The Six Delays Model: expanding the three delays model with evidence from Madagascar for maternal referrals in LMICs

Abstract

Maternal mortality remains disproportionately high in low-income and middle-income countries, where pyramidal health systems and inadequate referral processes often delay access to emergency obstetric care. The Three Delays Model has long been used to analyse delays in maternal care globally, but it fails to capture the full complexity of referral-related delays within pyramidal health systems. Drawing on the example of Madagascar, this paper reviews the limitations of the traditional model and proposes a revised ‘Six Delays Model’. This model expands the traditional three delays to include the following six stages: ‘delay in seeking initial care’, ‘delay in reaching initial care’, ‘delay in receiving initial care’, ‘delay in seeking referral care’, ‘delay in reaching referral care’, ‘delay in receiving referral care’.

This revised model improves granularity, integrates patient-provider and inter-provider dynamics that influence referral decisions and reflects both individual and community-level factors that influence decision-making. By mapping delays accurately along the patient journey, the revised model provides a more actionable model for policymakers and implementers seeking to reduce maternal mortality. While more complex, the expanded model offers necessary nuance and adaptability for pyramidal health systems and supports targeted intervention design to address systemic barriers to timely and adequate obstetric care.

SUMMARY BOX.

• Maternal mortality in low-income and middle-income countries (LMICs) remains high due to delayed access to emergency obstetric and surgical care, especially within pyramidal referral systems.

• The widely used Three Delays Model inadequately captures referral-related barriers in these systems.

• We propose a revised ‘Six Delays Model’ that adds referral decision-making, referral transport and receiving referral care as distinct, interconnected stages.

• This model provides a more detailed and adaptable model for analysing and understanding maternal healthcare journeys in LMICs.

• Policy-makers and implementers can use this revised model to prioritise interventions targeting the most relevant referral delays and reduce maternal mortality.

Introduction

Reducing maternal mortality to less than 70 per 100 000 live births is a key objective of the Sustainable Development Goals.¹ Currently, 90% of maternal deaths worldwide occur in low-income and middle-income countries (LMICs) and are primarily due to preventable complications such as haemorrhage, hypertensive disorders or sepsis.^{2 3} Likewise, most neonatal deaths in sub-Saharan Africa (SSA) are linked to intrapartum complications or complications of preterm delivery, many of which can be managed with timely access to functioning Emergency Obstetric and Newborn Care (EmONC) services.^{4 5} Basic EmONC (BEmONC) facilities can address some of these complications (eg, through administration of uterotonics, neonatal resuscitation, or management of infections), but the surgical interventions required for Comprehensive EmONC (CEmONC), particularly emergency caesarean sections, are typically only available at secondary or referral hospitals.^{6 7} As a result, women frequently require transfers to CEmONC facilities when complications arise that exceed the capacity of primary facilities.⁷ Delays during such referrals can lead to severe consequences, including maternal or neonatal death, extensive blood loss and long-term disability.⁸,¹¹ For example, for mothers, this may include disabilities such as obstetric fistula or incontinence following obstructed or traumatic labour, while for neonates, hypoxia during delayed delivery may result in lifelong neurological impairment.³⁸,¹²

Several interventions have been shown to reduce maternal mortality by addressing barriers identified in the original three delays model. For example, community-based intervention packages, including antenatal care promotion, clean delivery kits and the engagement of community health workers (CHWs), have demonstrated measurable effects on improving maternal and neonatal outcomes. ^{13 14 15 16} In addition, mobile health (mHealth) approaches—such as SMS reminders, appointment tracking and digital referral tools—have improved antenatal care attendance and timeliness of facility-based care.^{11 17} Previous studies have especially highlighted the critical role of referral timeliness once a complication arises to improve both maternal and neonatal outcomes.⁸,¹¹ Therefore, understanding, addressing and reducing delays in the access to CEmoNC services across LMICs are vital in improving access to lifesaving interventions and reducing maternal and neonatal mortality.

In many LMICs, healthcare delivery is organised as a pyramidal referral system.¹⁷ At the base of the pyramid are primary-level facilities, such as community health posts and health centres, which are the first point of contact for most patients.¹⁷,¹⁹ These facilities provide basic preventive and curative services, antenatal care, uncomplicated deliveries and first-line management of common conditions. Providers at this level are typically nurses, midwives or CHWs. Importantly, they lack surgical capacity and advanced obstetric interventions.^{18 19}

The next level consists of secondary facilities, which usually have general physicians or medical officers, sometimes supported by surgical or anaesthetic staff and, in such cases, can provide CEmONC services. However, the availability of skilled personnel, surgical equipment and consistent supplies is often limited, particularly in rural areas.^{7 11}

At the top of the pyramid are tertiary hospitals (eg, national or university referral hospitals). These institutions, usually located in major urban centres, provide speciality care, advanced surgical capacity and teaching functions.¹⁷ While the terminology and categorisation of levels differ by country and context, the underlying principle (a broad base of primary care facilities with limited scope and capacity, referring to fewer higher-level facilities with increasing capacity) is consistent across many LMICS.¹⁷,¹⁹

The pyramidal structure creates ratios of access—there are many more primary facilities than secondary, and only a handful of tertiary centres. For example, in Madagascar, the number of Centres de Santé de Base (CSBs) vastly outnumbers district or regional hospitals, and comprehensive obstetric surgical care is concentrated in very few urban referral hospitals.²⁰ This results in frequent referrals upward through the pyramid, often across long distances and poor transport infrastructure, leading to delays in timely management of obstetric emergencies.^{20 21} At each step, new decisions, transport barriers and systemic factors affect referrals and can cause delays.

In this analysis paper, we propose a revision of the existing Three Delays Model,²² to improve its granularity and pertinence in systematically analysing delays in referral processes and their downstream effects on maternal and child health, both in Madagascar, which we use as an example throughout the paper, and other LMICs more broadly.

Development of the Six Delays Model: methodological approach

The revised model presented herein is based on the illustrative example of Madagascar, where both authors were primarily working at the time of the manuscript conceptualisation, but has broader implications for LMICs globally.

We developed the example of Madagascar drawing on our combined professional experiences: one author as a physician working in maternal healthcare, and both authors through their work in non-governmental organisation (NGO) programming to improve maternal and neonatal health as well as in research on maternal health in Madagascar. One author in particular had previously published scientific work on access to EmONC services and obstetric referral care in Madagascar, which sparked the debate on the three delays model in the context of Madagascar.²³

In addition, this work is informed by (1) routine operational data from maternal health programmes in Southern Madagascar, (2) expert input from clinicians and maternal health and health systems strengthening implementers and researchers and (3) evidence from published studies on obstetric referral care in LMICs.

Internal programme data originated from NGO-supported maternal and neonatal health programmes implemented in southern and eastern Madagascar between 2016 and 2023. These data included routinely collected health facility reports on obstetric and neonatal care (eg, birth registers), referral and ambulance logs, and programme costing data related to emergency transport and referral support. These sources were reviewed descriptively to identify recurrent patterns in referrals for obstetric emergencies, following the analytical approach previously used to examine the cost-effectiveness of referral care in the same geographic location.²³ We further drew on programme data from the same NGO on obstetric ultrasound utilisation and findings²⁴ to enhance the understanding of key maternal and neonatal risk factors in the setting.

To expand on and contextualise these data further, we drew on national demographic data for Madagascar available through the 2021 DHS survey,²⁵ UNICEF MICS surveys²⁶ and the Malagasy National Institute of Statistics,²⁷ as well as published scientific evidence from Madagascar. The reports of these surveys were reviewed and analysed using content analysis to identify recurrent individual, geographic, financial and health system barriers affecting emergency obstetric referrals in Madagascar. We then compared and synthesised findings from these surveys with the patterns constructed from the programme NGO data.

To complement these analyses and improve our understanding of the applicability of our findings to the broader LMIC context, we conducted a targeted, non-exhaustive literature search using PubMed and Google Scholar. We designed search terms to capture literature on delays across the maternal care pathway, including barriers to timely obstetric emergency care, transport and referral challenges, and facility-level readiness to manage obstetric emergencies. The search strategy focused on studies examining components of the Three Delays Model, emergency obstetric and neonatal care, maternal morbidity and mortality, and health system factors influencing outcomes in LMICs. Search terms included but were not limited to “obstetric referral”, “three delays model”, “maternal mortality”, “emergency obstetric care”, “transport delays” and “LMICs”.

We considered articles published since the original three delays model was published in 1994²² and articles in both French and English. We considered peer-reviewed studies regardless of study design (quantitative, qualitative, mixed-methods, observational, interventional, quasi-experimental), as well as editorials, commentaries and analysis pieces. We excluded articles that did not focus on LMICs. We additionally screened the reference lists of key articles. Because the purpose of the search was to contextualise and validate the conceptual model rather than to conduct a systematic review, no quality appraisal of studies was conducted. The identified literature was used to (1) validate whether our conceptualisation of observed delays in Madagascar was consistent with findings from other LMICs and (2) identify additional determinants affecting obstetric referral care.

After synthesising the findings from internal programme data, national surveys and the literature, we constructed an initial draft of the revised and expanded Six Delays Model. We then refined the model through expert input. This input was obtained through semistructured consultations with six experts: two senior physicians with long-standing experience in maternal health in Madagascar, a representative of senior leadership of the implementing NGO, and three additional maternal health and health system strengthening researchers working across LMICs, based in Germany and the USA. All participants were chosen purposively due to either (1) their experience working directly in maternal healthcare in Southern Madagascar, to validate our interpretation and conceptualisation within the context of Madagascar or (2) their experience working in maternal health and health system strengthening in other LMICs, to validate the broader applicability of the suggested model to settings outside of Madagascar. We used preformed questions to obtain expert feedback on our critique of the Three Delays Model, the newly suggested Six Delays Model, the conceptual boundaries between delays, the distinction between levels and factors that influence delays, and validate the relevance of the proposed delays across LMIC settings.

The consultations took the form of focused discussions and iterative feedback rounds on preliminary versions of the expanded delay framework. Experts were provided with a written draft of the suggested model, as well as the background information that informed it, followed by a meeting with both researchers, either in person or online for further discussion. During those meetings, both researchers took notes on the feedback provided by the experts. The meetings were not recorded. Notes from these discussions, together with written feedback shared by experts, were summarised to identify points of agreement, clarification needs and areas of divergence. Following a first round of consultations, we used the input from all consultations to adapt the initial draft of the model. We then reshared the revised model with all experts, two of whom provided further feedback in writing, three in the form of further meetings. We repeated this process for one more consultation round until we received no further suggestions for improvement from the experts. Because the purpose was model refinement, we considered the consultation process complete once no new substantive feedback emerged across iterative rounds.

The example

In Madagascar, the maternal mortality rate in 2023 was 445 per 100 000 live births.²⁸ The bottom of the country’s pyramidal health system is comprised of CHWs who provide primary care at the community level. Moving up the pyramid, primary care is provided at the CSBs and secondary care is provided by the district referral hospitals (Centre Hospitalier de Référence de District, CHRD).^{20 29} Frequently, however, secondary surgical care is only available at regional hospitals, such as Centres Hospitalier de Référence Régionale (CHRR).^{20 29} At the top of the pyramid, tertiary care is usually only available at university teaching hospitals, Centres Hospitaliers Universitaires, which are limited to select urban centres of the country. Within this pyramidal structure, surgical care is only available at select CHRDs or CHRRs. Importantly, it is not available at the lower levels of the health system that women typically access first.^{20 21}

The original Three Delays Model

To analyse care-seeking delays and their relevance in preventable maternal deaths, especially in LMICs, Thaddeus and Maine proposed the ‘Three Delays Model’ in 1994.²² The original model includes the following three delays: (1) delay in the decision to seek care; (2) delay in arriving at a health facility and (3) delay in receiving adequate care at the facility.²⁰ 30 years later, this model continues to be widely used in academic literature to analyse causes of delays in maternal, neonatal and child healthcare journeys and to assess interventions addressing these delays.³⁰,⁴¹ Countries in which the model has been applied include India, Tanzania and Zambia.³¹,³³

In addition to the original three delays model, other authors have proposed expanded versions. MacDonald et al described a ‘fourth delay’ at the community level, capturing collective responsibility and social barriers beyond individual households.⁴² A ‘fourth delay’ has also been added in contexts of trauma and rehabilitation care, describing delays in remaining in care, that is, barriers to rehabilitation, follow-up and adherence after definitive treatment.⁴³ These examples demonstrate the adaptability of the model across health domains and underscore that expansion efforts vary according to the context under study. These previous expansions highlight the adaptability of the model to new contexts and care pathways, but fail to address the specific delays and influencing factors of patient journeys for emergency obstetric care, as we detail below.

Despite its utility, the original model as well as its existing adaptations inadequately address referral delays within pyramidal health systems.⁴⁴ In current practice, researchers either subsume referral delays under the second or third delay or treat referrals as separate processes, neglecting their interaction with earlier care-seeking stages.³⁷⁴⁴,⁴⁶ Each of these approaches carries significant disadvantages. In the first approach, specific factors that relate to maternal mortality during the referral process, for example, the availability of professional referral services^{47 46} or specialised care during referrals⁴⁶ are grouped with numerous other factors relating to care provision and, thus, are easily overlooked.²² Equally important and crucial contextual aspects and interactions between patients, healthcare providers (HCPs) and the healthcare system that affect care seeking, referral processes and maternal mortality more broadly are neglected.⁹ This leads to a lack of detail and granularity in identifying and evaluating key factors that delay maternal care seeking when referrals are necessary, limiting the model’s effectiveness in identifying potentially vital factors causing delays.

The second approach, which considers referrals separately from the initial healthcare seeking at the primary care level, fails to integrate vital information on maternal referral care, such as a delay in the decision for referral from primary to secondary care, comprehensively.^{46 48} Treating referrals as distinct from prior treatment seeking and care provision behaviours does not capture the interactions between these different stages of care seeking, hindering a comprehensive understanding of the patient journey.^{46 48} Consequently, a holistic appraisal of the patient journey and understanding of the relevance and respective weight of individual delays within this journey is impossible.^{46 48} This, in turn, makes it difficult for policymakers and implementers to focus on the most relevant delays along the patient journey or prioritise the most impactful interventions.

Building off our experience in Madagascar, we argue the current structure and use of the three delays model are inadequate for comprehensively and granularly analysing patient journeys for obstetric complications necessitating referral care. This limitation renders the model unable to holistically examine delays in maternal healthcare seeking in pyramidal health systems across LMICs, where understanding key delays in obstetric emergencies is crucial for guiding policy and practice to reduce maternal mortality.^{3 9 21 28}

The newly suggested Six Delays Model

To address these gaps, we propose revising the Three Delays Model to explicitly integrate referral delays as interconnected steps in the patient journey (figure 1). This expansion will facilitate a more comprehensive and nuanced analysis of the patient journey in the context of pyramidal health systems, which we will illustrate based on the example of Madagascar. By adopting this comprehensive model, implementers and policymakers can better prioritise interventions, ultimately reducing maternal mortality and improving maternal healthcare in LMICs.

Figure 1. Health system, patient and other factors influencing the occurrence of delays in cases needing referral care for obstetric emergencies in LMICs. The six delays, highlighted by the blue shaded central shapes, are: (1) ‘delay in seeking initial care’, (2) ‘delay in reaching initial care’, (3) ‘delay in receiving initial care’, (4) ‘delay in seeking referral care’, (5) ‘delay in reaching referral care’ and (6) ‘delay in receiving referral care’. Delays one and three are further split into individual and collective decision making, to highlight that referral decisions are not made by individual patients alone but often jointly with family and/or community members, health care providers and potentially other stakeholders, for example, religious leaders. The factors potentially contributing to delays at each step of the patient journey are grouped in five levels: (1) health system, (2) healthcare providers, (3) patient, (4) community and (5) other to illustrate that care journeys and their delays have multiple influencing factors that are not isolated but interact with one another. Examples of factors that may influence each delay at each level are highlighted in the figure in grey shapes, with some factors influencing multiple adjacent steps of the care journey. However, which factors apply to each step of the care journey are context-specific and may need adaptation depending on where the Six Delays model is applied.

This expanded model consists of six stages, three related to pre-referral care, three to referral care: (1) ‘delay in seeking initial care’, (2) ‘delay in reaching initial care’, (3) ‘delay in receiving initial care’, (4) ‘delay in seeking referral care’, (5) ‘delay in reaching referral care’ and (6) ‘delay in receiving referral care’. Delays one and three are further divided into individual and collective decision-making steps to better reflect health-seeking realities. For healthcare seeking, this includes the pregnant person’s decision and the influence of family and community dynamics—addressing prior critiques that overlooked these vital social factors.³⁸ Similarly, the referral decision involves both the patient’s and HCP’s choices and the collective input of family and other individuals.

For the ‘Delay in seeking referral care’ delay, this new model highlights the complex interactions between the health system, the patient and the patient’s family in deciding for referral, instead of focusing only on patient decision-making processes.

In our example context of Madagascar, referral decisions are shaped by systemic barriers such as staffing shortages and inadequate infrastructure at the primary care level,^{20 21} and by weak health financing mechanisms. Facilities, often dependent on out-of-pocket payments, may be reluctant to refer patients because referrals threaten facility revenue streams.^{20 49} At the same time, limited clinical expertise in rural areas can lead to both under-referral (when providers lack knowledge) and over-referral (when they lack confidence in their skills).^{20 21} These provider- and system-level dynamics interact with broader social and economic contexts. As in many LMICs, women’s limited decision-making power and inequitable social status in Madagascar can delay referral uptake.^{50 51} Financial constraints remain a dominant factor, with referral care—particularly surgical care—being prohibitively costly.⁵¹,⁵³ Patients and families frequently weigh these costs against the risk of catastrophic health expenditure, a decision intensified by widespread poverty and the absence of risk-pooling mechanisms.²⁰ None of these factors are unique to the Malagasy context, however. Instead, weak healthcare financing mechanisms that open families up to substantial financial risk from seeking medical care and make facilities dependent on out-of-pocket payments are common across SSA.^{9 53} Similarly, gender roles and cultural norms that negatively affect women’s agency and decision-making power are a common challenge across LMICs^{9 32 44 54 55} and intersect negatively with the economically vulnerable situations many women find themselves in globally.^{9 54 56} Lastly, challenges in patient-provider communication, communication between different levels of healthcare services and broader health systems constraints such as infrastructure, staffing and material are common across LMICs^{57 49 56 58 59}

Together, these factors show that referral decisions are not made by the patient alone, but in the interplay of the patient, their community and family, health care providers, the healthcare system and broader contextual factors such as security.

The second new delay, ‘delay in reaching referral care’, reflects challenges in accessing effective transport once a referral decision has been made. It focuses on the availability and functioning of formal referral systems, including ambulances, fuel and trained staff.^{5456 60},⁶² In rural Madagascar, most women rely on ox carts, rented vehicles or even walking to reach referral hospitals.²⁰ Seasonal rains worsen road conditions, and regional insecurity further restricts transport options.⁶³ Similar transport barriers are widely reported across LMICs, where inadequate ambulance coverage, poor roads and the high cost of private transport remain critical bottlenecks.⁶¹⁶³,⁶⁵ Evidence from both Madagascar and other LMICs highlights the importance of cost-effective referral systems and infrastructure improvements in mitigating these transport-related delays.^{66 2367},⁷⁰

The final additional delay, ‘Delay in receiving referral care’ echoes factors relevant to the third delay (‘Delay in receiving initial care’), but, importantly, includes additional factors essential for referral care, particularly centred on the health system. In Madagascar, constraints at referral hospitals include shortages of blood, equipment and surgical staff, as well as strained relationships and communication between primary and referral-level providers.^{20 21} Such bottlenecks are mirrored across LMICs, where weak inter-facility coordination, insufficient surgical capacity and under-resourced referral centres undermine timely lifesaving care, even if women reach higher-level facilities.⁵⁴⁷¹,⁷³

Discussion

By focusing on the specific delays and factors that occur during referral care and influence its outcomes within a pyramidal health system, our model addresses key gaps in existing approaches. This is especially relevant for LMICs, where pyramidal health systems mean referrals occur frequently and can significantly affect the patient journey. Our model avoids the loss of granularity that arises when too many factors and steps in a patient journey are grouped under one delay and instead allows for a detailed analysis of the factors at play at each stage. This, in turn, allows for a clearer understanding of which factors contribute most to delays at each stage and thus need to be addressed most urgently. Further, our model offers a comprehensive perspective on the entire patient journey, allowing further analysis of the relative weight and importance of the different delays in patient journeys, for example in the analysis of maternal or neonatal mortality or morbidity in a given setting. This more detailed understanding of the respective relevance of each delay in patient journeys improves priority setting in both evidence generation and intervention targeting.

In combination, this increased understanding of the relative importance of individual delays and their contributing factors renders the model more actionable than previous versions.

The proposed model also complements and extends existing work that differentiates between intrafacility and interfacility processes within emergency obstetric referral pathways. Existing evidence has examined intrafacility delays—such as triage failures, surgical readiness or breakdowns in communication—separately from interfacility referral challenges related to transport, coordination or geographic access.^{62 71 74 75} Other referral-specific models focus primarily on the transfer segment of care without considering how upstream delays shape referrals.^{61 64 69 76} By integrating both intrafacility and interfacility determinants into a single framework, the Six Delays Model captures the complexity of obstetric emergency care across all levels of the system. This integration helps identify how delays accumulate and interact along the patient journey. This revised model thus enables policymakers to systematically identify and address the delays that occur for women throughout their patient journey in LMICs.

The proposed model can be operationalised for programme monitoring and evaluation using a combination of indicators. Delays that are explicitly related to transport, such as the ‘delay in reaching initial care’ and the ‘delay in reaching referral care’, can be measured through time-stamped data, including travel times, ambulance dispatch and arrival logs, and assessment of transport availability.^{23 69} Other delays may require complementary sources of information, for example, assessments of facility preparedness—such as EmONC Service Readiness Assessments, staffing coverage and functioning operating theatres can be used to assess delays linked to the ability of facilities to provide timely emergency care (delay three and six).^{21 73} Similarly, quality-of-care indicators may be incorporated for delays related to the appropriateness or adequacy of care received, such as adherence to recommended timelines for emergency caesarean sections or the timely initiation of neonatal resuscitation.⁷⁰ Qualitative inquiry into healthcare seeking behaviours at the community level or into inter-provider communication can inform assessments of delays one and four.^{54 76} While the specific measurement approach will need to be adapted to the context of application, the ability to draw on diverse indicators for monitoring enhances the practical utility of the Six Delays Model.

In this process, the Six Delays Model can equally be an important tool to guide implementers and policymakers in identifying the ‘Goldilocks Zone’ of their specific health system context.⁷⁷ In this case, the ‘Goldilocks Zone’ would constitute a range in which the six delays each remain at levels that do not compromise maternal or neonatal outcomes in the given context. The Six Delays Model can help define what this optimal range looks like for each delay in a given setting by identifying baseline levels and guiding analysis on thresholds beyond which delays become harmful. When used as a monitoring tool, the model enables health systems to detect when delays begin to drift outside this optimal zone—for example, when transport times lengthen, surgical readiness declines or referral decision-making becomes slower. Such shifts may indicate mismatches between health systems demand and capacity and can thus warn of system strain before adverse outcomes such as maternal mortality become apparent. Coupled with routine measurement of the relevant time, coverage and quality indicators described above, the Six Delays Model thus offers a practical framework for ongoing system monitoring and timely action.

Key interventions emerging from such an analysis could include strengthening referral systems with formal transport options and dedicated funding for fuel and maintenance,^{45 58} or training HCPs on timely referral decision-making and improving communication channels between facilities.^{71 78} By differentiating between individual and collective decision-making at both the healthcare-seeking and referral stages, the model helps stakeholders pinpoint specific barriers and tailor interventions. This granular perspective on patient journeys thus aids in prioritising impactful strategies to reduce maternal mortality that holistically address delays along the patient journey.

It is important to note that, like the original Three Delays Model, the Six Delays Model does not assume that all maternal deaths are attributable to delays in care journeys, even if they are important contributors especially in LMICs.^{4 18 28} Certain events, such as amniotic embolism, may result in mortality irrespective of timeliness of care.⁷⁹ Likewise, in contexts where essential services (eg, comprehensive EmONC or surgical capacity for obstetric interventions) are entirely absent, the main challenge lies not with the delays within patient journeys, but the limited capacity of the health system more broadly.

Our model is not without limitations. Expanding the model from three to six delays renders the model, while more comprehensive, also more complex. This complexity, however, is necessary to reflect the actual complexity of patient referrals in pyramidal health systems. Additionally, some factors recur across delays and are listed at multiple time points in the model. However, this reflects the realities of patient journeys, where the same factors are relevant in influencing decision making at multiple stages, but have differing impacts at each stage. For example, the fear of treatment costs may be a more significant deterrent to care seeking at referral hospitals where expenses are typically higher than at primary care facilities.^{52 53} The model is specifically designed to capture these realities, recognising that the relative weight and relevance of each factor will differ across settings and stages of the referral process. The relative weight and relevance of factors will differ across settings, and it is up to policymakers, researchers and practitioners to determine which factors are most pertinent for their specific context and whether they recur across delays or not. In some applications, aggregating or collapsing overlapping determinants may be necessary; in others, retaining them separately will be more useful to capture and analyse local realities. This flexibility allows policymakers and researchers to adapt the model to their specific context, making informed decisions on which determinants are most relevant for analysis and intervention.

Further, our model currently incorporates only one step of referral. We acknowledge that for some patients, more than one referral may be necessary to access the care they require and that these additional steps of the patient journey are not reflected in our model. This was a conscious choice to reduce the model’s complexity and to reflect the most common patient journeys. Lastly, like all models, context-specific adaptations may be necessary to tailor the model to different settings. Local policies, such as free emergency transport,^{923 51},^{53 67} can significantly shape the factors influencing care-seeking behaviours and must be considered when applying the model in diverse contexts. However, the increased granularity of our model also renders the model more adaptable and flexible in its application to different contexts.

Lastly, in this article, we do not propose a specific programme theory or set of rules regarding the new model’s application. Instead, we propose our expanded model as a flexible tool that can be adapted to diverse contexts and health systems. While factors such as the duration of delays, financing elasticities or health system responses will vary across settings, the model offers a structured way to identify where in the patient journey delays occur and which actors are involved. Researchers and policymakers can then operationalise the model in their own contexts, for example to analyse factors influencing delays, such as availability of surgical staff, or to test hypotheses about the effects of specific determinants on maternal and neonatal outcomes. We position the expanded model primarily as a descriptive and analytic framework, designed to map where, how and why delays occur in patient journeys. It offers a structured lens through which researchers and policymakers can identify bottlenecks in obstetric referrals, map key determinants along care journeys, and design and evaluate interventions. In this way, the Six Delays Model provides a foundation for generating context-specific evidence and designing targeted interventions, across a variety of settings.

Conclusions

We used the illustrative example of maternal mortality in Madagascar to show how the conventional three delays model falls short in capturing the full complexity of pyramidal healthcare systems in LMICs. Using this example, our work extends the original three delays model to a Six Delays Model to improve its applicability to maternal healthcare across LMICs with pyramidal healthcare systems. Through this extension, the model can provide a more nuanced understanding of the barriers to accessing adequate care for obstetric emergencies in LMICs, encompassing the entire patient journey, offering a more accurate and actionable model for addressing barriers to maternal healthcare. This model equips policymakers and HCPs with the model to identify critical delays and implement targeted, context-specific interventions.

Future research should focus on examining the relative importance of these delays within care journeys and on identifying potential interventions to address delays at each stage of the referral process. Policymakers and implementers should use this evidence to design contextually relevant strategies that address each identified delay, thereby improving healthcare delivery and outcomes in LMICs.

Data availability statement

Data for this paper were drawn from publicly available sources and are available to others through those sources. Internal NGO data used in this study are available from the corresponding author upon reasonable request.