Sample size and power considerations for ordinary least squares interrupted time series analysis: a simulation study