The prevalence of NI in the present study was 29/1000 and the most common impairment was cognitive (24/1000), followed by physical impairment (5/1000). Other comparable studies using TQQ were conducted in Kenya
[12, 34], India
, South Africa
, Saudi Arabia
 and Pakistan
. In addition, TQQ cluster survey data was reported from 18 countries in the third round of UNICEF’s Multiple Indicator Cluster Survey 2005–2006 (MICS3)
. Within studies which assessed the validity of TQQ, the prevalence of NI ranged from 16 to 61 per 1000 and cognitive impairment was the most or second most common impairment. In a study in Kilifi, the prevalence of NI was 61/1000 and the most common impairment was epilepsy with prevalence 41/1000, followed by cognitive (31/1000), hearing (14/1000), motor (5/1000) and vision (2/1000) impairments. The prevalence of cognitive impairment in the present study was analogous to that in the Kilifi study
. However, the prevalence of cognitive impairment in other previous studies varies widely. The differences may be explained by the differences of definition of cognition and of age group, as children with severe disabilities will frequently die in infancy or early childhood. Moreover, the use of different assessment tools for cognitive impairment may considerably influence the prevalence of cognitive impairment. It is important to develop standardized assessment tools and definitions of cognitive impairment. However, the absence of standardized tools and lack of psychologists in our situation forced us to use a trained assessor and adapt the assessments for cognitive impairment that were culturally modified.
The prevalence of epilepsy and hearing impairment in this research was noticeably lower than in the research in Kilifi, Kenya
. It is well established that central nervous system (CNS) infection such as malaria and tuberculosis can lead to epilepsy
[39, 40] and the research site was one of the high malaria prevalence areas in Kenya. However, the prevalence of epilepsy was low in this study, compared with Kilifi study. It is estimated that the child who have epilepsy might be easy to die
, because of stigma from community member, poor treatment and poor care from caregivers
[41, 42]. The other possibility is that the method used to diagnose epilepsy. In this study, caregiver recall was used to assess epilepsy. However, in Kilifi study, a interview by clinical officer and electroencephalogram were used to diagnose epilepsy, so that more minor epilepsy might have been found. The prevalence of hearing impairment (0.8/1000) in the present study was considerably lower than that (14/1000) in the Kilifi study
. Schooling rate to special school may influence the prevalence. There was one residential special school focusing on hearing impairment in Mbita district, and the phase two survey was conducted during schooling term. Some children with hearing impairment in Mbita district may have stayed in a special school.
The risk factors related to moderate/severe NI were low monthly income, having more children, maternal age and no maternal antenatal tetanus shot. Poverty is regarded as both a cause and consequence of disability
. Poverty and disability reinforce each other, contributing to increased vulnerability and exclusion
. It was clear that low monthly income was significantly related with impairment but it is not clear whether low monthly income is a cause or a result of disability based on the results in this study, because the research design was cross-sectional. Having more children was also related with moderate and severe neurological impairments. This finding was similar to the results reported in Saudi Arabia
. With an increased number of children in poor countries, quality of care for a child is likely to be worse because of competing demands on mothers, while time and resource available to provide for each child become more limited
, so that severe disabilities are more likely to develop due to lack of care
. In addition, older maternal age was also related to have NI among children aged 6–9 years. As reported in other research, the older the woman, the greater the likelihood of miscarriage, stillborn or underweight baby, and likelihood of impairment also rises
. No maternal tetanus shot was another risk factor for NI. A study in Kenya reported that tetanus increased the risk of NI among survivors
. A mother who never received a tetanus shot was also less likely to go to antenatal care in this study. Moreover, antenatal care is related with child survival in terms of child fatal malnutrition
 and neonatal death
. Therefore, even when the child survives, poor neonatal care
, neonatal encephalopathy
 and neonatal insults
 increase the risk of NI among survivors.
Although neonatal insults was a risk factors of neurological impairment in other research
[12, 52], it was not significantly related in NI in this research. Further research on risk factors of NI would be better to ask more detail of neonatal insults.
There are four limitations in this research. There was a 1 year time-lag between phase one and phase two, because of ethical approval for physical assessment. Some children might have acquired impairments and died in the interim. However, since the age of first recognition of the impairment was mainly before 5 years old (86%) and the number of death was very few, the possibility of developing impairments or dying during the time-lag was supposed to be low. Moreover, around 100 households migrated during that time period and the households with CWDs might be likely to migrate. Hence, the years of staying in the community for households with CWDs were not different from Non-CWDs. Second, since our study design was cross-sectional, causal relationships between risk factors and impairments were not identified. A cohort study may be needed to illustrate the causal relationships. The third limitation was lack of a gold standard and standardized tools to assess cognitive impairment. The assessment for cognition in this research was an adapted and established cross-cultural assessment and definition of cognitive impairment were similar to the study in Kilifi. However, standardized tools and a gold standard for cognitive impairment are needed. Forth, since some CWDs might have stayed in special school during this research, there is a possibility that the prevalence of NI was underestimated.