Recent evidence suggests ever-increasing applications of statistical process control (SPC) in health data analysis. However, the diversity in numbers and types of included variables warrant new statistical control charts. This inclusion can be improved by profiles that monitor a describing functional relationship of the process. In this article, weproposed multiple generalized additive models (GAMs) for profile construction. GAMs permit complex fitting models with simultaneous inclusions of parametric and nonparametric terms. Therefore, GAMs can be applied in health data monitoring with a wide range of explanatory variables. We used two statistics to build control charts: (1) a commonly used univariate statistic in nonparametric profiles; (2) a new proposed multivariate statistic which enables the chart to track the role of each included element in the process changes. The statistics are compared according to their performance in monitoring monthly stroke types, including ischaemic and haemorrhagic strokes of patients with acute stroke in the Mashhad Stroke Incidence Study. Features of the proposed profile are discussed and suggestions are made about the utilized statistics in process monitoring. The results show the successful performance of GAMs in profile monitoring.

Keywords