Embedded survey study harms colorectal cancer screening attendance: Experiences from Finland 2010 to 2015

Organized screening programmes provide appealing settings for epidemiological surveys, but such additional investigations may reduce willingness of the target population to participate in screening. A population-based colorectal cancer (CRC) screening programme using biennial fecal occult blood testing (FOBt) for 60–69 year olds has been running in Finland since 2004. ¹ The ‘Psychosocial effects of colorectal cancer screening study’ was incorporated into the CRC screening programme in 2010–2012. Lifestyle and quality of life were assessed through a self-administered postal questionnaire at baseline one year before the first screening invitation, and again after two years.^2,3 We previously reported that participation after the first screening invitation was lower among those surveyed (56.6% vs. 60.2%). ² A similar baseline phenomenon has been reported from England, where CRC screening uptake was significantly lower in those who received a questionnaire with their FOBt kit than in those who did not (48.6% vs. 53.5%). ⁴ The purpose of this study was to examine CRC screening behaviour after subsequent invitations, to determine if the effect of surveys on screening attendance is long term.

The study population consisted of individuals born in 1951, residing in municipalities involved in the CRC screening programme and, therefore, eligible to be invited to CRC screening in 2011 (n = 31,951). In 2010 (i.e. in advance of screening), we sent a random sample of every sixth subject (n = 5312) a 7-paged lifestyle questionnaire, and to another random sample of every sixth subject (n = 5336) a 10-paged lifestyle and quality of life questionnaire (Figure 1). In 2011, 31,484 individuals of the original cohort (467 had moved to a non-screening municipality, or died) were independently randomized (1:1) to be invited for CRC screening (n = 15,748) or to the control group (n = 15,736), according to the invitation schema of Finnish CRC screening programme. ¹ In 2012, the survey cohort (n = 10,375; 273 had either died, refused 2010, or had unknown address) received a follow-up questionnaire with the same contents as in 2010. In 2013, 15,041 individuals were invited to CRC screening for the second time, following the original screening randomization, irrespective of their attendance at the previous screening round. The third screening round was in 2015 (14,096 individuals invited), with no questionnaires sent the previous year. OPEN IN VIEWER

To assess the association between CRC screening attendance and inclusion in the survey study, two-way tables were used. Screening attendance proportions were compared between those who were and were not sent the questionnaires, using a two sample test of proportions (z-test) with a 5% significance level. Attendance was also compared by sex, marital status, urbanization level of residential area and geographical area of residence. The study received ethical approval from the Helsinki and Uusimaa Hospital District Ethics Committee (15 December 2009, 420/13/03/00/2009).

In our sample, the overall CRC screening attendance was 62.1% in 2013. In those who had been sent the surveys earlier (in 2010, and again in 2012), the attendance was 58.4% (58.5% for short and 58.3% for long questionnaire), and in those not subjected to survey interventions it was 63.9% (p = 0.001) (see Table 1). Surveys were associated with significantly reduced attendance in both men and women, in married and non-married, and in urban and rural areas.

OPEN IN VIEWER

Table 1.

Colorectal cancer screening attendance 2013 and 2015 in those sent and not sent study questionnaires 2010 and 2012.

	Screening 2013						Screening 2015
	No Q		Q 2010 and 2012		Difference		No Q		Q 2010 and 2012		Difference
	inv	att	inv	att			inv	att	inv	att
	n	%	n	%	%	p	n	%	n	%	%	p
All	10,121	63.9	4920	58.4	−5.5	0.001	9499	66.2	4597	61.7	−4.5	0.006
Sex
Men	5013	56.1	2439	50.3	−5.8	0.009	4680	58.7	2258	53.8	−5.0	0.027
Women	5108	71.6	2481	66.4	−5.2	0.017	4819	73.5	2339	69.4	−4.2	0.050
Marital status a
Married	6369	68.1	3067	62.3	−5.7	0.005	5997	70.3	2869	66.2	−4.2	0.033
Divorced/single/widow	3741	57.0	1840	52.0	−5.0	0.039	3461	59.6	1709	54.4	−5.2	0.041
Municipality type
Urban	6026	62.8	2925	57.5	−5.3	0.009	5624	65.2	2758	61.2	−4.0	0.043
Rural	4095	65.5	1995	59.8	−5.8	0.017	3842	68.0	1839	62.3	−5.6	0.024
Area
Southern	1858	63.8	886	56.1	−7.7	0.030	1710	65.9	816	60.7	−5.2	0.109
Central	3553	63.9	1735	60.3	−3.6	0.109	3195	66.2	1546	62.5	−3.7	0.117
West-coastal	1957	64.2	951	56.9	−7.3	0.032	1889	66.9	925	61.2	−5.7	0.077
Eastern	1539	64.8	761	58.9	−5.9	0.091	1493	67.0	735	61.6	−5.3	0.118
Northern	1214	62.5	587	58.1	−4.4	0.189	1179	65.7	568	62.2	−3.6	0.241
CRC screening attendance on previous round b
Yes	6168	91.8	2828	89.1	−2.7	0.120	6121	94.0	2715	93.7	−3.2	0.444
No	3940	20.4	2079	16.6	−3.8	0.080	3345	15.8	1875	15.5	−2.9	0.460
Result of CRC screening on previous round
Negative	5947	92.5	2737	89.6	−2.9	0.104	5946	94.3	2643	94.3	0	0.475
Positive	221	71.5	91	73.6	2.1	0.568	175	82.9	72	70.8	−12.0	0.195

In 2015, screening attendance was still significantly lower in those who had received the questionnaires in 2010 and 2012 than among those who had not. Attendance was 61.7% (62.0% for short and 61.4% for long questionnaire) and 66.2%, respectively (p = 0.006).

Being subjected to a questionnaire study before screening invitation resulted in reduction in CRC screening attendance. We found that the detrimental effect of questionnaires on screening attendance was long term. The effect augmented with repeated surveys; the reduction in screening attendance in those subjected to questionnaires was greater in the second (2013) than in the first screening round (2011). The effect seemed to continue so that in the subsequent third screening round with no questionnaires the previous year, the attendance still did not reach the level of the group never exposed to questionnaires. Results were similar for men and women, by marital status, and by area. Unlike in 2011, the length of the questionnaire did not have an effect on screening attendance, except in Northern Finland, where screening attendance was decreased only among the long (10-paged) questionnaire recipients.

Screening attendance for CRC screening in the survey study population remained low in 2015, even though one could imagine that the detrimental effect of burdensome questionnaires would dilute after not receiving the questionnaires anymore. We believe that the long-term effect is partly due to attendance pattern, i.e. previous screening behaviour (especially participation in the first invitation round) is a predictor of subsequent participation. This has been observed previously in the context of breast cancer screening^5–7 and CRC screening.^8–10 Those disturbed by the original study questionnaires in 2010 established a pattern of non-participation, which did not abate with time.

Jäntti et al. ³ examined, in this same population, whether screening allocation in 2011 had any impact on follow-up survey response. Among CRC screening invitees in 2011 response to the survey increased in 2012 compared with that in 2010, whereas among controls (not invited for screening) the response to the survey decreased slightly. These results support the idea that it is the nature of the contact, not just the contact, which is essential. The screening invitation can be perceived as a service or offer, which could make the attitude towards the subsequent survey more positive. Rather laborious surveys, instead, can be seen as a burden, and the rejection of this burden also affects future screening.

Adequate attendance is fundamental for successful population screening, and careful monitoring and optimization of attendance must be ensured in any screening programme. Surveys are necessary to obtain information that health care or other registers cannot provide. When embedding surveys in cancer screening, questionnaire design should aim to include only relevant issues, in a compact form, and to present the questionnaire as advantageous for the recipient. There is an ethical issue in employing a research intervention endangering engagement; the benefit gained should outweigh the harm caused by reduced attendance.