Table 5.

Predicting chlamydia sample return.

	Sample returned in 30 days from day of kit request Round 1		Sample returned in 30 days from day of kit request Round 2		Sample returned in 30 days from day of kit request Round 1 + 2
	Odds ratio	z-stat	Odds ratio	z-stat	Odds ratio	z-stat
Incentive structures
GBP 5 voucher	1.03	0.13	1.35	1.55	1.20	1.31
GBP 10 voucher	–	–	1.28	1.31	1.20	1.04
Lottery EV GBP 5	0.96	−0.21	–	–	1.06	0.30
Choice EV GBP 5	0.96	−0.19	–	–	1.03	0.15
Choice EV GBP 10	–	–	1.06	0.32	1.02	0.13
GBP 5 endowment	1.12	0.53	–	–	1.17	0.79
GBP 10 endowment	–	–	0.89	−0.63	0.87	−0.82
GBP 5 charity	0.88	−0.60	–	–	0.99	−0.06
Socio-demographic characteristics
Male	1.03	0.21	0.83	−1.42	0.95	−0.64
IMD	0.99a	−3.23	0.99	−1.28	0.99a	−3.11
20–24 years	1.49a	3.16	1.22	1.41	1.35a	3.21
Black	0.88	−0.32	0.90	−0.21	0.84	−0.53
Asian	0.48c	−1.81	0.95	−0.09	0.61	−1.56
Mixed ethnicity	1.13	0.31	1.05	0.13	1.11	0.39
Sexual history
No new partner in the last 3 months	0.87	−0.85	0.84	−1.04	0.87	−1.29
Not had two or more partners in last 12 months	1.32	1.61	0.74c	−1.74	1.00	0.00
Yes, positive chlamydia test in the last 12 months	1.54a	2.80	1.13	0.80	1.34a	2.56
Yes, negative chlamydia test in the last 12 months	1.25	1.02	0.74	−1.13	0.94	−0.34
No condom use when last had sex	0.68b	−2.16	1.40c	1.85	0.98	−0.21
Same sex partner in the last 3 months	0.47b	−2.02	1.07	0.12	0.62c	−1.66
Partners of both sexes in the last 3 months	0.43b	−2.23	0.16a	−4.77	0.27a	−5.05
Study round 1	–	–	–	–	0.93	−0.49
Constant	2.76a	3.91	2.43a	2.88	2.51a	4.40
Number of observations	1489		1499		2988
Wald χ2 (prob > χ2)	82.71(0.000)		160.88 (0.000)		197.89 (0.000)
Log pseudolikelihood	−852.2		−788.8		−1678.1
Pseudo r2	0.046		0.129		0.067

Note: Reference categories are control group, age 16–19 years, white, new partner in last 3 months, no Chlamydia test in the last 12 months, condom use when last had sex, different sex partner in last 3 months and Study Round 2. Specifications also control for Strategic Health Authority in which individuals live, a proxy for region of England and ‘unknown’ responses to all independent variables (‘no answers’ and text message requests therefore questions were not asked). All models clustered by first three letters of the postcode with varying numbers of clusters based on sample size.

^aSignificant at 1%.

^bSignificant at 5%.

^cSignificant at 10%.