> the participants in the Q3 and Q4 groups exhibited lower cognition scores (Q3:β = -0.180, 95%CI -2.849~-0.860; Q4:β = -0.164, 95% -2.611~-0.666)

This seems wrong. If "β" is the estimate here (not sure), it should be inside the confidence interval, but is way outside...

Beta is dimensionless, in terms of standard deviations. In their results tables, you can find their unit versions under the column "B". These lie in the centers of the CIs, as expected.