Table 4.

Associations^a between medical necessity and other reasons for referral, stratified by GP sex

Reasons for referral	Referrals by male GPs			Referrals by female GPs			All referrals
	Medically necessaryb, = 0	Less medically necessaryc, = 1	OR (95% CI)	Medically necessaryb, = 0	Less medically necessaryc, = 1	OR (95% CI)	ORa (95% CI)
	n = 241	n = 135	n = 376	n = 142	n = 77	n = 219	n = 595
Perceived deficient medical knowledge			4.06d (1.89 to 8.72)			1.25 (0.60 to 2.60)	2.24d (1.34 to 3.76)
Agree, %	10.0	27.4		28.9	31.2
Disagree, %	90.0	72.6		71.1	68.8
To reassure the patient			13.44d (4.77 to 37.88)			1.35 (0.49 to3.70)	4.12d (2.23 to 7.63)
Agree, %	2.9	28.2		15.5	26.0
Disagree, %	97.1	71.8		84.5	74.0
Patient preference			3.28d (1.81to 5.96)			2.08 (0.97 to 4.49)	2.66d (1.69 to 4.19)
Agree, %	29.1	61.5		46.5	53.3
Disagree, %	70.9	38.5		53.5	46.8

^aAnalysed by multivariable multilevel logistic regression, allowing for clustering at GP level, adjusted for the reasons ‘perceived deficient medical knowledge’, ‘to reassure the patient’, ‘patient preference’, and ‘common practice’.

^bThe highest agreement level of medically necessary: ‘corresponds very well.

^cThe three lowest agreement levels of medically necessary ‘does not correspond’, ‘corresponds to a limited extent’, and ‘corresponds fairly well’ were merged.

^dP<0.05. OR = odds ratio.