Objectives Isolation and quarantine are core strategies in managing infectious disease outbreaks. In-depth analyses comparing adherence patterns over time during COVID-19 are lacking. Here we examine five distinct isolation and quarantine behaviours, the prevalence of low-versus high-risk violations, and predictors of (high-risk) violations. Study design We used survey data (>1 million survey records) from a dynamic cohort study throughout the COVID-19 pandemic in the Netherlands. Methods We depict isolation and quarantine adherence patterns, and the riskiness of violations over time. Using regression analyses we examined sociodemographic, contextual, and situational drivers of (high-risk) non-adherence. Results Full adherence was lowest (40%) for people returning from a high-risk country and highest (68%) for people testing positive. When we allowed for low-risk violations, adherence increased substantially for four isolation and quarantine behaviours (to 94% among people testing positive). Non-adherers returning from a high-risk country mainly displayed high-risk quarantine violations (e.g., going to work or the pub). The strongest adherence predictor was receiving a positive COVID-19 test. Conclusion Adherence varied considerably between quarantine and isolation behaviours, and was higher when the probability of actually being infected was greater. Moreover, many of the violations were (very) low-risk (e.g., a walk outside at a quiet time) and potentially functional for quarantine persistence. People thus behaved as active decision makers, weighing transmission risks against the costs of (full) adherence. For better adherence, public health professionals should clearly explain the relevance of each quarantine and isolation recommendation, offer rapid testing, and research the cost/benefit of allowing for low-risk violations.