ABSTRACT
Scott test is a simple, rapid, and low-cost preliminary test used extensively to suggest the presence of cocaine in drug seizures due to the development of a blue color. However, the presence of cutting agents can compromise the test result and may suggest the presence of cocaine when the drug is absent. This study evaluated the frequency of these results and the spectral behavior and color development of false positive substances. Furthermore, this study proposes the application of the partial least squares discriminant analysis (PLS-DA) method associated with photographic images obtained by a smartphone camera to increase the selectivity of the Scott test. For the first time, a study considered a diverse set of 173 samples, 126 of them from police drug seizures. The multivariate model presented a 100% hit rate for both the set of training samples and the test set. Thus, zero false positive (classified as positive in the absence of cocaine) and false negative (negative in the presence of cocaine) rates were achieved. Therefore, the proposed methodological alternative is promising, simple, low-cost, portable, and considerably increases the assertiveness of the preliminary test for researching cocaine.