Rethinking paediatric sepsis care through local provider voices and lived systems: a mixed-methods study in two hospitals in Ghana

Abstract

Background

Paediatric sepsis remains a significant cause of mortality in low- and middle-income countries (LMICs), where health systems are often resource-constrained. Global sepsis protocols, although effective in high-income settings, may not be well-suited to LMIC contexts.

Methods

We conducted a mixed-methods study in two hospitals in the Ashanti Region of Ghana: Komfo Anokye Teaching Hospital (KATH) and Presbyterian Hospital, Agogo (PreHA). Specifically, we conducted a retrospective chart review, followed by key informant interviews with clinical staff, and integrated our findings with a previously published situational analysis. Qualitative data analysis employed the Three Delays Framework and the Donabedian Model to identify locations and causes of delays in care.

Results

Seventy-one charts met the inclusion criteria for review, having a history of fever or hypothermia and complete vital signs documented (16 from PreHA, 55 from KATH). Despite KATH managing more severely ill patients with higher sepsis scores and longer stays, mortality rates were similar at both sites. The chart review highlighted gaps in documentation and inconsistent care processes. Key informant interviews revealed themes such as provider altruism, community financial support and the positive role of research collaborations, while also illustrating systemic delays linked to financial and resource constraints.

Conclusion

Paediatric sepsis care in Ghana is influenced by complex and interconnected structural, cultural and procedural factors. Our findings indicate that contextually adapted care pathways are crucial for improving sepsis outcomes in resource-constrained settings. Co-designed interventions, rather than wholly imported protocols, may offer a more sustainable approach to strengthening health systems in LMICs.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Paediatric sepsis is a major cause of morbidity and mortality globally, especially in low- and middle-income countries.

• Standardised sepsis bundles have improved outcomes in high-income countries; however, they pose challenges during implementation in resource-limited settings.

WHAT THIS STUDY ADDS

• Triangulated data show systemic and structural barriers drive delays in care-seeking, access and treatment.

• Health system limitations have a more significant impact on clinical outcomes than individual hospital capacity.

• Locally rooted education, infrastructure and participatory approaches are needed to improve sepsis care.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Encourages region-specific adaptations of sepsis guidelines.

• Highlights the role of research funding and community resilience in supporting resource-limited systems.

• Reinforces the need for implementation models grounded in the lived experiences of providers and patients.

Introduction

Sepsis remains a leading cause of paediatric mortality globally, disproportionately impacting low- and middle-income countries (LMICs).¹ Despite international efforts, such as the Surviving Sepsis Campaign,¹ and various risk stratification tools, including the quick Sequential Organ Failure Assessment (qSOFA)² and the Phoenix Sepsis Score,³ most protocols are based on evidence generated in high-income countries.⁴ These tools often assume the presence of advanced diagnostics, intensive care infrastructure and highly trained personnel, making their direct application in LMIC settings challenging.

Emerging evidence shows that efforts to address paediatric sepsis in LMICs have often focused on provider education and supporting clinical decision-making.^{5 6} However, such interventions may be insufficient when implemented in isolation. Structural issues, including financial barriers, referral gaps, resource limitations and sociocultural health beliefs, contribute to significant delays across the care continuum.⁷ As a result, children often present late in the disease process,⁸ and care delivery is constrained by fragmented health systems.⁹

To understand these overlapping barriers, we propose a unified conceptual model that combines the Three Delays Framework with the Donabedian Model of healthcare quality. The Three Delays Framework, originally developed to understand maternal mortality, identifies critical points where care may be delayed: (1) deciding to seek care, (2) reaching appropriate care and (3) receiving quality care.¹⁰ The Donabedian Model^{11 12} provides a complementary structure to categorise these delays as issues of structure (eg, resources, infrastructure), process (eg, provider behaviour, workflows) and outcomes (eg, morbidity and mortality). Together, these frameworks enable a more nuanced, systems-level understanding of how sociocultural and structural barriers interact across the entire care pathway.

This explanatory mixed-methods study explores how these barriers manifest in the care of paediatric sepsis at two hospitals in Ghana: a tertiary referral centre and a district-level hospital. By triangulating data from retrospective chart reviews, key informant interviews and a previously published situational analysis,¹³ we seek to move beyond algorithmic, imported models of care. Instead, we present an integrated, locally informed framework that emphasises the complex interaction of contextual factors in paediatric sepsis care and argues that provider education alone is insufficient to improve outcomes without addressing deeper structural and sociocultural determinants.

This study was originally conceptualised as a Type I hybrid effectiveness implementation trial designed to evaluate the impact of paediatric sepsis education intervention on provider knowledge and clinical outcomes. However, during the early stages of implementation, it became evident that substantial systemic barriers, particularly missing documentation and inconsistent quantitative records, limited the feasibility of outcome evaluation. Consequently, the design evolved into an explanatory mixed-methods study, drawing more heavily on qualitative and triangulated approaches to situate the education intervention and chart review findings within a broader systems framework. Positioning the study this way ensures transparency about its evolution and highlights the importance of contextual adaptation in health systems research.

Study design evolution

The study began as a Type I hybrid effectiveness-implementation trial¹⁴ to assess the impact of paediatric sepsis education intervention on provider knowledge and clinical outcomes. Early findings from a situational analysis, however, highlighted pre-hospital factors and systemic constraints that influenced care pathways. When a retrospective chart review was undertaken, significant missing and inconsistent data limited the ability to reliably quantify outcomes. In response, the quantitative component and key informant interviews were expanded to serve as the central explanatory elements of the study. This shift was reviewed and agreed on by the research team and local collaborators, and the study was restructured as an explanatory mixed-methods design guided by the Good Reporting of Mixed Methods Study (GRAMMS) framework.¹⁵ This evolution reflects the complexity of conducting sepsis research in resource-constrained settings and strengthens the study’s ability to capture systemic barriers across the care continuum. A visual representation of the study components is shown in figure 1.

Figure 1. Study components. CFIR, Consolidated Framework for Implementation Research; KATH, Komfo Anokye Teaching Hospital; PreHA, Presbyterian Hospital, Agogo.

Study sites

This study was conducted at the paediatric departments of two hospitals in Ghana: Komfo Anokye Teaching Hospital (KATH) and Presbyterian Hospital, Agogo (PreHA).

Ethical approval

As the chart review was retrospective, informed consent from patients was waived. Key informants gave verbal consent for the qualitative interviews.

Study components

Education

Modules were developed in collaboration with OPENPediatrics, peer-reviewed and co-designed by members of the study team.¹⁶ The training was delivered in real time with interactive sessions as well as pre-test and post-test to assess knowledge gains among frontline health staff (20 February and 26 March 2024, at PreHA, and on 10 October and 17 October 2024, at KATH). While not all key informants participated in the educational component, the sessions contributed to broader clinical awareness. They were integrated into routine clinical operations through engagement with the leadership of both paediatric departments. Pre-training and post-training assessments were conducted to evaluate immediate knowledge gains. These data will be presented in a separate manuscript focusing specifically on the educational intervention and are therefore not reported here. While the educational modules were available to a broad group of clinical staff, qualitative interviews did not include all participants. Instead, we purposively sampled across roles and cadres to capture a diverse range of perspectives on paediatric sepsis care, rather than conducting interviews with all module participants.

Situational analysis

This component has been previously published.¹³ Briefly, the situational analysis used a structured environmental scan to assess resource availability, staffing, infrastructure and system-level readiness for paediatric sepsis care at KATH and PreHA. It employed the Donabedian model and the Consolidated Framework for Implementation Research (CFIR)¹⁷ to evaluate inner and outer setting determinants. These findings informed the adaptation of the educational intervention’s implementation, as well as the interpretation of both quantitative and qualitative study components.

Retrospective chart review

The chart review was conducted at both PreHA and KATH for patients presenting to the paediatric departments between January and April 2023. The inclusion criteria included paediatric patients presenting within the period, aged 3 months to 14 years, who presented with a documented fever (greater than or equal to 38°C) or hypothermia (less than 36°C). They also must have complete initial vital signs documentation. Charts were excluded if they lacked key data such as temperature, heart rate, respiratory rate or mental status assessment. In total, 187 charts were screened across the two sites, of which 116 were excluded due to incomplete or missing vital sign documentation (temperature, heart rate, respiratory rate or mental status). The final analytic sample comprised 71 charts (16 from PreHA and 55 from KATH). Missing data were most frequently related to vital signs, particularly blood pressure and mental status assessments, which are known to be inconsistently documented in these settings. We acknowledge that this missingness may not have been random; it is possible that sicker children or those with more rapid deterioration were less likely to have complete documentation, while less critical cases may also receive less detailed recording. We therefore interpret missing data not only as a limitation but as a proxy indicator of structural and workload challenges in documentation practices.

Key informant interviews

Interviews were guided by a semistructured tool that drew on the Three Delays and Donabedian models with prompts covering provider role, clinical experiences with paediatric sepsis, enablers and barriers to care and reflections on system-level challenges. The qualitative approach was interpretivist, seeking to understand how frontline providers construct meaning around sepsis care within their clinical and cultural context. A hybrid-deductive process was employed. Twelve (12) key informant interviews with frontline healthcare staff were conducted between September 2024 and May 2025 across both sites following informed consent. Five (5) key informants at KATH included medical officers (2), resident physician (1), specialist paediatric nurse (1) and senior general nursing officer (1). The seven (7) interviews at PreHA involved staff nurses (3), senior staff nurses (2), physician assistant (1) and staff midwife (1) seconded to the department of child health. Participants were recruited through purposive and theoretical sampling to capture a diverse range of clinical roles, perspectives and experiences. Both paediatricians at PreHA were part of the research team and refrained from participating; therefore, those recruited at PreHA represented mid-level frontline staff. The inclusion criteria were employment in clinical teams at either study site and direct experience caring for paediatric patients admitted with sepsis. Interviews were conducted either in person or virtually, using a semistructured guide, lasting about 30 min. A token amount of 50 Ghana Cedis (US$3.5) was provided to compensate KIIs for their time. The interviews were conducted in English. Transcripts were subsequently coded using a hybrid approach that combined deductive and inductive thematic analysis.

Conceptual framework

To better understand how barriers to sepsis care manifest and interact in resource-constrained environments, we applied an integrated framework¹⁸ which draws on the Three Delays Framework and a health systems dynamics perspective to map paediatric illness trajectories across the care pathway, from symptom onset to facility-based treatment, while analysing how both individual/community decisions and health system functionality shape each phase.

We sought to understand how systemic and contextual elements shape clinical care. We adapted the Three Delays Framework, commonly used to evaluate access to maternal and emergency care, to represent the sequential barriers patients face: (1) the decision to seek care, (2) the ability to reach appropriate care, and (3) the quality and timeliness of care once at a facility. To avoid analysing these delays as isolated events, we embedded this pathway within a health systems lens informed by the Donabedian model of quality care: structure (eg, resources, staffing, infrastructure), process (eg, documentation, diagnostics, clinical protocols) and outcome (eg, mortality, recovery, delays in care).

This integrated approach enabled us to identify not only when and where delays occur, but why they persist, understanding challenges in sepsis care as emerging from the interaction of cultural norms, informal coping strategies, infrastructure limitations, provider workloads and financing models. Rather than treating patient-level decisions and facility-based care quality as distinct domains, we conceptualised sepsis care as a continuum in which barriers are cumulative, non-linear and systemically embedded.

The framework served as a guide for both data collection and analysis, organising our triangulated findings, that is, retrospective chart review, key informant interviews and a previously published situational analysis, into a cohesive systems narrative. A visual representation of this integrated care pathway is shown in figure 2.

Figure 2. Integrated framework.

Methods and collection

Quantitative data were obtained through retrospective chart reviews conducted at two health facilities, and the data were disaggregated by site. This provided descriptive insights into patient demographics, sepsis severity (as measured by the qSOFA score), and clinical outcomes. The qualitative component added critical context through triangulation, helping to interpret patterns and limitations in the chart data.

qSOFA was selected a priori as a pragmatic severity score, given its reliance on bedside clinical parameters and the availability of vital signs in most records. While qSOFA has limited sensitivity in paediatric populations,² it was feasible for use in this setting. During the study period, the Phoenix Sepsis Score was published,³ and we attempted to apply it retrospectively to enhance risk stratification. However, core components of the Phoenix score, such as lactate, D-dimer and mechanical ventilation status, were frequently undocumented. Due to these data limitations and infrastructure constraints, the Phoenix Sepsis Score was deemed unsuitable for this context.

Given the frequent lack of paediatric blood pressure monitoring due to a shortage of appropriate cuff sizes, we modified the qSOFA score by substituting hypotension with tachycardia, a well-established early marker of paediatric clinical deterioration. This modification is consistent with the Paediatric Sepsis Consensus Conference and Surviving Sepsis Campaign guidelines, which recommend adapting diagnostic tools for paediatric populations and local contexts.^{1 4} Fever was also added based on clinical input from local providers, as it was commonly used as a marker of sepsis in practice. The resulting modified qSOFA included four components: fever, tachycardia, tachypnoea and altered consciousness (as assessed by Glasgow Coma Scale). Each was scored as one (1) point, resulting in a total score ranging from 0 to 4. Scores were based on initial assessments, with a repeat measurement taken 24 hours after presentation. The highest score was used in the analysis.

Initially, the study aimed to collect a wide range of variables, based on input from a multidisciplinary team. These included ethnicity, distance from hospital, age, sex, mid-upper arm circumference, nutritional status, duration of admission, duration of illness prior to admission, presumed initial diagnosis, referral status, interventions prior to and within the first hour of admission, provider types, initial and follow-up sepsis scores, early diagnostics, timing of laboratory result return and hospitalisation outcomes. However, due to inconsistent documentation, many variables could not be reliably extracted. After a mid-point review and further discussion with the team, the analysis was limited to charts with sufficiently complete data. These included age, sex, sepsis severity score (highest in the first 24 hours), length of stay (LOS), duration of illness before presentation, duration of admission and outcome (death or discharge).

Qualitative data were collected through recorded semistructured interviews, which were conducted in English and subsequently transcribed. These key informant interviews explored provider knowledge, workflows and system-level barriers to care. Interviews were guided by a semistructured tool that focused on the provider’s role, clinical experiences with paediatric sepsis, and perceived enablers and obstacles to high-quality care. Participants were asked to describe both successful and challenging cases, as well as gaps in personnel, infrastructure or system processes. They were also encouraged to identify strengths in current sepsis care practices. These insights were essential for understanding real-world challenges and informing future improvements, especially in low-resource settings.

Reflexivity

Members of the research team occupied both insider and outsider positions. Ghanaian clinicians embedded in the study sites facilitated recruitment and contextual interpretation, while international collaborators brought external perspectives. We acknowledge that this positionality may have influenced data collection and interpretation, and hierarchical dynamics and expectations of professional performance may also have shaped participant responses.

Analysis

Quantitative analysis

All statistical analyses were performed using STATA V.16. Continuous data (age, sepsis severity score, LOS and duration of illness) were stratified by hospital, assessed for normality and summarised as means with 95% CIs. The mean difference between groups was estimated using an independent sample t-test with a two-sided significance level of 5%. Categorical variables (sex and hospitalisation outcome) were presented as counts and percentages.

The pooled quantitative results indicate that patients who died before discharge had a comparatively higher mean sepsis score than those discharged alive. The estimated mean difference of 0.9 (95% CI 0.5 to 1.3) was statistically significant. Although discharged patients had a relatively longer LOS than those who died, the estimated mean difference in LOS was not statistically significant. The proportion of male discharged patients was higher than that of female patients.

Qualitative analysis

The qualitative data were analysed using reflexive thematic analysis following Braun and Clarke’s approach. A hybrid inductive-deductive strategy was adopted: deductive codes were derived from the Three Delay Framework and the Donabedian Model, while inductive codes captured emergent concepts beyond the frameworks (provider altruism, research presence). The constant comparative method was applied to refine codes iteratively across transcripts. Two researchers (LR and CM-D) independently reviewed the transcripts to develop an initial codebook, which was refined through discussion and consensus. Transcripts were then coded using Taguette. Emergent themes and subthemes were identified and organised to capture patterns and insights relevant to the research questions. Discrepancies were resolved through team discussion until consensus was reached. Thematic saturation was considered achieved when no new themes emerged from the data, which was reached after four (4) interviews.

Triangulation

To enhance the rigour and credibility of the findings, we applied methodological triangulation by integrating quantitative chart review data with qualitative interviews through the lens of our previously published situational analysis. The quantitative data (including clinical characteristics and outcomes) were examined alongside qualitative themes (such as perceptions of delays, clinical care quality and systemic barriers) to identify areas of convergence, divergence and complementarity. Triangulation was conducted during the analysis phase by comparing quantitative results (eg, differences in sepsis scores, LOS) with qualitative narratives (eg, descriptions of delays and clinical decision-making). This approach enabled a more nuanced understanding of patient outcomes and the contextual factors that influence them. Discrepancies were discussed within the multidisciplinary research team until consensus was achieved, ensuring a more comprehensive and credible interpretation of the data.

Patient public involvement

This study was conducted with significant input from local healthcare providers and administrators at KATH and PreHA, ensuring that the research was grounded in local realities. The paediatric staff at both hospitals were involved in identifying key resource limitations and provided feedback on the study design, data collection tools and the interpretation of findings. While direct patient involvement in the study design was limited due to the clinical nature of the research, future work will incorporate patient and family perspectives through qualitative interviews to explore barriers to sepsis care. The findings of this situational analysis will be disseminated back to local communities and healthcare providers to foster dialogue around actionable improvements for paediatric sepsis care.

Results

A total of 931 paediatric admissions occurred across both hospitals between January and April 2023 (121 in PreHA; 810 in KATH). Charts from 71 children were included in the review (16 from PreHA, 55 from KATH) (see tables1 2). Data on reasons for exclusion were not collated. The mean duration of illness prior to presentation was 12 days. The mortality rate was similar across both sites (~32%), despite patients in KATH having higher average sepsis scores and longer hospital stays. The majority of participants were male (61%) with an average age of 43 months at PreHA and 39 months at KATH. Patients at KATH had longer hospital stays (mean 6.8 days) than those at PreHA (mean 3.9 days), and higher average sepsis severity scores (2.9 vs 2.4).

Table 1. Chart review findings.

	PreHA Mean (95% CI)	KATH Mean (95% CI)	Overall Mean (95% CI)
Age in months	43 (20 to 65)	39 (28 to 51)	40 (30 to 50)
Sepsis severity score (qSOFA)	2.4 (1.9 to 2.9)	2.9 (2.7 to 3.1)	2.8 (2.6 to 3.0)
Length of stay (in days)	3.9 (1.4 to 6.5)	6.8 (5.1 to 8.4)	6 (5 to 8)
Duration of illness prior to presentation (in days)	12 (−0.15 to 23.9)	13 (4 to 21)	12 (5 to 20)
	N (%)	N (%)	N (%)
Sex
Female	8 (50.0)	20 (36.4)	28 (39.4)
Male	8 (50.0)	35 (63.6)	43 (60.6)
Total	16 (100.0)	55 (100.0)	71 (100.0)
	N (%)	N (%)	N (%)
Outcome
Death	6 (37.5)	17 (30.9)	23 (32.4)
Discharge	10 (62.5)	38 (69.1)	48 (67.6)
Total	16 (100.0)	55 (100.0)	71 (100.0)

KATH, Komfo Anokye Teaching Hospital; PreHA, Presbyterian Hospital, Agogo; qSOFA, quick Sequential Organ Failure Assessment.

Table 2. Disaggregated outcome by gender, sepsis score and length of stay.

	Discharged	Death	Difference
Mean sepsis score	2.6 (2.3 to 2.8)	3.4 (3.1 to 3.8)	0.9 (0.5 to 1.3)
Mean LOS (in days)	6.8 (5.3 to 8.3)	4.7 (1.7 to 7.8)	1.5 (−5.1 to 0.9)
	N (%)	N (%)	Total
Outcome by gender
Male	32 (74.4)	11 (25.6)	43 (100.0)
Female	16 (57.1)	12 (42.9)	28 (100.0)
Total	48 (67.6)	23 (32.4)	71 (100.0)

LOS, length of stay.

Key informant interviews (n=12) revealed systemic challenges, including a reliance on herbal remedies, financial constraints, transportation barriers, fractured referral systems, supply shortages and documentation gaps. Emergent themes included the role of community support, physician altruism and the impact of research partnerships on care delivery (table 3).

Table 3. Qualitative results.

Parent theme	Subtheme	Quote (with citation)	Relation to the model
Delay 1: Delay in deciding to seek care	Herbal medication use	PreHA interview 3: “In fact, those herbs, those herbs around them, they mostly use them. They think it will help. By the time they get here, the situation is very worse.”	Three Delays Model: Cultural beliefs and a reliance on traditional medicine contribute to delays in recognising severity and seeking care. Donabedian: Structural limitation (lack of trust or access to formal care).
Delay 1: Delay in deciding to seek care	Pharmacies as a first stop	PreHA Interview 3: “Sometimes they… You attend to the pharmacy, the drug stores. All that they know, oh, my child is sick, it’s malaria. So they just go and buy this malaria (drugs). Maybe even it’s anemia, they don’t know. Something serious has happened, they don’t know. All that they know is malaria. They just go to the nearby clinic and go and buy malaria medication.”	Three Delays Model: Misdiagnosis and inappropriate early care delay escalation. Donabedian: Process failure in informal care, structural lack of accessible diagnostic services.
Delay 1: Delay in deciding to seek care	Financial strain	KATH Interview 4: “They mostly do over-the-counter medications, possibly have all medications, and when they realize that now it’s overwhelming, then they can report to a healthcare facility.”	Three Delays Model: Financial barrier delays initial engagement with the formal health system. Donabedian: Structural constraint related to affordability.
Delay 2: Delay in reaching the healthcare facility	Financial strain	PreHA Interview 4: “So sometimes, when they are in the villages and those far-far villages, what I know is that they might seek first intervention from the pharmacist and the drug store. So that will give them some minor, minor drugs to take. Like that (unclear drug name). So, when they take it and it’s not going, that’s whereby they might come. Or, as I was first saying, if I come to there, maybe they make me buy ceftriaxone and tomorrow they change ceftriaxone. And see that it’s waste of money.”	Three Delays Model: Economic burden deters consistent and escalated care-seeking. Donabedian: Process issues (changing prescriptions); structure (cost).
Delay 2: Delay in reaching the healthcare facility	Health centres giving inaccurate care	KATH Interview 1: “Yes. So, because KATH is a tertiary hospital, so sometimes, instead of the district hospital…… okay, so usually they refer cases from the district hospital to you. You may have a patient spending a lot of time or a longer time in the district hospital. With them doing the cocktail of things or not, maybe not picking up, the sepsis. And by the time they refer that case to us, it’s sort of too late.”	Three Delays Model - Delay 2: Referral delays due to under-recognition. Donabedian: Process failure in district hospitals, Outcome deterioration from late recognition.
Delay 3: Delay in adequate care at the appropriate facility	Burden of sepsis	PreHA interview 2: “Yeah, sometimes they will take the first two doses and then the third one, just miss it. Then you come and start with the fourth day. Yeah, so there’s been breakages” KATH Interview 2: “Ok. So, we all know we have specific parameters for I.O. (intraosseous access), so what we can do is improvise. But you know in our setting, lack of logistics and resources, we have to improvise to care for the patient. If you don’t have the needles for the IO, you have to use the 50cc needles. But it’s not always the IO, sometimes we do the scalp or the jugular. But this is improvising for the IO.”	Three Delays Model: A lack of resources due to burden (equipment, laboratory services, trained personnel, for example) Donabedian: Process failures occur because of a lack of resources

KATH, Komfo Anokye Teaching Hospital; PreHA, Presbyterian Hospital, Agogo .

Delay 1: deciding to seek care—caregiver perceptions and financial barriers

“They usually start with herbal medications and only come to the hospital when things have gotten really bad.”—KATH Interview 1

Providers noted that caregivers frequently delayed seeking biomedical care for children due to cultural norms, financial difficulties and reliance on herbal medicine. Key informants explained that herbal remedies were often the initial approach, commonly continued until a child’s condition deteriorated considerably. Chart review data corroborate this, showing an average duration of illness before presentation of 12 days at PreHA and 13 days at KATH.

Providers hypothesised that these delays were shaped by both household-level constraints (eg, transport costs, informal user fees) and broader health system limitations (eg, lack of accessible or affordable primary care options). Some families faced difficult trade-offs between seeking care and maintaining their income, while others delayed care due to distrust or past poor experiences with the formal health system.

Delay 2: reaching appropriate care—transport, referral and navigation challenges

“By the time they go to the health centre and then are referred to us, the child is often already in shock.”—PreHA Interview 2

Even when the decision to seek care was made, systemic gaps in transportation and referral networks impeded timely access to appropriate facilities. Participants described ambulance shortages, high transport costs and long travel times, particularly in rural areas. These barriers were compounded by delays at lower-level health centres, where initial recognition of sepsis was limited, and referral processes were slow or unclear.

This resulted in prolonged illness trajectories and, in some cases, further deterioration en route. The absence of integrated referral systems led to care fragmentation and an increased caregiver burden.

Delay 3: receiving adequate facility-based care—resource and process constraints

“Sometimes they will run away and not pay for the labs, or we can’t find the forms to request them, and we just have to guess.”—PreHA Interview 1

At both study sites, once patients arrived at the hospital, care was influenced by a combination of structural and procedural limitations. Key informant interviews revealed that providers possessed a strong understanding of sepsis and its treatment. They identified early recognition, prompt antibiotic administration and haemodynamic support as critical strategies. However, chart reviews showed inconsistent documentation of vital signs, sepsis indicators and treatment timelines, as well as irregular use of severity scoring tools. This disconnect was not due to a lack of clinical knowledge but rather stemmed from structural constraints such as heavy workloads, documentation burdens and the absence of standardised protocols. Furthermore, shortages of antibiotics, lack of paediatric-specific equipment and high provider workloads limited the effectiveness of sepsis management protocols.

Providers described both moral distress and improvisation in navigating resource constraints. Some relied on personal or community resources to support patient care. Interviewees also noted that despite clinical knowledge often being strong, the inability to act due to missing laboratory results, delayed diagnostics or inadequate staffing meant this knowledge could not be translated into timely care.

Emergent themes: system workarounds and provider resilience

“We do what we can. Sometimes our decisions are right, sometimes not—but we can’t wait for labs that won’t come.”—KATH Interview 3

Despite these systemic challenges, providers demonstrated resilience through informal solutions, including conducting bedside assessments without laboratory tests, relying on collective clinical experience and occasionally fundraising to support families. Notably, patients participating in research programmes gain access to essential laboratory services and medications, and they enhance their practice.

However, reliance on these workarounds was neither sustainable nor equitable, underscoring the need for more structural and systemic solutions to these problems.

(A framework-based triangulation matrix of qualitative and quantitative data is shown in table 4.)

Table 4. Framework-based triangulation matrix.

Framework domain	Chart review findings	Key informant interview findings	Situational analysis	Integrated interpretation
Delay 1: Deciding to seek care	Mean duration of illness before presentation: 12–13 days	PreHA Interview 3 Caregivers try home remedies or wait due to stigma, cost and misinformation “They mostly do over-the-counter medications… By the time they get here, the situation is very worse.”	At KATH, patients arrive after visiting other facilities; referral delays are likely PreHA patients often first access herbal medicine or pharmacies	Delay in seeking care is shaped by financial strain, when traditional medicine is seen as an alternative to Western medicine for sepsis and systemic distrust; education and community engagement are key
Delay 2: Reaching the facility	Not captured	Transport and funds are lacking; long journeys and weak referral systems noted	KATH has a more structured referral system PreHA depends on family-arranged/paid-for transport to KATH (2–3 hours’ journey); no PICU	Barriers to reaching care include geographic inaccessibility (long distances to care and road challenges), constrained emergency transport and cost burdens
Delay 3: Receiving adequate care	Incomplete documentation; delayed vital sign recording; care plans inconsistent	Delays due to supply shortages, workflow inefficiencies	KATH has a Paediatric Intensive Care Unit (PICU), Paediatric Emergency Unit (PEU) and triage systems. PreHA lacks a PICU; limited staff, poor coverage; frequent stock-outs	Systemic workflow obstacles due to resource constraints
Structure (Donabedian)	Not applicable	PreHA Interview 4 “We need more pulse oximeters… if more than three babies come…"	KATH has 69 paediatric beds and 75 physicians in the department of child health Pre-HA has 61 paediatric beds and 2 paediatricians; research funding supports diagnostics. Both have frequent power outages, but reliable generators.	Staffing, space, diagnostics and supply chain gaps compromise care. Differences in structural capacity, shaped by historical underfunding and reliance on research partnerships, inform interventions. Reliance on research also allows for increased capacity and diagnostics, however.
Process (Donabedian)	Extended length of stay (mean 6.8 days at KATH)	Triage varies; lack of standardised responses	Global guidelines are misaligned with local realities. Systemic financial gatekeeping. Workflow bottlenecks at PEU/PICU levels	Global protocols must be adapted for local contexts, reflecting logistical and financial constraints and hospital-specific workflow dynamics
Outcome (Donabedian)	Mortality ~32%; similar across both sites despite higher acuity at KATH	Providers believe deaths would be preventable with earlier presentation and support	Sepsis remains a leading cause of paediatric death; systemic inefficiencies are noted as likely contributors	Mortality is more reflective of systemic and structural challenges than clinical capability; improving structure and process may reduce preventable deaths
Emergent theme: Financial strain	Not applicable	Providers cite inability to pay as a significant barrier to care	Many diagnostics and medications require upfront payment; treatment is delayed as families raise funds	Financial barriers are deeply embedded and delay all stages of care; addressing cost structures and insurance coverage is essential
Emergent theme: Community support/physician altruism	Not applicable	Providers and families pool resources to support children	Community resilience is present, but physician-led resource support is not sustainable without system-level solutions	Informal support networks exist but are fragile; structural investment is needed to move from ad hoc to sustainable care models
Emergent theme: Perceived impact of research presence	Not applicable	Staff noted increased attentiveness and access during research engagement	Partnerships with PreHA (Kumasi Centre for Collaborative Research, German collaborators) improve clinical capacity; healthcare support due to research funding	Research involvement can strengthen systems but may create dependency and inequity across facilities; future partnerships must promote system-wide equity
Emergent theme: Education improves outcomes	Not captured	Training increases provider confidence and patient trust	Prior OPENPediatrics modules addressed gaps; future education modules planned	Provider and caregiver education are scalable strategies that support early recognition and appropriate care, especially in resource-constrained settings
Emergent Theme: Herbal/ traditional medication use and pharmacies as the first point-of-call	Not directly captured	PreHA Interview 3 "In fact, those herbs… they mostly use them. They think it will help. By the time they get here, the situation is very much worse.”	Seen as a primary care access point in rural areas, such as those served by PreHA	Traditional systems are part of the health-seeking pathway; integration and education are needed to promote earlier referral and trust in biomedical care

KATH, Komfo Anokye Teaching Hospital; PreHA, Presbyterian Hospital, Agogo.

Discussion

This study moves beyond traditional analyses of paediatric sepsis as a clinical or educational problem and instead frames it as a care pathway issue shaped by complex, interacting systems. By integrating chart reviews, provider narratives and environmental analyses within a unified framework, we demonstrate that delays in paediatric sepsis care in Ghana are not discrete failures, but rather the result of systemic fragmentation and contextual realities. Our findings demonstrate that barriers at the household level (eg, delayed decision-making due to financial constraints or reliance on herbal remedies) interact with logistical gaps (eg, poor referral networks, ambulance shortages) and facility-level constraints (eg, documentation gaps, medication stockouts, staff shortages) to produce compounding risk.

Our qualitative findings align with existing evidence from Ghana and the broader sub-Saharan African region, showing that caregivers frequently seek traditional healers or pharmacy-based care as their initial response to paediatric illness, motivated by perceived affordability, cultural preference and accessibility.¹⁹ Studies in Ghana revealed that nearly 50% of surgical patients initially consulted herbal or religious healers,²⁰ often resulting in delayed and advanced presentations.²¹ Similarly, a Ghana Demographic and Health Survey analysis found that caregivers often started with pharmacy or over-the-counter treatments for paediatric respiratory illnesses, delaying formal clinical care, especially when cost or accessibility issues were present.²² These patterns reinforce the notion that economic and cultural factors significantly drive delays in sepsis care, resulting in patients presenting in more critical condition.

Financial support for hospitalised children in Ghana is primarily provided through the National Health Insurance Scheme, which was implemented in 2003 as a step towards universal health coverage. An exemption policy exists to enable the poor and vulnerable to access healthcare.²³ Despite this social intervention, out-of-pocket payments for ambulance services, as well as some critical laboratory and pharmaceutical services, disincentivise health-seeking behaviours.²⁴

Although the introduction of improved ambulance services has been associated with increased utilisation of emergency transportation to health centres,²⁵ its benefit is unlikely to be substantial with population density and poor road networks.²⁶

Despite these challenges, broader cultural values in West African contexts, which prioritise interdependence and communal responsibility, nonetheless exist. As shown by Agulanna and colleagues, collectivistic orientations shape social behaviours around health and well-being,²⁷ while the concept of African communalism, ‘I am because we are’,²⁸ underscores an ethic of mutual support. These cultural norms help explain why providers and community members at PreHA engaged in personal and communal financial efforts as part of sepsis care. This is reinforced by historical examples, such as during the Ebola epidemic—where West Africans ‘stood by the ill’ even under risk.²⁹ Collectively, these data suggest that provider-level altruism and community solidarity are not isolated phenomena but are deeply embedded in local social and ethical systems.

This study reframes paediatric sepsis care in Ghana not as a failure of knowledge or guideline adherence, but as the outcome of a fragmented, under-resourced system shaped by structural and sociocultural forces. By integrating the Three Delays Framework with the Donabedian Model within a unified care pathway framework, we demonstrate how delays accumulate across the continuum of care, with provider capacity constrained by systemic forces that extend far beyond individual decision-making.

The findings reveal that clinical knowledge among providers was often strong; however, systemic limitations prevented its consistent translation into timely care. Delays in seeking care were deeply rooted in economic vulnerability, reliance on traditional healing systems and a lack of trust in biomedical interventions. Transportation failures and inadequate safety nets created barriers to accessing care, while infrastructure gaps, stockouts and referral bottlenecks hindered care delivery once patients arrived at facilities.

These insights challenge the prevailing assumption that provider education alone can improve paediatric sepsis outcomes in LMICs. While knowledge was frequently present, the ability to act on it was repeatedly undermined by cost barriers, resource constraints and system fragmentation. Sustainable improvements in paediatric sepsis outcomes will therefore require regionally adapted, systems-informed strategies that address both the structural foundations of care and the lived experiences of those delivering and receiving it, moving beyond educational interventions to comprehensive system strengthening.

Rather than isolated events, these barriers were cumulative and interconnected. For example, financial barriers influenced both the decision to seek care and the ability to follow through with treatment recommendations. Likewise, a delay in reaching appropriate care increased the clinical severity on arrival, placing further strain on already limited facility resources.

Yet, despite these challenges, providers demonstrated remarkable resilience and ingenuity. Their improvisation in the face of systemic constraints highlights both the need for structural investment and the risk of moral distress in contexts where the burden of compensating for system failures falls on frontline staff.

Ultimately, addressing paediatric sepsis in LMICs requires more than algorithms or educational workshops. It demands investments in the health system’s backbone: referral coordination, supply chains, financing mechanisms and supportive policy environments.³⁰ Future interventions must be locally adapted, co-designed with providers and grounded in the lived complexity of care.

This study’s evolution emphasises the importance of conducting in-depth qualitative exploration before implementing educational or behavioural interventions, particularly in low-resource health systems. Although the initial hybrid effectiveness-implementation design centred on evaluating an educational module, the subsequent methodological adaptation revealed that contextual and systematic barriers were more fundamental than knowledge deficits alone. This experience highlights how qualitative inquiry conducted beforehand could have identified these structural and process-related constraints earlier.

Limitations

This study has several limitations. First, the retrospective chart review was constrained by significant missing data. Of 187 charts screened, only 71 met the criteria due to missing or incomplete vital sign documentation. This exclusion raises the possibility of bias as documentation quality may have differed systematically between sicker and less sicker children, potentially skewing the analytic sample. We have treated missing data both as a methodological limitation and as a substantive finding reflecting structural and process barriers within the healthcare system. Second, the sequencing of the educational intervention prior to confirming the feasibility of the outcome evaluation may appear unconventional. Furthermore, interviews were conducted in English, which is the professional working language of healthcare in Ghana, but not always the first language of participants. This choice supported technical clarity but may have constrained the expression of culturally embedded understandings of illness. We also acknowledge that conducting the interviews after the educational sessions rather than beforehand may have shaped participants’ reflections. Additionally, data collectors varied over time, which may have introduced inconsistency in abstraction despite the use of standardised forms. Charts from only January to April 2023 were reviewed; therefore, there is a potential risk of seasonal bias. However, this was the same time frame across both sites, which mitigates this variation. The sample size was relatively small, particularly at PreHA, which limits the generalisability of the quantitative findings. Our focus, however, was not on the burden of sepsis but on the structures, processes and outcomes of sepsis care. Moreover, while the qualitative component captured rich insights from providers, it did not include the perspectives of patients or caregivers. Nonetheless, provider narratives frequently referenced barriers experienced before arrival at the hospital and between facilities, suggesting that future work should include community and pre-hospital perspectives to more fully understand the continuum of delays in sepsis care.

Conclusion

This study underscores the importance of contextualising paediatric sepsis care within the realities of health systems in low-resource settings. By triangulating chart review data with provider perspectives through the lenses of the Three Delays Framework and the Donabedian Model, we identified critical barriers that span both clinical and structural domains. Our findings highlight that delays in seeking, reaching and receiving care are not isolated events but rather interconnected challenges rooted in systemic constraints, inadequate infrastructure and limited resource availability.

Although our initial efforts focused on evaluating clinical outcomes through a traditional effectiveness lens, the study’s transition to an explanatory, mixed-methods approach revealed the deeper complexities that influence paediatric sepsis care. These insights emphasise the importance of developing facility-specific sepsis pathways that are not only evidence-informed but also grounded in the lived experiences of both providers and patients. Ultimately, improving sepsis outcomes in Ghana and similar settings requires moving beyond the importation of externally derived protocols toward co-designed, context-responsive interventions, pre-, peri- and post-hospital, that acknowledge both clinical science and local realities.

Recommendation

Future studies could build on these findings using narrative or systems-level approaches. Such methods may more effectively capture the interdependencies and feedback loops that shape sepsis care pathways, thereby offering deeper insights into how systemic barriers interact across cultural, structural and clinical domains.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.