Figure 2

The change of hospitalization rates before and after the HACCP implementation. Estimated rates of change of slope of pre- and post- Hazard Analysis and Critical Control Points Systems (HACCP) periods for each Census division derived from the 14-year analysis of a single time series are displayed with corresponding 95% confidence intervals. The outcome is weekly hospitalization rate in cases per 1,000,000 projected elderly, modelled using generalized linear models with Poisson distribution and controlled for annual oscillation. Points falling below (above) the horizontal line indicate a decrease (increase) of rates of change. The pattern of changes can be categorized into five groups: I: significantly decreasing in pre-, stable in post-HACCP; II: Stable in both periods; III: stable in pre-, significantly decreasing in post-HACCP; IV: significantly decreasing in both periods, with the post-HACCP decrease being of significantly greater magnitude; and V: significantly decreasing in both periods.