Int. J. Cancer: 122, 1618–1623 (2008) ' 2007 Wiley-Liss, Inc.
Toward understanding nonparticipation in sigmoidoscopy screening for colorectal cancer 3 Johannes Blom1*, Li Yin2, Annika Liden3, Anders Dolk1, Bengt Jeppsson4, Lars Pahlman ˚ , Lars Holmberg5 and Olof Nyren2 1 Division of Surgery, Department for Clinical Science, Intervention and Technology, Karolinska Institutet at Karolinska University Hospital, Huddinge, Stockholm, Sweden 2 Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden 3 Department of Surgical Sciences, Akademiska Sjukhuset, Colorectal Unit, Uppsala University, Uppsala, Sweden 4 Department of Laboratory Medicine and Surgery, Department of Clinical Sciences, Lund University, Malm€ o University Hospital, Malm€ o, Sweden 5 Division of Cancer Studies, King’s College, London, United Kingdom
Understanding the reasons for nonparticipation in cancer screening may give clues about how to improve compliance. However, limited cooperation has hampered research on nonparticipant profiles. We took advantage of Sweden’s comprehensive demographic and health care registers to investigate characteristics of all participants and nonparticipants in a pilot program for colorectal cancer screening with sigmoidoscopy. A population-based sample of 1986 Swedish residents 59–61 years old was invited. Registers provided information on each individual’s gender, country of birth, marital status, education, income, hospital contacts, place of residence, distance to screening center and cancer within the family. Odds ratios (ORs) with 95% confidence intervals (CIs), modeled with multivariable logistic regression, estimated the independent associations between each background factor and the propensity for nonparticipation after control for the effects of other factors. All statistical tests were 2-sided. Being male (OR 5 1.27, 95% CI 5 1.03–1.57, relative to female), unmarried or divorced (OR 5 1.69, 95% CI 5 1.23–2.30 and OR 5 1.49, 95% CI 5 1.14–1.95, respectively, relative to married) and having an income in the lowest tertile (OR 5 1.68, 95% CI 5 1.27–2.23, relative to highest tertile) was associated with increased nonparticipation. Living in the countryside or in small communities and having a documented family history of colorectal cancer was associated with better participation. Distance to the screening center did not significantly affect participation, nor did recent hospital care consumption or immigrant status. To increase compliance, invitations must appeal to men, unmarried or divorced people and people with low socioeconomic status. ' 2007 Wiley-Liss, Inc. Key words: mass screening; colorectal neoplasms; sigmoidoscopy; patient participation; registers
To be effective on a population level, screening programs have to be widely accepted. To minimize nonparticipation, it is important to understand the reasons for nonattendance and to identify groups that would benefit from extra recruitment efforts. Further, understanding of selection forces is necessary for valid interpretation of the effects of screening. Investigators of nonparticipation, whether in screening, public health campaigns or scientific investigations typically have to resort to interviews or questionnaires. As the motivation among nonparticipants generally is low, such studies are characterized by low response rates and questionable data quality. For example, in a recently published Norwegian study concerning a colorectal cancer screening program, only 11% of the nonparticipants responded to a follow-up questionnaire.1 To our knowledge, no comprehensive model explaining nonparticipation in cancer screening has yet gained widespread acceptance. The scarcity of unbiased background information among nonparticipants may have contributed to this lack of a firm conceptual basis for the understanding of underlying mechanisms. We took advantage of 8 national registers of demography and health care and collected information about participants and nonparticipants in a population-based pilot study of screening with flexible 60 cm sigmoidoscopy. Publication of the International Union Against Cancer
Material and methods The pilot colorectal cancer screening study has been described previously.2,3 Briefly, from the population register, we randomly selected 2,000 men and women 59-61 years of age residing in 2 areas in central and southern Sweden. Thirteen study subjects were excluded because they had moved out of the study areas and one because there was no match in the multiple record linkages owing to an erroneous national registration number (see later). No other exclusion criteria were used. A mailed invitation of approximately 1 and a half pages included a brief account of the descriptive epidemiology of colorectal cancer, a paragraph about the potential for reducing colorectal cancer mortality with screening and a description of the aims and design of the study. We also described the actual sigmoidoscopy examination and the options in case of a positive finding. Moreover, a questionnaire was included concerning, e.g., the invitee’s occupation, physical activity and diet. To make the appointment for sigmoidoscopy, half of the invitees were randomly selected to be called up by a nurse and half were asked to call themselves. All sigmoidoscopies were free of charge for the invitees and were done on an outpatient basis at 2 hospitals, one in each area. In Uppsala (in central Sweden), the catchment area was larger than in Lund (in southern Sweden); the maximum distance to the hospital was 90.0 versus 25.1 km. The design of the study of nonparticipants is illustrated in Figure 1. A computer file with the invitees’ national registration numbers—unique personal identifiers assigned to all Swedish residents and used in all of the registers utilized in this study—together with information about inviting hospital, invitees’ gender and participation status was sent to Statistics Sweden. There, data on country of birth (Sweden, other Nordic countries, Europe excluding the Nordic countries or outside Europe) and marital status were obtained from the Total Population Register.4 Income data were collected from the Register of Income and Wealth,5 based on the tax returns submitted to the National Tax Board of Sweden. Information on level of education was taken from the Register of Education,6 annually updated with the highest degree of education of all individuals in Sweden between 15 and 74 years of age. The Geodatabase95 provided information on the exact location of the place of residence and the character of the area. Further, since all domiciles have a map coordinate, we could use the coordinate in a GIS (Geographic Information System) program to calculate the distance to the screening center. The national registration numbers of all parents, siblings and children of the invitees were obtained from the Multigeneration Register.7 This register links all Swedish Grant sponsors: The Swedish Cancer Society, Karolinska University Hospital, Karolinska Institutet, Stockholm, Sweden. *Correspondence to: Division of Surgery, K53, Karolinska University Hospital, Huddinge, 141 86 Stockholm, Sweden. Fax: 146-8 585 823 40. E-mail:
[email protected] Received 25 March 2007; Accepted after revision 13 August 2007 DOI 10.1002/ijc.23208 Published online 6 December 2007 in Wiley InterScience (www.interscience. wiley.com).
SIGMOIDOSCOPY SCREENING FOR COLORECTAL CANCER
1619
FIGURE 1 – Structure of the nonparticipation study of sigmoidoscopy screening for colorectal cancer with registers and information collected.
individuals born from 1932 onward to their parents (biological or adoptive) and, thus also to their siblings. Record linkage with the Swedish Cancer Register,8 maintained by the National Board of Health and Welfare and in operation since 1958, provided dates and diagnostic codes according to the 7th revision of the International Classification of Diseases (ICD-7) for all cancer occurrences in the invitees and their first degree relatives. Likewise, information on hospital care during the preceding 5 years among invitees was collected from the Hospital Discharge Register.9 After having compiled all data in 1 file, Statistics Sweden deidentified the entire dataset before delivery to the investigators. The study was approved by the Regional Research Ethics Committees at Karolinska Institutet and Uppsala and Lund University. Statistical analyses Logistic regression was performed to model odds ratios (ORs) with 95% confidence intervals (CIs) of nonparticipation associated with investigated background factors, with and without adjustment for other cofactors. All statistical tests of significance were 2sided. Continuous variables were categorized before any analysis of effect. Income was categorized in tertiles, number of inhabitants in the area of residence into