The Preventive Effect of Outdoor Recreational Fishing on Anxiety Disorder

. Background: Anxiety disorder, a severe global public health problem, has caused many adverse e ﬀ ects. E ﬀ ectively preventing anxiety disorder is very important. This paper chose recreational ﬁshing as an intervention measure to explore its e ﬀ ect on preventing anxiety disorder. Methods: The prevention e ﬀ ect of recreational ﬁshing on anxiety disorder and the heterogeneity of this e ﬀ ect were analyzed through the ordinal logistic regression model. The robustness of the main results was tested through the multiple linear regression models. Results: “Fishing” signiﬁcantly a ﬀ ected the anxiety level of residents. Compared with those who did not participate in outdoor recreational ﬁshing, for those who participated there is a 3.494 decrease in the log odds of being in a higher severity of the anxiety level. With the increase in age, the negative e ﬀ ect of recreational ﬁshing on anxiety levels was signiﬁcantly lower. Con-clusion: We focused on the preventive e ﬀ ect of outdoor recreational ﬁshing on anxiety, contributing to add some evidence for non-medical measures to prevent anxiety disorder. Participation in outdoor recreational ﬁshing has a signiﬁcant prevention e ﬀ ect on anxiety disorder. Older people have less e ﬀ ect in preventing anxiety disorder by participating in ﬁshing.

disorder; (2) we discussed popular outdoor fishing activities' inhibitory and preventive effects on anxiety disorder.

Methods
We begin this section with the design of the questionnaire, the selection of the sample population and the collection of research data involved. Then, we explain the dependent and independent variables, and control the covariates and confounding variables. Finally, we sort out the effective data obtained, and make a descriptive statistical analyse to realize the research model.
The independent variable is participating in recreational fishing in the past month, which is treated as a binary variable. The dependent variable is anxiety. Here, the anxiety level value is used as the proxy variable of the dependent variable, which could be treated as either a continuous variable or a categorical variable. The control variables, treated as categorical variables, include other influencing factors, such as occupation, age and other sociodemographic characteristics (shown in table 1).
The first research content, the effect of recreational fishing on the prevention of anxiety, is analyzed by the ordinal logistic regression model. The second research content is the heterogeneity effect analysis. We stratify the population according to the information statistics of social demography characteristics. Age is taken as the moderator and is brought into the model to detect whether there is a difference in the effect of different age groups, which makes the results more universal. After that, the dependent variable is treated as a continuous variable for multiple linear regression analysis to test the robustness of the results.

Data
According to the survey on mental health data of NSDUH Institute in the United States, the occupations most vulnerable to anxiety and depression include service providers, social workers, doctors and nurses, artists, teachers, college students and other occupations [29,30]. Therefore, we takes the high-anxiety groups and the occupational population prone to anxiety in Sichuan province of China as the research object. Precisely, we choose university students, university teachers, doctors, lawyers, engineers, service workers and company employees in the region as the research population, and use a random sample for our analysis [31].
The main content of the questionnaire and scale includes four parts: basic information survey of social demography; measurement of anxiety level and stress level in various aspects such as work, study and social interaction; measurement of independent variables and regulatory variables used for research and analysis; investigation of relevant interference factors.
Since the level of anxiety is usually judged according to the individual's symptoms in the past two weeks, most of the data obtained are the data in the past one month or half a month [1]. Among them, the GAD-7 Generalized Anxiety Disorder Scale is used to measure the anxiety level. The higher the score is, the higher the anxiety level is. According to the score, the anxiety level is divided into five levels. From mild to severe, they are normal level, mild anxiety, moderate anxiety, moderate to severe anxiety, and severe anxiety [12].
We adopt the work stress questionnaire of Kimetal (1996) to measure working and learning pressure. Its four dimensions are workload, role conflict, vague role and insufficient resources. The higher the score is, the higher the working and learning pressure is.
The data are obtained mainly through two ways: one is to entrust a professional questionnaire company to fill in the questionnaire randomly, and the other is to collect data randomly by online questionnaire. Finally, 546 questionnaires are obtained. By the second method, 133 data outside the study population are obtained. Therefore, the part and 21 ineffective data with unqualified filling are excluded, and the final effective sample size for this study is 392.
These samples from high-anxiety occupations are representative in anxiety research. More importantly, the samples are independent of each other. Besides, the questionnaire company collected data as randomly as possible. Therefore, our sample can represent Sichuan Province.

Covariates and Confounding Variables
We mainly introduces the covariates and confounding variables that strongly interfere with the results. Basic information such as age, gender, occupation, marital status and other sociodemographic characteristics may interfere with the results [32]. In addition, we detect some influence factors according to the dependent variables that may have a great impact on the anxiety level , such as drug treatment, excessive drinking and smoking, work stress, social stress, family conflicts, etc. Moreover, we considered interference factors from the perspective of independent variables, such as participation in other entertainment activities (shown in table 1).
We control the covariates and confounding variables. We exclude these samples with unhealthy status, excessive smoking, excessive drinking, often staying up late, social stress, family conflicts, drug treatment, positive events, negative events, or other recreational activities that are more than 5 hours per day. At the same time, other samples in the marital status, such as widowhood and divorce, are excluded. The other two marital situations are brought into the model for analysis, and the remaining variables are also brought into the model for analysis and control. Among 392 effective data, we exclude 185 data interfered by some covariates and confounding variables. Eventually 207 data are brought into the model.

Descriptive Statistics
We make a descriptive statistical analysis of the 207 data that control a part of covariates and confounding variables in Section 2.2 (shown in table 2). The remaining variables are brought into the model calculation for analysis.

Model
We focus on exploring the preventive effect of recreational fishing on anxiety disorder. The degree of anxiety symptoms (anxiety level) is the dependent variable, which has five categories: normal level, mild anxiety, moderate anxiety, moderate and severe anxiety. Therefore, the ordinal logistic regression model can be used for analysis. Firstly, the preconditions of the model are tested and judged. Then a single-factor ordinal logical regression analysis is conducted to exclude variables without statistical significance from the model. Finally we conduct regression analysis. Its model is equation 1: Then, we explore the heterogeneity of our findings by a moderating effect model. "Age" is used as a moderator to analyze the different prevention effects of recreational fishing on anxiety disorders among samples of different ages. The moderating effect model is equation 2: Finally, we carry out the robustness test. The stability of the main results is tested through multiple linear regression analysis. The multiple linear regression models are equation 3 and equation 4: (The interpretations of variables are Shown in table 1).

Results
We begin this section with the conditional test of the ordinal logistic regression model. Then we present and analyze the main results of the model regression, and test the robustness of the main results.  Figure 1 shows the histogram of some data, making the data distribution more intuitive. The general distribution of the main data is presented. According to the histogram, the distribution of the independent variable Fishing and the dependent variable Anxiety is relatively scattered.

Conditional Examination
Besides, there are more young and highly educated people in the samples. The ordinal logistic regression model needs to meet preconditions. Therefore, we analyze and judge its condition 4 (no multicollinearity between independent variables) and condition 5 (meeting parallelism).   shown in table 3). The variance inflation factor (VIF) and tolerance of the independent variable are listed. The VIFs of all independent variables are lower than 10, and the tolerances are higher than 0.1, indicating that there is no severe collinearity between independent variables. Therefore, the effective data meets condition 4 of the ordinal logistic regression model.
Then we test the condition 5, the parallelism test (shown in table 4). The original hypothesis of the parallelism test is that the regression equations are parallel to each other. Brant = 31.12 and P = 0.072 > 0.05 represent accepting the original hypothesis, indicating that the parallelism hypothesis is true. The regression equations are parallel to each other. Thus it meets condition 5 of the ordinal logistic regression model. Therefore, it is feasible to use the ordinal logistic regression model for analysis.
Next, single-factor ordinal logistic regression is conducted (shown in table 5). Only "Sex" is not statistically significant (P = 0.781 > 0.05), which should be removed from the model. In contrast, other variables (P < 0.05) are statistically significant.
Finally, we briefly explain why the dependent variable meets the conditions of multiple linear regression: (1) there is a linear relationship between our independent variable and the dependent variable; (2) the observed values of anxiety level are independent of each other; (3) our residual obeys a normal distribution with a mean of 0 and a variance of σ 2 . Therefore, we can use the multiple linear regression model to do the robustness test.

The Prevention of Anxiety by Participating in Fishing
After the test of preconditions and the elimination of variables without statistical significance in the model, we analyze the main research content. Namely, we carry out ordinal logistic regression (shown in table 6 and table 7). The dependent variable "Fishing" is statistically significant. OR = 0.030 means that compared with those who did not participate in outdoor recreational fishing in the past month, the probability of the anxiety level of those who participated in outdoor recreational fishing increasing by one level is about 0.03 times of the original. Coefficient = −3.494 means that compared with those who did not participate in outdoor recreational fishing, for those who participated there is a 3.494 decrease in the log odds of being in a higher severity of the anxiety level. In other words, participation in outdoor recreational fishing has a significant inhibition and prevention effect on anxiety disorder. The same applies to the interpretation of other variables.

Heterogeneity in Prevention Effects
We analyze the heterogeneity effect by the moderating effect model (shown in table 8). However, compared with the results in table 6, the original significant "Age" becomes less significant after adding the moderator. It is because the highly collinearity between the interaction term (Fishing * Age) and the independent variable and the moderator makes the model estimation biased. Age itself is an influential factor in the choice of outdoor entertainment [33]. Then to solve this problem, we centralize the data (shown in table 9). According to the coefficient results of "Fishing" and "Age * Fishing", the results are statistically significant. "Age" weakens the negative effect of "Fishing" on "Anxiety1". Age plays a weakening or inhibiting role in the prevention effect of outdoor recreational fishing on anxiety. Compared with younger people, older people have less effect in preventing anxiety disorder by participating in fishing.

Robustness of The Main Findings
We conduct two robustness tests on the main results. The test method of the first robustness test is to conduct multiple linear regression analysis by treating the dependent variable as a  table 10). According to the calculation results, recreational fishing and anxiety levels directly show a significantly negative correlation. Compared with those who did not participate in outdoor recreational fishing in the past month, the anxiety level of those who participated in the past month may be reduced by about 6.921 points. Moreover, "Age" weakens the negative effect of "Fishing" on "Anxiety1". It means that age weakens or inhibits the prevention effect of outdoor recreational fishing on anxiety. It is consistent with the analysis results of the ordinal logistic regression model in subsection 3.2 and moderating effect model in subsection 3.3.
Next, we carry out the second robustness test. The control variable "Pressure" is deleted from the original multiple linear regression model. Then regression analysis is carried out (shown in table11). According to the comparison of the calculation results of the two multiple linear regression models, the negative correlation coefficient between "Fishing" and "Anxiety" changes very little. "Age" also weakens the negative effect of "Fishing" on "Anxiety". Additionally, the results are still statistically significant, which means that the main results pass the robustness test. It proves the stability of the main results: participation in recreational fishing has a significant effect on the prevention of anxiety, and the age of outdoor recreational fishers plays a weakening or inhibiting role in the prevention of anxiety.

Discussion
Our results support several previous studies on the prevention and cure of anxiety disorder through outdoor sports activities. These studies show that there is a significantly statistical relationship between the participation in outdoor recreational sports and the reduction of anxiety symptoms. Physical activities significantly affect and helps to control anxiety and improve overall health. Physical exercise is one of the most important, simplest, cheapest and most available treatment options for the elderly [34]. Moreover, many studies have proved that some specific forms of sports participation can significantly reduce participants' anxiety level, even depression. For example, yoga, walking, shadowboxing and other physical exercises can significantly reduce the anxiety level of participants [35,36]. We prove that recreational fishing also has a similar effect, adding to existing evidence. There is a significantly negative correlation between outdoor recreational activities and anxiety symptoms, which has the effect of preventing anxiety, even for depression [37]. However, there are still some research deficiencies. Firstly, the sample is from the general population of high-anxiety people in Sichuan Province, which may have some selective bias. We do not take into account the impact of geographical factors, which means lacking the effect analysis on other regions. Secondly, the control of the confounding variables and covariates is not comprehensive enough, leading to the model's inaccuracy. For example, some fishermen may reduce their anxiety level, not because of fishing, but because of an ordinary sport or exposure to sunlight, fresh air and beautiful outdoor environment. It may reduce their anxiety level and play a role in preventing anxiety disorders. Finally, the mechanism of the effect of recreational fishing on the prevention of anxiety deserves further study, and effective mediating effect analysis needs to be carried out. About heterogeneity research, population stratification can also be more detailed. We need further discussion and in-depth study.
Despite these limitations, this study has many advantages, including a representative sample of high-anxiety groups. At the same time, it also evaluated the heterogeneity of the preventive effect on anxiety disorder to understand the preventive effect of different populations more comprehensively, rather than simply paying attention to a mental health condition.

Conclusion
Taking the high-anxiety group in Sichuan as the population, we explore the effect of outdoor recreational fishing on the prevention of anxiety disorder, and whether this effect is heterogeneous. The results show that compared with the groups that did not participate in outdoor recreational fishing in the past month, the level of anxiety of the groups that participated in outdoor recreational fishing is significantly lower. Participation in outdoor recreational fishing has a significant inhibition and prevention effect on anxiety disorders. Moreover, age plays a weakening or inhibiting role in the prevention effect of outdoor recreational fishing on anxiety disorder. Compared with younger people, older people will have less effect in preventing anxiety disorder by participating in fishing.