Figure 1.

EPDS item response probabilities and total score trajectories of the 4-class model. A. Item response probabilities were estimated for each class where Answer 4 is the highest scoring response (i.e. highest severity/frequency). B. Observed EPDS scores split according to most likely class membership following LCGA using all data available (n=1369). Each line refers to a subject and violin plots illustrate the score distributions. Solid black/red lines are within group average trajectories (slopes different from 0 at p<0·05 are in red). Classes were defined as ‘Not Depressed’ and improving (n=575, 42%; mean intercept = 3·4 points (SE=0·22, p<0·001); slope =-0·9 points (SE=0·1, p<0·001)), ‘Mild’ and improving (n=513, 37·5%; intercept = 7·5 (SE=0·5, p<0·001); slope =-0·74 (SE=0·20,p<0·001)), ‘Moderate’ and stable (n=219, 16%, intercept = 11·3 (SE=0·55, p<0·001); slope = -0·34 (SE=0·46, p=0·47) and ‘Severe’ and stable (n=62,4·5%; intercept= 16·1(SE=0·87, p<0·001); slope =0·77 (SE=0·56, p=0·174)). C. Estimated mean score at each time point in each class, colors same as in B. Dashed lines in B. and C. are references to the 12/13 and 14/15 cutoff scores for all and major depression respectively (Murray et al., 1990). EPDS, Edinburgh Postnatal Depression Scale.