Spatial Autocorrelation Analysis of Nutrition Personnel as a Basis for Environmental Health Management in East Java

Nutrition professionals are essential for delivering effective nutrition interventions, yet their distribution is often uneven across regions. In East Java Province, Indonesia, evidence on the spatial structure of the nutrition workforce remains limited. This study aims to examine the spatial distribution of nutrition personnel across districts and municipalities in East Java, to (1) assess the presence of global spatial autocorrelation and (2) identify local clusters of relative surplus and deficit. A cross-sectional ecological design was applied using data on the number of nutrition workers in 38 districts/municipalities. Each unit was georeferenced using centroid coordinates. A K-Nearest Neighbors (KNN) spatial weights matrix was constructed based on geographic proximity and row-standardized. Global spatial autocorrelation was evaluated using Moran’s I with permutation tests, while Local Indicators of Spatial Autocorrelation (LISA) were used to detect significant local clusters and spatial outliers. Results were visualized through Moran scatterplots and LISA significance and cluster maps generated in R. Moran’s I indicated weak and statistically non-dominant global spatial autocorrelation, suggesting an absence of strong province-wide clustering. However, LISA revealed distinct local patterns. High–high clusters of nutrition personnel were concentrated in the metropolitan core, particularly around Surabaya, Sidoarjo, and Gresik, whereas several southern, western, and peripheral districts formed low–low clusters, indicating contiguous areas of relative deficit. Isolated high–low and low–high outliers were also identified. The distribution of nutrition personnel in East Java is characterized by modest global dependence but marked local inequalities. Integrating spatial analysis into human-resources-for-health planning can support more targeted and equitable allocation of nutrition workers, especially towards identified low–low clusters and spatial outlier districts.