Title: Help-Seeking for Elder Abuse Among Korean Adults: A Decision Tree Analysis

Background and Purpose: Elder Abuse is a major concern in the context of aging society, and the percentage of people seeking help remains low. Based on the data collected from the 2023 Living Profiles of Older People Survey (LPOPS),this study aims to analyze the multifaceted determinants of help-seeking behaviors of older survivors in Korea. Help-seeking for elder abuse may include formal reporting to police or elderly protection agencies, as well as informal disclosure to family, friends, or neighbors. Decision tree analysis was conducted to show the unique characteristics of the proactive help-seeking groups. Ultimately, this study aims to suggest priorities for practice and policies in reducing elder abuse and facilitating help-seeking.

Methods: The data were collected from the 2023 LPOPS and analyzed by using sampling weights to ensure the national representativeness. The variables included sociodemographic factors, health-related factors such as instrumental activities of daily living (IADL), social factors like social support, and abuse-related characteristics. First, the logistic regression method was conducted using STATA 18.0. Second, a decision tree analysis was conducted using the Classification and Regression Trees (CRT) algorithm within the SPSS 31.0. The prior probabilities were equalized since the number of help-seeking cases was small (8.7%). The model’s stability was maximized using 10-fold cross-validation, along with a minimum node size of 50,000 for parent nodes and 25,000 for the child nodes.

Results: The logistic regression results revealed the cumulative frequency of abuse is the strongest predictor of help-seeking (OR=3.21, p<.001). Moreover, social support level (OR=2.08, p=.001) and subjective health (OR=1.76, p=.012) were positively associated with the likelihood of help-seeking. Furthermore, lower levels of cognitive ability were associated with lower odds of help-seeking (OR=0.44, p=.030). The CRT results revealed an accuracy of 76.4%, and IADL to be the major criterion for the decision tree. Within the group featuring good IADL (>2.35) and high social support, help-seeking probabilities diverged based on abuse frequency and cognitive status. Specifically, the group with multiple abuse experiences and normal cognitive function recorded the highest help-seeking probability at 40.2%. This was approximately 2.6 times higher than the cognitively impaired group under identical conditions, which stood at 15.2%. These results indicate that cognitive judgment is an essential factor for determining help-seeking after experiencing abuse.

Conclusions and Implications: This study reveals that the process of elder abuse help-seeking is influenced by the complex interaction between accumulated abuse experiences and various health-related and social factors. One of the important findings from the CRT analysis was that IADL was the most important ‘split’ variable, while IADL was not significant in the logistic regression analysis. As IADL plays a role as a threshold for help-seeking, this study implies that in order to encourage them to seek help, a dual-level strategic approach should be developed, including the need to actively assist isolated and cognitively impaired older adults, as well as case management for those who are at high risk but with normal cognitive ability to fulfill the many diverse needs of older survivors.