You can use the three consecutive weeks, but your choice of base period weeks is not consistent with the model you want to execute. Let's try another tack.
If your goal is to forecast the number of visits per week, I would suggest applying a factor to the most recent week. You were coming up with a moving average factor, but them applying that to a base period that had changed by more than 10%. The math didn't make sense.
How about if you applied your moving average against the moving average of the visits per week?
Your average for the three weeks is (151+142+135)/3=143. That has already changed. Applying the average percentage change you computed, -3.81%, might imply that the week four target should be (143*(1-3.81%)=138. That forecasts the fourth week goal, based on the average of the prior three weeks, with a weighted average based on the prior three weeks. that way, you are forecasting a declining trend, and now can anticipate the change, and take management steps in advance of additional declines.
Financial modeling is a tool to analyze data, and create information estimates for future decision making. If you forecast low visits, this might be a good week to schedule no appointments on Friday, for example.