Skip to main content
. Author manuscript; available in PMC: 2016 Jul 1.
Published in final edited form as: J Psychiatr Res. 2015 May 5;0:95–104. doi: 10.1016/j.jpsychires.2015.04.021

Table 4.

Odds ratios, 95% confidence intervals, and probability values from cumulative logistic models of EPDS item 10: “The thought of harming myself has occurred to me” regressed on anxiety/disturbed sleep symptoms stratified by abuse measure, and the interaction between anxiety/disturbed sleep symptoms and abuse measure.

Main effect of symptoms (anxiety or
disturbed sleep) when specified abuse
is present
 Main effect of symptoms (anxiety or
disturbed sleep) when specified abuse
is absent
 Interaction
between
symptoms &
abuse
Measure N OR 95% CI p N OR 95% CI p N p
Anxiety symptoms
   Physical abuse as
a child		 102 1.011 (0.844,
1.211)		 0.9088 514 1.114 (1.009,
1.230)		 0.0327 616 0.3565
   Sexual abuse as a
child		 129 1.093 (0.920,
1.298)		 0.3112 487 1.090 (0.986,
1.205)		 0.0937 616 0.9767
   Physical abuse as
an adult		 186 1.147 (0.964,
1.365)		 0.1215 430 1.077 (0.973,
1.191)		 0.1526 616 0.5423
   Sexual abuse as
an adult		 68 0.980 (0.768,
1.250)		 0.8723 548 1.106 (1.007,
1.214)		 0.0348 616 0.3691
Symptoms of
sleep disturbance
   Physical abuse as
a child		 102 1.041 (0.844,
1.284)		 0.7080 511 1.148 (1.024,
1.286)		 0.0176 613 0.4313
   Sexual abuse as a
child		 129 1.043 (0.858,
1.268)		 0.6734 484 1.161 (1.033,
1.305)		 0.0122 613 0.3572
   Physical abuse as
an adult		 184 1.101 (0.933,
1.299)		 0.2552 429 1.154 (1.018,
1.308)		 0.0255 613 0.6489
   Sexual abuse as
an adult		 67 1.034 (0.799,
1.340)		 0.7981 546 1.144 (1.028,
1.274)		 0.0141 613 0.4699

How to read this table: thoughts of self-harm was regressed on each symptom measure (anxiety or disturbed sleep) as a bivariate model within each subsample of abuse (physical or sexual, as a child or adult) and then again as a multivariate model including the interaction between each symptom measure and abuse measure. For example, anxiety symptoms are positively associated with thoughts of self-harm among women who were physically abused as a child but the association is statistically non-significant (OR=1.011, p=0.9088). The same association is significant among women who were not physically abused as a child (OR=1.114, p=0.0327). However, in the full model the interaction between anxiety symptoms and physical abuse as a child did not reach statistical significance (p=0.3565).