From a quick scan of the article the chart is based on figure 1. p-value for the more disadvantaged group is 0.017 for the less disadvantaged it's 0.02. There's a 3rd group, 'medium disadvantaged', which has a horizontal line.

The trend is more obvious when the points for the individual groups are separated out, but it does seem to me that a small number of points at the extremes might account for slope - if you removed one or two key points from each group the slopes might go away.