Abstract
BACKGROUND: Street-migration of children is a global problem with sparse multi-level or longitudinal data. Such data are required to inform robust street-migration prevention efforts. OBJECTIVE: This study analyzes longitudinal cohort data to identify factors predicting street-migration of children - at caregiver- and village-levels. PARTICIPANTS AND SETTING: Kenyan adult respondents (n = 575; 20 villages) actively participated in a community-based intervention, seeking to improve factors previously identified as contributing to street-migration by children. METHODS: At two time points, respondents reported street-migration of children, and variables across economic, social, psychological, mental, parenting, and childhood experience domains. Primary study outcome was newly reported street-migration of children at T2 "incident street-migration", compared to households that reported no street-migration at T1 or T2. For caregiver-level analyses, we assessed bivariate significance between variables (T1) and incident street-migration. Variables with significant bivariate associations were included in a hierarchical logistical regression model. For community-level analyses, we calculated the average values of variables at the village-level, after excluding values from respondents who indicated an incident street-migration case to reduce potential outlier influence. We then compared variables between the 5 villages with the highest incidence to the 15 villages with fewer incident cases. RESULTS: In regression analyses, caregiver childhood experiences, psychological factors and parenting behaviors predicted future street-migration. Lower village-aggregated depression and higher village-aggregated collective efficacy and social curiosity appeared significantly protective. CONCLUSIONS: While parenting and economic strengthening approaches may be helpful, efforts to prevent street migration by children should also strengthen community-level mental health, collective efficacy, and communal harmony.