At the end of the follow up, there were 7.5% death and 5.8% lost from follow up. The mortality rate was 4/100 child-years of follow up. This was comparable to previous studies in Ethiopia
[10, 11] and other countries in sub-Saharan Africa
[12–16]. In Jimma University hospital
, 7.3% died and 71.9% were alive at the end of follow up. However, mortality was lower in this study compared to a study from Kenya (8.4%)
. This could be explained in three ways. Firstly, the difference in the study period as there were changes in the treatment and care of children on ART through time. Secondly, the Kenyan study was conducted on a lower sample size (n=149) and may be an inadequate estimate. Thirdly, it may be due to the health care system changes in Ethiopia such as decentralization, task shifting and delegation of HIV/AIDS services to low- and mid-level health care providers. Similar to many studies in sub Saharan Africa
[10–14, 16] where 69% - 89% of deaths occurred in the first six months after ART initiation, in this study more than three quarter of the deaths occurred in the first six months of ART.
This early peak in mortality might be due to delayed ART initiation which increases the risk of severe malnutrition, drug toxicity and immune reconstitution inflammatory syndrome. Moreover, as evidenced from this study, two third of the patients were at WHO clinical stage III or IV and half of them had absolute CD4 count below the threshold for severe immunodeficiency at ART initiation. The study setting was a referral hospital which provides ART and hence children with advanced stages of the disease may be referred in for initiation. The possible reason for increased survival with duration of ART could be the result of progressive increase in CD4 cell count which builds the immune system and the decrease in viral load across time.
As can be noted from the findings of multivariate Cox regression analysis, predictors of mortality were absence of cotrimoxazole preventive therapy, low baseline haemoglobin, absolute CD4 count below the threshold for severe immunodeficiency and delayed or regressing baseline developmental milestone(s). World Health Organization recommends starting cotrimoxazole preventive therapy for all exposed infants at four-to-six weeks of age even before the diagnosis of HIV-infection
 which is again supported by the result of this study where baseline cotrimoxazole preventive therapy was strongly related to lower risk of mortality. Similarly, data from a large randomized clinical trial in Zambia
 provided evidence that daily cotrimoxazole preventive therapy is effective in reducing morbidity, mortality, and hospitalizations in HIV-infected children regardless of CD4 value. Cotrimoxazole preventive therapy prevents the development of very serious and fatal OIs
 and hence it contributes to lower mortality and morbidity.
The other predictor of mortality was CD4 count below the threshold for severe immunodeficiency. This finding was also supported by several studies
[7, 11–13, 20–22].
Lower haemoglobin level (below 10gm/dl) at initiation of ART was another predictor of mortality. Several studies revealed anaemia as an independent predictor of mortality among children on ART
[7, 11, 12, 17, 21]. Anaemia, a common complication of pediatric HIV infection, is commonly associated with disease progression and death. Late presentation of children for treatment has several associated problems as evidenced from this study (68.4% presented with advanced clinical stage and 82.5% with underweight). Starting ART very early reduces disease progression and early mortality
Having delayed or regressing developmental milestone at initiation of ART was strong predictor of mortality. Delayed or regressing developmental milestone at ART initiation may further complicate the immunological and clinical recovery.
The main limitation of this study is that it includes fewer young children and over-represents children diagnosed at a late stage of the disease. This made comparisons with other similar studies difficult. As data were collected from secondary sources, incompleteness was inevitable and it was difficult to assess clinical and immunological responses. Mortality might be underestimated in this study, since lost to follow-up might also include those died without being reported. The study also failed to assess possible causes of death.