Objective. Several measures of variation have been used in SAVA. One study of DRGs found that the coefficient of variation from analysis of variance (CVA) had superior performance. That work is replicated here for ICD-9 surgical procedures, and extended to age/sex-standardized rates. Results are compared with those in the literature, and recommendations are made for assessing small-area variation in future studies.

Data Sources. Data were taken from Washington State's "Episode of Illness" file of hospital discharges in the State in 1987. Up to three ICD-9 surgical procedures and a unique patient identifier were available for each discharge.

Study Design. We calculated the usual small-area variation statistics for 153 different surgical procedures, among 28 counties in Washington state, with and without standardization for age and sex. We tested each variation statistic to determine whether it was correlated with the prevalence of the surgical procedure, and to see which statistic was most correlated with the true variation among counties. We used the empirical results to provide guidelines on how much of the observed variability among counties is likely to be due to chance alone.

Principal Findings. As in the previous study of DRGs, the CVA was uncorrelated with the prevalence and highly related to the true variance of a procedure, whereas the other statistics performed less well. The age/sex-standardized CVA's were usually smaller than CVA's based on the raw data. Confidence intervals for the CVA, either crude or standardized, can be calculated from the appropriate chi-square statistic. Typical CVs were similar to those published elsewhere, in the range of .3 to .5.

Conclusions. The CVA is the statistic of choice to measure small-area variation for procedures, with or without age/sex standardization. A CVA larger than .5 should be considered as "high" variation. In studies where fewer than 10 procedures are expected in the smallest area, the observed variation is likely to be primarily due to chance, and should be tested statistically before small-area analyses are conducted to determine the reason for the variation. The magnitude of the variation statistics here are similar to those published elsewhere, suggesting that these findings can be generalized to other settings.



Included in

Biostatistics Commons