Construct Validity

Multitrait-multimethod (MTMM) matrices are presented separately for waves one and two in Tables 9 and 10. The conceptual basis of the BDEPQ has assumed that dependence is a state rather than a trait, however, for simplicity the matrix is termed multitrait not multistate. The matrices presented do not conform to the strict definition of a MTMM matrix in that each trait has not been assessed with each method. Four traits (dependence, anxiety, depression, and neuroticism) were measured using a single method (self-report questionnaire). In addition the dependence `trait' was also measured using two other methods (CIDI interview and global self-rating).

Inspecting the coefficients in Table 9 shows that scores on the BDEPQ have a moderate to strong relationship with most other variables in the table. The correlations between BDEPQ scores and other measures of dependence indicate that the new scale has some convergent validity. However, BDEPQ scores are also correlated with BAI, BDI, and EPQ-N scores suggesting a lack of divergent validity. BDEPQ scores are more highly correlated with measures of anxiety, depression, and neuroticism than they are with CIDI diagnoses of dependence. Importantly BDEPQ scores are correlated with diazepam-equivalent dose and how frequently BZDs were used.

The information presented in Table 9 allows the BDEPQ to be compared with two other self-report measures of BZD dependence: the BWSQ and global self-rated dependence. Self-rated dependence is significantly shorter and simpler to administer than the BDEPQ however it is less reliable and it's relationship to CIDI diagnoses is not as strong. The divergent validity of self-rated dependence was shown to be greater by its correlations with BAI, BDI, and EPQ-N scores.

The BWSQ shows a pattern of relationships with other variables that is similar to the BDEPQ. BWSQ scores are related to other measures of dependence supporting convergent validity but are also correlated with BAI, BDI, and EPQ-N scores. Contrary to the predictions of simple physiological models of dependence, scores on the BWSQ are not highly related to measures of BZD dose, half-life, frequency or duration of BZD use. CIDI diagnoses of BZD dependence are not as highly related to BAI, BDI and EPQ-N scores suggesting that the correlations between these measures and BDEPQ scores is due to method variance.

Table 9: Multitrait-multimethod matrix: Wave one and phone interview

		Dependence					Anxiety	Depression	Neuroticism	Pattern of BZD use
		BDEPQ	BWSQ	SRD	DSM3R	ICD10	BAI	BDI	EPQ-N	Dose	1/2 life	Frequency	Duration
	BDEPQ	0.92
		(187)
	BWSQ	0.61	0.90
		(171)	(239)
Dependence	SRD	0.51	0.27	0.71
		(184)	(233)	(86)
	DSM3R	0.39	0.27	0.33	-
		(68)	(82)	(86)
	ICD10	0.37	0.40	0.28	0.50	-
		(68)	(82)	(86)	(88)
Anxiety	BAI	0.56	0.54	0.23	0.34	0.11	0.88
		(179)	(230)	(241)	(88)	(88)	(249)
Depression	BDI	0.50	0.55	0.16	0.20	0.13	0.52	0.84
		(187)	(238)	(254)	(87)	(87)	(248)	(247)
Neuroticism	EPQ-N	0.48	0.44	0.20	0.14	-0.00	0.53	0.59	0.88
		(181)	(232)	(244)	(82)	(82)	(240)	(250)	(253)
	BZD	0.31	0.23	0.24	0.29	0.29	0.18	0.12	0.06	-
	dose	(189)	(241)	(257)	(88)	(88)	(251)	(264)	(253)
	Half-	0.04	-0.02	-0.16	0.01	-0.06	0.06	-0.00	0.04	-0.38	-
Pattern of	life	(182)	(233)	(248)	(88)	(88)	(243)	(255)	(244)	(260)
BZD use
	Frequency	-0.25	-0.15	-0.50	-0.13	-0.05	-0.08	-0.07	-0.03	-0.22	0.15	-
		(188)	(238)	(256)	(87)	(87)	(246)	(259)	(248)	(264)	(255)
	Duration	0.06	0.03	0.24	-0.11	-0.02	-0.07	-0.09	-0.06	-0.04	0.03	-0.26	-
		(188)	(234)	(252)	(88)	(88)	(242)	(259)	(244)	(259)	(253)	(258)

Notes: The diagonal contains reliability coefficients which are coefficient alphas except for self-rated dependence which is a two to three week test-retest correlation. BDEPQ = Benzodiazepine Dependence Questionnaire, BWSQ = Benzodiazepine Withdrawal Symptom Questionnaire, SRD = Self-Rated Dependence, DSM3R = a current DSM3R diagnosis of Benzodiazepine Dependence made using the CIDI, ICD10 = a current ICD10 diagnosis of Benzodiazepine Dependence made using the CIDI, BAI = Beck Anxiety Inventory, BDI = Beck Depression Inventory, EPQ-N = Neuroticism scale of the Eysenck Personality Questionnaire, Dose = high estimate of diazepam-equivalent dose of BZDs, Half-life = dose weighted average half-life of BZDs taken, Frequency = how often were BZD taken, and Duration = for how long were BZDs taken.

Table 10: Multitrait-multimethod matrix: Wave two
 

		Dependence				Anxiety	Depression	Sleep Quality	Pattern of BZD use
		BDEPQ	BWSQA	BWSQB	SRD	BAI	BDI	PSQI	Dose	1/2life	Frequency
	BDEPQ	0.92
		(170)
	BWSQA	0.55	0.88
		(72)	(112)
Dependence
	BWSQB	0.50	0.72	0.83
		(168)	(109)	(240)
	SRD	0.47	0.21	0.22	-
		(168)	(110)	(236)
Anxiety	BAI	0.57	0.55	0.60	0.19	0.87
		(165)	(106)	(234)	(233)	(240)
Depression	BDI	0.40	0.55	0.45	0.14	0.61	0.88
		(163)	(101)	(227)	(225)	(225)	(229)
Sleep Quality	PSQI	0.28	0.27	0.29	0.09	0.36	0.39	0.83
		(135)	(86)	(186)	(186)	(187)	(179)	(189)
	BZD	0.10	0.17	0.07	0.26	0.14	0.09	-0.01	-
	dose	(170)	(110)	(240)	(246)	(237)	(229)	(189)
Pattern of	Half-	0.05	-0.01	-0.01	-0.13	-0.07	-0.03	-0.04	-0.47	-
BZD use	life	(160)	(78)	(202)	(207)	(200)	(194)	(163)	(209)
	Frequency	-0.20	-0.07	-0.21	-0.63	-0.11	-0.09	-0.07	-0.38	0.23	-
		(167)	(107)	(235)	(242)	(232)	(224)	(187)	(243)	(208)

Notes: BDEPQ = Benzodiazepine Dependence Questionnaire, BWSQA = Benzodiazepine Withdrawal Symptom Questionnaire during reduction or cessation of BZDs, BWSQB = Benzodiazepine Withdrawal Symptom Questionnaire when dose of BZDs was stable, SRD = Self-Rated Dependence, BAI = Beck Anxiety Inventory, BDI = Beck Depression Inventory, PSQI = Pittsburgh Sleep Quality Index, Dose = high estimate of diazepam-equivalent dose of BZDs, Half-life = dose weighted average half-life of BZDs taken, and Frequency = how often were BZD taken.

• Confirmatory factor analysis of MTMM matrices

10th April 2001