Pediatric emergency care in a low-income country: Characteristics and outcomes of presentations to a tertiary-care emergency department in Mozambique.

BACKGROUND: An effective pediatric emergency care (PEC) system is key to reduce pediatric mortality in low-income countries. While data on pediatric emergencies from these countries can drive the development and adjustment of such a system, they are very scant, especially from Africa. We aimed to describe the characteristics and outcomes of presentations to a tertiary-care Pediatric Emergency Department (PED) in Mozambique.
METHODS: We retrospectively reviewed PED presentations to the "Hospital Central da Beira" between April 2017 and March 2018. Multivariable logistic regression was used to identify predictors of hospitalization and death.
RESULTS: We retrieved 24,844 presentations. The median age was 3 years (IQR 1-7 years), and 92% lived in the urban area. Complaints were injury-related in 33% of cases and medical in 67%. Data on presenting complaints (retrieved from hospital paper-based registries) were available for 14,204 (57.2%) records. Of these, respiratory diseases (29.3%), fever (26.7%), and gastrointestinal disorders (14.2%) were the most common. Overall, 4,997 (20.1%) encounters resulted in hospitalization. Mortality in the PED was 1.6% (62% ≤4 hours from arrival) and was the highest in neonates (16%; 89% ≤4 hours from arrival). A younger age, especially younger than 28 days, living in the extra-urban area and being referred to the PED by a health care provider were all significantly associated with both hospitalization and death in the PED at the multivariable analysis.
CONCLUSIONS: Injuries were a common presentation to a referral PED in Mozambique. Hospitalization rate and mortality in the PED were high, with neonates being the most vulnerable. Optimization of data registration will be key to obtain more accurate data to learn from and guide the development of PEC in Mozambique. Our data can help build an effective PEC system tailored to the local needs.

Introduction

Over the last two decades, child mortality significantly decreased worldwide thanks to the development of the Millennium Development Goals (MDGs) and the Sustainable Development Goals (SDGs), elaborated by the United Nations to reduce healthcare disparities [1-5]. However, child mortality remains high in low-income countries (LICs) and, in particular, in Sub-Saharan Africa [6, 7].

The development of efficacious pediatric emergency care services has been identified as one of the crucial steps to reduce child mortality [8-10]. However, pediatric care services still represent the weakest links in the healthcare systems chain [11-13], with pediatric emergency medicine still being an understudied field [14, 15].

The study of the burden and profile of pediatric emergencies is important to understand how to optimize resource allocation and healthcare facilities to develop a structured emergency care system that could further reduce child mortality.

In Mozambique, as in other sub-Saharan countries, the health care system is extremely diverse. Mozambique has approximately 1,600 healthcare facilities (including health posts, health centers, district hospitals, provincial hospitals, and four referral/central hospitals) distributed in 11 provinces, 30 municipalities, and 157 provinces. Overall, 96% of these facilities only deliver primary care (i.e. essential preventive and curative health). Only two of the four tertiary-care referral hospitals have a Pediatric Emergency Department (PED), one of which is in Beira, the capital of the Province of Sofala. In 2017, the under-5 and neonatal mortality rates in Sofala were 75.6 per 1,000 and 25 per 1,000 live births, respectively [16-19].

As for other LICs, data on the epidemiology of pediatric presentations to the PED is important to develop tailored strategies to improve the management of acutely and critically ill children, in order to further reduce child mortality.

The present study aims to describe the profile and outcomes of pediatric presentations to a referral care PED in Mozambique over one year. We also aimed to identify predictors of hospitalization and death, eventually suggesting strategies to improve pediatric emergency care services.

Materials and methods

Study design and population

We retrospectively collected data on presentations of all children accessing the PED at Hospital Central da Beira (HCB, Beira, Mozambique), over a 12-month period, between April 2017 and March 2018. The upper age limit for patients' inclusion was 15 years. The study was approved by the ethical review committee of Hospital Central da Beira.

Study setting: Healthcare in Beira and the Sofala province

Beira is the second largest city of Mozambique (with approximately 530.000 inhabitants over an area of 633 km2) and the capital of the Sofala province (with approximately 2.3 million inhabitants, over an area of 68.018 km2), which includes 159 health centers and posts (13 in the urban area of Beira), four rural/district hospitals, and one referral Hospital (HCB). By merely averaging the overall distribution of healthcare facilities in the Sofala province, we obtain a health facility every 450 km2 in the extra-urban area and one every 45km2 in the urban area. At the time of the study an ambulance service was not available, and patients could reach health care facilities on foot, with private transport or with public transport, where available.

The HCB hosts a Pediatric Department with approximately 200 beds, and one of the two PEDs of the country. The wards are mostly staffed with generalists and pediatric residents, while only seven Pediatricians run the whole Pediatric Department. In the PED some beds are available for short-stay observation and a room is dedicated to Pediatric Intensive Care, where critical children are admitted once stabilized. The Pediatric Department also provides neonatal care to approximately 6000 newborns/year. The Neonatal Intensive Care Unit (NICU) counts about 30 beds and 2000 hospitalizations/year. Based on the most recent available data from 2017, in-hospital overall pediatric mortality was 13%, while in the NICU mortality was 33%.

Sources of data and data collection procedures

Demographic, clinical, and outcome data of all presentations to the PED were abstracted from three hospital paper registries: i) the “presentations registry”; ii) the “hospitalizations registry”; and iii) the “deaths registry”. Death and hospitalization registries were filled by PED medical personnel, while the presentation registry was filled by hospital administrative personnel. Details on data systematically recorded in each of the registries are reported in Fig 1. We were unable to collect data on interventions, treatments, or resuscitations in the PED because this information was reported on paper charts, which were not filed systematically and were therefore unavailable to either clinical or research staff. Also, reliable information on comorbidities was not available from the registries.

Fig 1

Characteristics of hospital registries from which study data were collected.

All the registries were reviewed by two of the Authors (VB and LNF) and patient identity was crossed-matched between registries. Data were entered into an electronic standard data collection system. Abstractors were trained locally based on the initial review of 200 registry records each. A two-month data abstraction overlap between the abstractors helped in ensuring consistency in data abstraction and coding, by training of the second data abstractor. No formal double entry of data by the two abstractors occurred during this time.

The following data were collected from the registries: sex, age, area of residence, modality of presentation, presenting complaint, outcome (discharge, hospitalization, or death, and time of death). Children's age ranged from 0 to 15 years, and the age variable was categorized in four age groups for analysis: from 0 to 28 days (neonates), from 29 days to 1 year (infants), from 1 to 5 years (preschoolers), and from 5 to 15 years (school-aged). The modality of presentation included self-presentations or referrals from other health care providers (i.e. health care centers, rural hospitals, private clinics, etc.). The area of residence variable was categorized in urban, when the child lived in the Beira urban area, and in extra-urban, which was sub-categorized into within the Sofala province and outside the Sofala province. Information on presenting complaints was categorized in the registries as medical or injury related. Medical complaints included the following locally predefined categories of non-traumatic complaints: fever of any origin, respiratory, gastrointestinal, cardiovascular, neurological, musculoskeletal, constitutional, sense organs (which included medical complaints to the eyes, ears, nose, throat and to the skin), and others (which included psychiatric disorders, genitourinary disorders, etc.). Injury-related presentations were locally classified into road accidents, falls, wounds, violence, inhalation/ingestion, burn, and drowning. This categorization system was maintained for data analyses in the current study. Time from presentation to death was categorized in early death (death on arrival or within four hours from arrival), and later death in the PED (after four hours from arrival). Unfortunately, the HBC did not have the facilities and resources (i.e., trained staff, information technology infrastructure) to code diagnoses according to the ICD 9/10 codes. Data on diagnosis were reported as per local documentation practices.

Statistical analysis

Descriptive statistics were reported as median and interquartile range (IQR) for continuous variables. Categorical variables were reported as proportions and percentages. The Wilcoxon test was used for comparison of continuous variables, while Chi-square and Fisher's exact tests, as appropriate, were used for categorical variables.

We then fit univariable logistic regression models specifying hospitalization, death, and death within four hours as the dependent variable, and clinical and demographic variables as independent variables. Subsequently, we fit multivariable logistic regression models to identify independent predictors of hospitalization, mortality, and mortality within four hours, with the variables that were found to be significant from the univariable models (p<0,05). Results of dependent variables analysis were tested again in a multivariable model for interaction with age, sex, area of residence, the modality of presentation (self vs referred presentation), the reason for presentation, and outcomes [20].

Given the high rate of missing data for the independent variable “presenting complaint” a sensitivity analysis was carried out to assess how missing values would affect the association of the independent variables with the dependent variable. With this respect we performed the following analyses:

A complete case analysis estimating the model on the valid case data (excluding records with missing data for the variable presenting complaint)

A missing data imputation analysis based on the model estimation on an imputed dataset. A Multiple Imputation by Chained Equations (MICE) procedure was used to handle the missing data for the variable presenting complaint.

An estimation of the multivariable model for the outcomes, but excluding the variable presenting complaint from the model.

When the p-value was <.05, the difference was regarded as statistically significant. All statistical tests were 2-tailed. All statistical analyses were performed using Stata Version 13.0 (StataCorp, College Station, TX) and R 3.6.2 together with caret, rms, and MICE packages [21-24].

Results

Patients characteristics

During the 12-month study period, 24,844 presentations were recorded. Of these, 14,448 (58.8%) were male, with a male to female ratio of 1.43 to 1. The median age at presentation was 36.5 months (IQR 12–85.2 months). The majority of patients (92%) came from the urban area, with 42% already been assessed by a health care provider in a health care center or at a countryside hospital. A summary of demographics and general characteristics of study presentations by age-group is presented in Table 1.

Table 1

Demographic characteristics of Pediatric Emergency Department presentations by age group.

Number of visits to the PED	Total	< 28 d	29 d–1 y	1–5 y	5–15 y
	(n= 24,844)	2.7% (n = 677)	22.7% (n = 5,634)	43.7% (n =10,845)	30.9% (n = 7,688)
Sex *
Male	58.8%	50.6% (260)	59.3% (3,315)	58.3% (6,293)	59.9% (4,580)
Residency *
Urban (Beira)	92.5%	87.0% (585)	92.0% (5,189)	94.0% (10,171)	91% (7,003)
Sofala Province	6.2%	12.0% (82)	6.0% (338)	5.0% (569)	7% (567)
Extra Sofala	1.3%	1.0% (9)	2.0% (105)	1.0% (102)	1.0% (115)
Modality of Access to the PED
Self-presentations	58.0%	54.0% (367)	62.0% (3,466)	59.0% (6,439)	45.0% (4,205)
Heath Care Center	36.0%	37.0% (249)	32.0% (1,827)	36.0% (3,885)	39.0% (2,976)
Peripheral Hospital	6.0%	9.0% (61)	6.0% (341)	5.0% (521)	7.0% (507)

*Data on sex available for 24,562/24,844 (98.9%); data on residency available for 24,835/24,844 (99.9%).

Data are reported in terms of percentages and absolute frequencies.

Data on presenting complaints specifications were available for 14,204 (57.2%) presentations. A medical issue was the reason for presentation in 67% of cases, while 33% were consequent to an injury. The most common medical presentations were respiratory diseases (29.3%), followed by fever (26.7%), and gastrointestinal disorders (14.2%). Among injury presentations, falls (63%) were the most common, followed by foreign body ingestion/inhalation (10.2%) and road accidents (9.8%). Data on presenting complaints by age group are reported in Table 2.

Table 2

Presenting complaints by age group.

PRESENTING COMPLAINT	Total	< 28 d	29 d–1 y	1–5 y	5–15 y
	n = 14,204	n= 297	n = 3,067	n = 6,232	n = 4,608
Injury	33.0% (4,682)	14.5% (43)	16.3% (500)	31.1% (1,941)	47.7% (2,198)
Fall	63.0% (2,949)	39.5% (17)	61.4% (306)	60.6% (1,178)	66.0% (1,448)
Ingestion/Inhalation	10.2% (480)	14.0% (6)	10.1% (52)	16.2% (314)	5.0% (108)
Road Accident	9.8% (457)	27.9% (12)	6.9% (34)	7.4% (144)	12.0% (267)
Wound	8.6% (401)	14.0% (6)	7.4% (37)	7.0% (135)	10.1% (223)
Burns	5.1% (237)	4.6% (2)	10.1% (52)	6.3% (123)	2.7% (60)
Violence	3.0% (143)	0.0% (0)	3.0% (15)	2.1% (41)	4.0% (87)
Drowning	0.3% (15)	0.0% (0)	0.8% (4)	0.3% (6)	0.2% (5)
Medical	67.0% (9,522)	85.5% (254)	83.7% (2,569)	68.9% (4,289)	52.3% (2,410)
Respiratory	29.3% (2,789)	27.2% (69)	35.6% (914)	26.3% (1,124)	28.3% (682)
Fever	26.7% (2,540)	22.0% (56)	20.9% (537)	31.0% (1,328)	25.7% (619)
Gastrointestinal	14.2% (1,355)	8.3% (21)	17.1% (440)	13.8% (593)	12.5% (301)
Sense Organs*	9.8% (936)	13.8% (35)	9.0% (231)	9.5% (410)	10.8% (260)
Constitutional**	8.4% (794)	20.1% (51)	8.3% (213)	8.5% (367)	6.8% (163)
Neurological	7.9% (752)	2.7% (7)	6.7% (171)	8.3% (355)	9.1% (219)
Musculoskeletal	2.4% (225)	3.9% (10)	1.5% (39)	1.7% (72)	4.3% (104)
Cardiovascular	0.8% (77)	1.6% (4)	0.6% (15)	0.4% (16)	1.7% (42)
Other***	0.6% (54)	0.0% (1)	0.4% (9)	.6% (24)	0.8% (20)

* sense organs are defined as the body organs by which humans are able to see, smell, hear, taste and touch or feel. This category includes medical complaints to the eyes, ears, nose, throat and to the skin.

**lethargy, weakness, loss of appetite, fatigue etc.

Data are reported in terms of percentages and absolute frequencies.

Outcomes

Overall, 4,997 (20.1%) of encounters resulted in hospitalizations and 396 (1.6%) in death in the PED. Data on outcomes by age group are described in Table 3. Data on length of stay in the PED for patients who were discharged were available for only 5,639 out of 19,451 visits (29.0%). Of these, 5,505 (97.6%) were discharged within 24 hours of arrival.

Table 3

Outcomes of pediatric emergency department presentations by age group.

OUTCOMES	Total	< 28 d	29 d- 1 y	1–5 y	5–15 y
	n = 24,844	n = 677	n = 5,634	n = 10,845	n = 7,688
Mortality	1.6% (396)	16.1% (109)	1.7% (96)	1.1% (118)	1% (73)
≤ 4 h	1.0% (247)	14.3% (97)	1% (57)	0.5% (57)	0.5% (36)
> 4 h	0.6% (149)	1.8% (12)	0.7% (39)	0.6% (61)	0.5% (37)
Hospitalization	20.1% (4,997)	29.7% (201)	25.2% (1,422)	19,7% (2,132)	16.1% (1,242)
Discharge	78.3% (19,451)	54.2% (367)	73.1% (4,116)	79,2% (8,595)	82.9% (6,373)

Data are reported in terms of percentages and absolute frequencies.

Hospitalization analysis

Of the 4,997 hospitalizations, 37 (0.7%) were direct admission from the PED to the PICU. Of these, 17 (45.9%) were for burns. Overall, the length of stay in the PED for visits that resulted in hospitalization was < 24 hours in 37%, between 24 and 48 hours in 56%, between 48 and 72 hours in 5% and > 72 hours in 2% of cases. Data on presenting complaints were available for 4,057 (81.2%) of visits resulting in hospitalization. Of these, 88.1% presented with a medical complaint, and 11.9% with an injury. A significantly higher proportion of medical presentations were hospitalized compared to injuries (37.9% vs. 9.6%, p-value < 0.001). Results of the univariable analysis assessing the association of available clinical variables with hospitalization is reported in S1 Table.

The multivariable analysis carried out on the subgroup of encounters with data on presenting complaint available and the multivariable analysis with missing data imputation showed similar high odds of being hospitalized if presentations to the PED were due to a medical problem rather than an injury (OR 12.19, 95% CI: 10.78 – 13.38 and 11.79, 95% CI: 10.62-13.1, respectively). (Table 4). A younger age, especially younger than 28 days, living in the extra-urban area and being referred to the PED by a health care provider were all predictors of hospitalization, with similar ORs at all the multivariable analyses performed.

Table 4

Determinants of hospitalization.

		Total of Data Available	Hospitalization	Multivariable analysis on valid cases only*	p-value	Multivariable analysis imputing missing data for the variable presenting complaint	p-value	Multivariable analysis excluding the variable presenting complaint	p – value
		N	n: 4997	OR (95% CI)		OR (95% CI)		OR (95% CI)
SEX	Male	14,448	2,893	0.97 (0.89 – 1.05)	0.43	0.99 (0.92 – 1.06)	0.75	0.92 (0.92 – 1.06)	0.02
	Female	10,114	2,054	Reference		Reference		Reference
AGE		24,844	4,997
	0 – 28 days	677	201	1.81 (1.34-2.43)	<0.001	1.68 (1.36-2.07)	<0.001	2.8 (2.28-3.43)	<0.001
	29 d – 1 year	5,634	1,422	1.53 (1.36-1.72)	<0.001	1.19 (1.08-1.31)	<0.001	2.07 (1.89-2.27)	<0.001
	1 – 5 years	10,845	2,132	1.12 (1.01-1.24)	0.03	1.05 (0.96-1.15)	0.27	1.42 (1.31-1.54)	<0.001
	5 – 15 years	7,688	1,242	Reference		Reference		Reference
RESIDENCY	Extra-urban	1,887	818	2.97 (2.55-3.45)	< 0.001	2.61 (2.31-2.95)	< 0.001	2.617 (2.31-2.95)	< 0.001
	Urban	22,948	4,177	Reference		Reference		Reference
MODALITY OF PRESENTATION	Health Care Provider referral	10,360	3,380	5.57 (5.08-6.11)	< 0.001	7.16 (6.63-7.73)	< 0.001	3.65 (3.4-3.91)	< 0.001
	Self-Presentations	14,477	1,617	Reference		Reference		Reference
PRESENTING COMPLAINT	Medical	9,522	3,607	12.19 (10.78-13.78)	< 0.001	11.79 (10.62-13.1)	< 0.001
	Injury	4,682	450	Reference		Reference

* Missing values for the variable presenting complaint were excluded from the analysis

The valid cases (total of data available) have been reported with the number of hospitalized patients. A sensitivity analysis has been performed reporting the results (OR, 95% Confidence Intervals (CI), and p-values) for 1) Multivariable Analysis on valid cases; 2) Multivariable Analysis on imputed data for variable presenting complaint; 3) Multivariable analysis excluding the variable presenting complaint.

S2 Table describes the univariable association between type of presenting complaint and hospitalization. The multivariable analysis (Table 5) showed that children presenting for a medical complaint had higher odds of being hospitalized if they presented for cardiovascular, constitutional, and neurological complaints compared to fever. Within the injury-related presentations children presenting for burns, road accidents, wounds or ingestion/inhalations had higher odds of being hospitalized compared to children presenting for falls.

Table 5

Association between presenting complaints and hospitalization.

		N^	Hospitalization (n)^^	Multivariable Analysis OR (95% CI)	p – value
PRESENTING REASON
	Medical	9,522	3,607
	Fever ‡	2,540	754 (21%)
	Respiratory	2,789	873 (24%)	1.17 (1.03-1.33)	0.019
	Neurological	752	465 (13%)	3.28 (2.72–3.96)	<0.001
	Gastrointestinal	1,355	461 (13%)	1.32 (1.13-1.65)	0.005
	Cardiovascular	77	64 (2%)	7.86 (4.08-15.13)	<0.001
	Musculoskeletal	225	81 (2%)	0.54 (0.39-0.73)	0.001
	Constitutional*	794	638 (18%)	6.98 (5.65 -8.62)	<0.001
	Sense Organs**	936	242 (7%)	0.72 (0.60-0.87)	<0.001
	Others***	54	29 (1%)	1.99 (1.08-3.69)	0.028
	Injury	4,682	450
	Drowning	15	1 (0%)	1.98 (0.25-15.70)	0.520
	Road Accident	457	84 (19%)	4.37 (3.26-5.87)	<0.001
	Fall ‡	2,949	175 (39%)
	Burn	237	109 (24%)	25.01(17.99-34.78)	<0.001
	Wound	401	36 (8%)	2.35 (1.60-3.47)	<0.001
	Violence	143	3 (1%)	0.42 (0.13-1.35)	0.150
	Ingestion/Inhalation	480	42 (9%)	2.24 (1.56-3.23)	<0.001

*lethargy, weakness, loss of appetite, fatigue, etc.

** sense organs are defined as the body organs by which humans can see, smell, hear, taste and touch or feel. This category includes medical complaints to the eyes, ears, nose, throat, and the skin.

***Psychiatric and genitourinary diseases.

‡ Reference Category.

Absolute number and percentages of hospitalized patients have been reported. Multivariable and Logistic Regression Model results (OR, 95% Confidence Intervals (CI), and p-values, adjusted for gender, age, residency and modality of presentation) are represented in the table.

The diagnosis distribution for hospitalized patients is described, stratified by age, in Fig 2.

Fig 2

Hospitalization diagnosis, by age and frequency.

Mortality analysis

Overall, mortality in the PED was 1.6%. The majority of deaths (81%) occurred in patients younger than five years, with the highest mortality found in the 0 – 28 days group (16.1%).

Results of the univariable analysis is reported in S1 Table, while those of the multivariable analyses are reported in Table 6. A younger age, especially younger than 28 days, living in the extra-urban area and being referred to the PED by a health care provider were predictors of mortality in the PED at all the multivariable analyses performed, although some variability in ORs for age, modality of presentation and presenting complaint was noticed between the models.

Table 6

Determinants of mortality in the Pediatric Emergency Department.

		Total of Data available	Mortality in the PED	Multivariable Analysis on valid cases only*	p-value	Multivariable Analysis imputing missing data for variable presenting complaint	p-value	Multivariable analysis excluding variable presenting complaint	p – value
		N	(n 396)	OR (95% CI)		OR (95% CI)		OR (95% CI)
SEX	Male	14,448	182	1.02 (0.63–1.65)	0.94	0.96 (0.78 – 1.18)	0.72	0.99 (0.79 – 1.25)	0.92
	Female	10,114	125	Reference		Reference		Reference
AGE		24,844	396
	0 – 28 days	677	109	5.34 (1.09–15.01)	<0.001	16.12 (11.69-22.22)	<0.001	5.91 (3.06–9.68)	<0.001
	29 d – 1 year	5,634	96	2.33 (1.25–4.34)	0.01	1.34 (0.99-1.83)	0.06	2.25 (1.65 -3.07)	<0.001
	1 – 5 years	10,845	118	0.98 (0.54–1.79)	0.95	0.95 (0.71 – 1.28)	0.74	1.31 (0.97 – 1.75)	0.08
	5 – 15 years	7,688	73	Reference		Reference		Reference
RESIDENCY	Extra-urban	1,887	43	2.40 (1.28-4.50)	0.01	2.14 (1.5–3.05)	<0.001	2.11 (1.49–2.98)	<0.001
	Urban	22,948	353	Reference		Reference		Reference
MODALITY of PRESENTATION	Health Care Provider referral	10,360	165	7.18 (3.97-12.98)	<0.001	2.36 (1.89 –2.95)	<0.001	1.95 (1.53 –2.47)	<0.001
	Self-presentation	14,477	231	Reference		Reference		Reference
PRESENTING COMPLAINT	Medical	9,522	48	2.68 (1.57-4.59)	<0.001	10.36 (6.94-10.45)	<0.001
	Injury	4,682	26	Reference		Reference

*Missing values for the variable presenting complaint were excluded from the analysis.

The valid cases (total of data available) have been reported with the number of deceased patients. A sensitivity analysis has been performed reporting the results (OR, 95% Confidence Intervals (CI), and p-values) for 1) Multivariable Analysis on valid case; 2) Multivariable Analysis on imputed data for variable presenting complaint; 3) Multivariable analysis excluding the variable presenting complaint.

Of the 396 deaths, 247 (62%) occurred within four hours from arrival (early deaths). Due to the low number of patients and the high rate of missing values for the independent variable “presenting complaint” we only performed a multivariable analysis excluding this variable from the model (Table 7).

Table 7

Determinants for early mortality in the PED (≤ 4h vs mortality> 4h).

		Total of Data available N 28,844	Early death (≤ 4 h) n: 247	Multivariable Analysis	p – value
				OR† (95% CI)
SEX	Male Vs	14,448	90	0.91 (0.66 – 1.26)	0.58
	Female	10,114	71
AGE		28,844	247
	0 – 28 days	677	97	2.71 (1.4 – 5.25)	< 0.001
	29 d – 1 year	5,634	57	1.36 (0.88 – 2.09)	0.160
	1 – 5 years	10,845	57	0.93 (0.61 – 1.43)	0.753
	5 – 15 years ‡	7,688	36
RESIDENCY	Urban Vs	22,948	230
	Extra sofala	1,887	17	0.85 (0.50 – 1.44)	0.727
MODALITY of PRESENTATION	Self-presentations Vs	14,477	178
	Health Care Provider referral	10,360	69	0.35 (0.25 – 0.49)	< 0.001

‡Reference category.

The number of valid cases (total of available data) and early death patients has been reported. Logistic Regression Model results (OR, 95% Confidence Intervals (CI), and p-values) are represented in the table.

A younger age, especially younger than 28 days, was a predictor of early mortality in the PED, while visits that were referred to the PED by a health care provider had lower odds of dying in the first four hours from arrival. Of patients who died after 4 hours, 34% died in the first 24 hours, 43% between 24 and 48 hours, 13% between 48 and 72 hours and 10% beyond 72 hours since arrival.

Limitations

The results of our study should be interpreted in light of its limitations, which are mostly related to its retrospective design and the available sources of data collection. First, approximately 40% of data on presenting complaints were missing in the “presentation registry”. This registry was filled by hospital administrative staff who were less aware of the importance of accurate data completion with respect to reporting and analysis purposes. The rate of missing information on presenting complaints was higher for patients who died in the PED (presenting complaint was not reported in 81.3% of deaths), followed by visits resulting in discharge to home (48.2% of missing information) and those resulting in hospitalization (missing information for 18.8%). While this lack of information affects the accuracy of our findings with respect to the description of presenting complaints, this was the best available data we could get access to at the time of the study. In addition, we performed a sensitivity analysis to report how missing values could have affected our results based on different scenarios. Missing information on ED visits records is a common challenge for many LICs and is inherently related to the limitations of the local data registration and repository system. Lack of human and information technology resources represent the major obstacles to the establishment of an accurate and long-lasting data recording and monitoring system in these Countries. Learning from local data is a valuable opportunity for a growing health system to improve the quality of care while optimizing resource use. Efforts towards establishing a robust data management and monitoring system should be made at an institutional and governmental level to best support the development of a pediatric emergency care system in Mozambique.

Second, data on comorbidities were reported inconsistently and only in the hospitalization registry and could not be analyzed. This would be extremely valuable information to include in the multivariable analysis, as underlying conditions such as malnutrition, HIV, and tuberculosis have shown to be associated with the need for hospitalization and mortality [11, 25–27]. Systematic collection of the main comorbidities should be pursued in order to be able to appropriately interpret data to improve care and optimize resource organization and use.

Third, we could not get access to data prior to April 2017 to identify possible biases in our results from random yearly variations. Based on the local clinical registry filing system, completed paper registries were temporarily filed and available for some months and then periodically burnt.

Fourth, this is a single center study and may not represent the rest of Mozambique or other LICs with different disease prevalence. However, there are only four tertiary care level hospitals in Mozambique, two of which (in Beira and Maputo) have a PED, and it is reasonable to believe our data may provide some useful insights into pediatric emergency care at a local and national level to help optimize distribution and use of resources, as well as plan the most appropriate feasible and effective interventions to improve pediatric emergency care within an integrated system of care.

Discussion

In this study we were able to provide the first, albeit limited, data on pediatric emergency visits to a tertiary care PED in the LIC of Mozambique. Our data represent a first important step to help the establishment of a pediatric emergency care monitoring system in Mozambique in order to guide the formulation of appropriate strategies to improve the management of the acutely and critically ill children and develop a structured and sustainable emergency care system. The first important finding of our study is that the current PED data registration system has many flaws and challenges, which hamper the provision of accurate and valid data to learn from and guide the development of pediatric emergency care tailored to local needs. Optimization of data registration is an important area on which to focus resources in order to obtain more accurate data for this purpose.

Our study found an overall high mortality rate in the PED setting of 1.6%. Although higher than reported in high-income countries (1.5/100 000 visits) [28], our result is in line or even lower compared to other sub-Saharan countries [29]. The majority of deaths occurred in patients younger than five years (81%), with the highest mortality found in the neonatal group (16%). Based on a recent systematic review [30], about a third of all neonatal deaths tend to occur on the day of birth, and approximately 75% die in the first week of life. These findings suggest that focusing on perinatal care, maternal education and improved access to healthcare is essential for saving newborn lives.

Our analysis also showed that living in extra-urban areas is a predictor of death. This may reflect the many challenges in transportation to the hospital that these children have to face even when severely ill. Fernandes and colleagues [18], highlighted the importance of health service availability, showing an overall improvement in child survival in Mozambique, associated with increased health workforce density, institutional birth coverage, and government health financing, despite the substantial disparity between provinces.

Physicians working at HCB's PED noticed that death occurred more frequently in children presenting late in their course of illness. As evidenced by Punchak and coworkers [31] there are many potential contributing factors to late presentations, including delays related to triage organization, bad tiered health care system, late care-seeking by families due to a lack of health education, and socioeconomic factors related to the geographic distribution of health centers and inadequate transportation infrastructures. Improving access to care, and further promoting health education, would likely result in an earlier presentation to the PED, eventually translating in better disease recognition and treatment.

The majority of deaths in our study occurred in the first four hours from arrival (62%). This group mostly included neonates and children living in the urban area who were brought in by parents. Our study also found that children who had already been evaluated by a health care provider (in a health care center or a peripheral hospital) had lower odds of dying early (within 4 hours) in the PED. In fact, these children died more often after four hours from arrival to the PED. Our findings may reflect the ability of health care centers or peripheral hospitals to stabilize severely ill patients for transport. However, late presentation to a health care facility and inability to provide effective care during transport may contribute to the unfavorable outcome of these children. Although WHO and UNICEF [6, 7] report that infectious diseases remain a leading cause of death for children under the age of 5 in sub-Saharan Africa, accurate local data would be paramount to better understand which interventions could be most effective to further reduce child mortality both in the PED and at a community level, within an integrated system of care.

We also found a high hospitalization rate of 20.1%, with neonates and infants showing higher odds of being hospitalized compared to older children, as previously described for other sub-Saharan regions [29]. Also, children who lived in the extra-urban area were at higher risk of being hospitalized compared to children who lived in the urban area, especially if already evaluated by health care center physicians or in a peripheral hospital. These results reflect a good organization of the health care referral system in treating pediatric critically ill presentations [11]. However, peripheral health centers have limited resources or lack of training for the management of the most severe presentations, and transfer conditions to the referral center remain challenging.

Our study showed that children presenting for a medical reason were more likely hospitalized, especially if admitted with cardiovascular, constitutional, and neurological diseases. Among injury presentations, burns, road accidents, wounds, and ingestion/inhalations were significantly associated with hospitalization, compared to fall. Based on our field experience, only severe burns with an intrinsic high risk of complications were referred to HBC, which justified the need for hospitalization [31-33]. The causes and risk factors behind the substantial number of severe burns should be further explored in order to implement effective preventive measures.

The analysis of hospitalization diagnosis by age group showed that sepsis, followed by lower respiratory infections (i.e pneumonia and bronchiolitis) and skin infections (i.e cellulitis, impetigo, piodermitis, etc.) were the most common in the neonatal age group. This highlights the need for interventions to improve perinatal care and parents' education. At HBC neonates and their mothers are usually discharged on the first day after delivery, without provision of any further assistance from health personnel in the out of hospital setting. Indeed, the implementation of specific protocols to educate mothers before discharge, such as providing informative graphic pamphlets and clear verbal instructions on when is necessary to present to a health care facility, could be an effective way to prevent clinical deterioration and delayed care, as demonstrated by Berhea and colleagues in Ethiopia [34]. We also found that severe malnutrition became a more frequent cause of hospitalization with increasing age, being the third cause of admission in preschoolers. This is an important finding, considering that it is estimated that malnutrition is the underlying cause of 45% of global deaths in children below 5 years of age [35, 36]. In the school-aged group, malaria was the leading cause of hospitalization followed by osteoarticular injuries, mainly due to falls and road accidents. This is consistent with previous studies in LMICs, reporting an increased frequency of injury in this age group [37-39]. Lastly, we also found an increased prevalence of haemato-oncological disorders in this age group compared to the others, mainly due to severe anemia.

Although data on presenting complaints were limited by the number of missing data, we found a similar distribution compared with previously published data from other low and middle-income countries (LMICs) [12, 40]. As expected, infection-related presenting complaints were the most frequent, with respiratory, fever, and gastrointestinal conditions being the most common [41, 42]. After analyzing presentations complaints by age groups, it becomes evident how injury-related presentations increased with age, reaching almost half of the visits (48%) in school-age children. This is consistent with a previous Mozambican report [43]. However, within the neonatal and infant age groups, we found a high rate of injury-related presentations (16% and 14%, respectively), in particular, due to falls. These rates are in contrast with reports from high-income countries and other LMICs [39, 40]. Our rates are concerning, considering the fragility of children in this young age group and the overall high mortality and morbidity associated with injury [39]. Although further investigations are necessary to identify the reasons behind the abnormally high prevalence of injury in non-ambulant children found in this study, our results suggest that there is an urgent need to develop injury-prevention programs and campaigns to reduce injury rates in young Mozambican children, supporting the needs evidenced by De Sousa Petersburgo and coworkers [43].

Other useful interventions to improve the quality of care provided to the acutely and critically ill children include the implementation of a triage system, training of healthcare personnel and the establishment of an efficient emergency call and transport system. Several studies have shown how the introduction of Emergency Triage and Treatment (ETAT) guidelines could be an easy and cost-effective strategy to improve emergency and overall care [44-46]. In Malawi, the implementation of ETAT halved the pediatric inpatient mortality rate [47]. Training of health care personnel to the early identification of critical diseases is another important step in the improvement of PEC, as shown by several studies [10, 18, 48, 49]. Training on the early recognition and management of conditions that most often result in death in the local setting should be prioritized. Although challenging and expensive, the establishment of an efficient emergency call and transport by ambulance service would be critical to ensure that severely ill children from both the extra-urban and urban areas have access to the PED in a timely manner [31].

Conclusions

Optimization of data registration is an important area on which to focus resources in order to obtain more accurate data to learn from and guide the development of pediatric emergency care in Mozambique, tailored to local needs. Our data provide insight into opportunities to reduce the high mortality in the pediatric emergency department and the high hospitalization rate, identifying the neonatal age group as the most vulnerable. Interventions such as maternal education, injury prevention, implementation of a triage system, training of health care personnel, and implementation of an emergency care transport system would be critical to improve the outcomes of acutely and critically ill children in Mozambique.

Determinants of hospitalization, mortality in the PED, and early death.

The valid cases (total of data available) have been reported with the number of hospitalized patients. Univariable analysis results are reported in the table.

(DOCX)

Click here for additional data file.

Absolute numbers and percentages of hospitalized patients have been reported. Univariable Analysis (OR, 95% Confidence Intervals (CI), and p-values, adjusted for gender, age, residency, and modality of presentation) are represented in the table.

(DOCX)

Click here for additional data file.

15 Jun 2020

PONE-D-20-12410

Pediatric emergency care in low income countries: characteristics and outcomes of presentations to a tertiary-care emergency department in Mozambique

PLOS ONE

Dear Dr. Brugnolaro,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Jul 30 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Itamar Ashkenazi

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

3. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Partly

Reviewer #4: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: No

Reviewer #4: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: Yes

Reviewer #4: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: - Materials and methods: The authors should specify if in registry of deaths are missing some data from the original 24844 and how many data are missed in the registry of hospitalization (we can assume that all the missed are discharged?)

- Statistical Analysis:

o The approval of the Ethics Committee is already written in Materials and Methods, where is more appropriated

o something in the data is incorrect: 396/24844 is 1,6% of mortality (1% is 247 <4h and 0,6% is 149 >4h) and hospitalization is 20,1% (4997/24844) and discharge 78,3% (19451/24844)

o It could be questionable that mortality rate is calculated in all population of 24844, because is not clear what’s happened to the children discharged (how long is the trip to home, what kind of assistance they have at home like general practitioner, nurse or other sanitary help, maybe some other information about the medical territorial system should be described in the introduction). Also, the description of the neonatology should be shortly written in the introduction (intensive unit is only pediatric? The hospital has a neonatal intensive care unit? How many beds if so?)

- Discussion:

o more emphasis about the missing records is necessary (only 14204 recorded at presentation means that more than 40% of the children arrived in ER could not have been properly evaluated and then deceased at home? Or only the children with more severe clinical presentation have been recorded?)

o Limitations in my opinion should be written before the discussion (mortality rate could be different for missing information i.e.)

Reviewer #2: The authors report a retrospective review of pediatric ED visits of >24,000 visits in Mozambique. They found that one third of ED visits were for injuries, ~20% of children were admitted from the ED, and 2% died in the ED. The strengths of this study are the large sample size, the clear writing, and the provision of the fundamental knowledge for future interventions for targeted improvements in care in Mozambique.

Despite the article’s strengths, there are several weaknesses that this reviewer thinks should be addressed. Some of these weaknesses include clearer interpretation of the data in the Tables, more of a discussion of how the large amount of missing data should affect their interpretation of the data, and the lack of data on interventions (or lack thereof) given in the ED. I have provided specific suggestions on ways to improve this important article below.

Title:

-I suggest changing “…in low income countries” to “in a low income country” as the study comes from a single low-income country.

Introduction:

-Consider avoiding discussing “critical care services” as they differ from emergency care.

-Final paragraph: sub-Saharan not “sub-Saharian”

-Minor point, but in the final paragraph the mortality rate (73 per 1,000) cited is a rate, not a probability.

-I suggest moving the final paragraph of the Introduction to the Methods as this is a description of the study setting and does little to emphasize the importance of the study. That being said, of course, the final two sentences in the final paragraph need to remain in the Introduction.

Methods:

-Did the authors gather data on average length of stay? I ask because some EDs in sub-Saharan Africa serve as hybrid EDs (in the Western sense) and ICUs as children who require critical care interventions stay in the ED because they lack stability for transfer to the actual ICU in the hospital.

-As the authors aim to understand emergency care and lay the groundwork for ways to improve care, was there any data on interventions, treatments, or resuscitations in the ED?

-A mention of the hospital mortality rate among children would be helpful to contextualize these findings that 2% of ED presentations resulted in death.

-Did the authors gather data on malnutrition, HIV, TB? These are important and well-documented associated factors with mortality among children.

Results:

-How does the fact that >1/3rd of all presenting complaints were missing influence the interpretation of the data?

-Table 2: I suggest listing the variables in descending order of frequency.

-Table 2: musculoskeletal is misspelled.

-Table 2: What are “sense organs” as a presenting complaint? I suggest including a footnote describing what those are and for all other disease categories in the Tables.

-I suggest including the exact number and exact percentage of patients who were admitted. “One-fifth of the presentations…” is not clear enough, especially since the following sentence cites a proportion of those who were admitted. It is helpful to have that actual number easily accessible.

-Table 3: It is unclear what the referent is for the multivariate analysis. For example, what exactly do the numbers on the same row as AGE mean? There should be a referent for each set of variables in the univariate and multivariate analysis.

-Table 3: aren’t the ORs under multivariate analysis adjusted ORs?

-In general, in the Results, the authors should be careful to clearly state the comparator for any comparison. For instance, “Children presenting for any kind of injury had a higher chance to be discharged except if they presented for burns.” Higher chance than what? Also, this should state higher odds.

-While almost 2/3rds of deaths occurred in the first 4 hours, what was the median time to death? This will help get at how long patients tend to stay in this ED.

-I suggest double checking the wording or stats cited in the sentence that contains, “significantly higher for children in the 0-28 days and in the 29 days - 1 year groups (OR: 3.58, 95% CI: 2.30 - 5.59, p-value < 0.001 and OR: 1.18, 95% CI 0.88 – 1.58, p-value = 0.283, respectively)” as the second OR is not significant.

-Same comment on Tables 6 and 7 regarding the need for a referent when reporting ORs.

-Table 7, no need to repeat column of Mortality in the PED as this was shown in a previous table.

-Why do the authors not take into account underlying medical problems in the multivariate analysis? Surely HIV, malnutrition, TB, cancer, etc. contribute to hospital admissions and death but these are not taken into account in the current analysis.

Discussion:

-Please define LMIC the first time it appears in the paper.

-Only 19% of reason for presentation among children who died? This should be discussed in the Limitations specifically as much of the Introduction and Discussion circles around mortality.

-I suggest adding to the Limitations that this was a single hospital and may not represent the rest of Mozambique or other LICs with different disease prevalences.

Reviewer #3: In the context of improving pediatric emergency care services among low-income countries, authors aimed at studying profiles and outcomes in a tertiary care PED in Mozambique.

As highlighted by the authors, even if results from the study could not be generalized, the research question is of interest because these results from the field of a PED would help in leading strategies to improve their PED.

Introduction: Well written, correctly documented and referenced. The research question is well exposed and argued. It could be added some additional description about Mozambique and his healthcare facilities in general (if possible) in order to better understand how is organized the patient pathway. Is there any difference between Beira and Sofala about healthcare facilities?

Materials and Methods: it is a retrospective study. Because there were 3 registries, data were abstracted by 2 investigators with a high risk of missing data. Because there are many subset of data with which statistical analysis were performed, it would be helpful if authors could add a flow chart on the way they obtained the different subsets.

Statistical analysis: results were exposed from many different subsets of data (Table 1, n = 24,844 / Table 2, n =14,204 / Table 3, n = 14,448 / Table 4, n = 14,204 / Table 5, n = 24844, Table 6, n = 14,448 / Table 7, n = 28,844).

Even if it is understandable to perform analysis using the largest size of available data according to type of analysis, wouldn’t it more readable and more understandable if results were coming from a unique data set that contain all the data?

Could the authors indicate how missing data were handle for statistical analysis?

Results section: According logistic regression analysis: odds ratio are exposed according to a reference category.

In many tables, could the authors explain why reference categories were not specify? As we could understand, authors exposed results using dichotomized variables (Yes versus No). For example, in Table 3, for the variable “Residency”, wouldn’t it be more readable if the item “urban” was set as a reference category? If yes, Authors should consider this question for all dichotomized variables (in table 3, 4, 5, 6, 7).

In table 4, all the presenting complaints are at higher risk to be hospitalized that is unusual. Did the authors test interaction between them? Authors should not be limited in presenting univariate analysis. In order to better understand, I would suggest the authors to complete those results by performing multivariate analysis including demographic variables.

In table 6 &7, it may be frustrating not to go further in the analysis. Replacing “Presenting complaints” (“Medical” versus “Injury”) by “Presenting Reason” (“Fever”, “Respiratory”, “neurological” etc…) would be helpful in order to point out a more detailed profiles of those patients at higher risk of mortality.

Discussion section: If possible, to understand the trend of the results, could authors explain why their results were compared with data from Nepal, Pakistan, Malawi, Ethiopia, etc…? Are the patient profiles comparable? What about comparing their healthcare facilities?

Reviewer #4: PONE-D-20-12410

Review

General thoughts:

- This is a retrospective chart review with a primary objective of determining systems-level interventions and design features for an effective pediatric emergency medicine system in a low-income (Mozambique) country. The manuscript is well written in general, with more description required primarily in the methods section.

- With such a broad perspective, it puzzles me as to why the authors only went back one year (April 2017-2018). Internal validity of data is difficult to determine as there is a possibility of bias from random yearly variations (as our current COVID-19 era has surely shown us), that cannot be excluded without more than one year of data for comparison.

- Would interesting to know of acuity of presentation, and proportion of children requiring resuscitation, given the (fortunately) lower prevalence of mortality in the PED.

- The authors have taken a systems framework to the implications of their results. It would seem appropriate to further analyze and present data from a QI/systems lens to truly capture the complexity of their data and how they interact. (i.e. SEIPS 2.0/3.0 system). This “next level” analysis would provide much greater context for their prevalence data.

- The novelty of this study lies in their data source (low income African PED). As written this manuscript is more akin to an annual report. What concrete suggestions do these data make leadership focus upon when presented with the data? What is the generalizability of these data to countries in similar situations?

Introduction:

- Would state Millennium Development Goals/Sustainable Development Goals are sourced from the UN, to give context and international standards to these statements.

Methods:

- A key threat to the external validity of your study is the quality of your data registries? Please provide more data on who is responsible for keeping registry data. Are ICD 9/10 codes used? How is data stored/accessed? Is there a possibility that data could be altered after the fact? Who cureates/owns the databases (ie government, hospital, etc).

- How were charts identified for review?

- Was there a research ethics board that approved this study? Please state in the manuscript earlier, instead of in the last line.

- Please provide more of a description of your data abstraction process. Were there standard forms? What data was abstracted? Who were the abstractors? Where they trained? Was their intermittent overlap of data abstraction to ensure interrelater reliability?

Results

- There is a substantial proportion (43%) where presenting complaint data was not available. Please provide rationale of missing data (this is an interesting finding in and of itself and should be embraced as a finding of the study).

- What is the rationale for the presenting reason categories? Are they ICD9/10 based?

- To help with the generalizability of the data, it would be useful to get information on how the PED is structured, size of division, training background.

- Table 4 – Injury – column 2 – would add % data in brackets.

Discussion

- Interesting points are contained. Somewhat disorganized and would take a systems analysis perspective lens. Consider SEIPS 2.0 or 3.0 models

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: Yes: Antoine TRAN

Reviewer #4: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

18 Aug 2020

Dear Editor in Chief, Prof. Itamar Ashkenazi,

Thank you for your thorough and timely review of our manuscript "Pediatric emergency care in a low-income country: characteristics and outcomes of presentations to a tertiary-care emergency department in Mozambique".

We have responded point by point to all the comments made by the reviewers in our rebuttal letter below. As per instructions we also uploaded a marked-up copy of our manuscript that highlights changes made to the original version as a separate file labeled 'Revised Manuscript with Track Changes'. We uploaded the unmarked version of our revised paper without tracked changes as a separate file labelled 'Manuscript'. However, following the reviewers’ suggestions, we have almost completely re-written the manuscript based on the input from co-authors, making it very challenging to submit a precise marked-up copy of our manuscript that highlights every changes made to the original version. For this reason, we ask the editors and reviewers to consider primarily the clean version of the revised manuscript. Moreover, we endeavored to be as specific as possible in our response to the reviewers below.

We are looking forward to hearing from you.

Sincerely yours.

Best regards,

Dr. Valentina Brugnolaro

(corresponding author)

Reviewer #1:

- Materials and methods: The authors should specify if in registry of deaths are missing some data from the original 24844 and how many data are missed in the registry of hospitalization (we can assume that all the missed are discharged?)

Authors’ reply: We thank the Reviewer for the opportunity to clarify this point. The death and hospitalization registries of the pediatric emergency department (PED) only recorded data of patients who died in the PED and were hospitalized, respectively. This means that all the visits that were included in the presentations registry but were not recorded either in the hospitalization registry, or in the death registry were discharged to home. We hope we have been able to better clarify this point in the revised version of the Material and Methods section (subsection “Sources of data and data collection procedures”)

- Statistical Analysis:

o the approval of the Ethics Committee is already written in Materials and Methods, where is more appropriated

Authors’ reply: The Reviewer is right; this information was reported twice in the original manuscript. Following the Reviewer’s suggestion, we have removed the sentence on Ethics Committee approval from the Statistical Analysis section.

- something in the data is incorrect: 396/24844 is 1,6% of mortality (1% is 247 <4h and 0,6% is 149 >4h) and hospitalization is 20,1% (4997/24844) and discharge 78,3% (19451/24844)

Authors’ reply: We thank the Reviewer for noticing this inconsistency in rounding decimal digits. We made the necessary adjustments, and further checked all the results throughout the revised manuscript.

- It could be questionable that mortality rate is calculated in all population of 24844, because is not clear what's happened to the children discharged (how long is the trip to home, what kind of assistance they have at home like general practitioner, nurse or other sanitary help, maybe some other information about the medical territorial system should be described in the Introduction). Also, the description of the neonatology should be shortly written in the Introduction (intensive unit is only pediatric? The hospital has a neonatal intensive care unit? How many beds if so?)

Authors’ reply: We appreciate the Reviewer's comments on mortality rate. However, we aimed to describe the mortality in the PED, rather than the overall mortality of children accessing the PED after they have been discharged. This would be very valuable information to have to identify further areas of improvement in pediatric emergency care. Unfortunately, the current local system does not allow to retrieve data on mortality after discharge. Based on the Reviewer’s suggestion we have added more detailed information on the territorial healthcare system and on the different types of care facilities for children, such as the neonatal and intensive care settings (Introduction Page 3,, Paragraph 4 and Material and Methods Page 4, Paragraph Study Setting: Healthcare in Beira and the Sofala province)

-Discussion:

o more emphasis about the missing records is necessary (only 14204 recorded at presentation means that more than 40% of the children arrived in ER could not have been properly evaluated and then deceased at home? Or only the children with more severe clinical presentation have been recorded?)

Authors’ reply: We thank the Reviewer for the opportunity to better clarify this important point. Although death and hospitalization registries were filled by ED medical personnel, allowing for accurate reporting of data, the presentation registry was filled by administrative staff who were less aware of the importance of complete information recording. This led to demographic data to be reported in the registry for all visits, while presenting complaints were reported for only 14,204 visits. The rate of missing information on presenting complaints was higher for patients who died in the PED (presenting complaint was not reported in 81.3% of deaths), followed by visits resulting in discharge to home (48.2% of missing information) and those resulting in hospitalization (missing information for 18.8%). As reported in the response to a previous Reviewer’s comment above, we could not retrieve information on death after discharge. We agree this would be valuable information to have for a broader understanding of gaps in pediatric emergency care management. We have revised the limitations section to give more emphasis to this aspect (Limitation section Page 18, Paragraph 1, Line 281)

o Limitations, in my opinion, should be written before the discussion (mortality rate could be different for missing information i.e.)

Authors’ reply: We do understand the Reviewer’s point and following their suggestion we have moved the limitation section before the discussion in order to help the reader correctly interpret our findings before the start of the discussion. The missing information on presenting complaints was inherently related to the local data registration and collection system. As reported above, data on mortality are accurate and there is no missing information with respect to PED visits that resulted in death. The limitations section has been revised to better clarify these points.

Reviewer #2:

The authors report a retrospective review of pediatric ED visits of >24,000 visits in Mozambique. They found that one third of ED visits were for injuries, ~20% of children were admitted from the ED, and 2% died in the ED. The strengths of this study are the large sample size, the clear writing, and the provision of the fundamental knowledge for future interventions for targeted improvements in care in Mozambique.

Despite the article's strengths, there are several weaknesses that this reviewer thinks should be addressed. Some of these weaknesses include clearer interpretation of the data in the Tables, more of a discussion of how the large amount of missing data should affect their interpretation of the data, and the lack of data on interventions (or lack thereof) given in the ED. I have provided specific suggestions on ways to improve this important article below.

Title:

-I suggest changing "…in low income countries" to "in a low income country" as the study comes from a single low-income country.

Authors’ reply: We agree with the comment made by the Reviewer and we have changed the title accordingly.

Introduction:

-Consider avoiding discussing "critical care services" as they differ from emergency care.

Authors’ reply: Based on the Reviewer’s comment we replaced “critical care services” with “emergency care services”

-Final paragraph: sub-Saharan not "sub-Saharian"

Authors’ reply: Thank you for noticing this typo. We have made the suggested change

-Minor point, but in the final paragraph the mortality rate (73 per 1,000) cited is a rate, not a probability.

Authors’ reply: The Reviewer is right, we have amended the manuscript accordingly

-I suggest moving the final paragraph of the Introduction to the Methods as this is a description of the study setting and does little to emphasize the importance of the study. That being said, of course, the final two sentences in the final paragraph need to remain in the Introduction.

Authors’ reply: We thank the Reviewer for this suggestion. We have revised the Introduction according to the Reviewer's advice (paragraph moved to Material and Methods, Study setting, Page 4, Paragraph Study Setting, Line 79).

Methods:

-Did the authors gather data on average length of stay? I ask because some EDs in sub-Saharan Africa serve as hybrid EDs (in the Western sense) and ICUs as children who require critical care interventions stay in the ED because they lack stability for transfer to the actual ICU in the hospital.

Authors’ reply: We thank the Reviewer for their suggestion to expand on our analysis including the average length of stay. We have now included in the results section the data on the length of stay in the PED, which we could abstract from the hospitalization registry only. (Page 10, Paragraph Outcomes, Line 189). The PED at the HCB includes a room dedicated to Pediatric Intensive Care, where critical children are admitted once stabilized. This information is now included in the methods section of the revised manuscript. (Page 4, Paragraph Study Setting, Line 90). However, patients who were admitted in the Pediatric Intensive Care area were classified as hospitalized to PICU in the hospitalization registry.

-As the authors aim to understand emergency care and lay the groundwork for ways to improve care, was there any data on interventions, treatments, or resuscitations in the ED?

Authors’ reply: Unfortunately, we were unable to collect data on interventions, treatments, or resuscitations in the PED because this information was reported on paper forms, which were not filed systematically and were therefore unavailable to either clinical or research staff. We have added a paragraph in the Methods section to better clarify this point. (Page 4, Paragraph Sources of data and data collection procedures, Line 102).

-A mention of the hospital mortality rate among children would be helpful to contextualize these findings that 2% of ED presentations resulted in death.

Authors’ reply: We thank the Reviewer for this suggestion. Based on the most recent available data from 2017, in-hospital overall pediatric mortality was 13%, while in the NICU mortality was 33%. We have now added this information in the Materials and Methods section, under the “Study Setting” paragraph.

-Did the authors gather data on malnutrition, HIV, TB? These are important and well-documented associated factors with mortality among children.

Authors’ reply: We agree with the Reviewer that it will be very important to have data on malnutrition, HIV and TB. Unfortunately, these data were reported inconsistently in the hospitalization registry only and cannot be accurately summarized to reflect the actual frequency of these underlying conditions in patients presenting to the PED. We have added this as an additional limitation to the study in the Limitations section.

Results:

-How does the fact that >1/3rd of all presenting complaints were missing influence the interpretation of the data?

Authors’ reply: We thank the Reviewer for this comment. We have now better specified this point in the limitation section, which, based on the suggestion of Reviewer 1, has now been moved before the discussion in order to help the reader correctly interpret our findings before the start of the discussion. In addition, we re-run the multivariable analysis and performed a sensitivity analysis to better address how missing information affects our results.

-Table 2: I suggest listing the variables in descending order of frequency.

-Table 2: musculoskeletal is misspelled.

-Table 2: What are "sense organs" as a presenting complaint? I suggest including a footnote describing what those are and for all other disease categories in the Tables.

Authors’ reply: We thank the Reviewer for their suggestions to improve the quality of Table 2. We have made the suggested changes. We added the following footnote to the Table “Sense organs are defined as the body organs by which humans are able to see, smell, hear, taste and touch or feel. This category includes medical complaints to the eyes, ears, nose, throat and to the skin.”

-I suggest including the exact number and exact percentage of patients who were admitted. "One-fifth of the presentations…" is not clear enough, especially since the following sentence cites a proportion of those who were admitted. It is helpful to have that actual number easily accessible.

Authors’ reply: We have amended the Results section according to the Reviewer’s suggestion (Page 10, Paragraph Outcomes, Line 188)

-Table 3: It is unclear what the referent is for the multivariate analysis. For example, what exactly do the numbers on the same row as AGE mean? There should be a referent for each set of variables in the univariate and multivariate analysis.

Authors’ reply: Based on the Reviewer’s comment we have involved experienced professional statisticians (Dr. Danila Azzolina and Prof. Dario Gregori) to re-run the multivariate analysis. All the Results section has been revised based on the new analysis. Given the substantial contribution of Dr. Danila Azzolina and Prof Dario Gregori in the analysis and revisions of the manuscript they have been included as authors in the manuscript.

-Table 3: aren't the ORs under multivariate analysis adjusted ORs?

Authors’ reply: Please see our response to the comment above.

-In general, in the Results, the authors should be careful to clearly state the comparator for any comparison. For instance, "Children presenting for any kind of injury had a higher chance to be discharged except if they presented for burns." Higher chance than what? Also, this should state higher odds.

Authors’ reply: Based on the Reviewer’s comment we have revised the whole Results section to add clarity to the reporting of our findings, making sure the comparator is specified for any comparison.

-While almost 2/3rds of deaths occurred in the first 4 hours, what was the median time to death? This will help get at how long patients tend to stay in this ED.

Authors’ reply: Unfortunately, the exact time of death was not reported in the death registry, as data were collected according to the predefined categories of early death < 4 hours and later death in the PED >4 hours. This was an intrinsic limitation related to the local data collection system.

-I suggest double checking the wording or stats cited in the sentence that contains, "significantly higher for children in the 0-28 days and in the 29 days - 1 year groups (OR: 3.58, 95% CI: 2.30 - 5.59, p-value < 0.001 and OR: 1.18, 95% CI 0.88 – 1.58, p-value = 0.283, respectively)" as the second OR is not significant.

Authors’ reply: We thank the Reviewer for noticing this imprecision in reporting of our results. As reported in the response to a previous Reviewer’s comment above we have thoroughly revised the whole Results section to ensure more clarity and consistency in data reporting.

-Same comment on Tables 6 and 7 regarding the need for a referent when reporting ORs.

Authors’ reply: As reported above all the Results section, including all the analyses and Tables have been thoroughly revised. All the multivariate analyses now include a reference category for each of the independent predictor variable.

-Table 7, no need to repeat column of mortality in the PED as this was shown in a previous table.

Authors’ reply: Table 7 has been revised taking into account the Reviewer’s suggestion.

-Why do the authors not take into account underlying medical problems in the multivariate analysis? Surely HIV, malnutrition, TB, cancer, etc. contribute to hospital admissions and death but these are not taken into account in the current analysis.

Authors’ reply: As reported in the response to a previous Reviewer’s comment above, unfortunately data on underlying medical problems were reported only in the hospitalization registry and inconsistently. This information cannot be accurately summarized to reflect the actual frequency of these underlying conditions in patients presenting to the PED and cannot be used in the multivariate analysis. We have added this as an additional limitation to the study in the Limitations section.

Discussion:

-Please define LMIC the first time it appears in the paper.

Authors’ reply: We thank the Reviewer for noticing the acronym was not defined in the text. We have amended the manuscript accordingly (discussion section).

-Only 19% of reason for presentation among children who died? This should be discussed in the Limitations specifically as much of the Introduction and Discussion circles around mortality.

Authors’ reply: We agree with the Reviewer and we have expanded the limitation section to include this point.

-I suggest adding to the Limitations that this was a single hospital and may not represent the rest of Mozambique or other LICs with different disease prevalences.

Authors’ reply: We agree with the Reviewer and we have expanded the limitation section to include this point.

Reviewer #3: in the context of improving pediatric emergency care services among low-income countries, authors aimed at studying profiles and outcomes in a tertiary care PED in Mozambique.

As highlighted by the authors, even if results from the study could not be generalized, the research question is of interest because these results from the field of a PED would help in leading strategies to improve their PED.

Introduction: Well written, correctly documented and referenced. The research question is well exposed and argued. It could be added some additional description about Mozambique and his healthcare facilities in general (if possible) in order to better understand how is organized the patient pathway. Is there any difference between Beira and Sofala about healthcare facilities?

Authors’ reply: We thank the reviewer for their positive comment. As suggested by the Reviewer we added a dedicate paragraph in the Material and Methods section “Study Setting: Healthcare in Beira and the Sofala district”

Materials and Methods: it is a retrospective study. Because there were 3 registries, data were abstracted by 2 investigators with a high risk of missing data. Because there are many subset of data with which statistical analysis were performed, it would be helpful if authors could add a flow chart on the way they obtained the different subsets.

Authors’ reply: We thank the reviewer for his advice. We added a figure to better describe the characteristics of data sources. (Figure 1)

Statistical analysis: results were exposed from many different subsets of data (Table 1, n = 24,844 / Table 2, n =14,204 / Table 3, n = 14,448 / Table 4, n = 14,204 / Table 5, n = 24844, Table 6, n = 14,448 / Table 7, n = 28,844). Even if it is understandable to perform analysis using the largest size of available data according to type of analysis, wouldn’t it more readable and more understandable if results were coming from a unique data set that contain all the data?

Authors’ reply: The Reviewer is right in that using different subsets of data may lead to confusion. However, presenting complaints were reported for only 14,204 visits. For analyses where data on presenting complaints were included, the subset of analysis was limited to this number of patients. We made sure there were only two subsets of data used for analysis throughout the manuscript, i.e. the total number of records 24,884 and those where information of presenting complaints was reported, namely 14,204 records.

Could the authors indicate how missing data were handle for statistical analysis?

Authors’ reply: We thank the reviewer for his comment. Based on the Reviewer’s comments we have involved experienced professional statisticians (Dr. Danila Azzolina and Prof Dario Gregori) to re-run all the analyses. All the Results section has been revised based on the new analysis. Given the substantial contribution of Dr. Danila Azzolina and Prof Dario Gregori in the analysis and revisions of the manuscript they have been included as authors in the manuscript. Missing data were not included in the statistical analysis. However, in the analysis on the determinants of hospitalization and death we ran a sensitivity analysis for missing data imputation on the variable “presenting complaint” classified as “medical” versus “injury”, and found similar results to the analysis run on the subset of 14,204 patients for which information on presenting complaint was available. Details of this analysis are now reported in the Statistical analysis section and results are included in the manuscript Tables.

Results section: According logistic regression analysis: odds ratio are exposed according to a reference category.

In many tables, could the authors explain why reference categories were not specify? As we could understand, authors exposed results using dichotomized variables (Yes versus No).

Authors’ reply: Based on the Reviewer’s comment we have re-run the multivariate analyses. All the Results section has been thoroughly revised based on the new analysis. Please see our response to the Reviewer’s comment above.

For example, in Table 3, for the variable "Residency", wouldn't it be more readable if the item "urban" was set as a reference category? If yes, Authors should consider this question for all dichotomized variables (in table 3, 4, 5, 6, 7).

Authors’ reply: We re-run all the multivariate analyses following the Reviewer’s suggestion.

In table 4, all the presenting complaints are at higher risk to be hospitalized that is unusual. Did the authors test interaction between them? Authors should not be limited in presenting univariate analysis. In order to better understand, I would suggest the authors to complete those results by performing multivariate analysis, including demographic variables.

Authors’ reply: Based on the Reviewer’s comments we have re-run the multivariate analyses. We have also analyzed in a separate multivariate analysis the influence of presenting complaints according to the medical and injuries sub-categories, as suggested (Table 5).

In table 6 &7, it may be frustrating not to go further in the analysis. Replacing "Presenting complaints" ("Medical" versus "Injury") by "Presenting Reason" ("Fever", "Respiratory", "neurological" etc…) would be helpful in order to point out a more detailed profiles of those patients at higher risk of mortality.

Authors’ reply: We agree with the Reviewer that it is frustrating not to be able to proceed further in the analysis of the association between presenting complaints and mortality due to the very high number of missing data on presenting complaints for visits that resulted in death (only 19% of these records had data on presenting complaints reported). Unfortunately we could not go further with the analysis in this sense.

Discussion section: If possible, to understand the trend of the results, could authors explain why their results were compared with data from Nepal, Pakistan, Malawi, Ethiopia, etc…? Are the patient profiles comparable? What about comparing their healthcare facilities?

Authors’ reply: We understand the Reviewer’s point. However, as data on pediatric emergencies from low and middle income countries are very limited, we compared our findings with the available published data from other low-middle income countries. While comparison of healthcare facilities between different low-income countries goes beyond the purpose of our study, we focused our comparison on similarities on presenting complaints/outcomes between available pediatric datasets from the emergency department settings from different low-income countries. We have revised the discussion to better clarify the meaning of comparing our findings with data from other countries.

Reviewer #4:

General thoughts:

- This is a retrospective chart review with a primary objective of determining systems-level interventions and design features for an effective pediatric emergency medicine system in a low-income (Mozambique) country. The manuscript is well written in general, with more description required primarily in the methods section.

- With such a broad perspective, it puzzles me as to why the authors only went back one year (April 2017-2018). Internal validity of data is difficult to determine as there is a possibility of bias from random yearly variations (as our current COVID-19 era has surely shown us), that cannot be excluded without more than one year of data for comparison.

Authors’ reply: We agree with the Reviewer’s comment. Unfortunately, based on the local clinical registry filing system, completed paper registries were temporarily filed for some months, and then periodically burnt. For this reason we did not have access to registries including data before April 2017. We have now specified this in the limitation section.

- Would interesting to know of acuity of presentation, and proportion of children requiring resuscitation, given the (fortunately) lower prevalence of mortality in the PED.

Authors’ reply: Unfortunately, we were unable to collect data on interventions, treatments, or resuscitations in the PED because this information was reported on paper forms, which were not filed systematically and were therefore unavailable to either clinical or research staff. We have added a paragraph in the Methods section to better clarify this point. (Page 4 Paragraph Sources of data and data collection procedures, Line 102).

- The authors have taken a systems framework to the implications of their results. It would seem appropriate to further analyze and present data from a QI/systems lens to truly capture the complexity of their data and how they interact. (i.e. SEIPS 2.0/3.0 system). This "next level" analysis would provide much greater context for their prevalence data.

Authors’ reply: We thank the Reviewer for this suggestion. Having complete and accurate data to be presented from a QI/system lens perspective would be ideal. However, our data specifically focus on the PED and, at this initial stage, they are able to offer insight into possible interventions at different stages of care within an integrated system of care that may be effective in reducing mortality in the PED and hospitalization. We hope that a more accurate and structured data monitoring system, linking the different stages and areas of care will be able to provide the appropriate data for a more complex and informative “next level” analysis such as using the SEIPS framework.

- The novelty of this study lies in their data source (low income African PED). As written this manuscript is more akin to an annual report. What concrete suggestions do these data make leadership focus upon when presented with the data? What is the generalizability of these data to countries in similar situations?

Authors’ reply: Based on the Reviewer’s comment we have thoroughly revised the discussion section and partially the conclusions in order to better report concrete suggestions for improvement in care. We have also expanded the limitations section to address the issue of generalizability.

Introduction:

- Would state Millennium Development Goals/Sustainable Development Goals are sourced from the UN, to give context and international standards to these statements.

Authors’ reply: We have amended the Introduction according to the Reviewer’s suggestion.

Methods:

- A key threat to the external validity of your study is the quality of your data registries? Please provide more data on who is responsible for keeping registry data. Are ICD 9/10 codes used? How is data stored/accessed? Is there a possibility that data could be altered after the fact? Who cureates/owns the databases (ie government, hospital, etc).

Authors’ reply: We thank the Reviewer for the opportunity to better clarify this point. Death and hospitalization registries were filled by ED medical personnel, while the presentation registry was filled by hospital administrative personnel. Data were collected according to local pre-defined categories for presenting complaints and hospitalization diagnosis. ICD 9/10 codes were not used. Based on the local clinical registry filing system, completed paper registries were temporarily filed in the PED for some months, and then periodically burnt. The registries were owned by the hospital. All this information has been reported in the revised version of the manuscript under the Material and Methods and Limitations sections.

- How were charts identified for review?

Authors’ reply: Data were not abstracted by medical patients’ charts but only from the registries. Clinical notes and patients’ charts were not systematically filed for PED visits and therefore were not available for review. We hope the additional information on data collection we included in the revised version of the paper helps better clarify this point.

- Was there a research ethics board that approved this study? Please state in the manuscript earlier, instead of in the last line.

Authors’ reply: The study was indeed approved by the hospital ethics board. This piece of information is reported at the start of the Materials and Methods section.

- Please provide more of a description of your data abstraction process. Were there standard forms? What data was abstracted? Who were the abstractors? Where they trained? Was their intermittent overlap of data abstraction to ensure interrater reliability?

Authors’ reply: Data were abstracted by the paper registries in an electronic standard data collection system according to a predefined coding. All data systematically reported in the registries were abstracted. Abstractors were trained locally based on the initial review of 200 registry records each and a two-month data abstraction overlap between the abstractors helped in ensuring consistency in data abstraction and coding. However, formal interrater reliability was not calculated. This information has been better detailed in the Material and Methods section of the revised manuscript.

Results

- There is a substantial proportion (43%) where presenting complaint data was not available. Please provide rationale of missing data (this is an interesting finding in and of itself and should be embraced as a finding of the study).

Authors’ reply: We agree with the Reviewer that this point deserves further attention. Based on the Reviewer’s suggestion we have now discussed this as a finding of the study, early on in the discussion section. We have also expanded the Methods and Limitations sections to provide the rationale of the high proportion of missing data and their implications for the study and for the development of a pediatric emergency care system, which should be tailored to the local setting and needs, as reflected by local data, staff experience and patients’ perspectives.

- What is the rationale for the presenting reason categories? Are they ICD9/10 based?

Authors’ reply: Data in the registries were collected according to local predefined categories. This information has been included in the revised version of the Material and Methods section. Unfortunately the HBC did not have the facilities and resources (i.e., trained staff, IT infrastructures) to collect data according to the ICD 9/10 codes.

- To help with the generalizability of the data, it would be useful to get information on how the PED is structured, size of division, training background.

Authors’ reply: We thank the Reviewer for this suggestion. We have revised the Material and Methods section to include this information.

- Table 4 – Injury – column 2 – would add % data in brackets.

Authors’ reply: We have amended Table 4 according to the Reviewer’s suggestion.

Discussion

- Interesting points are contained. Somewhat disorganized and would take a systems analysis perspective lens. Consider SEIPS 2.0 or 3.0 models

Authors’ reply: We thank the Reviewer for their suggestion. We have revised the discussion to improve its flow. Please see, our response to the comment above about SEIPS 2.0 or 3.0 models.

Submitted filename: Response to Reviewers_.docx

Click here for additional data file.

14 Sep 2020

PONE-D-20-12410R1

Pediatric emergency care in a low income country: characteristics and outcomes of presentations to a tertiary-care emergency department in Mozambique

PLOS ONE

Dear Dr. Brugnolaro,

Thank you for submitting your manuscript to PLOS ONE. Three reviewers submitted their reviews and all three favored publishing your manuscript.  There still some minor comments the reviewers thought are appropriate and if dealt with will improve your manuscript further.  Therefore, we invite you to submit a revised version of the manuscript that addresses these minor points raised during the review process.

Please submit your revised manuscript by Oct 29 2020 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Itamar Ashkenazi

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #2: All comments have been addressed

Reviewer #3: All comments have been addressed

Reviewer #4: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #2: The authors have been very responsive to reviewers’ comments. They have performed additional analyses and have been very thorough in assessing the limitations of their study. I applaud the authors for their responsiveness. I think the article is much strengthened at this point and I feel this is a good beginning to further elucidate what type of patients and outcomes occur in a single PED in Mozambique. My comments are extremely minor at this point.

-If the authors had a “two-month data abstraction overlap between the abstractors helped in ensuring consistency in data abstraction and coding”, should there be a kappa calculated to determine the inter-rater agreement?

-What was the p-value cut off the authors used to determine “significant” in the univariate model to then include in the multivariable model? Could be added around line 142.

-Line 166: “already been” could be reworded. Consider “42% had been assessed by a health care provider…”

-Consider presenting the variables in Tables 4 and 6 on different rows to make it clearer which variable is being reported on some lines. The authors appear to use bold text to do this, but I think splitting variables like “Health care provider referral” and “Self presentation” as separate rows. This will help with the multivariate reporting to as it will emphasize what the referent is.

Reviewer #3: The author has well answered point by point to all the comments. Since, the manuscript has been substantially improved and I don't have any additional comments to the author.

Reviewer #4: General thoughts:

- This is a retrospective chart review with a primary objective of determining systems-level interventions and design features for an effective pediatric emergency medicine system in a low-income (Mozambique) country. The authors responses were thorough and appreciated.

Abstract:

- A major finding in this paper is the lack of a “paper trial” or system for data records (i.e 43% of presenting complaint data is missing, the lack of multiple years of data to review). This finding shouldn’t simply be relegated to a limitation of the study. It should be a finding that is embraced as a major barrier to improving the care of children presenting for PED care in this country. I would love to see this finding reported on the abstract.

Materials and Methods:

- This revised section flows much better. The inclusion of information about the ward set-up and layout is of the PED and the Pediatric Healthcare system is much appreciated, and flows well.

- There seem to be two captions for Figure 1 (place clarify where the location of this figure is meant to be placed. The title for Figure 1 seems like it could be more like a result was reported. I would include the keyword: registries (i.e. something to the effect of Figure 1: Study register characteristics.

- We are unable to see Figure 1, and hence cannot comment on it.

- Please include in the manuscript your rationale for why ICD9/10 codes are not used.

- The new sensitivity analysis is a nice touch.

o Typo in point 3. Should read “outcomes”, not “otcomes”

Results:

- Please check spacing of the first row “mortality” in Table 3. Please make sure all the chart lines up (in admission section)

- Data is presented well an in an organized fashion. Tables are excellent.

- Limitations is presented in the Results section (page 19). Confusing to same it the same name as a section usually found in the discussion. Would rename this subtitle according to the primary finding regarding limitations of data sources (which is not a limitation of the methods and study design per se, but a separate stand alone finding).

Discussion:

- Rewritten discussion and focus on recommendations from data is much better understood. The relation to WHO and UNICEF guidelines is a good direction to take.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #2: No

Reviewer #3: No

Reviewer #4: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

2 Oct 2020

Dear Editor in Chief, Prof Itamar Ashkenazi,

Thank you for your thorough and timely review of our revised manuscript "Pediatric emergency care in a low-income country: characteristics and outcomes of presentations to a tertiary-care emergency department in Mozambique".

We have responded point by point to all the comments made by the reviewers in our rebuttal letter below. As per instructions, we also uploaded a marked-up copy as a separate file labeled 'Revised Manuscript with Track Changes'. We uploaded the clean version of our revised paper without tracked changes as a separate file labelled 'Manuscript'.

We are looking forward to hearing from you.

Sincerely yours.

Best regards,

Dr. Valentina Brugnolaro

(corresponding author)

Reviewers' comments:

Review Comments to the Author

Reviewer #2: The authors have been very responsive to reviewers’ comments. They have performed additional analyses and have been very thorough in assessing the limitations of their study. I applaud the authors for their responsiveness. I think the article is much strengthened at this point and I feel this is a good beginning to further elucidate what type of patients and outcomes occur in a single PED in Mozambique. My comments are extremely minor at this point.

Author’s reply: We thank the reviewer for the supportive comment.

-If the authors had a “two-month data abstraction overlap between the abstractors helped in ensuring consistency in data abstraction and coding”, should there be a kappa calculated to determine the inter-rater agreement?

Author’s reply: We thank the reviewer for this comment. The two-month data abstraction overlap was used for training purposes of the second data abstractor, and for clarification of possible doubts. There was no double entry of all data, so unfortunately, we cannot determine the inter-rater agreement. To clarify this point, we added the following wording to the sentence "...consistency in data abstraction and coding, by training of the second data abstractor. No formal double entry of data by the two abstractors occurred during this time.”

-What was the p-value cut off the authors used to determine “significant” in the univariate model to then include in the multivariable model? Could be added around line 142.

Author’s reply: The p-value cut-off we used to include significant variables at the univariate model into the multivariable model was 0.05. We have added this information to the manuscript, as suggested by the Reviewer.

-Line 166: “already been” could be reworded. Consider “42% had been assessed by a health care provider…”

Author’s reply: We have revised the sentence accordingly.

-Consider presenting the variables in Tables 4 and 6 on different rows to make it clearer which variable is being reported on some lines. The authors appear to use bold text to do this, but I think splitting variables like “Health care provider referral” and “Self-presentation” as separate rows. This will help with the multivariate reporting to as it will emphasize what the referent is.

Author’s reply: We have revised tables 4 and 6 according to the reviewer suggestions.

Reviewer #3: The author has well answered point by point to all the comments. Since, the manuscript has been substantially improved and I don't have any additional comments to the author.

Author’s reply: We thank the reviewer for the supportive comment.

Reviewer #4: General thoughts:

- This is a retrospective chart review with a primary objective of determining systems-level interventions and design features for an effective pediatric emergency medicine system in a low-income (Mozambique) country. The authors responses were thorough and appreciated.

Author’s reply: We thank the reviewer for the supportive comment.

Abstract:

- A major finding in this paper is the lack of a “paper trial” or system for data records (i.e 43% of presenting complaint data is missing, the lack of multiple years of data to review). This finding shouldn’t simply be relegated to a limitation of the study. It should be a finding that is embraced as a major barrier to improving the care of children presenting for PED care in this country. I would love to see this finding reported on the abstract.

Author’s reply: As suggested, we have now included this finding in the Results section of the abstract and commented on it in the conclusion section of the abstract.

Materials and Methods:

- This revised section flows much better. The inclusion of information about the ward set-up and layout is of the PED and the Pediatric Healthcare system is much appreciated and flows well.

Author’s reply: We thank the reviewer for the supportive comment.

- There seem to be two captions for Figure 1 (place clarify where the location of this figure is meant to be placed. The title for Figure 1 seems like it could be more like a result was reported. I would include the keyword: registries (i.e. something to the effect of Figure 1: Study register characteristics.

Author’s reply: The Reviewer is right, in the marked-up copy of the manuscript the caption appears twice. We realized that the paragraph referring to Figure 1 is duplicated in this version of the manuscript, while the clean version is correct and the Figure caption appears only once. As suggested, we have reworded Figure 1 caption as follows: "Characteristics of hospital registries from which study data were collected".

- We are unable to see Figure 1, and hence cannot comment on it.

Author’s reply: We are sorry that the Reviewer was unable to see figure 1. We followed the instructions for the upload of figures and could see it in the pdf of the manuscript for download. We hope there are no technical issues preventing the inclusion on the figure in the revised version of the manuscript.

- Please include in the manuscript your rationale for why ICD9/10 codes are not used.

Author’s reply: As recommended by the Reviewer we have included the following sentence in the revised version of the manuscript, at the end of the section on "Sources of data and data collection procedures": "Unfortunately the HBC did not have the facilities and resources (i.e., trained staff, information technology infrastructure) to code diagnosis according to the ICD 9/10 codes. Data on diagnosis were reported as per local documentation practices".

- The new sensitivity analysis is a nice touch.

Author’s reply: We thank the reviewer for the positive comment.

o Typo in point 3. Should read “outcomes”, not “otcomes”

Author’s reply: Thank you for noticing it, we have corrected the typo accordingly (line 154).

Results:

- Please check spacing of the first row “mortality” in Table 3. Please make sure all the chart lines up (in admission section)

Author’s reply: We thank the reviewer for noticing this, we have corrected the spacing accordingly (Table 3)

- Data is presented well an in an organized fashion. Tables are excellent.

Author’s reply: We thank the reviewer for the positive comment.

- Limitations is presented in the Results section (page 19). Confusing to same it the same name as a section usually found in the discussion. Would rename this subtitle according to the primary finding regarding limitations of data sources (which is not a limitation of the methods and study design per se, but a separate stand-alone finding).

Author’s reply: We thank the Reviewer for the suggestion. The limitation section was adjusted and reported just after the Results and before the Discussion following a previous recommendation from another Reviewer. We are aware some Journals prefer for the Limitations section to appear before the discussion rather than at the end of the discussion. We are happy to follow the Editors' recommendations as to where the Limitations section should be placed in the manuscript.

Discussion:

- Rewritten discussion and focus on recommendations from data is much better understood. The relation to WHO and UNICEF guidelines is a good direction to take.

Author’s reply: We thank the reviewer for the supportive comment.

Submitted filename: Response to reviewers 2 DEF.docx

Click here for additional data file.

12 Oct 2020

Pediatric emergency care in a low income country: characteristics and outcomes of presentations to a tertiary-care emergency department in Mozambique

PONE-D-20-12410R2

Dear Dr. Brugnolaro,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Itamar Ashkenazi

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

22 Oct 2020

PONE-D-20-12410R2

Pediatric emergency care in a low-income country: characteristics and outcomes of presentations to a tertiary-care emergency department in Mozambique

Dear Dr. Brugnolaro:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Itamar Ashkenazi

Academic Editor

PLOS ONE