Purpose: This study was done to analyze variations in unit staffing and recommend policies to improve nursing staffing levels in intensive care units (ICUS). Method: A cross-sectional study design was used, employing survey data from the Health Insurance Review Agency conducted from June-July, 2003. Unitstaffing was measured using two indicators; bed-to-nurse (B/N) ratio (number of beds per nurse), and patient-to-nurse (P/N) ratio (number of average daily patients per nurse). Staffing levels were compared according to hospital and ICU characteristics. Result: A total of 414 institutions were operating 569 adult and 86 neonatal ICUs. Tertiary hospitals (n=42) had the lowest mean B/N (0.82) and P/N (0.76) ratios in adult ICUs, followed by general hospitals (B/N: 1.34, P/N: 0.97). Those ratios indicated that a nurse took care of 3 to 5 patients per shift. Neonatal ICUs had worse staffing and had greater variations in stafnng ratios than adult ICUs. About 17% of adult and 26% of neonatal ICUs were staffed only by adjunct nurses who had responsibility for a general ward as well as the ICU Conclusion: Stratification of nurse staffing levels and differentiation of ICU utilization fees based on staffing grades are recommended as a policy tool to improve nurse staffing in ICUs.