The current opioid epidemic in the US presents a great problem which calls for policy supervision and regulation. In this work, the opioid cases of five states were used for trend analysis and modeling for the estimation of potential policy effects. An evaluation model was established to analyze the severity of the opioid abuse based on the entropy weight method (EWM) and rank sum ratio (RSR). Four indexes were defined to estimate the spatial distribution of development and spread of the opioid crisis. Thirteen counties with the most severe opioid abuse in five states were determined using the EWM-RSR model and those indexes. Additionally, a forecast of the development of opioid abuse was given based on an autoregressive (AR) model. The RSR values of the thirteen counties would increase to the range between 0.951 and 1.226. Furthermore, the least absolute shrinkage and selection operator (LASSO) method was adopted. The previous indexes were modified, incorporating the comprehensive socioeconomic effects. The optimal penalty term was found to facilitate the stability and reliability of the model. By using the comprehensive model, it was found that three factors-VC112, VC114, VC115-related to disabled people have a great influence on the development of opioid abuse. The simulated policies were performed in the model to decrease the values of the indicators by 10%-50%. The corresponding RSR values can decline to the range between 0.564 and 0.606. Adopting policies that benefit the disabled population should inhibit the trend of opioid abuse.
Keywords: autoregression model; entropy weight method; opioid crisis; policy evaluation; rank sum ratio method.