Identifying delays to emergency laparotomy in a low-resource setting using the Three Delays Model: a prospective observational study (DEL study)

Abstract

Background

Over the last decade, there has been a global emphasis on prioritising disparities in surgical care in low- and middle-income countries. The Three Delays Model, originally developed for maternal mortality, has been increasingly applied to surgical and injury populations, yet few studies have quantified delay predictors for emergency abdominal surgery in sub-Saharan Africa. This study aims to identify where the longest delays in accessing emergency laparotomy are, determine factors associated with delays and identify delays associated with mortality.

Methods

We conducted a prospective observational study of 240 consecutive emergency laparotomy patients at a tertiary referral centre in Lusaka, Zambia (April 2024 to April 2025). Data were collected from patients or caregivers and medical record review by the research team during the postoperative phase of care using a questionnaire. Using the Three Delays Model, variables were collected for each of the following delays: (1) seeking care, (2) reaching care and (3) receiving care. The primary outcome was time spent in each delay. Multivariable linear regression on log₁₀-transformed delay times was used to identify predictors of the first and second delays across the full analytic cohort. Logistic regression assessed delays associated with in-hospital mortality.

Results

Among 240 patients, the median total time from symptom onset to emergency laparotomy was 34 hours. The longest delay occurred in the seeking care phase, with a median of 18.5 hours (IQR: 2–72 hours), followed by reaching care with a median of 7.5 hours (IQR: 3–29.5 hours) and receiving care with a median of 8 hours (IQR: 6–11 hours). On multivariable log10 linear regression, believing symptoms would resolve and being referred from another facility predicted prolonged first delay. Predictors of prolonged second delay included less urgent surgical indication, education status and referral status. There were 28 in-hospital deaths (11.7%). Time in the first delay was associated with mortality (OR 1.01, 95% CI 1.001 to 1.008, p=0.006).

Conclusions

The longest delays to emergency laparotomy in Zambia occurred before patients entered the health system. Symptom misperception and referral from other facilities were the strongest modifiable predictors of delay.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• This study provides the first prospective patient-level report of delay predictors across multiple phases of emergency surgical care in Zambia.

• Symptom misperception and referral from other facilities emerged as the strongest modifiable drivers of prehospital delay and the care-seeking phase accounted for the largest share of total time to surgery.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• With these data, interventions for emergency surgery patients can be developed where they will have the most impact on decreasing the time to surgery and perioperative mortality.

• The quantitative predictor models identified here could be combined with complementary methodologies, such as process mapping across facility networks to design and target context-specific interventions along the surgical care pathway.

Background

Around the world, millions of people who need surgery still cannot get it, and the gap continues to widen. Recent analyses estimate that more than 160 million additional operations are needed each year to prevent death and disability, an increase from the 143 million reported by the original Lancet Commission on Global Surgery in 2015.1,3 Much of this unmet need lies in sub-Saharan Africa, where deaths from treatable surgical diseases outnumber those from HIV/AIDS, tuberculosis and malaria combined.4,9

In response, countries such as Zambia have made the expansion of surgical services a national priority. Over the past decade, regional initiatives and National Surgical, Obstetric, and Anaesthesia Plans (NSOAP) have sought to increase surgical capacity, integrate surgery into universal health coverage and strengthen referral networks.¹⁰ Yet, progress has been uneven and slow. Limited funding and workforce shortages persist, and reliable data to guide investments remain scarce.¹¹ The Three Delays Model, which categorises obstacles to care into delays in seeking, reaching and receiving treatment,¹² has driven major improvements in maternal mortality reduction and has been proposed for surgical and trauma populations.¹³ A systematic review of 119 studies mapped to this framework found that delay 3 (receiving care) dominated the research literature (79.3% of studies) and only 10.8% examined delay 1 (seeking care), despite evidence that prehospital barriers may account for up to 40% of avoidable trauma mortality in low- and middle-income countries (LMICs).¹³ In the emergency laparotomy population specifically, a recent multinational study of 187 hospitals across 51 countries identified lack of emergency medical services, limited health literacy and patient financial constraints as the leading causes of prehospital delay in lower-Human Development Index (HDI) settings.¹⁴ Prospective, patient-level data on where, why and how long delays occur along the surgical care pathway remain scarce in sub-Saharan Africa.

To address this evidence gap, we applied the Three Delays Model to study patients undergoing emergency laparotomy at Zambia’s national referral hospital. Our objectives were to determine where the longest delays occur, identify factors associated with the first two delays and determine which delay is linked to mortality. By tracing steps from symptom onset to surgery, this study aims to uncover where delays might be reduced to improve outcomes in resource-limited settings.

Methods

Study design and setting

We conducted a prospective observational study of patients undergoing an emergency laparotomy at the University Teaching Hospital (UTH) in Lusaka, Zambia. UTH is the country’s largest tertiary referral centre and receives emergency referrals from district and provincial facilities nationwide. Its role as the primary destination for complex surgical emergencies made it an appropriate setting to study system-level delays. This study was developed by the Center for Surgical Healthcare Research (CSHR) located in UTH’s department of surgery. The CSHR is comprised of physicians, community members and patients who collaborated on the development of this research. Specifically, the context-specific questions on delays were developed from the literature and from input from local surgeons, patients and community members. Input from local CSHR members helped identify cultural norms to seek counsel from traditional healers and religious leaders before seeking care at a healthcare facility for medical ailments. CSHR patient and community members will also be involved in the dissemination of this work through The Republic of Zambia’s NSOAP working group.

The CSHR and foreign university have a formal collaboration agreement. The primary author, who lived and worked in Zambia during the study period, and the local authors are Institutional Review Board and National Health Research Authority approved for reviewing patient sensitive data. The data were de-identified for non-local authors for subsequent analysis. This study adheres to the STROBE guidelines.¹⁵

Participants and data collection

Consecutive patients aged 18 years and older who underwent emergency laparotomy between 8 April 2024 and 8 April 2025 were eligible for inclusion. Patients were excluded if they underwent elective operations, were pregnant or were prisoners. Data were collected from patients or primary caregivers and from medical records in the postoperative phase of care and prior to discharge. Family members were allowed to consent if the patient was unwell or preferred their family member answer questions. The primary author and two junior resident medical officers collected data using a questionnaire built in Research Electronic Data Capture (REDCap). Consent forms were translated to the most common local languages and the questions were administered in the patient’s preferred language and dialect. All information was entered into REDCap hosted by the University of Zambia. Data verification for completeness and accuracy was performed weekly by the study team to ensure interrater reliability. Data currently resides in the REDCap web application and can be requested from the primary or last author.

Measurement of delay times

Using the Three Delays Model, variables were collected for each of the following delays: (1) seeking care, (2) reaching care and (3) receiving care. Seeking care was defined as the time from symptom onset to the participant’s decision to seek healthcare. Reaching care was defined as the time from the participant’s decision to seek healthcare to arrival at UTH (definitive care hospital). Receiving care was defined as the time from UTH presentation to surgery start.

Variable definitions and analytic approach

To identify factors associated with delays in seeking and reaching surgical care, we conducted multivariable linear regression analyses using the full analytic cohort. The outcome variable for each model was delay time in hours. Delay distributions were right-skewed; we therefore log₁₀-transformed delay times prior to analysis to satisfy normality assumptions for residuals and stabilise variance. Regression coefficients represent the estimated change in log₁₀ (delay hours) per unit change in each predictor. Back-transformation (10^coefficient) yields a multiplicative effect on the geometric mean.

Two separate models were constructed for the first delay (seeking care) and second delay (reaching care). Covariates were selected a priori based on clinical relevance and prior literature and included demographic factors (age, sex, education, income, employment, insurance status, HIV status, smoking, alcohol use), prehospital behaviours (consulting religious leaders or traditional healers, perception of illness severity, prior bad experience with the health system, distance and transportation barriers, financial barriers and family obligations), referral characteristics (referral status, prior facility visit) and surgical indication classified by urgency. Emergency laparotomy indications were separated into most urgent (incarcerated hernia, gastric perforation, typhoid perforation, volvulus and trauma) and less urgent (appendicitis, small bowel obstruction, large bowel obstruction and other). The full list of variables collected is listed in online supplemental appendix 1. Model fit was assessed by inspection of residual plots for linearity, homoscedasticity and normality of residuals. Binary versions of the traditional healer and alcohol variables were retained after coded categorical versions were excluded for significant missingness causing model failure.

Associations between delay duration and in-hospital mortality were tested with multivariable logistic regression, adjusting for age, sex, HIV status and referral status. Statistical analyses were performed in Stata V.18. Significance was defined as p<0.05.

Results

During the 1-year study period, 248 patients underwent an emergency laparotomy at the UTH. After exclusion, 240 patients were included in this study. Seven patients had incomplete data and one patient did not consent. The median age was 36 years and 77.5% were male. Most participants had no health insurance (76.7%) and reported a monthly income below the national average wage (approximately US$36 per month). Nearly 9 in 10 had attained a primary or secondary education, and just over 15% were living with HIV. The majority (80%) resided within Lusaka Province, though a substantial number were referred from other regions (table 1). Common operative indications included gastroduodenal perforation (27%), intestinal volvulus (18%), adhesive bowel obstruction (12%) and trauma (11%).

Table 1. Baseline demographics.

Variable		Median (range interval)
Age		36 (17; 88)
Variable		n (%)
Sex	Male	186.00 (77.50%)
	Female	54.00 (22.50%)
Province	Lusaka	193.00 (80.42%)
	Other	47.00 (19.58%)
Insurance	Public	52.00 (21.67%)
	None	184.00 (76.67%)
	Private	4.00 (1.67%)
Education	None	14.00 (5.88%)
	Primary	80.00 (33.61%)
	Secondary	119.00 (50.00%)
	University or above	25.00 (10.50%)
Employment	Unemployed	71.00 (29.58%)
	Employed	52.00 (21.67%)
	Self-employed/farmer	87.00 (36.25%)
	Student	18.00 (7.50%)
	Retired	12.00 (5.00%)
Income (ZMK)	<1000	153.00 (64.02%)
	1000–3000	55.00 (23.01%)
	>3000	31.00 (12.97%)
Tobacco use	Yes	74.00 (30.83%)
	No	166.00 (69.17%)
Alcohol use	Yes	106 (44.20%)
	No	134 (56.80%)
HIV status	Unknown	86 (36.80%)
	Yes	36 (15.4%)
	No	32 (30.5%)
Prior abdominal surgery	Yes	25.00 (10.68%)
	No	209.00 (89.32%)
Referral	Yes	193.00 (80.42%)
	No	47.00 (19.58%)
Healthcare visit and discharge prior to admission	Yes	132 (55.7%)
	No	105 (44.3%)

ZMK, Zambian Kwacha.

The median total time to operation was 34 hours. The longest delay occurred before patients decided to enter the health system, with a median of 18.5 hours (IQR: 2–72 hours) from symptom onset to the decision to seek care. Seeking care variables are listed in table 2. The median time to reach definitive care was 7.5 hours (IQR: 3–29.5 hours) and the median in-hospital delay (from UTH arrival to the surgery start) was 8 hours (IQR: 6–11 hours).

Table 2. Seeking care.

Seeking care variable	Yes (n, %)	No (n, %)
Sought the advice of a traditional healer	21 (8.8%)	219 (91.2%)
Sought the advice of a religious leader	58 (24.2%)	182 (75.8%)
Prior bad experience with the health system	89 (37.1%)	151 (62.9%)
Concerns about distance to the hospital	73 (30.4%)	167 (69.6%)
Concerns about finding transportation to the hospital	90 (37.5%)	150 (62.5%)
Concerns about financing healthcare	120 (50%)	120 (50%)
Competing family obligations	33 (13.9%)	205 (86.1%)
Thought symptoms would resolve	102 (42.7%)	137 (57.3%)

More than half of all participants (56%) reported being sent home after their first healthcare visit, and one in four (24%) sought advice from a religious leader before deciding to present to a hospital. Consultation with traditional healers was less common but still present. Most transfers (80%) to UTH originated from referral hospitals where participants spent a median of 7 hours before transfer, highlighting persistent interfacility bottlenecks.

On multivariable linear regression of log₁₀-transformed delay times across the full cohort, two variables predicted prolonged first delay (seeking care). Patients who believed their symptoms would resolve without medical treatment had a 2.28-fold longer geometric mean delay compared with those who did not hold this belief (coefficient 0.358, 95% CI 0.165 to 0.552, p<0.001). Referred patients had a 2.30-fold longer first delay (coefficient 0.362, 95% CI 0.102 to 0.622, p=0.007). Age, HIV status and consultation with a religious leader showed directional trends consistent with longer delays but did not reach statistical significance in this model (age: coefficient 0.006 per year, p=0.162; HIV negative vs positive: coefficient −0.169, p=0.276; religious advice: coefficient 0.162, p=0.192).

For the second delay (reaching care), four variables emerged as predictors. Less urgent surgical indication was associated with a 1.69-fold longer geometric mean compared with the most urgent cases (coefficient 0.227, 95% CI 0.062 to 0.392, p=0.007). Education level showed a counterintuitive pattern: patients with primary education had a 2.60-fold longer delay (coefficient 0.416, 95% CI 0.189 to 0.644, p<0.001) and those with secondary education a 2.00-fold longer delay (coefficient 0.301, 95% CI 0.064 to 0.538, p=0.013), each compared with patients with no formal education. University-level education was not significantly associated with delay (coefficient 0.072, p=0.630). Referral from another facility again predicted a longer reaching care delay (2.11-fold; coefficient 0.324, 95% CI 0.133 to 0.516, p=0.001). The distribution of delay times across all three phases is illustrated in the box and whisker plots (figure 1), which show wide variation in the prehospital period compared with later stages.

Figure 1. Time in each delay.

There were 28 in-hospital deaths (11.7%). On multivariable logistic regression, the duration of the first delay (time to decision to seek care) was independently associated with mortality (p=0.006; OR 1.01, 95% CI 1.001 to 1.008). The Kaplan-Meier curve (figure 2) shows the cumulative probability of surgery over time among patients who survived and those who died (p=0.004, HR 0.56, 95% CI 0.37 to 0.84).

Figure 2. Probability of surgery over time from symptom onset to surgery start.

Discussion

This prospective study of 240 emergency laparotomy patients in Lusaka, Zambia, found that the longest delays to surgery occurred before patients entered the formal health system. The care-seeking phase accounted for a median of 18.5 hours of the 34-hour total. Two factors predicted prolonged first delay on multivariable regression: the belief that symptoms would resolve (2.28-fold longer geometric mean) and referral from another facility (2.30-fold longer). For the second delay, surgical urgency, education level and referral status were significant predictors. In-hospital mortality was 11.7% (28/240), and first delay duration was associated with mortality after adjustment for age, sex, HIV status and referral status (OR 1.01 per hour, p=0.006). The per-hour effect is modest, but the clinical significance lies in cumulative exposure: a patient delayed by the median 18.5 hours faces a meaningfully different risk profile than one who presents promptly.

The dominance of the care-seeking phase is consistent with work applying the Three Delays Model to surgical and trauma populations across LMICs. In Tanzania and Timor-Leste, studies have shown that delay 1 contributes the largest share of total delay, driven by reliance on traditional or faith-based healers and uncertainty about symptom severity.^{16 17} In Somaliland, caregivers cited financial constraints, lack of available care and cultural preferences for home remedies as major barriers to timely presentation.¹⁸ A systematic review by Whitaker et al found that only 10.8% of 119 injury care system assessments in LMICs examined delay 1, despite evidence that prehospital barriers contribute substantially to avoidable mortality.¹⁹ The Global Outcomes After Laparotomy for Trauma (GOAL) Collaborative, surveying 187 hospitals across 51 countries, identified a lack of emergency medical services, limited health literacy and patient financial constraints as the top-ranked causes of prehospital delay in lower-HDI settings.¹⁴ The companion multinational study of trauma laparotomy outcomes demonstrated that patients in lower-HDI settings presented with less severe injuries yet experienced higher mortality, a pattern attributed in part to system-level delays and resource gaps rather than injury severity alone.²⁰ Our findings quantify these dynamics at the patient level in Zambia and reinforce that the first delay is both the longest and the most amenable to targeted intervention.

Counterintuitively, education predicted longer delays to reaching care. Patients with primary or secondary education waited at least two times as long as those with no formal education, though this association did not extend to university-educated patients. One explanation is that patients with some education may be more likely to initially present at lower-level facilities, attempt self-medication or navigate a more complex health-seeking pathway before reaching tertiary care. The significance of referral status in both delay models supports this interpretation: the transfer system between facilities functions as a bottleneck regardless of patient characteristics, and patients who enter the referral chain at a peripheral facility accumulate delay at each step. More than half of participants in this study were sent home after their first healthcare encounter with what proved to be a time-sensitive surgical condition, a finding that parallels prior evidence from Malawi and other sub-Saharan African settings where diagnostic uncertainty at first-level hospitals has been linked to poor outcomes.²¹ Whitaker et al, using process mapping across 11 facilities in northern Malawi, identified similar referral barriers and proposed that pairing quantitative predictor models with system-level process analysis could generate more actionable intervention targets.¹⁹ Our data and theirs together suggest that both the patient decision to seek care and the system’s capacity to respond without unnecessary transfers are critical intervention points. Also of interest, over 75% of participants were male which may be due to cultural trends in healthcare seeking or lifestyle choices including alcohol consumption. Alcohol use may play a part in the large number of young male patients presenting with gastric perforations and future qualitative work would help identify some of these trends.

This study has several limitations. It was conducted at a single urban referral hospital, and delay patterns may differ in rural areas where transport times and access barriers are greater. Additionally, while UTH is the largest hospital, there are surrounding private, lower-levelled public and mission hospitals that provide surgical services. According to Zambia’s perioperative capacity survey, 71% of levelled hospitals are able to perform an exploratory laparotomy and there are 17 levelled hospitals in the Lusaka Province²²; this may limit generalisation of our findings to the population. Data on symptom onset and decision times were based on patient or caregiver recall, introducing potential measurement bias. The study was also limited to patients who ultimately reached surgical care and does not capture those who died before arrival at UTH or who were managed at secondary-level facilities, a group whose delay profiles may differ substantially. As a public hospital, UTH offers discounted or free healthcare but patients are still required to purchase medications, supplies and nutrition during their hospitalisation. There may be a selection bias of lower income patients presenting to UTH instead of private hospitals. Patients under 18 years were excluded. Children comprise a large proportion of the sub-Saharan African population, and paediatric delay profiles warrant separate investigation; consent requirements, care-seeking dynamics and decision-making patterns differ from those of adults. Emergency laparotomy encompasses heterogeneous surgical conditions, from acute perforations to subacute bowel obstructions to trauma. Surgical indication was included as a covariate in the regression models, but the sample size limited our ability to perform condition-specific subgroup analyses. The observation that 56% of patients were sent home at first presentation may partly reflect the subacute nature of some conditions where initial discharge is clinically appropriate. An earlier version of this analysis used a tertile-based comparison (early vs late groups). More than one-third of participants reported a prior bad experience with the health system, but this variable did not predict first delay duration in our regression. Whether prior care quality shapes future care-seeking decisions is a question this quantitative design could not fully answer. Complementary qualitative work would be valuable to explore the cultural and behavioural drivers of delayed presentation.

These findings have implications across the care continuum. For patients and communities, the strong association between symptom misperception and delayed care-seeking points to a need for public education campaigns targeting recognition of acute abdominal emergencies. In a community with strong religious affiliations, there is an opportunity to partner with churches and faith-based organisations for surgical disease education. Community healthcare workers, a strategy with established evidence for raising disease awareness in LMICs,²³ could be deployed for early symptom recognition and referral. For providers and frontline facilities, the number of patients with time-sensitive pathologies who were sent home at first presentation underscores the need for surgical education outreach in a system where task-sharing is common.²⁴ For the health system and policymakers, referral status predicted longer delays in both the first and second phases and 80% of participants required transfer between facilities. Per the 2025 perioperative capacity survey, only 52% of Zambian hospitals report reliable ambulance availability.²² Decentralising surgical services from tertiary to district hospitals, a priority outlined in the revised NSOAP, has the potential to decrease transfer burden, improve access and shorten delays.²⁵ The delay-mortality association, while modest on a per-hour basis (OR 1.01), accumulates across the observed distribution; strategies that compress the median 18.5-hour first delay by even a few hours could translate into a population-level survival benefit. We did not formally assess care quality at UTH and cannot isolate how perioperative factors contributed to the 11.7% mortality rate, which falls within the range reported across comparable sub-Saharan African settings.^{26 27} While prior work has demonstrated a significant mortality correlation with prehospital delay there are also prospective studies evaluating in-hospital mortality in LMICs. An international, 7-day cohort study of adults undergoing inpatient surgery by the International Surgical Outcomes Study group identified failure to rescue and less critical care unit admissions as possible reasons for the mortality rates in younger patients with lower American Society of Anesthesiologist (ASA) scores.²⁸ A full accounting of emergency surgical outcomes in Zambia will require attention not only to prehospital access but also to perioperative care quality, including resuscitation, sepsis management and critical care capacity.

Conclusion

In Zambia, the greatest delays to emergency laparotomy occur before patients reach the hospital and are associated with mortality. Symptom misperception and referral from other facilities are the strongest modifiable predictors of prehospital delay. Addressing these delays will require strategies that extend beyond hospital capacity to strengthen community awareness, early diagnosis and referral efficiency. Pairing quantitative delay measurement with system-level process analysis could inform targeted interventions along the surgical care pathway. Improving outcomes will demand parallel investment in both prehospital access and the quality of perioperative care.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data availability statement

Data are available upon request from the authors.