Table 2 presents the results of the stepwise regression analysis on variable risk perception. The predictor variables included in the regression model are gender, age, education, monthly income, infection information obtained from online social networks (oi) and infection information obtained from physical social networks (\(\pi \)). The results have shown that variables infection information coming from both physical (M2) and online social networks (M3) significantly contribute to individuals’ risk perceptions when they are separately included in the model. However, when both predictors are included together in the regression model (M4), infection information obtained from physical social networks \((\beta = 0.29,\ p < .01)\) is proven to be a better predictor of risk perception scores than online social networks \((\beta = 0.14,\ p < .01)\). In the analysis, demographic characteristics such as gender, age, education level or monthly income, were not found to be significant predictors \((all \ p > .05)\) of risk perception. The overall model fit was statistically significant \((F= 8.796,\ p < 0.01)\), indicating that the considered predictors together accounted for a significant amount of variance in risk perception. The proposed regression model explained \(11.6\%\) of the variance in conformity buying \((R^2 = 0.116,\ p < 0.01)\).

As previously mentioned, panic consumption in this paper is measured through variables conformity buying and uncontrolled self-medication. Results in Table 3 pertain to the variable conformity buying as the dependent variable, whereas the proposed models in Table 4 are associated with uncontrolled self-medication as the dependent variable.

Conformity buying In the stepwise regression analysis predicting conformity buying, several predictors were considered. The predictor variables included in the model are gender, age, education, monthly income, infection information obtained from physical social networks (oi), infection information obtained from physical social networks (\(\pi \)), health change (hc), risk perception (rp), and the interaction effect between health change and risk perception (hc \(\times \) rp).

Among the individual predictors, risk perception (rp) has proven to be a significant and best predictor of conformity buying among all variables in model \((\beta = 0.79,\ p < 0.01)\), indicating that higher levels of risk perception positively impact the increase in conformity buying behaviors.

The variable health change (hc) did not show a significant relation with conformity buying \((\beta = -0.02,\ p > 0.1)\), but a significant negative correlation between conformity buying and the interaction effect of health change and risk perception (hc \(\times \) rp) was found \((\beta = -0.26,\ p < 0.05)\). This suggests that scores on the variable risk perception affected the connection between health change and conformity buying, and these two predictors together impact the results on the variable conformity buying. Regarding the variables of infection information, physical social networks (\(\pi \)) had a significant positive association with conformity buying \((\beta = 0.27,\ p < 0.01)\), suggesting that the information about infection rates coming from physical social networks is associated with higher levels of conformity buying. On the other side, infection information coming from online social networks (oi) did not show a significant association with conformity buying \((\beta = 0.027,\ p > 0.1)\). Surprisingly, demographic variables such as age, gender, education, and monthly income also are not significant predictors of the score on variable conformity buying \((all\ p > 0.1)\). The overall model fit for the regression analysis on the variable conformity buying was statistically significant, as indicated by the F-values \((F = 24.14,\ p < 0.01)\), suggesting that the included variables collectively contributed to explaining the variance of the dependent variable. The proposed regression model explained \(35.3\%\) of the variance in conformity buying \((R^2 = 0.35,\ p < 0.01)\) with all considered variables included in the model.

Uncontrolled self-medication The regression analysis of scores on uncontrolled self-medication includes the same set of variables as in conformity buying. The results have shown that gender was a significant predictor of the uncontrolled self-medication \((\beta = -0.153,\ p < 0.01)\), indicating that women are more prone to take drugs on their own than men. Other demographic characteristics; age, education level and monthly income, had no significant predictor value when it comes to uncontrolled self-medication \((all\ p > 0.1)\). Regarding the transmission of infection information, in the initial model (M1) where only two variables related to infection information and demographics were considered, infection information obtained from physical social networks (pi) emerged as a significant predictor of uncontrolled self-medication \((\beta = 0.14,\ p < 0.01)\). In the same model, infection information obtained from online social networks (oi) was not a significant predictor \((\beta = -0.06,\ p > 0.1)\). But, in the final model (M5), when all variables are included in the regression analysis, infection information obtained from physical social networks (\(\pi \)) was not a significant predictor of uncontrolled self-medication anymore \((\beta = -0.01,\ p > 0.1)\), while infection information obtained from online social networks (oi) had a negative significant correlation with uncontrolled self-medication \((\beta = -0.12,\ p < 0.01)\), indicating that receiving information about infection from online social networks cause a decrease in the amount of uncontrolled self-medication. The results of conducted regression analysis have shown that the significant predictor of uncontrolled self-medication is also the variable (rp) risk perception \((\beta = 0.33,\ p < 0.05)\), implying that the higher risk perception of infection with COVID-19led to a higher uncontrolled self-medication. As expected, health change (hc) also had a significant negative correlation with uncontrolled self-medication \((\beta = -0.09,\ p < 0.05)\), indicating that the worsened health state led to an increase in uncontrolled self-medication. The interaction between risk perception and health change (hc \(\times \) rp) on uncontrolled self-medication was not found to be significant \((\beta = 0.11, p > 0.1)\). The overall model fit for the regression analysis variable uncontrolled self-medication was statistically significant, as indicated by the F-values \((F = 13.61,\ p < 0.01)\), suggesting that the considered predictors collectively contributed to explaining the variance of the dependent variable. The proposed regression model explained \(23.5\%\) of the variance in uncontrolled self-medication \((R^2 = 0.24,\ p < 0.01)\) when all considered variables were included in the model.

Based on the results of the regression analysis presented earlier, it was found that when infection information is obtained through physical social networks, it has a positive effect on panic consumption by influencing risk perception. To further examine the mediating role of risk perception regulated by health change in the relationship between infection information and panic consumption, an intermediary model test procedure was conducted. The moderated mediation model test is conducted using the bootstrap method and 5000 sample groups at a 95% confidence interval. The results indicate that the variable risk perception has an intermediary effect in the impact of infection information on panic consumption. Considering the mean value and the mean value plus or minus one standard deviation, the low, medium, and high levels of health change are stratified, and the intermediary effect of risk perception under different health change levels is analyzed. The analysis reveals a significant indirect effect of infection information on conformity buying through risk perception, with the indirect effect gradually decreasing as health levels improve. However, the moderated mediation effect of infection information on uncontrolled self-medication is not significant. The analysis reveals that infection information significantly affects both conformity buying and uncontrolled self-medication intake through risk perception, with a notably greater indirect effect on conformity buying. Changes in health play a regulatory role in the impact of risk perception on conformity buying but do not significantly regulate its impact on uncontrolled self-medication. The decline in health levels leads to increased concern, promoting conformity buying driven by risk perception. Consequently, the decline in health levels during an epidemic does not significantly enhance the impact of risk perception on uncontrolled self-medication. However, uncontrolled self-medication intake may be influenced by other factors beyond health change.