This study has assessed feeding practices of HIV exposed infants using a cross sectional summary index which summarize key complementary feeding practices in to a composite index by considering the current infant feeding recommendations. The mean (±SD) CS-ICFI score was 9.09 (±2.59) (95% CI: 8.69-9.49). This finding is significantly higher than the mean score reported in Rwanda which is 8.04
. The difference might be due to the difference in age category of study subjects. The study done in Rwanda included younger infants (6–15 months) as compared to the current study (6–17 months). The mean CS-ICFI score of the youngest age group infants was significantly different from mean CS-ICFI score of oldest age group (p = 0.002). Other studies also reported that the older age groups had higher mean index score
[21, 22]. This implies that complementary feeding practices among HIV exposed infants around initiation of complementary feeding practices are less optimal as compared with complementary feeding practice on the older age time.
There was significant difference in CS-ICFI mean scores between urban and rural infants (9.34 Vs 8.44) (p = 0.02). Similar finding was reported from China in which the mean CS-ICFI scores between urban and rural infants were significantly different (p < 0.05)
[12, 23]. The internal consistency of the index for the whole sample was a little below the acceptable limit (0.67 Vs 0.70)
. But removing either bottle feeding or breast feeding from the index increased the Cronbach’s α coefficient to the acceptable range. This indicated that breast feeding and bottle feeding had weak or negative correlation with other complementary dimensions. This again implicate that bottle feeding and breast feeding practice displace other complementary feeding practices. This finding is consistent with the study done in rural Senegal that omitting breast feeding component from the index increase value to 0.82
. In the current study the internal consistency of CS- ICFI was good for children aged 9–11 months (α = 0.70) and for those aged 12–17 months (α = 0.71), but it was lower for infants aged 6–8 months (α =0.68). This showed that CS-ICFI is a reliable measure of complementary feeding practices for infants aged 9 months and above. The lower internal consistency of the CS-ICFI for infants of age 6–8 months resulted from higher prevalence of breast feeding practices in this age group as compared to older age groups (p = 0.003). However, this finding is not consistence with the finding from rural Burkina Faso which showed that the internal consistency was good for the youngest infants (6–11 months) (α = 0.79) and lower among children aged 12–23 months (Cronbach’s α = 0.63)
. This contradiction might be due to the difference in the practices of breast feeding and the socio cultural difference between the two communities. Breast-feeding was almost ubiquitous and was prolonged (even above two years) in Burkina Faso.
In this study, 36.6% of HIV exposed infants were in high CS-ICFI tertile while 31.4% of them were in lower CS-ICFI tertile. However, the age specific ICFI analysis showed that most of the infants (48.4%) in the youngest age group (6–8 months) were found in the lower CS-ICFI category while most (48.5%) of the infants in the oldest age groups (12–17 months) found in the medium CS-ICFI tertile. This finding is consistent with a study done in rural Burkina Faso which showed that among infants of aged 6–11 months, 52%, 23% and 25% were found in the poor, average and good feeding index while among children of aged 12–23 months 35%, 37%, 28% were found in poor, medium and good feeding index category respectively
This study showed that there was a statistically significant difference in mean LAZ and WAZ even after controlling the potential confounders. A similar finding was reported from India which revealed that LAZ and WAZ had showed significant association with ICFI
. Another study from India showed that complementary feeding index was associated with LAZ score but not with WAZ and WLZ scores
. On the other hand a study conducted in Bangladesh reported that the mean LAZ score of children aged 12–23 months was significantly higher among those who were at the upper ICFI tertile compared to those who were at the middle or lower ICFI tertile (-2.01 and -3.20 respectively)
. A similar study done in Rwanda indicated that ICFI was positively associated with WLZ and WAZ scores. However, neither the ICFI nor any of its components were associated with the LAZ score
. There was also a study from rural Senegal reported that feeding index was not associated with either height-for-age or with linear growth
. But a study from China indicated that ICFI was associated with both WAZ and WLZ scores, and did not show statistically significant association with children’s LAZ score
 which could be due to the differences in the participants’ age group in which the study participants in China were infants of aged 6–11 months. This justification is further supported by a study conducted in Latin America which conclude that the association between feeding practices and HAZ score of children was generally weaker and less consistent among children in 12 months of life but increased gradually with age
. The Latin American study explained the observed statistical association between feeding practices and LAZ in older infants as compared to their younger infants could be explained by the clustering/cumulative effects of previous feeding practices. In the current study the difference in mean WAZ and LAZ scores between the lowest and the highest tertile was 0.69 and 1.39 respectively. This mean difference in Z score between the two extreme CS-ICFI tertiles was statistically significant and biologically important for both WAZ and LAZ
. In this study, WLZ score was not associated with CS-ICFI which is consistent with other studies
Presence of statistically significant association between CS-ICFI and WAZ and LAZ but not with WLZ may implicate that CS-CFI has the ability to reflect chronic malnutrition among infants. Thus the CS-ICFI summarizes information on feeding practices and can be used to illustrate the strength and magnitude of associations between adequate complementary feeding practices and infant nutritional outcomes in long term.
None of the two-way interactions between the CS-ICFI and the child, maternal, and household characteristics included in the model were statistically significant. Thus, it appears that the magnitude of differences in LAZ and WAZ between feeding tertiles was not conditioned by any of the child, maternal, and household factors. But a Chinese study showed that better feeding practices were more important for children of lower socioeconomic status
. Finally the findings of this study support the existing literature despite the fact that the index constructed in this study include hygiene and psychosocial variables as index component. The index is applicable for measuring complementary feeding practices comprehensively which include both feeding behavior and diet quality among HIV exposed infants. However further study is recommended on best approaches of assessing hygiene and psychosocial practices during complementary feeding practices.
Limitation of the study
Interpersonal measurement error, recall bias, and absence of validated questionnaire to assess hygiene and psychosocial care were the possible limitations of this study. There might be also bias that could be introduced by the data collectors. Equal weights during scoring were given for all feeding practices especially who have binary responses. But the actual effect of those feeding practices on nutritional status might not be similar. Since the study used cross-sectional design so that it is difficult to establish cause and effect relationship between nutritional status and summary index.