Preferably, you need a Reference population, e.g., results for those who were treated With the "Old" treatment. There are several online sample size calculators/descriptions for one sample continous variable compared With a Reference mean or CI.
Much better would be improving the design to an RCT and use sample size calculations for two Groups (the old treatment expecatations (Guess you have some data) versus the New treatment (expectations or results from a pilot study).