Carefully characterized DMMs are usually specified for 24h, 90 days and 1 year accuracy, within a certain temperature range. Long term drift is rarely specified, but in good designs (eg. Fluke 87), this can be expected to be close to the 1 year figure, since drift tends to level off with time as SeanB mentions. By far the most drift will be within the first year.
In the case of Uni-T, you'd be lucky if it meets the 1 year spec, all bets are off for long term stability. People with the equipment to test DMM stability are not usually interested in cheap handhelds. Cal labs don't have much data on them either, since people rarely send them in for calibration.