We examined the variation in childhood stunting in Nigeria by using within-data analysis triangulation method and found that there is a wide variation in the childhood stunting in Nigeria, with twenty-eight states (78%) out of the thirty-seven states in the country showing evidence of special-cause variation which merits further investigation to identify possible causes. However, about eight states representing 22% are consistent with common-cause variation which is a variation that is consistent with and within the national average (stable process). We found that childhood stunting is highest in the northern part of Nigeria.
Looking at the league table, it appears that childhood stunting in Nigeria is highest in Kebbi, Plateau, Katsina, Zamfara, Yobe, Sokoto, Jigawa, Gombe, Kaduna, Kwara, Bauchi, Borno and Kano states. These states were also the outliers identified through the funnel plots. This implies from the control chart that the variations identified in these states were due to a ‘special cause’.
The ESDA identified Kebbi, Niger, Katsina, Zamfara, Yobe, Sokoto, Jigawa, Gombe, Bauchi, Borno and Kano states as the ‘hot spots’. These are states that share similar neighbourhood or area characteristics and have high childhood stunting correlating with each other. The findings of this study support those of other similar studies which indicated that living in certain geographical areas has a negative effect on health outcomes [1–3, 20, 21]. Results from ESDA were not significant for Kaduna, Plateau and Kwara states. Kebbi, Niger, Katsina, Zamfara, Yobe, Sokoto, Jigawa, Gombe, Bauchi, Borno and Kano states were consistently identified from all methods used. From this study, it seems that childhood stunting is a key issue in the northern states and this should call further investigation to understand the special cause associated with these states when compared with others.
However, Anambra, Lagos, Imo, Enugu, Abia, Rivers, Bayelsa and Akwa Ibom states appear to have the lowest childhood stunting in Nigeria—all three analytical methods identified these states. The inference that could be drawn from this study is that there are ‘special practices’ from these states that could be identified as good practice. This should also warrant further studies to identify what practices or process that are particular to these states. However, one likely special practice that may be responsible for the low level of childhood stunting in some states in the southern part of Nigeria could be the preschool/school feeding programmes of some state governments in which proteinous foods are given freely to pupils. Deeply rooted in poverty and deprivation, stunting is a nutritional problem that affects mainly developing countries like Nigeria with varied levels of poverty rate and health deprivation index across different geopolitical zones. Northern states have higher poverty rates and health deprivation index than their southern counterparts which may explain for the observed variation seen in this study. Future studies should examine the observed special-cause variation using the generic pyramid model . The generic pyramid model is adapted from industry and comprised checking the data first, then the case-mix, the resources and the work process, and finally check the individual health care practitioners . Thus, the vast majority of problems (95%) may be due to work system and not from individual healthcare practitioners .
A number of methods have been used for assessing the performance and exploring the variation in healthcare outcomes in the past [23–28]. These methods, individually, have their shortcomings and consequently lead to concerns on reliability and acceptance of results from each of these methods. Most literatures have mentioned insufficient risk adjustments as a problem with some methods [7, 26, 28]. Marshall and Mohammed  explored the use of control charts in monitoring mortality outcomes across hospitals in the United Kingdom. The authors summed up their findings that case-mix adjustments may not be essential for longitudinal monitoring.
The study can be criticized for using an indirect measure (synthetic life table) for measuring childhood stunting. However, due to the fact that in low- and middle-income countries, it is hard to obtain reliable data from birth registers; synthetic life table generated from cross-sectional data could be considered a good proxy for childhood stunting data. Teller et al.  were able to verify the validity and reliability of the DHS for monitoring childhood stunting in developing countries. These analyses give a snapshot of childhood stunting in Nigeria using the 2008 DHS; in the course of further investigation of risk factors, there is the need to report on trends with (statistical significance) as well.
Despite these limitations, the study strengths are significant. It is a large, population-based study with national coverage. In addition, data of the DHS are widely perceived to be of high quality, as they were based on sound sampling methodology with high response rates of about 98%. To the best of our knowledge, this is the first study to use ‘within data triangulation’ to explore childhood stunting as a health outcome in sub-Saharan Africa. The within-data triangulation was developed using three different analytical methods, namely league table, funnel plots and spatial cluster analysis. This has potential to enhance the validation and reliability of our analysis through cross verification of methods used.