I don't entirely agree with Qshake (that may be a first, by the way).
If you have a site which is liable to significant foundation movements (mining subsidence or the like) then you have to make a choice between adopting fully statically determinate superstructures or designing for the effect of differential settlements in continuous structures (or locating the bridge somewhere else entirely

).
Statically determinate structures basically require one deck joint for every pier. There is a school of thought that aims to minimise deck joints wherever possible (because of ongoing joint and bearing maintenance, poor road surface geometry, etc).
Continuous structures may offer some gains in structural economy and will reduce the number of deck joints, but at the risk of some sensitivity to settlements.
The choice between the two basic alternative structure types is not really all that simple. (Two of us spent about 6 weeks analysing a possible continuous design for Medway Bridge, UK [500 foot span] in the late 1950's before it had to be abandoned because of uncertainty regarding possible settlements in the founding chalk stratum. This was before there was any Code of Practice for prestressed concrete in the UK, and we were working mostly to allowable stress design - a different decision might be made today).
If you have a mining-settlement-prone site which also is subject to seismic activity and your bridge spans a known fault line, like one I worked on in New South Wales in 1977, shortly before Ingenuity's example, then all this gets just that much worse.
The amount of differential to be designed for must be considered for each particular site; this is where a good dose of sound engineering judgement is called for. If there is really good site data, and the behaviour of the foundation strata is well known, then maybe about 25% of the calculated maximum settlement would be reasonable.
If your structure is sufficiently ductile, (permitting all of the necessary redistribution of load before collapse), you MAY be able to ignore the effects of settlement in the ultimate strength condition.
However, for serviceability limit states (crack widths, etc), you should treat the adopted differential settlement effects as just another permanent load, to be combined with all other design loads, factored according to your local Code of Practice.
Just a reminder if you are designing a concrete superstructure. Since settlement is likely to be a very slow and long term effect, creep will be significant. You should base your settlement analyses on a reduced (long-term) value of Elastic Modulus for your concrete.