Standard voltage in Switzerland is 230VAC, not 220VAC. The data sheet shows in Figure 2 that the device more or less clamps its voltage at a level small compared to what you are dealing with; so you basically only need to consider the operating current you want to achieve. As others mentioned, you can use a capacitive dropper for that. However, you still incur the same voltages and currents for operation, it's just that they are out of phase and thus the capacitor will not heat up like the resistors would. Input threshold current is 3.11mA max (I am not considering AC here because I am lazy: essentially you have to stay above this threshold when taking into account the ripple after rectification and integration on the capacitor across 2&3); if you subtract a generous 10VAC for voltage drop (and assume a minimum guaranteed mains voltage of, well, 220V), you get a capacity of 3.11mA/(210V*50Hz) = 300nF. You'll need a foil capacitor of that size (not electrolytic) and about 400V voltage tolerance.
One problem is inrush current if the voltage is switched on outside of zero crossings. There is a maximum rating for surges of 3ms of 140mA, so you can put a resistor of 340V/140mA (the peak voltage of 240Vrms is 340V), something like 2.4kOhm or so in series. In sustained operation, it would waste 2.4kOhm*(3.11mA)^2 = 23.2mW which is still considerably less than what you started out with. The time constant RC will then be 2.4kOhm*300nF = 0.72ms so you stay reasonably safely below the surge duration of 3ms.
Of course you need to redo everything with actually available values and take into account their tolerances for the worst case, too.
While you could "save" even more energy by using an inductor for inrush surge limitation, LC combinations for taking care of voltage drops are a lot less well-behaved than RC so I wouldn't bother.
See how fiddly things get once you try avoiding the energy loss in a series resistor?
