Enhancing hand hygiene compliance in healthcare environmental services staff: a systematic approach to indicator development

Abstract

Objective

The aim was to develop a comprehensive system of hand hygiene (HH) indicators for environmental services (EVS) staff in medical institutions, thereby providing clear guidelines on the appropriate moments for EVS staff to perform HH, offering monitoring and feedback metrics for their HH practices, and utilizing the collected monitoring data to evaluate the effectiveness of these practices and serve as a basis for implementing improvement measures.

Methods

We conducted non-participant observations to document the workflows of EVS staff across 38 clinical departments within a single tertiary hospital in China, creating a textual corpus. Utilizing the Latent Dirichlet Allocation (LDA) modeling, we identified thematic work tasks for EVS staff in medical settings. We analyzed HH protocols based on standard operating procedures for each task and synthesized these with literature insights to derive HH guidelines for EVS staff. The Delphi method was employed to refine these guidelines and establish their relative importance through hierarchical analysis.

Results

Our research identified and labeled twelve themes of janitorial tasks. Through a meticulous examination and extraction process based on detailed standard operating procedures for each task, we delineated seven HH moments for EVS staff: before handling clean items, before cleaning or disinfection, before donning personal protective equipment (PPE), before doffing PPE, after cleaning or disinfection, after touching highly contaminated surfaces or items, and after doffing PPE. Following two rounds of Delphi consultation, experts reached a consensus and five indicators were retained based on importance, feasibility, and coefficient of variation. The final HH indicators for healthcare EVS staff, ranked by importance, included: after touching highly contaminated surfaces or items, before handling clean items, after cleaning or disinfection, before cleaning or disinfection, and after doffing PPE.

Conclusion

The formulation of HH indicators for cleaning personnel not only clarifies when and under what circumstances HH should be performed but also fosters further advancements in HH management for EVS staff.

Introduction

Among the myriad practices implemented in medical institutions to mitigate the spread of infections, the correct execution of hand hygiene (HH) remains the most fundamental, straightforward, and economical means of effectively controlling pathogen transmission and reducing the incidence of healthcare-associated infections (HAIs) [1–3]. In 2009, the World Health Organization (WHO) introduced the "Five Moments for HH" as a standardized framework to guide HH practices among healthcare workers. These five moments included: before contacting the patient, before performing aseptic tasks, after contacting the patient's body fluids, after contacting the patient, and after contacting the patient's surroundings [4–6]. This approach aimed to clearly define critical moments for HH, thereby improving compliance and reducing the transmission of infections in healthcare settings.

However, despite this framework, overall adherence remains low across all five moments. While HH after exposure to bodily fluids or direct patient contact may show slightly better compliance due to self-protection motives, moments such as before patient contact or after touching patient surroundings are still frequently neglected [7, 8]. Such lapses in compliance significantly increase the risk of microorganisms present in the environment being introduced to the patient's body via the hands of healthcare workers, potentially leading to individual infections and even outbreaks of HAIs [9–11]. Studies showed environmental multidrug-resistant organism(MDRO) positivity rates of 13.11–28.2% [4, 12], with genomic similarities between MDRO strains in patients and hospital environments, highlighting environmental contamination as a key factor in nosocomial outbreaks [9, 11, 13]. Environmental services (EVS) staff are crucial in maintaining hygiene, and the efficacy of their cleaning efforts hinges on the standardization of their operations. A particularly crucial aspect is the consistent practice of HH, given the frequent interactions of EVS staff with various areas within the hospital. In essence, when HH is not meticulously practiced, the hands of EVS staff can serve as significant vectors for pathogen transmission [3, 6]. Consequently, the enactment of HH protocols by EVS staff in medical institutions is paramount to ensuring a safe healthcare environment.

HH indications denote the specific instances when healthcare workers are required to perform HH, thereby fostering adherence among these workers [5]. However, among the WHO's five moments, only two —"after contacting the patient's body fluids" and "after contacting the patient's surroundings"—are applicable to EVS staff. The remaining indicators primarily pertain to clinical staff who engage in direct patient care and aseptic procedures. Even so, these two moments inadequately reflect the daily workflows of EVS staff, neglecting critical moments when their hands may become contaminated or require sanitization.

Cross et al. [14] demonstrated that EVS staff often experience uncertainty about appropriate HH opportunities. This confusion likely contributes to the consistently low compliance rates (9.9–38.9%) observed across studies employing the WHO's five Moments framework [15–18]. Notably, these investigations failed to delineate compliance rates for EVS-specific indications, with only one study reporting a 20.7% compliance rate after patient contact—an infrequent occurrence in routine EVS duties. Currently, research on HH primarily focuses on improving compliance among medical personnel, as reflected in systematic reviews of intervention efficacy [19] and assessments of HH technique quality using the WHO six-step method [20], with scant investigation into the evidence-based rationale for HH guidelines tailored for EVS staff. In the absence of foundational research, enhancing compliance among EVS staff remains an elusive goal. Given that EVS staff are the frontline workforce tasked with maintaining the safety of healthcare environments and curtailing infection spread, it is crucial to provide them with scientifically grounded HH guidelines.

To date, only Wenbin et al. [21] have undertaken a study concerning HH indicators for EVS staff, identifying six indicators through clustering HH behaviors within the workflows of EVS staff. However, as many EVS staff lack clarity on when to perform HH, directly clustering their behaviors may inadvertently overlook some critical indicators. Moreover, the indicators identified in their study have not been validated or refined by experts, resulting in a lack of scientific credibility. The Latent Dirichlet Allocation (LDA) model serves as an unsupervised text analysis method capable of unveiling latent dimensions from extensive unstructured text, facilitating thematic clustering based on topic distribution [22–24]. This approach is particularly valuable for examining non-systematic or non-standardized text corpora, mitigating subjective bias in classifications. Thus, our study employs the LDA model to systematically, objectively, and scientifically identify the themes surrounding the work tasks of EVS staff. We meticulously reviewed the standardized operating procedures (SOPs) of each task to extract moments necessitating HH, thereby minimizing the risk of omission or subjective bias. The identified moments were then synthesized with HH indicators reported in the literature to establish an initial indicator system. Experts were subsequently invited to evaluate these indicators, providing revisions to enhance the feasibility and practicality of the HH indicator system. Finally, hierarchical analysis was employed to determine the relative weights of these indicators, thereby refining the overall system.

Research methodology

Research design

Workflow texts from 38 departments in a tertiary hospital in Wuhan, China, were collected through non-participant observation, encompassing eight internal medicine departments, eight surgical departments, eight obstetrics and gynecology departments, three pediatric departments, two intensive care units (ICUs), three infectious disease departments, five outpatient departments, and one operating room. The LDA theme model was utilized to cluster themes within the workflow text, thereby elucidating the primary tasks performed by the EVS staff. Detailed and standardized steps for each task were compiled, and HH moments were extracted from the SOPs. These moments were then integrated with literature-reported indicators to create HH Indicator System 1.0. Experts in relevant fields were invited to evaluate, modify, and refine this system into HH Indicator System 2.0, with the weights of each indicator established through hierarchical analysis.

1. Text acquisition and preprocessing: Work activities of EVS staff were observed and recorded through non-participatory shadowing across 38 clinical departments. These observations were transcribed into workflow texts and preprocessed into a clean corpus suitable for topic modeling through segmentation, stopword removal, and synonym merging.

2. Theme clustering and recognition: LDA was applied to the preprocessed corpus to extract latent work task themes. The optimal number of topics was determined based on perplexity and coherence scores. Each topic was manually reviewed and labeled to identify distinct cleaning tasks.

3. HH indications analysis: The SOPs were reviewed to extract HH moments associated with each task. These were combined with evidence from the literature to form the preliminary set of indications (HH Indications 1.0). Using the Delphi method, expert panelists evaluated the indicators for importance and feasibility. The refined indicators (HH Indications 2.0) were finalized through consensus and weighted via hierarchical analysis to determine their relative importance.

4. Sampling Frame and Department Selection

To ensure representative coverage of key clinical workflows, we employed a stratified purposive sampling strategy based on the major departmental categories commonly found in Chinese tertiary hospitals. Clinical departments were first grouped into eight strata according to their functional classifications: internal medicine, surgery, obstetrics and gynecology, pediatrics, intensive care, infectious diseases, outpatient services, and surgical operating units. Within each stratum, departments were selected proportionally based on the number of constituent subspecialties. This approach yielded a total of 38 departments, including eight internal medicine departments, eight surgical departments, eight obstetrics and gynecology departments, three pediatrics departments, two ICUs, three infectious diseases departments, five outpatient departments, and one central operating room.

Observer training and observation design

A total of 12 infection control trainers employed by the hospital's contracted cleaning services company served as observers. All had at least 5 years of professional experience, held associate or higher degrees, and received biannual training and examinations in hospital infection control organized by the hospital's infection prevention and control (IPC) department. They were also responsible for quarterly infection control training and practical skills instruction for EVS staff in their assigned departments.

All observers underwent a standardized 3-h training session, including a review of the observation protocol, scenario-based simulations, and calibration exercises to ensure consistency in identifying both HH moments and janitorial task sequences. Two of the observers also served as senior SOP trainers and participated in observations across all 38 departments. The remaining 10 observers participated in observing departments where they were already responsible for training, with each department being jointly observed and recorded by three observers.

To reduce the risk of omission and bias, each department was observed continuously for at least two days. EVS staff were informed that workflow was the primary focus, while actual HH observations were only recorded when staff were unaware of being observed (to minimize Hawthorne effects). HH observations were limited to 30-min intervals, while workflow observations were conducted from 6:00–11:30 and 14:00–17:00 for two full days per department. Observations across 38 departments were completed over 4 months. No audio or video recordings were used during the observations.

Given that two experienced SOP trainers directly participated in all departmental observations and oversaw data aggregation, inter-rater reliability statistics were not calculated. However, consistency was ensured through training, joint observations, and standardized data recording procedures.

LDA topic model implementation

Employing the workflow of EVS personnel as the analytical corpus, the text preprocessing was conducted with Python's Pandas, tokenization and stopword removal were executed with Jieba, and synonym aggregation was performed. Only Chinese characters were retained, and part-of-speech filtering retained nouns, verbs, and adjectives with more than one character. The LDA topic model was implemented using Gensim, with hyperparameters set to alpha = 'auto', eta = 'auto', passes = 30, and random_state = 42. The dictionary was constructed using no_below = 5, no_above = 0.5, and keep_n = 5000. After pre-processing and dictionary pruning, the final analytical corpus comprised 12,305 total tokens, representing a vocabulary of 226 unique terms. Perplexity and coherence scores (c_v) were calculated across a range of topic numbers to determine the optimal number of work task themes. To assess the robustness of the model, sensitivity analysis was conducted by applying four different random seeds (24, 42, 66, 88). Topic coherence and top-word consistency were evaluated to determine the stability of topic structures under different initializations.

Experts in infection control, in collaboration with EVS managers, proceeded to designate appropriate titles for each work task theme. They meticulously reviewed the workflow of the EVS staff to ascertain that no pivotal or significant task themes had been overlooked.

Delphi method for expert consultation

Twenty experts from IPC and nursing management were consulted regarding HH indicators. Inclusion criteria mandated a title of associate senior or above and a minimum of eight years of work experience. Experts evaluated the importance and feasibility of each indicator using Likert scale scores and provided additional feedback until consensus was reached, with a two-week interval between rounds.

The Delphi survey instrument was organized into the following distinct sections to ensure clarity and comprehensiveness. (a) Introduction: This section outlined the study's purpose and importance, provided participation guidelines, explained the theoretical framework, and introduced related assessment tools. It also specified the timeline for distributing and collecting the questionnaire. (b) Demographic Information of Participants: This part collected key background details about the participating experts. (c) Expert Evaluation: A 5-point Likert scale was used to measure the perceived importance of each item, with scores ranging from 5 ("Extremely Important") to 1 ("Not Important at All"). Similarly, the feasibility of each item was rated on a scale from 5 ("Highly Feasible") to 1 ("Not Feasible at All"). Experts were encouraged to provide feedback on the appropriateness of the items and propose modifications such as additions, deletions, or reorganization. (d) Delphi Methodology and Expert Consultation: This phase evaluated the authority and expertise of the experts by examining the sources of the questionnaire's content across four dimensions: theoretical analysis, practical experience, literature review, and intuitive judgment. These sources were categorized into three influence levels (high, moderate, and low), with theoretical analysis assigned weights of 0.3, 0.2, and 0.1, respectively; practical experience given weights of 0.5, 0.4, and 0.3; and literature review and intuition uniformly scored at 0.1 across all levels. This weighted approach facilitated the quantification of expert input, ensuring a systematic and accurate evaluation. Additionally, the familiarity of experts with the questionnaire content was assessed on a 5-level scale, ranging from "Very Familiar" to "Very Unfamiliar".

The positive coefficient of experts was calculated as the number of returned questionnaires divided by the total distributed. The average score of experts' judgments (Ca) and familiarity (Cs) was assessed, with the expert authority coefficient (Cr) defined as (Ca + Cs)/2, where Cr≥0.7 indicated a high level of authority. The concentration of expert opinion was determined by importance and feasibility scores, with scores ≥3.5 indicating consensus. The coordination of expert opinions was measured by the coefficient of variation (CV) and Kendall's coefficient of concordance (W). The inclusion criteria for indicators were importance and feasibility scores ≥3.5 and CV≤0.25. Indicators with mean importance and feasibility scores ≥3.5 but CV exceeding 0.25 were excluded due to insufficient expert consensus.

Hierarchical analysis method for indicator weights

The HH indicator system designed in this study comprised two hierarchical levels, necessitating the comparison of various indicators at the second level. A pairwise comparison matrix was constructed, with twenty experts assigning scores to each indicator based on Saaty's relative importance scale. The relative weights of the indicators were determined by calculating the eigenvalues and eigenvectors of the pairwise comparison matrix. To ensure logical consistency, the consistency of the pairwise comparison matrix was evaluated using the formula , where CR is the consistency ratio, CI is the consistency index, and RI is the random consistency index. A consistency ratio below 0.1 was considered acceptable. Finally, the indicators were ranked based on their respective weights (Fig. 1).

Fig. 1.

The research framework

Statistical analysis

Measurement data were presented as Mean±S.D., while categorical data were expressed as percentages and frequencies. Statistical analyses were performed using SPSS 22.0 and Python 3.11, with significance set at P≤0.05.

Results

HH compliance and theme identification results of janitorial tasks

A total of 122 HH opportunities were recorded, with a compliance rate of 31.97% (39/122) and a correctness rate of 30.77% (12/39). To minimize human bias, we determined the optimal number of themes through a perplexity and coherence line curve. As illustrated in Fig. 2, perplexity declined as the number of topics increased, whereas coherence scores peaked at 13 topics (c_v = 0.542). Although inflection points were also noted around topics 3, 8, and 12, coherence at those points was lower. Hence, 13 was selected as the optimal number of topics. Among them, one was manually identified as non-task related and excluded, resulting in twelve labeled work-task themes. The corresponding theme-word probability distributions are summarized in Table 1.

Fig. 2.

Graphical representation of perplexity and coherence under different topic numbers

Table 1.

Themes and stemmed words generated by LDA

No	Themes	Stemmed terms (20 high-frequency words)
1	Cleaning and disinfection of general patient care areas	ward, mopping, floor cleaning, protocol, waste, room, disposal, stairwell, stains, area, floor, handling, staircase, inspection, handrail, corners, wall, localized area, lobby
2	Cleaning tools management	cleaning tools, janitorial staff, placement, cleaning, inventory check, mopping, inspection, recording, tidiness, wiping cloths, ward, corridor, organizing, contaminated items, garbage, shift rotation, food delivery cart, completion, replacement, water marks
3	Personal protective measures	disinfectant preparation, concentration, available chlorine, disinfectant solution, patients, PPE, cleaning equipment, donning PPE, task completion, cleaning zones, proper use, differentiation, mopping, signage, workflow guidance, reception support, cleaning tasks, ward, windowsill, equipment
4	Public area cleaning	key zones, general cleaning, changing rooms, demand-based tasks, coordination, air conditioning units, public zones, departments, area patrol, corridor, high-touch surfaces, assistance, environmental hygiene, public lobby, seasonal focus, implementation, scrubbing, waste bin, stains, anti-slip mat
5	Cleaning and disinfection of special infection patient care areas	wet wipes, sequential steps, high-priority, severely contaminated, bedside cabinet, patient bed headboard, bedrail, foot of bed, infusion stand, hand towels, medical equipment, oxygen bag, decontamination, preparation, procedure reference, proper use, air mattress, cleaning sequence, protocol, one cloth per use
6	Trash collection and transfer	medical waste, double-layer packaging, clinical waste, temporary storage, centralized disposal, double-bagging, waste generation, patients, waste transport, waste retrieval, patient areas, disinfection, infectious waste container, shift rotation, used items, yellow-labeled bags, storage room, mop cloths, wiping cloths, disposal tools
7	Terminal cleaning and disinfection of patient care areas	terminal cleaning, task execution, decontamination, discharged patients, mirror surfaces, cleaning tasks, responsible personnel, examination room, disinfectant concentration, chemical preparation, administrative surfaces, terminal disinfection, chlorine-based disinfectant, windowsill, tabletops, faucets, janitorial staff, offices, patients
8	Restroom cleaning	disinfectant concentration, chlorine-based solution, surface disinfection, chemical preparation, mopping, placement, safety signs, caution notices, restroom areas, flooring, scrubbing, disinfection, waste, tidying, corners, cleaning cloths, surface maintenance, record-keeping, inventory, rest area
9	Patrol and stain removal	item replacement, garbage bag handling, trash removal, cleaning patrol, sweeping, wards, patient areas, waste retrieval, transport, soiled items, temporary waste station, proper placement, prohibited actions, dropped items, execution, tools, scrubbing, cleaning standards, work attire, implementation protocol
10	Cleaning and disinfection of high-frequency contact surfaces	control panels, chlorine-based disinfectant, surface disinfection, concentration setting, disinfectant prep, switches, wiping, door frames, electric kettle, door handles, elevators, handrails, windowsills, waiting chairs, frequently touched surfaces, completion, trash bins, surface tasks, registration area, stainless steel
11	Operating room cleaning and disinfection	surgical area, flooring, soiled items, operating table, waste removal, mopping, sterile corridor, medical waste, inspection, disinfection, PPE changes, designated personnel, garbage bag handling, surface wiping, disinfectant, chlorine-based agents, water stains, rotating table, surgical mattress, coverage range
12*	Principles of cleaning and disinfection	patient zone, ward, cleaning workflow, preparation, traffic control, PPE protocol, cleaning equipment, movement routing, restricted access, personal protection, surface decontamination, procedure reference, organizing tools, item placement, workstation positioning, utility cart, item fixation, hot water usage, baseboard, utilities
13	Cleaning and disinfection of semi-contaminated and clean zones	task completion, ward, routine task, staff office, corridors, meeting rooms, nurses’ station, treatment room, physicians’ offices, switches, staff-only zones, cleaning process, sequence, countertops, scrubbing, toilet bowl, bed frames, meal tray, cleaning zone, environmental condition

^*for removed

PPE, personal protective equipment

Sensitivity analysis using four random seeds (24, 42, 66, 88) showed coherence variation < 0.1 and an average Jaccard similarity of 0.083. This value indicates variability in the specific keywords representing the topics under different model initializations, a known characteristic of LDA models whose stability is sensitive to corpus characteristics and random seeds [25]. However, it is widely recognized in the academic community that quantitative metrics alone are often insufficient for evaluating topic models; the ultimate measure of a model's success is its human interpretability and applicability for a specific task [26]. For example, a topic might be represented by 'cleaning, ward, floor' in one run and by 'disinfection, room, floor' in another; although the keyword overlap is low, the experts unanimously identified this as the same task. Therefore, our subsequent qualitative assessment by the expert panel confirmed that the core thematic interpretations of these different run results remained consistent. This indicates that despite the keyword volatility, the model reliably captured the same meaningful, underlying work tasks, confirming its suitability for the purpose of this study.

HH moment extraction results for EVS staff

Detailed SOPs for each task were meticulously reviewed to extract relevant HH moments, which were initially categorized into seven distinct types. They were before handling clean items, before cleaning or disinfection, before donning personal protective equipment (PPE), before doffing PPE, after cleaning or disinfection, after touching highly contaminated surfaces or items, and after doffing PPE. The moment "after touching highly contaminated surfaces or items" encompassed scenarios such as the necessity for HH following interactions with patient belongings or contact with waste receptacles. The moment "before donning PPE" encompassed scenarios such as the imperative for HH before the application of gloves, masks, head coverings, and isolation garments. The moments "before removing PPE" and "after removing PPE" aligned with the scenarios included in "before donning PPE".

Basic information of experts

Twenty experts, averaging years in age, predominantly held master's degrees (65.00%) and were mostly deputy seniors (80.00%), with 45.00% possessing over 20 years of experience, split evenly between IPC and nursing management specialties (50.00% each). Refer to Table 2 for details.

Table 2.

Sociodemographic characteristics of consulting expert panel (n = 20)

Characteristic	Category	Frequency	Percentage (%)
Gender
	Male	2	10.00
	Female	18	90.00
Age (years old)
	30–39	5	25.00
	40–49	9	45.00
	50 or older	6	30.00
Education
	Bachelor	4	20.00
	Master	13	65.00
	Ph.D	3	15.00
Professional title
	Senior	4	20.00
	vice-senior	16	80.00
Years of work experience
	> 20	9	45.00
	16 ~ 20	3	15.00
	10 ~ 15	7	35.00
	8 ~ 9	1	5.00
Professional background
	Infection prevention and control	10	50.00
	Nursing management	10	50.00

Expert authority assessment

In this study, all 20 questionnaires distributed during each round of expert correspondence were returned as valid, yielding a positive expert coefficient of 100%. The experts' familiarity scores averaged 0.82 and 0.84 across two rounds, while their overall scores were 0.95 and 0.96, resulting in authority coefficients of 0.89 and 0.90, respectively. The Kendall coefficients of concordance for the two rounds were 0.47 and 0.62 .

Questionnaire revisions and hierarchical analysis results

Following statistical analysis of the first round of inquiries, four indicators were eliminated per the screening criteria: "before donning PPE", "before doffing PPE", "after preparing tools", and "after environmental sorting of waste". The remaining descriptions received no recommendations for modification. Following the second round, all indicators met screening criteria, with no proposed changes. Based on Delphi screening criteria, three final HH indicators were retained. Hierarchical analysis using pairwise comparisons was then conducted to determine the relative weight of each indicator, ranked in descending order as follows: after touching highly contaminated surfaces or items, before handling clean items, after cleaning or disinfection, before cleaning or disinfection, and after doffing PPE (see Table 3).

Table 3.

The results of two round-survey of experts' opinion

Indications	Round 1()					Round 2()					AHP weight
	Importance (Mean±S.D)	CV	Feasibility (Mean±S.D)	CV	Result	Importance (Mean±S.D)	CV	Feasibility (Mean±S.D)	CV	Result
After touching highly contaminated surfaces or items	5.00 ± 0.00	0.00	4.95 ± 0.22	0.05	√	5.00 ± 0.00	0.00	5.00 ± 0.00	0.00	√	0.245
Before handling clean items	4.80 ± 0.52	0.11	4.65 ± 0.67	0.14	√	4.90 ± 0.31	0.06	4.90 ± 0.31	0.06	√	0.233
After cleaning or disinfection	4.55 ± 0.83	0.18	4.75 ± 0.64	0.13	√	4.90 ± 0.31	0.06	4.85 ± 0.37	0.08	√	0.231
Before cleaning or disinfection	4.50 ± 1.00	0.22	4.45 ± 1.05	0.24	√	4.20 ± 0.41	0.10	4.15 ± 0.49	0.12	√	0.146
After doffing PPE	4.95 ± 0.22	0.05	4.85 ± 0.37	0.08	√	4.15 ± 0.49	0.12	4.05 ± 0.51	0.13	√	0.144
Before donning PPE	3.85 ± 1.39	0.36	3.45 ± 1.23	0.36	×	–	–	–	–
Before doffing PPE	3.95 ± 1.19	0.30	3.45 ± 1.15	0.33	×	–	–	–	–
After preparing tools	3.30 ± 1.38	0.42	3.00 ± 1.08	0.36	×	–	–	–	–
After environmental sorting of waste	3.95 ± 1.39	0.35	3.70 ± 1.42	0.38	×	–	–	–	–

"√" for retention, "×" for deletion

Abbreviations: PPE, personal protective equipment

Discussion

In this single-center study, we developed and refined HH indicators for EVS staff in medical institutions utilizing the LDA thematic model, Delphi method, and hierarchical analysis. The resulting HH indicator system encompasses five key indicators: before handling clean items, before cleaning or disinfection, after cleaning or disinfection, after touching highly contaminated surfaces or items, and after doffing PPE, summarized as "two before and three after". This framework offers explicit guidance on when and under what circumstances EVS staff should perform HH in healthcare settings. Furthermore, our study addresses the limitations of prior research, broadening the scope of HH studies to include EVS staff.

The HH indicators proposed herein establish a foundation for enhancing HH management among EVS staff, thus playing a crucial role in preventing HAIs and promoting safety within healthcare environments. The cleaning and disinfection responsibilities of EVS staff vary across different clinical departments, as each department has its own specific needs. The substantial data and complexity gleaned from observing 38 clinical departments were effectively analyzed through the application of the LDA model, enabling comprehensive and objective insights while conserving time and resources.

Although the extracted HH moments from the standardized operating procedures were thorough, their cohesiveness and feasibility require enhancement through expert input. These moments can only be utilized as HH indicators once they have been critically reviewed and endorsed by experienced professionals, ensuring that they are scientific, systematic, reasonable, and practicable. The evaluations from the two rounds of expert correspondence indicate that the experts possess considerable experience and authority in environmental cleaning and disinfection, resulting in consistent and reliable evaluations. The five final HH indicators not only garnered high mean scores but also exhibited a strong consensus among expert opinions.

The HH indicators for EVS staff in this study align with those proposed by the WHO, encapsulating five indicators easily memorable for EVS staff. While the WHO emphasizes the importance of HH before and after patient contact and medical procedures, the indicators for EVS staff focus on HH before and after interacting with various objects and during specific cleaning and disinfection tasks. This distinction arises from the differing occupational contexts and responsibilities of these groups. Indicators that do not align with the workflow of EVS staff impede their understanding and implementation of proper HH practices. Kigozi et al. [27] also underscored the necessity for HH indicators to be tailored to the specific environments of EVS staff.

Moreover, the experts consulted for this study noted that while most EVS staff recognized that PPE was clean before use, they often failed to acknowledge that such equipment, including masks and hats, becomes contaminated after use. A plethora of research has meticulously dissected the donning and doffing procedures of [28–30] PPE by healthcare workers, revealing a pervasive issue of cross-contamination. This underscored the imperative for stringent HH post-PPE removal, crucial for curbing pathogen spread and safeguarding the potency of infection control protocols in healthcare settings [31, 32]. Since EVS staff frequently don and doff gloves, masks, and hats throughout their work, the indicator "after doffing PPE" should be explicitly stated. This assertion is corroborated by the findings of Wenbin et al. [21], where similar indicators were proposed. Both studies agree on three shared indicators "before cleaning or disinfection", "after cleaning or disinfection" and "after doffing PPE". However, this study determined that indicators such as "after preparation of tools" and "after environmental sorting of waste" should be omitted due to their limited applicability and lack of clarity.

Our findings on HH moments should be interpreted within the broader context of evolving cleaning practices and appropriate personal protective equipment (PPE) use, particularly concerning gloves. A crucial distinction must be made between heavy, chemical-resistant gloves and disposable gloves. For tasks involving chemical disinfectants, occupational safety standards mandate the use of chemical-resistant gloves, which are less susceptible to micro-tears than disposable gloves and offer superior protection [33, 34]. We acknowledge the practical reality that EVS staff are unlikely to change these durable gloves after every contact in a standard patient room. In workflows where these gloves are kept consistently moist with a disinfectant, the risk of cross-contamination between rooms or bed-units may be reduced (excluding, of course, rooms for patients under special isolation precautions). Conversely, the use of disposable gloves with chemical disinfectants should be discouraged, both for the safety of EVS workers and because these gloves are not designed to maintain their integrity during prolonged contact with such agents.

Furthermore, as many hospitals transition to cleaning with microfiber and water instead of chemical agents, the role of HH becomes even more critical. In such scenarios, where a disinfectant is not continuously applied, the hands of EVS staff (gloved or not) are a primary vector for pathogen transmission. This shift heightens the importance of adhering to a "clean-to-dirty" workflow. If this principle is consistently followed, changing gloves and performing HH after cleaning each patient room or bed-unit may be sufficient. However, if the workflow is not strictly followed (e.g., cleaning a toilet after a cleaner surface), it is crucial to change gloves and perform HH even within the same room after touching highly contaminated surfaces to break the chain of transmission [5, 35]. Regardless of the glove type or cleaning method, due to the risks of contamination during removal, HH is universally recommended after any glove use [5]. The standard doffing sequence remains a key protocol designed to prevent self-contamination during this process [36].

In the hierarchical analysis of HH indicators, the weights were ranked as follows: after touching highly contaminated surfaces or items, before handling clean items, after cleaning or disinfection, before cleaning or disinfection and after doffing PPE. This ranking reflects the severity of microbial contamination risks associated with inadequate HH practices across different indicators and directs the attention of EVS managers and IPC staff. Notably, the highest-ranking indicator—"after touching highly contaminated surfaces or items"—underscores the critical importance of post-contact decontamination, as EVS staff frequently handle surfaces or materials with a high microbial load. Failure to perform HH at this moment may facilitate pathogen transfer to other areas, tools, or even oneself, thereby undermining environmental hygiene efforts. HH is a critical factor in the prevention of HAIs and the emergence of antimicrobial resistance [37–39]. While considerable efforts have been made in many healthcare facilities to improve compliance with recommended HH practices among healthcare personnel [40–42], the management of HH for EVS staff requires more focused attention and should evolve into a stage of refined management.

Despite recognizing the importance of HH [14, 43, 44], many EVS staff report not receiving explicit training regarding when to perform it. This gap may stem from the absence of clear HH indicators that align with the workflow of EVS staff, hindering the capacity of IPC staff to provide appropriate training. Additionally, the lack of established indicators complicates the assessment of HH compliance among EVS staff, stalling initiatives aimed at continuous quality improvement [18]. The indicators proposed in this study delineate precise moments for HH, forming a solid basis for training.

This research offers quantitative tools and methodologies for monitoring and enhancing the adherence to HH practices among EVS staff, enabling the identification of low-compliance moments and guiding targeted training and supervisory measures. The scientifically grounded indicators provide a robust foundation for the further development of HH standards and policies for EVS staff.

Our study has several limitations that should be acknowledged. First, all observations and data were derived from a single tertiary hospital in China. Given the potential diversity in janitorial workflows, task granularity, and HH opportunities across different countries, healthcare systems, and outsourcing models, the generalizability of our findings may be limited. Second, our Delphi consultation panel did not include frontline EVS staff. Although we mitigated this by involving experienced EVS infection control trainers in the initial workflow analysis to ensure the extracted moments were grounded in practice, we recognize that direct input from cleaners would have provided a more nuanced perspective on end-user practicality.

To address these limitations and advance the field, several avenues for future research are recommended. Foremost, multi-centre validation studies should be prioritized to test and adapt these HH indicators across varied healthcare systems and operational structures. Furthermore, future research should systematically include frontline EVS personnel, for instance through focus groups and qualitative interviews, to gather direct feedback on the practicality of these indicators and to co-design effective implementation strategies. Beyond the timing of HH, future work should also investigate the quality of its performance, as highlighted by ISO guideline on HH, such as rub coverage and dosing consistency [45, 46]. Finally, future research could also investigate the critical infrastructure supporting HH, such as the impact of the ABHR formulation itself [47] and the working characteristics of hand rub dispensers [48].

Conclusion

This exploratory study on HH indicators establishes a measurement standard, training framework, and practical strategy for effectively enhancing HH compliance among EVS staff. It represents a significant advancement toward precise and scientific HH management, emphasizing the necessity for increased attention to the HH practices of cleaning personnel as they transition into a phase of refined management.

Author contributions

Y. L. was responsible for funding acquisition, project administration, resources, and writing – review & editing. F. J. contributed to project administration, supervision, validation, and writing – original draft. L. Y. contributed to conceptualization, data curation, investigation, methodology, validation, visualization, and writing – original draft. H. N., H. W., F. R., and Y. Z. were involved in investigation, methodology, and writing – review & editing. All authors read and approved the final manuscript.

Funding

This study was funded by the Ministry of Education 2021 Humanities and Social Sciences Fund in China (Grant number 21YJA630062); the Research on the Development Path of Smart Healthcare and Health Management in Jiangxia District Communities, Wuhan City (Grant number H20230163); the Research on the Collaborative Mechanism of Primary Healthcare Services under the Background of Health Needs (Grant number H20220099); and the Hubei Provincial Soft Science Research Program(Grant number RKX202500151).

Data availability

Additional datasets analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Informed consent

Informed consent for the data collection and the use of the data was obtained from all subjects..

Ethical approval

Methods in this study were reviewed and approved by the Institutional Review Board (IRB) of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology (No. 20220134) and were in accordance with the 1964 Helsinki declaration and its later amendments, or comparable ethical standards.