Table 2.

MFQ clinical cutoff point at T3 predicted by propensity score weighted mental health service contact at T1

	Propensity score weighted only*		Propensity score weighted and adjusted for post-baseline confounds		Post-baseline confounds
	OR (95% CI)	p value	OR (95% CI)	p value
Adolescents with a T1 mental disorder: service contact vs none
Full propensity score sample (n=119)	3·70 (1·40–9·82)	0·0086	5·23 (1·47–18·63)	0·011	T2 MFQ; T3 family dysfunction, stressful life events
Common support sample (n=98)	4·36 (1·41–13·47)	0·011	7·38 (1·73–31·50)	0·0069	T2 MFQ; T3 stressful life events, family dysfunction, living with biological parents
All with T1 mental health service contact vs T1 mental disorder but no services
Full propensity score sample (n=134)	1·78 (0·81–3·92)	0·15	2·41 (0·92–6·32)	0·073	T1 MFQ;† mental health service contact after T1; T2 friendships; T3 stressful life events, living with biological parents
Common support sample (n=94)	2·36 (0·93–6·02)	0·072	2·65 (0·88–7·97)	0·085	T1 MFQ; mental health service contact after T1; T3 stressful life events, living with biological parents, family dysfunction

OR=odds ratio. T1=timepoint 1 (age 14·5 years). T2=timepoint 2 (age 16 years). MFQ=Mood and Feelings Questionnaire. T3=timepoint 3 (age 17·5 years).

^*Variables used in the propensity score model are ethnic origin, gender, pubertal status, mental health referrals aged 0–13 years, past Schedule for Affective Disorders and Schizophrenia for School-Age Children diagnosis, current behavioural diagnosis, and environmental factors (current friendships and family dysfunction, past-year stressful life events).

^†T1 MFQ was used if more strongly related to predictor and outcome than T2 MFQ.