OBJECTIVE: To assess the effect of lockdown during the 2020 COVID-19 pandemic on daily all-cause admissions, and by age and diagnosis subgroups, and the odds of all-cause mortality in a hospital in rural KwaZulu-Natal (KZN). DESIGN: Observational cohort. SETTING: Referral hospital for 17 primary care clinics in uMkhanyakude District. PARTICIPANTS: Data collected by the Africa Health Research Institute on all admissions from 1 January to 20 October: 5848 patients contributed to 6173 admissions. EXPOSURE: Five levels of national lockdown in South Africa from 27 March 2020, with restrictions decreasing from levels 5 to 1, respectively. OUTCOME MEASURES: Changes and trends in daily all-cause admissions and risk of in-hospital mortality before and at each stage of lockdown, estimated by Poisson and logistic interrupted time series regression, with stratification for age, sex and diagnosis. RESULTS: Daily admissions decreased during level 5 lockdown for infants (incidence rate ratio (IRR) compared with prelockdown 0.63, 95% CI 0.44 to 0.90), children aged 1-5 years old (IRR 0.43, 95% CI 028 to 0.65) and respiratory diagnoses (IRR 0.57, 95% CI 0.36 to 0.90). From level 4 to level 3, total admissions increased (IRR 1.17, 95% CI 1.06 to 1.28), as well as for men >19 years (IRR 1.50, 95% CI 1.17 to 1.92) and respiratory diagnoses (IRR 4.26, 95% CI 2.36 to 7.70). Among patients admitted to hospital, the odds of death decreased during level 5 compared with prelockdown (adjusted OR 0.48, 95% CI 0.28 to 0.83) and then increased in later stages. CONCLUSIONS: Level 5 lockdown is likely to have prevented the most vulnerable population, children under 5 years and those more severely ill from accessing hospital care in rural KZN, as reflected by the drop in admissions and odds of mortality. Subsequent increases in admissions and in odds of death in the hospital could be due to improved and delayed access to hospital as restrictions were eased.