Would you be happy to lose one of your other filters in order to create a 4th order filter? That’s an easy implementation possibility. You’d just gang the filter coefficients to its neighbour, and they’d already be cascaded. To make a 3rd order filter, a 1st order filter could instead be calculated, and that would run at no extra DSP cost. I am of course making the assumption that we’re using biquad filters here 
eg: you want a 4th order HPF:
your MF bell filter would disappear, and its coefficients would mirror that of the LF filter, creating a cascade of two biquad filters.
eg: you want a 3rd order LPF:
your HM bell filter would disappear, and its coefficients would be recalculated for a first order LPF with the same frequency parameter as the HF filter. This creates a cascade of a 1st and 2nd order filter.
While on the subject, I don’t think you need higher order filters to create shelving filters with an adjustable transition bandwidth (or slope).
EDIT: After a quick look at this in MATLAB, the downside of cascading two 2nd order filters would be that the centre frequency would sit at -6dB, not -3dB, which is what standard 4th order Butterworth polynomials would give. Same applies between a 1st/2nd order cascade and 3rd order Butterworth polynomials.
