259
Within groups	1833.925	749	2.448
Total	2011.158	814
ABO blood type
Between groups	9.589	3	3.196	1.295	0.275
Within groups	2001.569	811	2.468
Total	2011.158	814

The cleaned dataset does not explain the relationship between demographic variables and haemostatic clip usage.

Main results

After clustering and merging, sample sizes were 100, 453 and 262 for y=1, y=2 and y=3, respectively. Thus, the valid number of variables was 10. The positivity rates, ranked from high to low, were as follows: the presence of acute cholecystitis, severe adhesions, scarring/fibrotic atrophy, anatomical variation, liver cirrhosis, associated injuries and portal hypertension. Considering the EPV constraint, the first four variables with the highest positivity rates were selected for modelling. The preliminary model (premodel) included the presence of acute cholecystitis and scarring/fibrotic atrophy. The full model added anatomical variation and severe adhesions, two intraoperative variables with determinable values, to the premodel. The collinearity check observed no multicollinearity (tolerance >0.99, variance inflation factor >1).

The ordered multinomial logit model (online supplemental file 1) revealed that the presence of acute cholecystitis decreased the odds of using four or more clips (OR: 0.625, 95% CI 0.504 to 0.775, p<0.001), a counterintuitive result warranting further investigation. Conversely, the presence of anatomical variation (OR: 4.142, 95% CI 2.180 to 7.867, p<0.001), scarring/fibrotic atrophy (OR: 5.992, 95% CI 3.155 to 11.380, p<0.001) and severe adhesions (OR: 3.143, 95% CI 2.593 to 3.809, p<0.001) increased the odds of using four or more clips (see online supplemental file 2 for calculations). Both models met the parallel lines assumption, and the two-stage model outperformed the thresholds-only model (omnibus test). Figure 2 shows the consumption references based on the full model. The full model suggested using no more than three clips in 17.30% of cases, precisely four clips in 81.72% of cases and five or more clips in 0.98% of cases, depending on the co-occurring factors. The consistency between model outputs and actual usage was 0.56 for both models, measured the deviation between the clip usage in the 815 cases used for modelling and the reasonable usage represented by the model output, indicating relatively consistent clip usage behaviour in LC during the study period, but with room for improvement in attaining rational and compliant usage.

Figure 2. Haemostatic clip usage recommendations based on the full model.

In sensitivity analyses that additionally adjusted for sex and aged separately, the results remained largely unchanged. Online supplemental file 3 presents the relevant details of the sensitivity analyses, where there was insufficient evidence that more clips were needed for elderly patients, and no significant gender differences were observed in clip usage.

Discussion

Key results and interpretation

Our study established haemostatic clip consumption references for LC patients with good prognoses on discharge (patients who met the requirements for discharge). The findings suggest that in most situations, four clips are recommended as the standard usage. However, this can vary based on the presence of certain factors such as acute cholecystitis, scarring/fibrotic atrophy and severe adhesions.

The unexpected negative beta coefficient for acute cholecystitis might be due to milder symptoms in the early stages, leading to less bleeding, shorter surgery times and faster recovery, as suggested by existing studies.^{9 10} Despite variations in haemostatic clip usage among LC patients, sensitivity analyses adjusting for sex and age did not alter the primary conclusion that four clips are recommended in most situations, supporting the robustness of our findings. Ultimately, this investigation aimed to optimise clip usage without compromising patient safety, aligning with the principles of value-based healthcare.

Comparison with previous studies

Previous studies have explored predictive factors and risk factors associated with difficult LC, as well as cost variance and supply utilisation variability among surgeons.^{13 19 30 31} However, there are limited reports on the fine-grained management of high-value consumables used in LC, particularly the appropriate use of haemostatic clips. This represents an underexplored area with opportunities for value-based healthcare. Accordingly, this study aims to address this knowledge gap by providing an in-depth analysis of haemostatic clip application during LC. Other studies have identified poor performance status and history of common bile duct stones as predictors of conversion to subtotal cholecystectomy,^{32 33} proposed assessing surgical difficulty via intraoperative video findings³⁴ and suggested various preventive strategies (eg, confirming Calot’s triangle, minimising clip use) for the known rare complication of postoperative clip migration into the bile duct.³⁵ Nonetheless, reports on optimisation of haemostatic clip usage during LC remain lacking.

Clinical implications

Based on our findings, we propose the following references for haemostatic clip usage in LC: if only acute cholecystitis is present, three clips should suffice; when only one factor other than acute cholecystitis is present, or when acute cholecystitis co-occurs with any other factor, four clips should suffice; if two factors other than acute cholecystitis are simultaneously present, or if acute cholecystitis, scarring/fibrotic atrophy and severe adhesions co-occur, five or more clips may be used. Managers should exercise discretion and tailor the solution accordingly. The model outputs can inform performance evaluations to reflect inappropriate consumable usage and enable comparisons of surgeon competencies. Importantly, this investigation was conducted without compromising patient safety, aligning with the overarching principles of value-based healthcare.

Strengths and limitations

The strengths of our study include the use of a two-stage model that met EPV requirements and produced interpretable and presentable results. The group coding of haemostatic clip usage helped mitigate the impact of extreme values in the dataset. Sensitivity analyses adjusting for sex and age did not substantially change the main findings, providing reassurance about the robustness of the results. However, several limitations should be noted. The use of single-centre data may limit the generalisability of the findings, and retrospective data may introduce information biases like under-reporting. Potential influencing factors such as liver cirrhosis, associated injuries and portal hypertension were not verified in this study. As patients were further stratified by combinations of their clinical and demographic characteristics, the role of acute cholecystitis showed differences that warrant further clarification.

Future research directions

Additional multicentre prospective studies could help validate the results and improve representativeness. To enhance interpretability, information regarding whether the surgery time is within 72–96 hours of disease onset³⁶ should be supplemented. Extended studies with more detailed information are needed for comprehensive understanding. Future research should also focus on validating the model’s relevance across diverse populations and settings, particularly in regions with varying healthcare resources and practices.

Implications for practice

This study provides an exploratory framework for high-value consumable management in LC, demonstrating potential for extrapolation to other consumable types and surgical combinations. Continuous monitoring and timely adjustment will be imperative as medical products and practices evolve. In developed countries with abundant medical resources and higher guideline adherence, physician practice variability is lower.³⁷ However, enhancing governance and steering healthcare towards value-based models is particularly crucial in underdeveloped countries. Ultimately, this research contributes to the optimisation of consumable administration and achieves value-based healthcare, especially in resource-constrained regions.

Conclusions

Fine-grained consumable control can catalyse the shift from fee-for-service to value-based payment. Nevertheless, customisable implementation suited to local contexts will be key. This study established consumption references for haemostatic clips in LC patients discharged with good prognoses and could serve as a foundation for medical consumable governance to enable policy compliance, protect patients and support insurance reforms that benefit all stakeholders.

Data availability statement

Data are available in a public, open access repository.