Chronic wasting disease (CWD), a prion disease affecting North American cervids, has been discovered in at least 12 states and provinces throughout the continent. Since 2002, a number of states and provinces have initiated surveillance programs to detect CWD in native cervid populations. However, many questions remain about the appropriate methods, geographic scope, and number of samples required for an effective CWD surveillance program. We provide an improved statistical method to calculate the probability of detecting CWD in primary sample units (e.g., county or deer management unit) that also considers deer abundance and the nonrandom distribution of CWD and hunter harvests. We used this method to analyze data from a statewide CWD detection program conducted in Wisconsin during the autumns of 2002 and 2003 to determine the distribution of CWD in white-tailed deer (Odocoileus virginianus). Deer heads were collected at hunter registration stations, and brainstem (obex) and retropharyngeal lymph nodes were removed for disease testing. Our analysis includes samples from >35,000 deer collected outside the known affected area. The probability of detecting chronic wasting disease at a prevalence of 1% varied from 0.89 to ≥0.99 among the 56 primary sample units. Detection probabilities for 1% CWD prevalence were >0.9 in 55 primary sample units, and >0.99 in 10. Detection probabilities will be higher in areas where CWD prevalence exceeds 1%. CWD-positive deer were detected in eight primary sample units surrounding the known affected area during surveillance activities. Our approach provides a novel statistical technique to accommodate nonrandom sampling in wildlife disease surveillance programs.