Patterns, Predictors and Outcome of Time to Presentation Among Critically ill Paediatric Patients at Emergency Department of Muhimbili National Hospital, Dar es Salaam, Tanzania

Alphonce N Simbila Muhimbili University of Health and Allied Sciences Said S. Kilindimo (  skilindimo@yaho.com ) Muhimbili University of Health and Allied Sciences Hendry R. Sawe Muhimbili University of Health and Allied Sciences Zawadi E. Kalezi Muhimbili University of Health and Allied Sciences Amne O. Yussuf Muhimbili University of Health and Allied Sciences Hussein K. Manji Muhimbili University of Health and Allied Sciences Germana Leyna Muhimbili University of Health and Allied Sciences Juma A. M nanga Muhimbili National Hospital Ellen J. Weber Muhimbili National Hospital

poor family were independent predictors of delay. Infants and those referred from other facilities had 2.2 (95% CI 1.3-3.8) and 1.7 (95% CI 1.1-2.7) times increased odds of presenting late to the ED respectively. The overall 30-day in-hospital mortality was 26.5% in which those who presented late were 1.3 more likely to die than those who presented early (RR=1.3, CI: 0.9-1.9). Majority died >24 hours of ED arrival (P-value=0.021).
Conclusion: Delayed ED presentation of more than 48 hours from onset of illness was associated with inhospital mortality. A larger study is needed to evaluate the care pathway of critically ill paediatric patients to identify preventable course of delay to tertiary care facility.

Background
Critical illness is a life-threatening process which, without timely medical or surgical intervention, is highly likely to result in death (1). Half of the deaths among children under ve years globally occurred in sub-Saharan Africa in 2017. Unlike in high income countries where 1 in 185 children died before the age of ve years, 1 in 13 died before the age of ve in Sub-Saharan Africa. The risk of dying for a child in sub-Saharan Africa is 15 times higher than in Europe. Most of these children die due to treatable and preventable causes such as complications of during birth, pneumonia, diarrhea, malaria and neonatal sepsis (2). A 2010 survey by the ministry of health in Tanzania showed high mortality rate in under-ves with 75% of deaths occurring in the rst 24 to 48 hours after admission (3). The unfolding of events in the process of critical illness is in uenced by multiple factors which can potentially modify and affect the outcome of critical illness. In addition to improvements in the quality of care for children reducing delays in critical illness is among the main approaches to reducing mortality.
Paediatric critical illness in the LMICs differ from the developed countries in that children tend to be younger and suffer more from infectious causes of illnesses. Late presentations to the hospital due to referrals and travels through long distances to reach hospitals are common occurrence which contribute to an increased disease severity and mortality upon admission (4).
Delayed presentation to the hospital in the course of paediatric critical illness has been shown to be one of the factors that negatively in uence health outcomes (5). Delayed health care seeking of more than 48 hours has been observed in 35% of paediatric patients in the acute phase of illness in Kigali, Rwanda(6).
At a tertiary hospital in Ethiopia about 4. 1% of children died at the Paediatric Emergency Department, which translates to a mortality of 8.2 per 1000 children. Delay of more than 48 hours since the onset of symptoms was among the top causes of early mortality (5). In Southwestern Uganda, 50% of paediatric patients who presented late to the health facility and were admitted died within 24 hours of admission (7). In Dodoma Tanzania, the median time of delay to seek care among children under ve years with fever was 2 days(8).There is an association between delayed ICU admission and mortality. Fraction of mortality attributable to ICU delay was 30% (9). Many studies in this area have looked at timeliness at the level of the family which affects the rst phase in the Thaddeus and Maine's delay model (10).
Presence of a full capacity ED, Paediatric ICU and de nitive paediatric services at MNH have provided early resuscitation, stabilization and other management for critically ill paediatric patients. The objective of this study was to describe patterns, determine factors associated with delay and nd out whether delay made a difference in mortality outcome among critically ill children who sought care at the ED of a national tertiary referral hospital.

Study design
This was a prospective observational cohort study of paediatric patients aged 28days to 14 years presented to the MNH ED from September 2019 to January 2020 and triaged level 1 according to Emergency Severity Index (ESI).

Study setting
The study was conducted at ED of MNH which is a national tertiary referral government hospital located in Dar es Salaam with 1500 bed capacity. The MNH serves an annual average of 60,000 patients who are referred from all over the country. The ED at MNH is full capacity public ED in Tanzania and the only training site for Emergency Medicine residency program in the country. On a daily basis the MNH ED attends about 150-200 critically ill patients, among them approximately 25% are children excluding neonates who present directly to the maternity unit. The MNH ED uses a triage system modelled on the ESI but with only 3 levels, with emergent being equivalent to levels 1 and 2 of ESI, and they are assigned to the resuscitation room. Apart from ability to provide full resuscitation including continuous cardiopulmonary monitoring, the department also has a full range of point of care tests like ultrasound, chemistry, blood gas analyzer and portable x-ray. After resuscitation and stabilization, the critically ill children get admitted to general pediatrics ward or pediatric intensive care unit (PICU).

Study Participants
We included all pediatric patients (aged 28 days to 14 years) triaged as emergent whose parents/guardians consented to participate in the study.

Study protocol
Consecutive sampling technique was employed to enroll patient s where by data collected for 24 hours/day on alternate days. Demographics, clinical presentation, initial management, and ED outcomes were observed and documented using information given by the parent/guardian, the treating physician, and the electronic medical record (Wellsoft TM ) using structured case report form. All patients were asynchronously followed up in hospital wards (if admitted) to determine their in-hospital outcome (discharge/mortality) in 24 hours and later weekly for a maximum period of 1 month after admission.

Measurements:
Each caretaker of a critically ill paediatric patient was asked about the date and time of onset of an index illness of the child. Time of presentation to the MNH ED was therefore calculated by nding the difference between these. Delay was de ned as presentation to MNH ED after 48 hours from onset of illness. Wealth was determined based on household characteristics and asset ownership hence relative household wealth index was constructed using principal component analysis (PCA). Households were then ranked in ascending order. The scores were separated into quintiles; each representing 20% of the population. Those in the highest quintile might not have been rich but were in higher socioeconomic status than 80% of the participants in this study.

Outcomes
The primary outcome was the association of delay with mortality and secondary outcomes was predictors of delay among critically ill paediatric patients.

Data analysis
Data were imported into the Statistical Package for Social Science for analysis (SPSS) (version 26.0, IBM, LTD, North Carolina, USA) from the Research Electronic Data Capture (RedCap version 7.2.2, Vanderbilt, Nashville, TN, USA). Relevant frequencies and tables were generated for categorical variables (injury and referral factors). Medians/inter-quartile ranges were calculated for continuous variables. We calculated a proportion of children with delayed presentation and contingency tables were constructed for univariate analysis to explore differences between children who had a delayed vs. timely presentation using the Chisquare test. Multivariate logistic regression analysis was completed on variables with p value ≤0.20 in the univariate analysis to identify independent predictors of delayed presentation. Relative risks were computed for association of delay with overall, early and late mortality. The odd ratios and 95% con dence intervals were estimated for each studied factor. Statistical signi cance was set at p-value<0.05.

Results
A total of 3616 paediatric patients attended the ED during the study period, of whom 745 (20.6%) were triaged "emergent". A total of 440 (59.1%) patients were eligible and consented to participate in the study.
Of the 99(26.5%) who died within 30 days of presentation, 64(29.5%) presented late (after 48 hours) and 35 (22.3%) presented early (before 48 hours). (Figure 1) Socio-demographic characteristics of critically ill paediatric patients at the MNH ED and their caretakers In the studied cohort, majority 281(63.9%) were males, the median age was 12 [IQR =9-60] months and majority 321(73.0%) were referred from a lower-level health facility. Most of the caretakers of the critically ill paediatric patients were parents 410(93.2%) and more than half of them 260 (59.1%) were between the age of 25 and 34 years. Slightly more than half of them 236(53.6%) had primary education, 145(33.0%) were unemployed, 46(10.5%) were peasants and 59 (67.8%) were the poorest in the socioeconomic status. (Table 1) Of the paediatric patients, 72% were admitted to the ward, 19% were admitted to the PICU and 9% (33/374) died while receiving care at the ED.

Discussion
Delay to seek care has been reported in several studies as a cause of avoidable morbidity and mortality in children and many of these studies was done in higher income countries with limited information in lowand middle-income countries. In our study over fty percent of critically ill paediatric patients presented late (after 48 hours from onset of illness) to the ED. This is similar to ndings by a study in Ethiopia but higher than that observed in Rwanda with delay of thirty ve percent of paediatric patients. (5,6). A possible explanation could be differences in sociodemographic characteristics and lifestyle of the study settings. The hierarchical referral system in the setting of this study could be contribute to the observed delay as this study was conducted in tertiary hospital which is the highest destination in the chain of referral (11).
In this cohort almost a quarter of the critically ill children who presented to the ED after 48 hours from onset of illness died with almost two-thirds of the deaths occurred beyond 24 hours of hospital admission. This is similar ndings to the study done at a tertiary hospital in Ethiopia (5). This may signify the impact of resuscitation and stabilization at the ED before admission that presumably prolonged their lives beyond 24 hours but later decompensated (13). However, limited number of beds in paediatric ICU could explain the occurrence of death beyond ED stabilization We also found that critically ill children below one year of age were more likely to present late to the ED.
Non-speci c symptoms especially in infants and caretakers' poor knowledge on danger signs of critical illnesses might contribute to delay in this age group (14). Our study also found coming from the poorest households independently doubled the odds of a critically ill paediatric patient being late to the tertiary hospital during a critical illness. This is in keeping with a study conducted in Ethiopia which had similar ndings (15).
About two thirds of the critically ill paediatric patients who delayed were referred from primary health care facilities with no capability to care for critically ill children. Failure to recognise critical illness with series of evaluations by primary health care providers before they are referred to de nitive care signi cantly lead to delay with limited management (16). The existing referral system in place (11) doesn't take into account the urgency to de nitive care hence more delay with poor outcome observe.

LIMITATIONS
This study was conducted in urban settings and single center tertiary hospital with full capacity to resuscitate and stabilize critically ill paediatric patients. However, the MNH ED receives referrals from all over the country, the patients sampled are likely to provide a true representation of the Tanzanian population of critically ill paediatric patients.
Patients were only enrolled when researchers were in the department; some of potential participants might have been missed. However, as the researchers worked different shifts, there is no reason to expect that the missed patients were different from those that were enrolled.
There was loss to follow up encountered in this study. However, it was mitigated by including 10% loss to follow up rate during the estimation of the sample size of the study participants who were to be included during the development of the study protocol.

Conclusion
The proportion of delayed presentation to tertiary hospital among critically ill paediatric patients is substantially high. Age below one year, being referred from a primary healthcare facility and being in the poorest category of socioeconomic status predicted delay to appropriate care facility. Future studies should focus on the evaluation of care pathway among critically ill paediatric patients to identify preventable delays in care.

Competing interests
The author declares no con icts of interest Ethics approval and consent to participate The study was conducted after obtaining permision from the MUHAS Institutional Review Board and MNH. All patients were entered into the study after an informed consent was signed either by the patients themselves or their guardians in the case where patients were not able to do so. The investigator received the waiver of consent from the MUHAS Institutional Review Board incase patients weren't able to consent such as unconscious patients with absence of relatives to provide consent. Only participants who on their free will consent were included in the study. Con dendiality was observed on the questionnaires lled with participants information. All patients received treatment as per standard hospital policies. The data obtained during the study was kept anonymous.

Consent to publish:
Not applicable Availability of data and material The dataset supporting the conclusion of this article is available from the authors on request

Funding
This was a non-funded project; the principal investigators used their own funds to support the data collection and logistics Authors' contributions ANS was involved in the study design conceptualization, data collection, analysis and interpretation, drafted the manuscript, and made all necessary changes to the manuscript. SSK was involved in the study design conceptualization, review of the data analysis and interpretation, and critical review of the manuscript. HRS was involved in the study design conceptualization, review of the data analysis and interpretation, and critical review of the manuscript. JAM was involved in review of the data analysis and interpretation, and critical review of the manuscript. ZEK and GL were involved in the study design conceptualization, data analysis and interpretation and revision of the manuscript. AOY were involved in study design conceptualization and revision of manuscript. HKM were involved in study design conceptualization and revision of manuscript. EJW was involved in the study design conceptualization, review of the data analysis and interpretation together with revision of the manuscript. All authors have read and approved the nal manuscript.  Distribution of early and late mortality vs early and late presentation among paediatric patients with ESI triage level 1 *(p-value=0.021)