The paper examines the value of Czech National Cancer Registry (NCR) for the health care assessment. More than 1.3 millions of records collected since 1977 were audited from the viewpoint of comprehensiveness and correctness. The audit proved steadily increasing quality of NCR in time, diagnostic error rate (including wrongly unstaged cases) decreased below 6% after 1990. Most recent NCR records are therefore fully usable for the evaluation of health care results. Reference data set with > 330000 valid records was defined as a source of population standards for overall survival modeling. The reference data set is actual (it covers recent period 1995-2003) and clinically relevant (it contains only fully diagnosed and then treated patients). So called "Complete analysis" of 5-yr survival was proposed as optimal for benchmarking against population-based reference. All calculated reference standards should not be applied for mutual comparisons of hospitals, but for self-benchmarking.