Purpose To develop a behavior-weighted stalking classification system that guides emergency intervention decisions by examining how different patterns of stalking behaviors influence victims' willingness to seek formal protection and police officers' approval of emergency measures, ultimately identifying system vulnerabilities and informing targeted protection strategies. Design/methodology/approach Latent class analysis identifies stalking patterns from administrative police case records. Logistic regressions and mediation models then link class membership to (1) victims' formal requests for emergency protective measures and (2) police approvals of those measures. Findings A three-class solution best fits the data: Class 1 (low risk, 62.2%) shows limited behavioral diversity and short persistence, Class 2 (moderate risk, 26.6%) combines multiple modalities with greater continuity and Class 3 (high risk, 11.2%) concentrates explicit threats, proximity-based approaches, and longer persistence. Class membership predicts help-seeking and police decisions. For Class 2, the direct association with approvals becomes no longer significant once requests are included (full mediation), revealing a protection gap–elevated objective risk alongside comparatively low help-seeking and approvals. For Class 3, a strong association with police decisions remains, but is substantially mediated by victims' requests. Overall, victim requests are the primary pathway by which heterogeneous stalking patterns translate into formal protection. Originality/value This study provides an empirically derived, behavior-weighted stalking typology linked to a clear mechanism from risk to protection. It pinpoints a mid-risk class where persistence and multimodality do not reliably trigger protection, informing class-specific practice (risk-triggered safeguarding for high-risk cases, proactive, burden-shifting outreach for persistence-driven cases and proportionate boundary-setting for lower-risk cases).
Yu et al. (2026) studied this question.