Survival status and predictors of mortality in severely malnourished children admitted to Jimma University Specialized Hospital from 2010 to 2012, Jimma, Ethiopia: a retrospective longitudinal study

A retrospective longitudinal study was conducted at Jimma University Specialized Hospital (JUSH). The hospital is the only teaching and referral hospital in the southwestern part of Ethiopia. It provides services for approximately 9000 inpatient and 80,000 outpatient attendances a year coming to the hospital from the catchment population of about 15 million people. Severely malnourished children are directly admitted to NRU and treated by Interns (Medical and Health Officer), Nurses, Residents and/or Pediatricians. NRU (sub-section of pediatrics ward of the hospital) is just one of traditional treatment centers for SAM. Admission, treatment and discharge of severe acute malnutrition were as per the Protocol for the Management of Severe Acute Malnutrition, Ethiopia – Federal Ministry of Health, March 2007 [9] which is the update of guideline for the management of severe malnutrition endorsed by the Ministry of Health in May 2004.

Study participants consist of all 947 eligible (out of total 997 SAM patients admitted to NRU from September 11, 2010 to September 10, 2012 ) severely malnourished children . The power of the study was calculated by EpiInfo for all of potential predictors identified in this study. The minimum power calculated was 83.5 % indicating adequate size of the study population was included into the study.

Admission criteria for the ward were as follows: Infants less than 6 months or less than 3 kg being breast-fed were admitted if too weak or feeble to suckle effectively (independent of weight-for-length) or if they had Weight-for-Length (W/L) less than 70 % or bilateral oedema. Children 6 months to 18 years were admitted if they had W/H or W/L < 70 % or MUAC < 110 mm with a length > 65 cm or bilateral pitting oedema. Admissions to intensive care unit (ICU) and surgical ward were excluded as these were not taken to NRU ward. Children with unknown treatment outcome and whose records were not found were also excluded from the study.

Data were collected after permission to conduct the study was obtained from the ethical clearance committee of College of Public Health and Medical Sciences, Jimma University. Permission to use the data was obtained from JUSH and department of Pediatrics, Jimma University. Confidentiality was assured by collecting data anonymously using just the card number of each record.

Data were collected by 5 BSc nurses who had experience in data collection. They also received a one day training to ensure common understanding of the data collection process. The data collection instrument was pre-tested and modified in terms of order and content. Collected data were sorted and checked for errors and completeness onsite daily by supervisors. Reviewed cards were boldly marked to avoid re-review. Data were extracted first from children’s registers and then from records (card and multi-chart). Finally, data from two sources were linked by patient’s card number. Data on variables such as patient’s card number, age, anthropometry at admission (weight, height, mid-upper arm circumference (MUAC)), length of stay in the hospital and treatment outcome were collected from SAM children’s registers. Length of stay in the hospital was also cross-checked by calculating the difference between date of admission for the current problem and date at which the patient died/lost/discharged and corrections were made where inconsistencies were found. Assessments such as vital signs and  presence or absence of clinical signs/symptoms that were made by clinicians caring for children according to hospital standards (protocol for the management of severe acute malnutrition from the Ethiopian Ministry of Health) were extracted from medical records by study nurses. Clinical conditions were evaluated and categorized according to the protocol for the management of severe acute malnutrition from the Ethiopian Ministry of Health. Co-morbidity/complication at admission was defined as co-existence of any other disease(s) with severe malnutrition at the time of admission to the hospital or manifestation of new disease(s) in addition to severe malnutrition during the first 48 h of patient’s admission to the hospital.

According to Ethiopian Protocol for the Management of Severe Acute Malnutrition, nutritional cure is when it is clear that the child is gaining weight on breast milk alone after the Supplemented Suckling technique has been used, there is no medical problem and the mother has been adequately supplemented with vitamins and minerals (for infants less than 6 months or less than 3 kg being breast-fed); when they reach 85 % weight for length and they can be switched to infant formula (for Infant less than 6 months or less than 3 kg with no prospect of being breast-fed). For children 6 months to 18 years, it is W/L ≥ 85 % or W/H ≥ 85 % on more than one occasion and absence of oedema for 10 days [9]. However, in this study, nutritional improvement, rather than nutritional cure, was considered as the endpoint because patients whose medical co-morbidities were stabilized, whose oedema disappeared and who started to gain weight were referred to nearby health facilities (health center or health post) for completion of malnutrition management. Vital signs (RR and PR) were categorized according to Advanced Paediatric Life Support [11] to enable comparison with previous literatures which have used the same classification.

Data were edited, entered into EpiData 3.1, exported to SPSS version 20 for Windows and cleaned to check for completeness, extreme and missing values. All statistical analyses were done using SPSS version 20 for windows. Univariate (descriptive) analyses were performed and presented by tables and graphs. Chi-square test was conducted to determine if there were adequate cell counts for each categorical variable. Kaplan-Meier and Cox regression were used to assess the association of independent variable with outcome. Before modeling, Cox regression model assumption of proportional hazards was checked by Kaplan-Meier hazard plots and testing an interaction of covariate with time. Multi-collinearity among independent variables was checked and did not found any that was significant.

During modeling, multivariable Cox regression was preceded by bivariate Cox regression. P-value of less than 0.2 and clinical importance were used to identify candidates for multivariable analysis. Multivariable Cox regression was run using Forward Wald method to identify best independent predictors of death. The possibilities of interactions (effect measure modification) among independent variables were explored by including interaction terms in the multivariable Cox regression. However, neither statistically significant interaction nor violation of proportional hazards assumption was found. P-value of less than 0.05 was considered as a statistical significance to identify independent predictors of earlier death in multivariable analysis. Hazard ratio (HR) was used as a measure of association (effect).

More than half (58.6 %) of the children enrolled into the study were males and 68.1 % were in the age group of 6–59 months with median age of 24 months. Most (60.8 %) of the children enrolled into the study had oedematous malnutrition (kwashiorkor or marasmic-kwashiorkor). A larger proportion (42.8 versus 27.3 %) of marasmus (non oedematous malnutrition) was observed among 0–59 months old than >59 months old children whereas more oedematous malnutrition (72.7 vesus 57.2 %) was observed among >59 months old than 0–59 months old children (Fig. 1). More than half [58.5 % (87 % of non oedematous and 43 % of oedematous)] of 6–59 months old children had MUAC less than 11.5 cm; the cutoff point for severe acute malnutrition.

Most of children were in critical condition at the time admission. Out of total 947 children, 1.8 % were hypothermic (axillary temperature ≤ 35 °C), 69.2 % had deranged respiratory rate and 21.8 % had deranged pulse rate. Pale conjunctiva and palmar pallor were present in 23.6 and 18 % of the children respectively. Dehydration was present in 11.8 % of the children of which 67.9 % were severely dehydrated. Shock was present in 6.4 % of the children. Children with impaired level of consciousness (lethargic or comatose) account for 12.5 % of the total. The majority (66.0 %) and more than half (51.6 %) of the children had diarrhea and vomiting respectively where 88.8 % had watery diarrhea. Ninety eight (15.7 %) patients with diarrhea were dehydrated (of which 66 (67.3 %) were severely dehydrated). Skin lesions were present in 31.3 % of the children (Table 1).

Concerning distribution of clinical conditions by type of malnutrition, the majority of children with deranged respiratory rate (55.4 %), pale conjunctiva (68.3 %), palmar pallor (68.2 %), dehydration (57.1 %) and shock (52.5 %) were those with non oedematous type of malnutrition whereas most of the children with hypothermia (64.7 %), deranged pulse rate (55.6 %), impaired consciousness level (60.2 %), acute gastroenteritis (diarrhea or vomiting) (64.0 %) and skin lesion (81.8 %) were those with oedematous type of malnutrition.

The majority (69.5 %) of the children had co-morbidity/complications on admission. More than half (54.9 %) of those with co-morbidities on admission had oedematous malnutrition. Pneumonia (32.5 %), anemia (24.6 %), disseminated TB (15.8 %) and conjunctivitis (9.9 %) were the most frequent co-morbidities/complications. Twenty-three (2.4 %) of the children were reactive for HIV test and HIV status was not known for 37 % of the children.

Of the children without co-morbidities/complications on admission, 17.0 % had developed co-morbidity/complication after admission. Pneumonia (49 %), UTI (16.3 %), acute gastroenteritis (16.3 %), oral thrush (14.3 %), conjunctivitis (10.2 %) and anemia (10.2 %) were the leading co-morbidities/complications after admission. Twenty-nine (3.3 %) of the patients without shock on admission developed shock after admission. The majority (85.7 %) of children who developed co-morbidity after admission were also those with oedematous type of malnutrition.

Characteristics at admission for the group that absconded were compared to the overall population. The groups were similar in most admission characteristics; however, there was a statistically significant (p = 0.006) difference in terms of resuscitation (infusion) at admission (Table 2). Only 8 % of overall population were resuscitated with IV fluid at admission, whereas twice as many (15.6 %) were resuscitated from absconded group.

Of 947 children whose records were reviewed, 737 (77.8 %) were discharged with improvement, 88 (9.3 %) died during treatment and 122 (12.9 %) absconded (left the NRU before completing treatment). Of 88 deaths, 27.3 % occurred in the first 48 h and 60.2 % by the end of the first week. The average length of stay in the hospital was 17.4 days (16.7 for children with non oedematous and 17.9 for children with oedematous malnutrition) and the average weight gain was 10.4 g/kg/day (12.9 g/kg/day for children with non oedematous and 7.6 g/kg/day for children with oedematous malnutrition).

Larger proportions of discharges had occurred in the second (35.8 %) and third (24.0 %) weeks of admission (Fig. 2). The mean and median duration from admission to discharge with improvement were 19.5 and 16 days respectively. Of 737 children discharged with improvement, only 226 (30.6 %) of the children achieved a target weight of 85 % weight for height. Larger proportions of deaths (60.2 %) and absconds (43.4 %) occurred in the first week of admission (Fig. 3). The mean and median duration from admission to death were 9.5 and 7 days respectively whereas mean and median duration from admission to abscond were 10.6 and 8 days respectively.

Of 947 children whose records were reviewed, 119 (12.6 %) were infused (resuscitated with IV fluid) and 14 (1.5 %) were transfused. The majority of (66.7 %) deaths among infused children occurred within first day of infusion (Fig. 4). The mean and median duration from infusion to death were 4.1 and 1 days respectively while from transfusion to death were 1.4 and 2 days respectively.