BACKGROUND: Hospital-based contact tracing aims to limit spread of COVID-19 within healthcare facilities. In large outbreaks, this can stretch resources and workforce due to quarantine of uninfected staff. We analysed the performance of a manual contact tracing system for healthcare workers (HCW) at a multi-site healthcare facility in Melbourne, Australia, from June-September 2020, during an epidemic of COVID-19. METHODS: All HCW close contacts were quarantined for 14 days, and tested around day 11, if not already diagnosed with COVID-19. We examined the prevalence and timing of symptoms in cases detected during quarantine, described this group as proportions of all close contacts and of all cases, and used logistic regression to assess factors associated with infection. RESULTS: COVID-19 was diagnosed during quarantine in 52 furloughed HCWs, from 483 quarantine episodes (11%), accounting for 19% (52/270) of total HCW cases. In 361 exposures to a clear index case, odds of infection were higher after contact with an infectious patient compared to an infectious HCW (aOR: 4.69, 95% CI: 1.98-12.14). Contact with cases outside the workplace increased odds of infection compared to workplace contact only (aOR: 7.70, 95% CI: 2.63-23.05). We estimated 30%, 78% and 95% of symptomatic cases would develop symptoms by days 3, 7, and 11 of quarantine, respectively. CONCLUSION: In our setting, hospital-based contact tracing detected and contained a significant proportion of HCW cases, without excessive quarantine of uninfected staff. Effectiveness of contact tracing is determined by a range of dynamic factors, so system performance should be monitored in real-time.