Outcome in Children Admitted to the First PICU in Malawi*

Abstract

OBJECTIVES: 

Dedicated PICUs are slowly starting to emerge in sub-Saharan Africa. Establishing these units can be challenging as there is little data from this region to inform which populations and approaches should be prioritized. This study describes the characteristics and outcome of patients admitted to the first PICU in Malawi, with the aim to identify factors associated with increased mortality.

DESIGN: 

Review of a prospectively constructed PICU database. Univariate analysis was used to assess associations between demographic, clinical and laboratory factors, and mortality. Univariate associations (p < 0.1) for mortality were entered in two multivariable models.

SETTING: 

A recently opened PICU in a public tertiary government hospital in Blantyre, Malawi.

PATIENTS: 

Children admitted to PICU between August 1, 2017, and July 31, 2019.

INTERVENTIONS: 

None.

MEASUREMENT AND MAIN RESULTS: 

Of 531 included PICU admissions, 149 children died (28.1%). Mortality was higher in neonates (88/167; 52.7%) than older children (61/364; 16.8%; p ≤ 0.001). On univariate analysis, gastroschisis, trachea-esophageal fistula, and sepsis had higher PICU mortality, while Wilms tumor, other neoplasms, vocal cord papilloma, and foreign body aspiration had higher survival rates compared with other conditions. On multivariable analysis, neonatal age (adjusted odds ratio [AOR], 4.0; 95% CI, 2.0–8.3), decreased mental state (AOR, 5.8; 95 CI, 2.4–13.8), post-cardiac arrest (AOR, 2.0; 95% CI, 1.0–8.0), severe hypotension (AOR, 6.3; 95% CI, 2.0–19.1), lactate greater than 5 mmol/L (AOR, 4.2; 95% CI, 1.5–11.2), pH less than 7.2 (AOR, 3.1; 95% CI, 1.2–8.0), and platelets less than 150 × 10⁹/L (AOR, 2.4; 95% CI, 1.1–5.2) were associated with increased mortality.

CONCLUSIONS: 

In the first PICU in Malawi, mortality was relatively high, especially in neonates. Surgical neonates and septic patients were identified as highly vulnerable, which stresses the importance of improvement of PICU care bundles for these groups. Several clinical and laboratory variables were associated with mortality in older children. In neonates, severe hypotension was the only clinical variable associated with increased mortality besides blood gas parameters. This stresses the importance of basic laboratory tests, especially in neonates. These data contribute to evidence-based approaches establishing and improving future PICUs in sub-Saharan Africa.

Over the past decades, childhood mortality has drastically decreased in low- and middle-income countries (LMICs). Nevertheless, global mortality is immense and millions of children die annually from preventable or treatable causes (¹). Recently, there has been growing interest in pediatric and surgical critical care and dedicated pediatric critical units are starting to emerge in these settings.

Establishment of these new PICUs presents numerous challenges including identification of: 1) the variables which can be used to help identify children with poor prognosis; and 2) the children who are most likely to benefit from this new type of care. As the PICU capacity will be small compared with the high disease burden, it is important to use the resources where they will provide most benefit (^2–5). Data on PICU admission outcome from sub-Saharan Africa is scarce, while the conditions and determinants may differ from data from elsewhere due to the different population, diagnosis, comorbidities, and care systems. The opening of a new PICU in 2017 in Malawi provided the opportunity to get more insight on the outcome of the children admitted there. To help answer these important questions for our and other LMIC settings, we studied PICU outcome and explored associated diseases and demographic, clinical, and laboratory variables on admissions during the first 2 years since opening.

METHODS

Setting

The Queen Elizabeth Hospital (QECH) is a tertiary governmental hospital in Blantyre, one of the poorest countries in sSA (^6,7). QECH serves 5.5 million people and is the national referral center for pediatric surgery. At any one time, 200–350 pediatric in-patients are hospitalized, 60–70 patients in the pediatric surgical ward. In 2017, the first Malawian six-bed PICU was opened as part of the Mercy James Centre for pediatric surgery and intensive care within QECH, serving both surgical and medical patients. Prior to PICU opening, critically ill children were admitted to high-dependency areas in hospital wards or to the four-bed adult ICU. Surgical children were mostly admitted to the pediatric surgical ward, which had a mortality of 1.9% in the 2 years prior to opening, while overall pediatric mortality was 3.5% (⁸). Admission guidelines of the new PICU included: 1) reversible conditions with expected good outcome; 2) no history of a chronic illness associated with poor prognosis; and 3) anticipated limited stay.

Staffing and Equipment

The PICU was staffed by 25 largely locally trained nurses, four to six working per shift. A concise PICU nurse training program was established by Kamuzu College of Nursing, supported by several international collaborators. The PICU was run by two experienced pediatric intensivists in collaboration with the Malawian anesthesiologists, pediatricians, pediatric surgeons, and trained anesthetic clinical officers. The unit had essential equipment, including ventilators, monitors, infusion pumps, a mobile X-ray machine, an ultrasound machine, and a radiometer ABL80 basic blood gas machine. Full blood counts were analyzed at QECH. Microbial cultures (blood, cerebrospinal fluid) were taken by the discretion of the clinician and according to the sepsis protocol and were analyzed by the Malawi-Liverpool-Wellcome Trust laboratories (⁹).

Study Type and Data Collection

Patient characteristics were analyzed using a prospectively designed database existing of a form with admission and discharge information, completed at time of discharge by the attending doctor. Variables were determined by consensus between intensivists based on literature research and existing models. As the form was adjusted during the study period, added variables were extracted retrospectively from PICU charts and laboratory databases. Blood gas and full blood count were collected on admission to PICU and exported to an excel file that was used to complete the forms. If specimens were not collected on admission, these results were not included. Blood culture (BC) results were cross-checked in the laboratory database, and all results obtained during admission were included. However, they only counted as “positive BC at admission” if positive prior to or at the day of admission. Table 1 lists the definitions used. Medical ethical approval was applied for and waived by the local medical ethical committee (College of Medicine Research Ethics Committee of the University of Malawi, P.11/19/1023).

TABLE 1. - Definitions and Cutoff Values

Demographics and diagnoses
Neonate	Child < 28 d old (1)
Older children	All children of 28 d old or more
Underweight	Weight for age < –2 sd according to World Health Organization Child Growth Standard (10). This applied only to children < 10 yr old
Low birth weight	Birth weight of < 2,500 grams (10). This was only collected for children of < 6 mo old
Planned admission	Elective admissions to PICU, i.e., for perioperative support for major surgery
Unplanned admission	Nonelective admissions to PICU, including emergency surgeries and all medical patients
Readmission	Patients that had any prior admission admitted to PICU during the study period
Intubation	Patients that were intubated prior to or directly on PICU admission
Post-cardiac arrest	CPR prior to PICU admission/CPR reason for PICU admission
Physical examination
Hypothermia	Axillary temperature < 36.0 degrees Celsius (11)
Severe tachycardia	Patients < 1 yr old HR > 180 breaths/min; 1–4 yr old HR > 160 breaths/min; > 5 yr old HR > 140 breaths/min (12)
Severe hypotension	Patients < 1 yr old SBP < 50 mm Hg; 1–4 yr old < 60 mm Hg; > 5 yr old SBP < 70 mm Hg (12)
Increased respiratory rate	Patients 0–3 mo old RR > 60 breaths/min; 3–12 mo old RR > 50 breaths/min; 1–4 yr old RR > 40 breaths/min; 4–12 yr old RR > 30 breaths/min; > 12 yr old RR > 18 breaths/min (13)
Decreased saturation	Oxygen saturation < 90% (11)
Decreased mental state	Blantyre Coma Score < 5 (out of 5) or Glasgow Coma Score < 15 (out of 15 or score less than Alert on Alert, Voice, Pain, Unresponsive score). Patients that were sedated were not scored as having a decreased mental state
Abnormal pupil reaction	Unilateral or bilateral abnormal pupil response
Laboratory results
Acidosis	pH < 7.2 (11)
Lactatemia	Lactate > 5 mmol/L (12)
Thrombocytopenia	Platelet count < 150 × 10^9/L (12)
Severe anemia	Hemoglobin < 6 g/dL (14)
Leukopenia	WBC count < 2.0 × 10^9/L (15)
Hypoglycemia	Random blood glucose < 2.5 mmol/L (11)

CPR = cardiopulmonary resuscitation, HR = heart rate, RR = respiratory rate, SBP = systolic blood pressure.

Study Period and Patient Selection

All patients admitted during the first 2 years after opening (August 1, 2017–July 31, 2019) were included, except patients admitted as part of a clinical trial. All admissions were entered as new entries. Primary diagnosis was coded according to the Australian and New Zealand Paediatric Intensive Care registry. In case of readmission, the reason was listed as primary diagnosis.

Statistical Analysis

Data was entered into an anonymized OpenClinica database and analyzed using SPSS Statistics (IBM Version 26.0 [IBM SPSS Statistics for Windows; IBM Corp, Armonk, NY]). Weight-for-age z scores were computed using Stata (StataCorp, College Station, TX). Continuous variables were dichotomized using clinical cutoff values adapted from the literature (Table 1). Primary outcome was defined as death in PICU. Univariate correlations between variables and mortality were examined using chi-square or Fisher exact test. Stratification was applied to explore differences between neonates and older children and to limit the possibility of interaction between age and clinical variables. Variables that were correlated with mortality (p < 0.10) on univariate analysis were entered in unconditional logistic regression models built in consultation with pediatric intensivists. The first model contained only demographic and clinical variables; the second model added laboratory results. If more than one indicator of an organ system failure was associated with outcome, the strongest associated variable was chosen to prevent confounding. Missing observations were included as a separate category and compared with the dummy variable. Goodness of fit of the models was assessed by the Hosmer-Lemeshow test and Nagelkerke R².

RESULTS

During the study period, 573 PICU admissions occurred. Forty-one were excluded as they were admitted for a clinical trial and may otherwise not have qualified for PICU, and one because no data on outcome could be retrieved. Table 2 summarizes the characteristics, interventions, and outcomes of the final 531 patients. One-hundred sixty-seven admissions were neonates (31.4%), of which 95.2% had a surgical condition. Median length of stay (LOS) was 3 days (interquartile range, 1–5 d). Twenty-five patients were long-stay patients (LOS > 15 d), 10 of which were neonates. The long-stay patients accounted for 5% of the total admissions and 35% of all bed days (953/2704).

TABLE 2. - Demographics, Interventions, and Outcome of All Children Admitted to PICU

Demographics	Count (n = 531)	Percentage (%)
Gender
Female	241/528	45.7
Age (mo)
Median (IQR)	9.0 (0.33–57.0)
Age groups
Neonates (< 28 d)	167/531	31.4
Birth weight (< 6 mo)
Low birth weight	89/138	64.5
Weight for age
Underweight	169/442	38.4
HIV status
Infected/exposed	19/313	6.1
Admission details
Type of condition
Surgical	446/531	84.0
Readmission
Yes	58/531	10.9
Type of admission
Unplanned	410/516	79.5
Reason for admission
Perioperative support	349/530	65.5
Respiratory failure	116/530	21.8
Cardiovascular failure	39/530	7.3
CNS failure	16/530	3.0
Metabolic	4/530	0.8
Other	6/530	1.1
PICU interventions
Ventilation
Intubation performed in PICU	157/528	29.7
Invasive respiratory support	313/527	59.4
Blood products
Blood (product) transfusion	196/500	39.2
Cannulation
Central venous catheter	153/512	29.9
Arterial line	31/510	6.1
Cardiovascular support
Inotropes	142/526	27.0
Resuscitation (cardiopulmonary resuscitation)	99/521	19.0
Do not resuscitate	60/478	12.6
Outcome and complications
Length of stay (d)
Median (IQR)	3.0 (1.0–5.0)
Length of stay
> 15 d (long-stay)	25/531	4.7
Death in PICU	149/531	28.1
PICU outcome
Uncomplicated	307/508	60.5
Complications in PICU
Auto-extubation	11/508	2.2
Prolonged hypoxia	11/508	2.2
Sepsis	93/508	18.3
Drug error	11/508	2.2

IQR = interquartile range.

Admission Diagnosis

The main reason for PICU admission was perioperative support (65.5%), followed by respiratory failure (21.8%), and cardiovascular failure (7.3%). Most prevalent diagnoses were gastroschisis (n = 71; 13.3%), foreign body aspiration (n = 41; 7.7%), and sepsis (n = 38; 7.2%). In the 85 medical patients, respiratory failure due to bronchiolitis (n = 16; 18.8%) and pneumonia (10; 11.8%) were the most common reason for admission. Sepsis was the presenting problem in 5.4% of surgical and 16.5% of medical patients. In neonates, patients were most admitted for gastroschisis (69/167; 41.3%), intestinal atresia (20/167; 12.0%), and trachea-esophageal fistula (17/167; 10.2%). In older children, foreign body aspiration (41/364; 11.3%), Wilms tumor (27/364; 7.4%), and sepsis (23/364; 6.3%) were most common (Supplemental Digital Content Fig. 1, https://links.lww.com/PCC/C333).

Mortality in PICU

Of the 531 admissions, 382 patients (71.9%) were alive on discharge from PICU and 149 died (28.1%) in PICU. The mortality rate in neonates (88/167; 52.7%) was higher compared with older children (61/364; 16.8%; p ≤ 0.001) and did not differ between medical (19/85; 22.4%) and surgical admissions (130/446; 29.1%; p = 0.20). Admission diagnoses associated with mortality were gastroschisis (odds ratio [OR], 3.6; CI, 2.2–6.0), sepsis (OR, 3.5; CI, 1.8–6.9), and trachea-esophageal fistula (OR, 4.3; CI, 1.8–10.2) (Fig. 1). Patients with Wilms tumor had lower mortality in comparison with patients with other conditions (OR, 0.1; CI, 0.0–0.7). Patients with vocal cord papilloma, foreign body aspiration, and malignancies other than Wilms tumor were all discharged alive.

Figure 1.: Main diagnoses at admission and corresponding mortality rates. N/A = not available.

Demographic, Clinical, and Laboratory Variables Predicting Mortality

Fifteen predictors were found to be associated with death in PICU in univariate analysis (Table 3). Five variables were associated with mortality in the first multivariable model: neonatal age, congenital anomaly, severe hypotension, cardiac arrest, and decreased mental status. In the second model, introducing laboratory variables, all variables from the first model except for congenital anomalies remained associated and acidosis and thrombocytopenia were added (Table 3; and Supplemental Digital Content Fig. 2, https://links.lww.com/PCC/C334).

TABLE 3. - Unadjusted and Adjusted Odds Ratios of Mortality for Two Multivariate Models

Category	Risk Factor	Total, n/N (%)	Death in PICU, n/N (%)	Univariate Unadjusted OR (95% CI)	Clinical Model (1) Adjusted OR (95% CI)	Clinical and Laboratory Model (2) Adjusted OR (95% CI)
Demographics and admission diagnosis
Gender	Female	241/528 (45.6)	77/241 (32.0)	1.4 (1.0–2.1)	1.2 (0.8–1.9)	1.3 (0.8–2.1)
Age	Neonate	167/531 (31.5)	88/167 (52.7)	5.5 (3.7–8.3)	3.7 (1.8–7.3)	4.0 (2.0–8.3)
Low birth weight	Yes	89/138 (64.5)	48/89 (53.9)	1.3 (0.7–2.7)
Nutritional status	Underweight	169/442 (38.2)	67/169 (39.6)	2.2 (1.5–3.4)	1.3 (0.7–2.2)	1.4 (0.8–2.5)
Congenital anomaly	Yes	195/531 (36.7)	89/195 (45.6)	3.9 (2.6–5.8)	1.9 (1.0–3.7)	1.6 (0.8–3.2)
Readmission	Yes	57/531 (10.7)	18/57 (31.6)	1.2 (0.7–2.2)
Patient type	Medical	85/531 (16.0)	19/85 (22.4)	0.7 (0.4–1.2)
Admission type	Unplanned	410/516 (79.5)	136/410 (33.2)	3.9 (2.1–7.3)	1.9 (0.9–4.1)	1.4 (0.6–3.1)
Post-cardiac arrest	Yes	29/376 (7.7)	16/29 (55.2)	3.6 (1.7–7.7)	4.9 (1.9–12.2)	2.9 (1.0–8.0)
HIV status	Infected/exposed	19/313 (6.1)	10/19 (52.6)	2.8 (1.1–7.2)	1.8 (0.6–5.4)	1.6 (0.5–5.6)
Respiratory support	Intubated	207/519 (39.9)	57/207 (27.5)	1.0 (0.6–1.4)
Physical examination
Temperature	Hypothermia	190/442 (43.0)	60/190 (31.5)	1.6 (1.0–2.4)	1.1 (0.7–1.8)	1.1 (0.7–2.0)
Heart rate	Severe tachycardia	67/482 (13.9)	16/67 (23.9)	0.9 (0.5–1.6)
Systolic blood pressure	Severe hypotension	25/362 (6.9)	16/25 (64.0)	7.0 (3.0–16.6)	7.2 (2.6–19.9)	6.3 (2.0–19.1)
Saturation	Decreased	26/482 (5.4)	12/26 (46.2)	2.6 (1.2–5.7)	1.5 (0.6–3.8)	2.0 (0.7–5.4)
Respiratory rate	Increased	104/413 (25.2)	24/104 (23.1)	0.7 (0.4–1.2)
Mental status	Decreased	47/414 (11.4)	18/47 (38.3)	2.0 (1.1–3.8)	5.6 (2.5–12.5)	5.8 (2.4–13.8)
Pupillary reaction	Abnormal	13/109 (11.9)	5/13 (38.5)	2.7 (0.8–9.3)
Respiratory effort	Increased	213/425 (50.1)	63/213 (29.6)	1.4 (0.9–2.2)
Laboratory results
pH	Acidosis	43/161 (26.7)	25/43 (58.1)	2.9 (1.4–6.0)		3.1 (1.2–8.0)
Lactate	Lactatemia	40/139 (28.8)	26/40 (65.0)	5.2 (2.4–11.5)		4.2 (1.5–11.2)
Hemoglobin	Severe anemia	9/346 (2.6)	4/9 (44.4)	1.9 (0.5–7.2)
WBC count	Neutropenia	4/297 (1.3)	2/4 (50.0)	2.5 (0.3–18.0)
Platelet count	Thrombocytopenia	61/294 (20.7)	35/61 (57.4)	4.7 (2.6–8.5)		2.4 (1.1–5.2)
Glucose	Hypoglycemia	5/397 (1.3)	4/5 (80.0)	10.8 (1.2–97.7)		7.1 (0.5–95.8)
Positive blood culture	Yes	52/531 (11.7)	30/52 (57.7)	2.8 (1.6–4.7)		1.2 (0.6–2.6)

OR = odds ratio.

Definitions: Age: neonatal age < 28 d, older children ≥ 28 d; underweight: z score weight for age < –2; post-cardiac arrest: cardiopulmonary resuscitation (CPR) prior to PICU admission/CPR reason for PICU admission; hypothermia: axillary temperature < 36 degrees Celsius; severe tachycardia: age-dependent cutoff; severe hypotension: age-dependent cutoff; decreased saturation: oxygen saturation < 90; decreased mental status: Blantyre Coma Score < 5 (out of 5) or Glasgow Coma Score < 15 (out of 15 or score less than Alert on Alert, Voice, Pain, Unresponsive score); acidosis: pH < 7.2; lactatemia: lactate > 5 mmol/L; severe anemia: hemoglobin < 6 g/dL; leukopenia: WBC < 2.0 × 10^9/L; thrombocytopenia: platelet count < 150 × 10^9/L; and hypoglycemia: glucose < 2.5 mmol/L.

Mortality in Older Children Versus Neonates

Three factors were associated with mortality in neonates in a univariate analysis stratified by age: severe hypotension (OR, 4.9; CI, 1.3–18.9), acidosis (OR, 5.2; CI, 1.4–18.5), and lactatemia (OR, 4.8; CI, 1.3–18.3) (Fig. 2). Neither low birth weight, present in 65% of neonates, nor thrombocytopenia were associated with mortality. In older children, eleven variables were associated with mortality (Fig. 2).

Figure 2.: Unadjusted odds ratios of mortality stratified for older children and neonates. Definitions: Age: Neonatal age less than 28 d, older children greater than or equal to 28 d; underweight: z score weight for age less than –2; post-cardiac arrest: cardiopulmonary resuscitation (CPR) prior to PICU admission/CPR reason for PICU admission; hypothermia: axillary temperature less than 36 degrees Celsius; severe tachycardia: age-dependent cutoff; severe hypotension: age-dependent cutoff; decreased saturation: oxygen saturation less than 90; decreased mental status: Blantyre Coma Score less than 5 (out of 5) or Glasgow Coma Score less than 15 (out of 15 or score less than Alert on AVPU score); acidosis: pH less than 7.2; lactatemia: lactate greater than 5 mmol/L; severe anemia: hemoglobin less than 6 g/dL; leukopenia: WBC less than 2.0 × 10^9/L; thrombocytopenia: platelet count less than 150 × 10^9/L; and hypoglycemia: glucose less than 2.5 mmol/L.

DISCUSSION

In one of the first PICUs in sub-Saharan Africa, we report a mortality rate of 28% which is much higher than 2–3.5% reported in PICUs in high-income countries (HICs) (^16,17). However, it is consistent with a newly opened PICU in Mozambique (25%) (¹⁸) and lower than reported from Kenya (37.5%) (¹⁹). There are several potential explanations as to why the mortality rate is higher in new PICUs in LMIC. Although some beyond the scope of this study, other etiologies, comorbidities, delayed presentation, advanced disease, and resource limitation have all been identified to contribute to mortality in these settings (^2–4,20,21). We will briefly discuss our findings in the next paragraphs.

Surgical and Medical Patients

The percentage of surgical admissions was 84%, consistent with a PICU in South Africa (²²) but substantially more compared with PICUs in Mozambique (¹⁸) and Kenya (¹⁹), which reported 5–10% surgical admissions. Relative priority was given to surgical patients as the PICU is part of a pediatric surgical center and patients were admitted postoperatively for observation to the PICU. It has been reasoned that surgical patients can be good candidates for PICU with limited beds as they often require a short duration of (perioperative) intensive care with great benefit from admission (²³). This applied in the older surgical patients admitted for oncological surgery, vocal cord papilloma, and foreign body aspiration in which PICU survival was close to 100%. Our data thus supports the recommendation to prioritize PICU admission for older surgical patients. Given the scarce bed capacity, neonatal surgical patients may not be prioritized for PICU admission as this group should (first) receive a more specialized approach to improve their outcome.

Neonatal Admissions and Congenital Anomalies

The 52% neonatal mortality in our study is higher than the 36% reported in a neonatal ICU population in South Africa (²⁴). A possible explanation is the higher prevalence of congenital anomalies in our study (83% vs 53%), as these were significantly associated with mortality. The mortality rate of gastroschisis, accounting for more than 40% of the neonatal admissions in our PICU, was approximately 50% in both studies (²⁴). This is still substantially lower compared with 75–100% previously reported from other African countries (^25,26), which could indicate the additional benefit of PICU care. Worldwide, congenital anomalies cause half a million annual deaths, many of which could potentially be prevented by the improvement of surgical and perioperative care (²⁷). Our data show that this may be possible yet challenging and important elements such as parenteral feeding and pre- and post-PICU care should be in place to further improve survival in this population (^{2–4,21,26–28}). Without access to parenteral nutrition, one should develop strategies to utilize breast milk enterally as soon as possible (e.g., use of post-pyloric tubes). If this is not possible, one should carefully consider if there is a realistic benefit of a surgical procedures and PICU admission.

Sepsis and Malnutrition

Sepsis and malnutrition were identified as potential areas of attention in our setting. Sepsis was a common problem in our study and the associated mortality was high. Sepsis occurred in medical and surgical patients and occurred both as an admission diagnosis and during PICU admission. Most isolates were Gram-negative bacteria (Supplemental Digital Content Fig. 3, https://links.lww.com/PCC/C335), some becoming multiresistant during the study. The high prevalence and poor outcome findings corroborate with other studies in LMIC (^27,29). They may be explained by late referral and delayed start of antibiotics, the high vulnerability in neonates and children with central lines, catheters, or those being ventilated. However, these data underline the importance of preventive and therapeutic approaches to infections when setting up a PICU and should include hygienic and antibiotic approaches and implementation of sepsis bundles (^30–33). To prevent these nosocomial infections, evaluation of entire “pathways to care” of patients are needed, which should include core facilities such as storage areas; central cleaning and sterilizing departments; laundry services; and water and sanitation services. Malnutrition is a well-recognized risk factor for adverse outcome, yet we did not confirm that in this study. Low birth weight were very common as 65% of the neonates had low birth weight and a nonsignificant downward trend in survival with decreasing birth weight was observed. Our unit did not have access to parenteral nutrition and enteral tube feeding options were limited. This will have impacted PICU outcome; however, the exact impact cannot be determined from this study. Inequity in nutritional resources has been highlighted in studies showing access to parenteral nutrition in, respectively, 19% and 29% of centers in LMIC and 100% in HIC (^28,29). Optimizing (parenteral) nutritional care is likely to improve PICU outcome, especially when taking care of neonates with intestinal anomalies (^28,29,33).

Variables Associated With Mortality

The second aim of our study was to identify factors associated with mortality. The clinical and laboratory parameters identified in this study variables have all been adapted in the most commonly risk predictor scores (Pediatric Risk of Mortality-III, Pediatric Logistic Organ Dysfunction-3) in HIC. Our findings support that these variables are also important in a low-resource setting. Severe hypotension, decreased mental state, lactatemia, and acidosis are all indicators of pediatric shock and severe illness. Severe hypotension is considered a late sign of deterioration in children and showed the highest mortality, as 65% of patients admitted with severe hypotension died in our PICU. Decreased mental state was likewise associated with a high mortality rate of 40%, consistent with a study in a general pediatric population in sSA (¹²). The mortality of children admitted with cardiac arrest prior to admission or at admission was 55.2%, in line with data from South Africa (²²), and slightly higher compared with the United States (49%) and Europe (50%) (³⁴). The fatality rate of pediatric (septic) shock continues to be high globally and consistently higher in LMIC compared with HIC (³⁵). Laboratory parameters can be important to assess the clinical condition of the child, however, access to basic arterial blood gasses can be limited in LMIC. In this study, acidosis and lactatemia were both associated with mortality, which confirms the findings of other studies in HIC (^36–38) and LMIC (¹²). Full blood counts may be more feasible and are also relevant as thrombocytopenia was found to be associated with mortality which corresponds with data from HIC (^38,39). Noteworthy, neonatal mortality was high, yet less variables associated with mortality were identified in this subgroup. All three variables, severe hypotension, lactatemia, and acidosis, are indicators of poor perfusion. The absence of (clinical)variables highlights the difficulty in recognizing the critically ill neonate and emphasizes the added value of laboratory in this vulnerable population.

Limitations

This study had several limitations. Firstly, our database had missing data. Missing data imputation was considered yet discarded due to the explorative design, heterogeneity of the patient group and the possibility of nonrandom missing data. Secondly, the study focused on PICU mortality and therefore did not include late in-hospital or post-discharge mortality such as neurological sequelae. To determine the value of PICU care more information this information on long-term outcome is required. Thirdly, our models could only explain mortality to some extent as the R² (Nagelkerke) of the final model was 0.429. Fourthly, blood gas samples, BCs, and FBC were not routinely taken on admission but taken by the discretion of the clinician and according to our sepsis protocol. This may have led to patient selection which may apply most to BCs. Fifthly, cutoff values for blood pressure and heart rate were not stratified further for patients younger than 1 year old. Sixthly, the PICU was a starting unit, so we acknowledge that our findings may change over time. Our data, however, may be applicable to other new PICUs. Lastly, our study does not represent patients that died in the wards prior to PICU consultation/admission nor those that were deferred from PICU admission. These data were not routinely collected, which may have affected our results.

CONCLUSIONS

In one of the first public PICUs in sub-Saharan Africa, we report 28% mortality. We identified that older surgical patients have a good outcome and may benefit from PICU admission, while neonates, especially with congenital anomalies, have a poor outcome. Areas to improve PICU care could include specific PICU infection preventive and curative plans and (par)enteral feeding protocols, alone or as part of care bundles.

We further identified seven factors associated with mortality in our population, including four clinical/demographical variables (neonatal age, decreased mental status, post-cardiac arrest, and severe hypotension) and three laboratory variables (acidosis, lactatemia, and thrombocytopenia). In neonates, the clinical signs associated with poor outcome were limited and additional laboratory markers including blood gas analysis are important. These data can be used to identify children at risk of deterioration, establish evidence-based admission criteria, allocate resources and to improve the quality of (critical) care in our setting and future PICUs in sub-Saharan Africa.

ACKNOWLEDGMENTS

We acknowledge all children and guardians of patients admitted to PICU. We further acknowledge the efforts of all involved in setting up and running the PICU including the staff, the hospital, the ministry of health, and all Malawian and international partners that have contributed.