Cancer patients’ participation in population-based health surveys: findings from the HUNT studies

After adjustment for age and sex CaPts’ likelihood to participate was significantly reduced in HUNT1 and HUNT2, but not in HUNT3. However, any differences in participation rates were small with the largest differences emerging in CaPts diagnosed within the last 2 years prior to a HUNT survey invitation. Respectively, one-third and approximately 10 % of the CaPts from HUNT1 participated in HUNT2 and HUNT3, the comparable figures more than tripled for NonCaPers.

During the last decades health survey participation rates have generally varied between 60 and 80 %, but have declined over time. [8] This trend was also demonstrated for all invited persons in the HUNT surveys [13], and confirmed for CaPts in the present study. Morton et al. [18] list some factors which may explain the variability of participation rates: differences in method of contact (mail, telephone), use of incentives, mandatory versus optional collection of biological material, and variability of the disease to be studied. The length and eventual sensitive nature of questionnaires may also be reasons why invited persons decline to participate [4], together with the opportunity of extensive data linkage. Full time work and rural residence have also been discussed as reasons for non-participation in general [11].

Analyses of multiple health surveys including the HUNT surveys have documented elevated non-participation rates for identifiable subgroups as socio-economically deprived individuals [1, 2, 6], and for those with reduced physical and mental health [3–6, 8, 9]. Associations with non-participation have been documented for cardiovascular diseases [9], chronic lung diseases [6, 19], but not for diabetes mellitus [10, 13]. The participation rates among males are often lower than among females [4, 19, 20] though Jacobsen et al. observed the opposite in the Rochester Epidemic Project [21]. In the HUNT surveys males, in particular younger men, had particularly low participation rates [13]. However, among CaPts participating in the HUNT studies the sex differences were much less, with more male than female CaPts participating in HUNT3.

The percentages of participating CaPts seem particularly low in reports if the patients are invited by cancer registries without involvement of the institution or the medical team responsible for the patients’ treatment. In Geller et al.’s study only 33 % of the invited cancer survivors participated [11], this percentage being in agreement with that observed by Jegou et al. [12]. These authors identified patients with colorectal cancer in a regional cancer registry and invited them to participated in future research. Only 37 % of all persons with rectal cancer were willing to be included in such a research-based cancer registry. This contrasts to the experience of Downing et al. [22] of a participation rate of 63 % among British persons with colorectal cancer. These authors included the names of the hospital director and of the chair of the patient’s responsible medical team in the head of the invitation letter. With this background, participation rates of 90 % (HUNT1), 70 % (HUNT2) and 58 % (HUNT3) among invited CaPts are surprisingly high. Among surveys specifically inviting CaPts [11, 12] non-awareness of the malignant diagnosis may be a reason for non-participation. In Nord et al’s [23] analysis of HUNT2, 26 % of non-participating CaPts identified by the CRN, claimed never to have been diagnosed with cancer.

Our findings support the suggestions by Langhammer et al. [9] of no or only minor differences between CaPts and NonCaPers as to participation in the HUNT studies, even though these authors’ observations were based on non-participants’ self-report by a questionnaire sent to them 9 months after the original invitation. In contrast, our study uses the CRN as basis of diagnostic validation. Therefore we are able to present more patient details than that study.

In agreement with Ness et al. [20] we observed the influence of the diagnostic interval on CaPts’ participation rate. In their US survey reduced “participation limitations” were reported among cancer patients whose cancer diagnosis preceded the survey for >5 years as compared to shorter intervals. In our study the reduced participation rates in CaPts with a diagnostic interval of <2 years may be explained by particularly poor health due to an aggressive malignancy or treatment-related side effects. Further, recently diagnosed cancer patients regularly have multiple contacts with the health care services during the first 2 post-diagnosis years, and many of them have not completed their initial treatment before 2 years have elapsed since diagnosis. They probably anticipate few benefits from participation in population-based HUNT surveys, and they therefore may be less motivated to participate. Researchers using the HUNT studies, and probably also other population-based health surveys, should therefore be aware of possible participation bias due to short diagnostic interval in cancer patients. Any statements comparing CaPers with NonCaPers may be less valid for cancer survivors less than 2 years after diagnosis.

Large differences in participation rates of CaPts and NonCaPers became more evident in the longitudinal assessment of participation. Approximately 60 % of the NonCaPers, but only one-third of the CaPts participated in two HUNT surveys, the comparable differences for attendance in all three surveys being respectively 37 and 11 %. The differences in repeated participation among persons with or without cancer leave comparisons of CaPts and NonCaPers problematic in longitudinal studies which aim to assess cancer patients’ health in subsequent surveys. On the other hand, the HUNT surveys allow valid analyses of health changes in long-term cancer survivors, if investigators are aware of the participation bias in the relatively few CaPts with repeated participation: Persons with a prior cancer diagnosis participating in more than one health survey are probably particularly healthy, the majority of them being without tumor activity of their first life-time cancer.

Our study has several limitations. First, the definition of cancer patients and thus the estimation of their participation rates can be debated since patients with a first life-time malignant diagnosis within the 2 months preceding the date of diagnosis as reported by the CRN were not included. The aim of our study was, however, to compare participation rates among NonCaPers with that of CaPts who were aware of their malignant diagnosis. Clinicians know that up to 2 months may elapse between performance of a diagnostic biopsy and patient information. In our view this justifies the exclusion of persons with a first life-time malignant diagnosis within 2 months from the cohort of CaPts, when participation rates were calculated. Retrospectively we could identify 185 persons who most probably were diagnosed with their first life-time cancer within 2 months prior to a HUNT invitation (HUNT1 47, HUNT2 61, HUNT3 77). Of these 105 individuals (57 %) did not participate in the relevant HUNT survey. From these figures we can stipulate that the non-participation rate for CaPts diagnosed within the two preceding would have been slightly higher, but without any principal change, if these patients had been reported as CaPts. Second, though we present results from three population-based studies the findings may not be valid for surveys in countries with different health care services. In countries without easy and non-expensive access to health care services, more persons will possibly participate in public health studies, than this was the case for the HUNT surveys. Third, we have only considered the first life-time cancer in CaPts whose records from the CRN could contain up to five malignant diagnoses. If the last date of cancer diagnosis had been taken into account, the participation rates within the diagnostic intervals of less than 5 years would have increased. Finally, adjustment for socio-economic status would have been desirable, but this information was not available during the performance of the analyses. The individualized data linkage with Norway’s populations-based registries is viewed as an essential advantage.