There is a toolbox in SPM called "Artifact Detection Tool" (ART) (you can download it from NITRC), which dose the same thing as fsl_motion_outliers.
It can read .par file generated from FSL and your preprocessed images to identify outliers. Then, it will automatically generate two .txt files (one includes only outliers and the other includes both motion parameters and outliers) and you can input .txt in the SPM model as multiple regressors. Actually, the .txt file it generates is exactly what Mark described, which contains a big matrix with time points of outliers as "1". However, outliers from these two tools (fsl_motion_outliers and ART) may be different because the criteria used to identify outliers are different.
As to the potential issues of using preprocessed images from FSL and inputing them in SPM for stats, you really need to consult SPM experts.