Utilizing Process Mapping as a Framework for Identifying and Prioritizing Barriers to Antibiotic Delivery in Resource-limited Settings

Abstract

In resource-limited settings, healthcare providers face unique structural barriers to antibiotic delivery. Process mapping (PM) is a low-cost, low-technology approach used to understand the antibiotic delivery process and systematically identify barriers to timely antibiotic initiation. In this paper, we will use a 5-phase PM framework to provide the readers with a guide to developing antibiotic delivery process maps, identifying barriers to antibiotic delivery, and prioritizing the order in which these barriers should be addressed. We will then use our experience at the Uganda Cancer Institute in Kampala, Uganda as a case study to describe how we used PM to identify barriers to antimicrobial delivery for patients with neutropenic fever. We will also provide information about how to use low-cost supplies and open-access software to develop and analyze the process map. By the end of the paper, the reader will have the necessary tools to develop and analyze their own antibiotic delivery process maps. This will allow the reader to systematically identify barriers to antimicrobial delivery and understand how to prioritize which barriers to address using targeted interventions.

Process mapping is a low-cost technique used to understand the antibiotic delivery process, identify barriers to care, and create interventions to improve timely antibiotic delivery in resource-limited settings.

Among patients with severe infections, the ability to rapidly initiate appropriate antibiotics is critical for reducing morbidity and mortality [1–3]. For patients with severe infections, international guidelines recommend antibiotic initiation within 1–2 hours of presentation. However, delays of more than 6 hours are common throughout low- and middle-income countries (LMICs) [4–6]. This contributes to the high rate of infection-related mortality seen in LMICs. For example, 83–87% of sepsis-related mortality occurs in LMICs [7, 8]. Many factors contribute to delayed antibiotic initiation, including a shortage of healthcare workers per capita [9, 10], a lack of access to effective antibiotics [11], inadequate surveillance structures to guide antibiotic selection [12], higher rates of antimicrobial resistance [13], and high antibiotic costs [14]. While many of these challenges occur throughout LMICs [15], the specific factors that contribute to antibiotic delay vary across settings. Even within the same country, factors may differ significantly between urban versus rural settings or public versus private hospitals. To successfully improve healthcare delivery, interventions must be adapted to the local context and account for available resources [16, 17]. Engaging local stakeholders increases the chance of successful program improvement [18–20]. Process mapping (PM) is an important collaborative approach used to holistically understand the antimicrobial delivery process, identify areas for improvement, and prioritize, which improvement areas to address.

Originally developed in the early 1900s to improve efficiency and safety in business and manufacturing [21], PM has been adapted by various quality improvement methodologies (eg, Lean Six Sigma) to improve healthcare delivery [22]. PM uses visual techniques to create a “map” of a healthcare-related process [23]. PM is task- and system-oriented, focusing on how work is actually done, as opposed to what people think or feel, which is common in traditional focus groups or interviews. It can be used to create an “as-is” or “current state” process map that breaks the process into a series of consecutive steps, showing who is responsible, what is done, when it is done, where each task occurs, the inputs needed for each task, and areas of inefficiency (eg, delays, redundancies, or resource wastages). This map can then be used as a framework to systematically identify ways to improve the quality and efficiency of the process. A recent study of healthcare workers identified 8 potential benefits, including developing a shared understanding of the reality of the process, identifying gaps and improvement opportunities within the system, and engaging stakeholders [24].

Although PM is an important tool for improving healthcare delivery [22], relatively few studies have used PM to improve infection diagnosis and treatment [25–35]. Most of these studies were conducted in high-resource settings and did not provide step-by-step instructions for creating and analyzing a process map. In this paper, we will use our experience at the Uganda Cancer Institute (UCI) to provide a step-by-step guide for using PM to understand barriers to antimicrobial delivery for patients with neutropenic fever (NF) within a healthcare system in a resource-limited setting. NF is a model condition to illustrate how PM can improve antimicrobial delivery, since it provides a predictable and high-risk clinical scenario that is associated with high mortality when antibiotics are delayed [3]. We will organize our guide according to the 5 phases of the conceptual PM framework developed by Antonacci et al (Table 1 [19]) [22]. In the first part of this manuscript, we will provide the readers with an overview of evidence-based practices to develop their own process maps, focusing on the steps to creating and analyzing the map. In the second part of the manuscript, we will describe how we developed and analyzed a process map and identified barriers to antimicrobial delivery for patients with NF at the UCI.

Table 1.

Evidence-based Framework for Successful PM Developed by Antonacci et al With Corresponding Steps

Phase of Process Map	Criteria/Standards(Antonacci et al)	Steps
Phase 1: Preparation, planning, and process identification	Cleary identify the process	1. Determine whether PM is an appropriate tool
		2. Work with key stakeholders to set project goals and appropriate level of process map
		3. Define the start and the end of the process
		4. Select the appropriate data gathering technique(s)
	Include the patient perspective	5. Work with key stakeholders to identify members of the PM team
		6. Be sure include the patient perspective during data collection
	Educate/train the team to use the PM tools	7. Select and train the facilitator
		8. Train the PM team on the PM tools
Phase 2: Data and information gathering	Gather information to inform the PM exercise	9. Collect data using the selected techniques
Phase 3: Process map generation	Gather different perspectives from multiple stakeholder groups	10. Use the data to create the initial draft of the process map
		11. Review the process map to ensure no additional perspectives are required
Phase 4: Process map analysis	Create a clean map	12. Create a clean process map, using PM software if possible
	Gather additional information to include in the final map	13. Review the semi-final process map with the PM team
	Validate the final map with key stakeholders/experts	14. Disseminate the process map to the larger community for review and feedback
		15. Incorporate suggestions and finalize the process map
	Analyze the process map	16. Select data analysis method(s) and use them to identify areas for improvement (eg, barriers, bottlenecks, sources of delay)
Phase 5: Taking it forward	Use the knowledge gained from the PM exercise to develop, implement, and test improvement ideas	17. Use consensus to identify which barriers to address
		18. Work with key stakeholders to design and test an intervention strategy

STEPS TO SUCCESSFUL PM

Phase 1: Preparation, Planning, and Process Identification

PM is only successful if it is preceded by thorough preparation and planning. This includes setting discrete goals and training the members of the PM team. This phase occurs in 3 stages:

1. Clearly identify the process

   • Step 1. Determine whether PM is the appropriate tool: PM is best used when the process of interest is not well understood and there are multiple complex issues leading to process delays.

   • Step 2. Work with key stakeholders to establish project goals: It is important to work with key stakeholders to develop goals that align with departmental and institutional priorities [36]. This ensures local buy-in and increases the likelihood of successful process improvement. While establishing goals, the team will also determine the appropriate level of detail to include in the process map. Supplementary Table 1 explains process map levels and their applications.

   • Step 3. Define the start and end of the process: Healthcare processes are complex and interconnected. Without clearly defining the start and end of the process, it is easy for the PM team to become distracted by related processes that do not affect timely antibiotic administration (eg, drawing blood cultures and patient resuscitation). By defining the beginning and end of the process, PM team can focus on identifying the relevant tasks.

   • Step 4. Select the data gathering technique(s): Numerous data gathering techniques can be used to create the process map. This includes direct observation, focus group discussions (FGDs), in-depth interviews (IDIs), and review of medical records or administrative databases. To create a robust map that accurately reflects the process, a combination of techniques is often used (eg, FGDs and direct observation). Supplementary Table 2 describes the benefits and drawbacks for each technique.

2. Include the patient perspective:

   • Step 5. Work with key stakeholders to identify members of the PM team: The number of people included in the PM team depends on the complexity of the process and data gathering techniques used (Supplementary Table 2). Individuals involved in each aspect of the process should be included (eg, nurses, doctors, pharmacists). If the process is being evaluated at multiple sites (eg, different wards and different healthcare systems), the PM team should include representatives from each. Remember that the people most qualified to analyze the workflow are those who use it every day. Those in leadership may not understand the day-to-day realities of patient care. However, they are more likely to have a “big picture” understanding of ways in which a process integrates into the healthcare system [37]. Working with administrators and clinical leaders to identify and recruit PM team members can also improve leadership buy-in. Table 2 shows various sampling methods that can be used to recruit study participants.

     It is important to consider power dynamics when forming the PM team [38]. Frontline healthcare workers may not feel comfortable speaking openly in front of administrators or clinical leaders. If it is important to gather the perspectives of both frontline healthcare workers and leaders, the team can use IDIs or conduct separate FDGs. Other techniques to reduce power imbalances include training in appropriate facilitation techniques [39] and employing “scaffold focus groups,” which use surveys and progressive FGDs to build comfort in discussing sensitive issues [40].

   • Step 6. Include the patient perspective: In LMICs, patients and their families are key members of the healthcare delivery team [41]. They may be responsible for taking temperatures, filling prescriptions, and mobilizing resources to pay for services. IDIs and surveys are good ways to obtain patient and family perspectives.

Table 2.

Sampling Methods Used to Identify Participants and Cases Involved in a PM Study

Sampling Method and Associated Namesa	Description	Pros	Cons
Convenience sampling Volunteer sampling	Participants volunteer to participate in the research	Easy and efficient Participants who respond are more likely to be engaged in the research	May not recruit the most knowledgeable participants Less likely to generate a representative sample
Snowball sampling Chain sampling	Current participants suggest other people who might be willing to join the study	Practical and efficient Researchers can work with the current participants to recruit participants with specific knowledge of the antimicrobial delivery process Less time to gain trust	Quality of the participants depends upon the quality of the referrals
Purposive sampling Purposeful sampling Judgement sampling Selective sampling	The researcher intentionally chooses individuals with desired knowledge about different aspects of the antibiotic delivery process. Subsets of purposive sampling include: Maximum variation sampling: Participants with a wide variety of experiences selected (eg, different clinical roles, work at different sites). Goal is to understand the full range of experience. Homogenous sampling: Participants or cases with similar experiences selected (eg, a group of nurses) Goal is to analyze the experience of a specific subgroup in depth. Typical case sampling: Cases are selected that are considered “average” or “representative” of the process. Goal is to describe the typical experience. Extreme (deviant) case sampling: Cases are selected that are exceptional or unusual and not representative of the process (eg,. Goal is to uncover patterns/barriers that do not occur during the typical samples. Critical case sampling: Cases are strategically selected that can provide the most information regarding the process (eg, a case with a bad outcome). Goal is to maximize knowledge gathering.	Allows the researchers to select participants with specific knowledge of the antibiotic delivery process Can use a variety of sampling strategies to explore variations in the antibiotic delivery process	May be challenging to find participants with the desired type of knowledge regarding the antibiotic delivery pathway.

^aMultiple sampling techniques can be used within the same study to recruit a multidisciplinary group of participants from each cadre (eg, patients, healthcare providers, administrators, etc.).

Adapted from Gil et al. Qualitative Sampling Methods. J Human Lact. 2020; 36(4): 579–581.

(c) Educate/train the team to use the PM tools:

• Step 7. Select and train the facilitator: Ideally, the facilitator should have a clinical background with relevant experience to understand the process and ask appropriate clarifying questions. The facilitator should have formal or informal training in PM (eg, Lean Six Sigma coursework, online tutorials, and how-to manuals). Regardless of the level of experience, we suggest conducting several practice sessions before beginning data collection.

• Step 8. Train the PM team: Prior to collecting data, the facilitator should orient the PM team. The facilitator should emphasize that participation in the PM exercise is voluntary, responses will remain anonymous, and responses will not jeopardize employment. The facilitator should then explain the project goals and orient team members to the tools and techniques that will be used. This includes explaining any PM notation (Figure 1) [42]. These trainings should also be used to establish meeting expectations, including meeting frequency and duration.

Figure 1.

Summary of symbols commonly used for PM in healthcare. Adapted from Mclaughlan et al. Neurosurgery 2014;75(2): 99–109 [42].

Phase 2: Data and Information Gathering

In this phase, the facilitator should work with the PM team to gather information to create the preliminary process map.

• Step 9. Collect data using the selected technique(s): The amount and type of data collected depends upon the level of process map that is being developed. Supplementary Table 1 provides details regarding the 4 levels of PM and their associated purpose, uses, pros, and cons. The tools used to collect this data depend upon the data collection technique (Supplementary Table 2). Data collection should continue until saturation is reached, which means that no new steps in the process are identified. In our case study below, we describe how we used FGDs and low-cost office supplies to develop our process map. Supplementary Table 3 provides step-by-step instructions with corresponding photographs.

Phase 3: Process Map Generation

In this phase, the team should use an iterative process to ensure that different perspectives from multiple stakeholders groups are included.

• Step 10. Use the data to create the initial draft of the process map: Once the data has been collected, the facilitator should create an initial draft of the process map. This will be refined and validated in Phase 4.

• Step 11. Review the process map to ensure no additional perspectives are required: To gain an accurate understanding of the process, it is important to include a wide variety of perspectives. When developing the initial draft of the process map, the data collection team should regularly review it to identify the missing voices and collect additional data as needed.

Phase 4: Process Map Analysis

In this phase, the team should validate and analyze the process map in 4 distinct stages:

1. Create a clean process map

   • Step 12: Create a clean process map, ideally using PM software: A semi-final copy of the process map should be created for dissemination and validation. The easiest way to do this is using PM software. When selecting the appropriate PM software, people working in LMICs should consider licensing costs and internet access requirements. For example, Microsoft Visio has a higher up-front cost but only requires a one-time licensing fee. It also has a desktop application that can run offline. Other software, including Lucidchart and SmartChart, use a subscription model and are only accessible online. There are also several free software options (eg, Draw.io, Camunda, Valtimo). Many of the free options are only accessible online, although Draw.io can be downloaded and used offline. If the team does not have access to computers, they can use butcher paper, sticky notes, and markers to create a clean map that can be physically displayed or photographed and disseminated via smartphones. For standardization purposes, conventional PM notation should be used (Figure 1).

2. Gather additional information to include in the final map

   • Step 13. Review the semi-final process map with the PM team: After creating the clean copy of the map, the facilitator should review and finalize the map with the PM team to ensure that it accurately and completely reflects the current antimicrobial delivery process.

3. Validated the final map with key stakeholders/experts

   • Step 14. Disseminate the process map to the larger community for review and feedback: The final map should be disseminated to a broader group of healthcare workers for validation and feedback. Ways to disseminate the map include hanging it in high-traffic areas, sending a copy using electronic methods (eg, email, WhatsApp), and presenting it at clinical and research conferences.

   • Step 15. Incorporate suggestions and finalize the process map: The feedback received in Step 14 should be incorporated into the final version of the process map.

4. Analyze the process map

   • Step 16. Select data analysis method(s) and use them to identify areas for improvement

     The process map should be used as a framework to identify barriers, bottlenecks, sources of delay, duplications, and unnecessary steps in the antibiotic delivery process [43]. It is also important to look for “facilitators” to antibiotic delivery, things that are going well and should not be changed. The most common analysis methods include failure modes and effects analysis (FMEA) [44–46], Fishbone (Ishikawa) diagrams [47], and semi-structured IDIs or FGDs. Table 3 provides instruction on each of these methods.

Table 3.

Common Methods for Identifying Barriers to Antimicrobial Delivery After PM has Been Completed

Healthcare Failure Modes And Effects Analysis (HFMEA) (a.k.a. Fail Modes Effects Analysis (FMEA)/Failure Modes, Effects, and Critical Analysis (FMECA))
Description	When to Use	Pros	Cons	Examples
Uses a standardized step-by-step approach to identify “failure modes” (ways in which the process can fail) and assess their effect upon the antimicrobial delivery process as follows: The facilitator walks the team through the process map For each step, the team generates a list of failure modes. Each failure mode is scored according to a) severity and b) probability of occurrence. The scores are multiplied to produce a hazard score for each failure mode. The failure modes are ranked by severity according to hazard score Decision tree analysis is used to determine which failure modes and potential causes should be addressed. Note: HFMEA has a formal process that can be used to identify actions and outcomes	When a comprehensive analysis of a complex, multistep process is desired	Standardized data collection technique that is easy to follow Multifaceted analysis of the ways in which the failure modes affect timely antibiotic delivery Risk Priority Number (RPN) ranking system makes it easy to identify which failure modes contribute most to antibiotic delay. Can be combined with standardized process to address problems (HFMEA)	Can be tedious and time-consuming for complex processes Difficult to use with large, multidisciplinary groups	Howard I, et al. Understanding system-focused barriers to the identification and reporting of medication errors and adverse drug events in emergency services. Drugs & Therapy Perspectives. 2019. 35:285–295 [48] Training/Guides VA Center for Safety. Healthcare failure mode and effect analysis (HFMEA): A simple step-by-step guide. 2021. Available online at: https://www.patientsafety.va.gov/docs/joe/Step-by-Step-Guidebook-HFMEA-January2021.pdf58 Centers for Medicare & Medicaid Services. “Guidance for performing failure mode and effects analysis with performance improvement projects.” 2020 [49]

Example of an HFMEA data collection tool with associated hazard scoring system

Fishbone diagram (a.k.a. Ishikawa diagram, Root cause diagram)
Description	When to use	Pros	Cons	Examples
A diagraming technique used to identify the root causes of the problem but grouping these causes into different categories. Common categories used in healthcare-related diagrams include staff/people, policies/procedures, equipment/supplies, and environmental factors.	Use when a thorough analysis of a simpler process is desired	Can identify the root cause of the problem Promotes collaboration Doesn’t require extensive training or supplies Visual representative of complex relationships Uses a structured approach	Technique doesn’t work well on complex processes Doesn’t have a quantitative component to measure impact of each cause Difficult to use with large, multidisciplinary groups	El Feghaly, R, et al. “A quality improvement initiative: reducing blood culture contamination in a children's hospital.” Pediatrics. 2018; 142(4): e20180244 [27]. Training/Guides American Society for Quality (ASQ). “Fishbone.” Available online at https://asq.org/quality-resources/fishbone47 Minnesota Department of Public Health. “Fishbone Analysis” Available online at https://www.health.state.mn.us/communities/practice/resources/phqitoolbox/fishbone.html [50]

Example of an Sample of a fishbone diagram for the antimicrobial delivery process

Semi-structured in-depth interview or focus discussion groups
Description	When to use	Pros	Cons	Examples
The research team uses a semi-structured interview guide to solicit information regarding barriers and facilitators to antibiotic delivery. Prior to data collection, the research team should develop an interview guide. If desired, the team can use structured frameworks (eg, CFIR and TDF) to ensure that the appropriate categories are addressed. The research team conducts a series of IDIs or FDGs using the interview guide. During data collection, the process map can be used as a resource for participants to refer to while they consider barriers to care.	When you would like input from a larger number of participants and multidisciplinary groups in a timely manner	Research team can use knowledge gained to develop targeted interview guides May be less time and labor intensive than other techniques Easier to incorporate feedback from more individuals and from multidisciplinary groups	Less structured than other approaches May not identify as many barriers to care, as it relies on predeveloped questions and direction of interview Depending on how the questions are structured, may not get to the root cause of the underlying barrier Doesn’t have a quantitative component to measure impact of each cause	Aaronson E. et al. Mapping the process of emergency care at a teaching hospital in Ghana. Healthcare: The journal of delivery science and innovation. 2021; 5(4): 214–220 [51] Training/Guides: Bradley, EH, et al. Qualitative data analysis for health services research: Developing taxonomy, themes, and theory. Health Services Research. 2007; 42(4): 1758–1772 [52]

Example of semi-structured interview questions based on CFIR domains

In your own words, please tell me what you know about [your hospital's] Guidelines for Febrile Neutropenia. Where can you find these guidelines? (CFIR Domain/Construct: Inner Setting: Access to Knowledge and Information) Based on your experience, how often do you think availability of supplies is a major issue that holds up timely antibiotic administration? Which supplies seem to have the most issues with shortages? (CFIR Domain/Construct: Inner setting/Available resources) What do you think are the reasons for these shortages? Are there factors outside of [your institution] that contribute to these shortages? (CFIR domain/Construct: Outer setting/Environmental factors) How accurately do you think patients and their family members can identify when a patient is febrile? (CFIR Domain/Construct: Individuals/Capability) From your experience, which actions have taken place at your institution to prioritize neutropenic fever management or to make neutropenic fever management easier? If you haven’t seen any, why do you think this is the case? (CFIR Domain/Construct: Inner setting/Relative priority) Examples of prioritizing would include goal setting by institutional leadership (eg, key performance indicators selected), tracking of neutropenic fever management, rewards or punishments tied to rapid identification and treatment of neutropenic fever etc. Based on the process that was described in the antimicrobial delivery process map, we understand that patients and their family members must sometimes travel off campus to purchase antibiotics that are not in stock at the hospital. Tell me a little bit about ways in which patients get the money to purchase these supplies. (CFIR Domain/Construct: Inner Setting/Available resources) a. How often are patients able to get enough money to purchase the recommended antibiotics? What happens if they cannot? (CFIR Domain/Construct: Inner Setting/Available resources)

Phase 5: Taking it Forward

Successful PM must include a plan for further action [22]. In this phase, the team should use the knowledge gained from the PM exercise to develop, implement, and test strategies to improve the antimicrobial delivery process.

• Step 17: Use consensus to identify which barriers to address: There are various ways of determining which barriers to address, ranging from informal meetings with stakeholders to formal ranking exercises. Table 3 includes an example of using Healthcare FMEA analysis to rank the barriers according to severity and determine whether an intervention should take place. Another formal method for ranking barriers is the Prioritizing Implementation Barriers Toolkit (available at https://impscimethods.org/toolkits/prioritizing-implementation-barriers-toolkit) [53]. Regardless of the method used, researchers should engage with key stakeholders, including patients and their families, to reach a consensus regarding which barriers to address. An overview of commonly used consensus techniques is contained in the paper by Arakawa et al [54].

• Step 18: Work with key stakeholders to design and test an intervention strategy: After identifying the highest priority barriers, the research team should work with key stakeholders to design and test an appropriate strategy to improve timely antibiotic delivery. These improvement strategies often include a bundled group of interventions designed to address the highest-priority barriers. Antimicrobial stewardship teams should use evidence-based frameworks, such as quality improvement frameworks or implementation science frameworks [55–57], to design and test these strategies. While designing and testing improvement strategies is beyond the scope of this manuscript, we would like to highlight the importance of including implementation outcomes when evaluating real-world interventions (Supplementary Table 4) [58]. Many interventional studies done in LMICs were designed by researchers from high-resource settings, focus on efficacy/effectiveness endpoints, and do not assess whether the intervention is likely to succeed within the local healthcare setting. Collecting implementation outcomes from the full range of people involved in the intervention (eg, frontline healthcare workers, administrators, patients, and caregivers) increases the chance that interventions will be integrated into clinical practice. It also provides researchers with information about how to tailor interventions to improve uptake and sustainability.

PM CASE STUDY: GUIDELINE-RECOMMENDED ANTIBIOTIC DELIVERY AT THE UCI

As an illustration of how to develop and analyze a process map, we present the following case study of a process map that we developed to understand barriers to antibiotic delivery for patients with NF at the UCI.

Study Setting

Our project took place from February 2023 to December 2024 at the UCI. Located in Kampala, Uganda, the UCI is the National Cancer Referral Center and African Development Bank's East African Center of Excellence in Oncology. More than 10 000 patients are treated annually in the outpatient clinics and in the 100-bed inpatient hospital. Previous studies at the UCI found that patients with NF had a case fatality ratio of 43–66%, which is 4–6 times higher than among patients in high-resource settings [6, 59]. These studies also found that it took a median of 3 days for guideline-recommended antibiotics to be initiated [6]. To better understand these antibiotic delays, we created and analyzed an antimicrobial delivery process map for patients with NF.

Phase 1: Preparation, Planning, and Process Identification

Prior to creating the process map, we surveyed UCI healthcare workers [60] and conducted informal discussions with UCI leadership to understand whether improving antibiotic delivery was a priority. We found that the UCI antibiotic delivery process and the barriers to guideline-recommended antibiotic initiation were poorly understood (Step 1). We worked with UCI leadership to develop 2 project goals: (1) understand the current process of antibiotic administration for inpatients with NF and (2) determine the barriers and facilitators to rapid guideline-recommended antibiotic administration (Step 2). Based on these goals, we decided to create a Level 3 swim lane process map beginning at the time that an inpatient developed NF and ending when the first dose of guideline-recommended antibiotics was administered (Step 3) (Supplementary Figure 1). We also decided to create a Level 2 process map to provide a high-level overview of the antibiotic delivery subprocesses for administrators and health systems researchers (Figure 2). We used FDGs to encourage maximum group consensus and efficiency of data collection (Step 4).

Figure 2.

Level 2 process maps summarizing the subprocess of antibiotic delivery among inpatients with NF at the Uganda Cancer Institute (Kampala, Uganda) during the daytime (A) and night time (B). An asterisk (*) indicates a step in which patient family members are involved.

To recruit the PM team, we worked with the in-charge nurse and the physician head of each inpatient ward (Step 5). We used maximum variation sampling, which is a type of purposive sampling, to select the PM team. Our goal was to recruit participants with separate roles in the process (eg, doctors, nurses, pharmacists, etc.) and different experiences within each role (eg, experience working day vs night shifts and weekdays vs weekends). We recruited 13 participants to our PM team: 1 doctor and 1 nurse from each of the 4 inpatient wards, 1 emergency department doctor, 2 pharmacists, 1 laboratory worker, 1 phlebotomist, and 1 information technologist. Given time constraints, we did not include patients in these FGDs. However, in the future we will conduct IDIs with patients and families (Step 6). The FGDs were conducted by EAG, a board-certified infectious diseases physician and consultant at the UCI (Step 7). EAG received training in PM through online workshops and direct observation of PM FGDs. During the first FGD session, the facilitator oriented the PM team to the purpose of the study and standard notation used (Step 8).

Phase 2: Data and Information Gathering

We conducted a series of 90-minute FGDs, in which we used sticky notes and butcher paper to create antimicrobial delivery process maps (Step 9) (Supplementary Table 3). To capture process variation, we developed 3 maps: one showing the process during weekdays, one showing the process during nights, and one showing the process on weekends. Given the complexity of the process, we broke our PM team into 2 separate groups. Group 1 generated the process map from fever onset to antibiotic prescription. Group 2 generated the process map from antibiotic prescription to antibiotic administration. We used a combination of individual exercises and group consensus to develop each process map in an iterative process as described in Supplementary Table 3. When disagreement arose, the facilitator worked with group members to verbally reach a consensus. If the group could not reach a consensus, the facilitator asked for input from the other PM group. It took a total of 10 FGDs (5 with each group) to create the initial process maps. Supplementary Table 3 shows the methods and Supplementary Appendix 1 contains the FGD guide we used to generate the process map.

Phase 3: Map Generation

After collecting the data, we created the semi-final process map (Step 10). The facilitator walked each PM team through each step in the process to ensure that the final process map was accurate and to determine where additional input was needed (Step 11). When additional input was needed, the facilitator worked with the other PM group to clarify the steps in the process map.

Phase 4: Process Map Analysis

We used Lucidchart (lucid.co) to create a digital copy of our antimicrobial delivery process map (Step 12) and conducted one additional FGD session to review and finalize this map (Step 13). To validate the process map, we disseminated it to UCI healthcare workers by hanging copies in high-traffic areas, sending an electronic version through email and WhatsApp groups, and presenting the map at UCI meetings and conferences (Step 14). We used this feedback to create the final process map (Step 15), which can be viewed at https://lucid.app/lucidchart/ee088eab-188b-4d58-8f6c-cd37b47c8d4c/view or Supplementary Appendix 2.

To analyze the process map, we used a modified version of the healthcare failure modes effect analysis (HFMEA) (Step 16). We conducted another series of 5 FGDs, one each with UCI doctors, nurses, pharmacists, laboratory personnel, and administrators. To avoid desirability bias, we grouped the participants by role. The facilitator walked the participants through the process map to create a list of “failure modes” (ways each step could fail) and a list of “facilitators” (factors that help antibiotic delivery). We used deductive qualitative analysis to group the failure modes and facilitators into categories according to the theoretical domains framework (TDF) [61] and Consolidated Framework for Implementation Research (CFIR) [62].

Overall, we identified >50 unique barriers to antibiotic delivery. Many of these failure modes fell under the categories of knowledge (eg, knowledge of the antibiotic recommendations in the UCI NF guidelines, knowledge regarding the critical importance of starting antibiotics for patients with NF), local conditions (eg, high prevalence of multidrug-resistant bacteria that do not respond to locally available antibiotics), communication (eg, lack of communication between pharmacists and doctors regarding available antibiotics, language barriers between patients and healthcare workers), available resources (on-site antibiotic stock-outs, available funding to procure antibiotics), physical infrastructure (eg, the on-site pharmacy is not open at night; limited antibiotics available on the inpatient ward), UCI policy (eg, antibiotics can only be procured after a physician has written a prescription) and work infrastructure (high patient-to-provider ratios). Nursing staff tended to identify barriers that were more related to staffing shortages, communication issues (eg, challenges in contacting doctors), and on-site supplies. Doctors tended to identify challenges with understanding which medications were in-stock and issues with procuring the appropriate antibiotics. Pharmacists tended to identify challenges with prescribers understanding the appropriate NF guidelines and issues with antibiotic stock-outs. A detailed discussion of the barriers and facilitators to antibiotic delivery that we identified during our PM exercise will be described in a future manuscript.

Phase 5: Taking it Forward

While developing and testing strategies to improve timely antibiotic delivery is beyond the scope of this manuscript (Steps 17 and 18), one key finding we would like to highlight from our process map is the role of the patient's family as a core member of the healthcare delivery team. Family members were involved in 69% (9/13) of antimicrobial subprocesses during the daytime (Figure 2). They were responsible for measuring temperatures, alerting the nurse when there was a fever, filling prescriptions, and procuring money to purchase antibiotics at local pharmacies. Although the patient's family is integral to healthcare delivery in LMICs, the role that family members play in the antibiotic administration process is rarely highlighted. Our findings underscore the importance of including patient family members when developing strategies to improve timely antibiotic delivery in resource-limited settings. The next steps in our study include using HFMEA to prioritize the identified barriers and using this to develop strategies to improve timely antimicrobial delivery (Steps 17 and 18).

DISCUSSION

PM is a tool that can be used by antimicrobial stewardship teams to understand infection diagnosis and treatment workflows and determine how to improve these processes. In this paper, we have provided readers with a discrete example of how our group used PM to comprehensively map out the antimicrobial delivery for patients with NF at the UCI and to understand the specific factors associated with delayed antimicrobial initiation. This framework can be adapted and used by antimicrobial stewardship teams working in healthcare systems throughout LMICs. The PM framework described in this manuscript is a simple, low-cost tool that can be added to the antimicrobial stewardship arsenal and used to improve the prevention, diagnosis, and treatment of other infectious syndromes (eg, pneumonia, sepsis, and preoperative infection prophylaxis). Its strengths include (1) the ability to incorporate the viewpoint of multiple stakeholders, (2) adaptability to the appropriate level of detail needed to understand the process, (3) low-cost, and (4) reliance on local experience to understand systems-based problems. In our experience, we found that PM was feasible within the context of the local healthcare system using locally available resources. Thus, PM is a tool that antimicrobial stewardship teams can easily and sustainably embed within hospital quality improvement structures, allowing the stewardship team to respond to local challenges that lead to suboptimal care. It is important to note that the PM framework is limited to identifying and prioritizing barriers to care; it does not provide guidance regarding best practices in developing and implementing improvement strategies. The antimicrobial stewardship team should use alternative frameworks (eg, quality improvement, implementation science frameworks) to develop and implement process improvement strategies. The team should evaluate the success of these strategies using both traditional effectiveness/efficacy outcomes and the implementation outcomes described in Supplementary Table 4.

CONCLUSION

In this paper, we discussed ways in which PM can be used to systematically understand the antimicrobial delivery system, identify barriers to antibiotic administration, and use this to determine which barriers to address using targeted interventions. We provided practical guidance about how to conduct this process using low-cost supplies based on our example from the UCI. It is our hope that using this stepwise process will allow antimicrobial stewardship teams to work with key stakeholders to identify areas for systems improvement and develop appropriate strategies to address barriers to antibiotic delivery within the local healthcare setting. Ultimately, this will allow healthcare teams to sustainably and effectively improve timely antimicrobial delivery within the context of resource-limited settings. The framework described in this manuscript provides a scalable and adaptable approach for improving the antibiotic delivery process across diverse healthcare settings in LMICs.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.