Rapid review of evaluation of interventions to improve participation in cancer screening services

Objective Screening participation is spread differently across populations, according to factors such as ethnicity or socioeconomic status. We here review the current evidence on effects of interventions to improve cancer screening participation, focussing in particular on effects in underserved populations. Methods We selected studies to review based on their characteristics: focussing on population screening programmes, showing a quantitative estimate of the effect of the intervention, and published since 1990. To determine eligibility for our purposes, we first reviewed titles, then abstracts, and finally the full paper. We started with a narrow search and expanded this until the search yielded eligible papers on title review which were less than 1% of the total. We classified the eligible studies by intervention type and by the cancer for which they screened, while looking to identify effects in any inequality dimension. Results The 68 papers included in our review reported on 71 intervention studies. Of the interventions, 58 had significant positive effects on increasing participation, with increase rates of the order of 2%–20% (in absolute terms). Conclusions Across different countries and health systems, a number of interventions were found more consistently to improve participation in cancer screening, including in underserved populations: pre-screening reminders, general practitioner endorsement, more personalized reminders for non-participants, and more acceptable screening tests in bowel and cervical screening.


Introduction
To achieve their desired public health impact, population cancer screening services require high levels of participation. While it is agreed that decisions to participate in cancer screening should be free from undue pressure, and should be well-informed, it is also frequently observed that there are considerable inequalities in participation in cancer screening. [1][2][3] In the United Kingdom (UK), screening participation rates are lower in areas of deprivation and among certain ethnic groups. [3][4][5] There is a wide range of potential interventions to improve access to cancer screening services and, therefore, increase participation. Given the perceived need to address health inequalities, 6 one tactic would be to improve uptake of public health measures such as screening programmes in currently underserved populations. We review the evidence on effects of interventions to improve screening participation, with particular reference to effects on inequalities. To inform policy and practice, such a review should identify those measures most and least likely to be effective, including any findings with respect to inequalities, or effects of interventions in deprived or otherwise underserved populations.

Methods
We specified in advance that the studies of relevance to this review had to report on interventions aimed at participation in cancer screening services (not randomized trials of screening). The focus of interest was population screening programmes (as opposed to surveillance of specific genetic or other high-risk groups). To be eligible, a study had to report a quantitative estimate of the effect of the intervention on participation rates. We only included studies published since 1990. Studies which assessed the effect of personal invitation against no invitation were not included, as the NHS Cancer Screening Programmes in the UK would always send personal invitations in any case.
The commissioners of this research required results in a relatively short time and resources were not available for the traditional systematic review. Therefore, rather than the customary search strategy which begins as comprehensively as possible and frequently identifies tens of thousands of populations, we started with a narrow search and expanded successively (by adding 'OR' terms) until the number of new publications eligible on title review comprised less than 1% of the total. The major assumption here was that if successive expansions of the search yield diminishing numbers of potentially eligible publications, and if the most recent expansion yields a relatively small addition to the pool, stopping the expansion at this point is unlikely to lead to a major loss of information. The successive searches and their outcomes are shown in Table 1. As this strategy is less comprehensive than the standard systematic review procedure, as a further safeguard, we also specified that the publications yielded by the final search had to include four recent studies which were clearly relevant. [7][8][9][10] The other restriction was that only peer-reviewed results obtainable in PubMed searches were considered. Searches took place in late September 2015.
Papers passing title review underwent abstract review. Those remaining eligible after abstract review underwent full paper review. In addition to original papers, our searches identified five reviews. [11][12][13][14][15] From these, and from the reference lists of the papers eligible after full paper review, we identified further potential papers which, in turn, were subject to abstract and, if eligible, full paper review. Finally, colleagues identified evaluations of eight interventions published in three papers since the searches took place. [16][17][18] Papers were reviewed for the effect of the intervention by the cancer for which they were screening and by intervention type (reminders in addition to usual invitation; primary care endorsement; additional interventions in non-participants; enhanced invitation materials or varying invitation strategy; direct contact; varying the screening test). Interventions can occur at various stages of the process, from advance notice communications sent prior to first invitation to screening, to the offer of a different screening test at the screening episode some years following non-participation in a previous episode (see Figure 1). We deal with the arbitrary nature of the classification below (see Discussion section). In addition, interventions were studied to identify effects by different socioeconomic or ethnic grouping, if any, and to identify any other inequality dimension in the work.
In the description of results of interventions below, percentage increases in participation refer to absolute increases; for example, a 5% increase would refer to a difference between 20% and 25% (an absolute increase of 5%) rather than 20% and 21% (a 5% relative increase).

Results
After abstract review, 62 papers were deemed eligible. On full paper review, two were excluded, leaving 60. From reference lists and reviews, after full paper review, further five papers were added. As noted above, colleagues identified another three papers published since the searches, bringing the total to 68 papers included, although one reports on four separate trials and is therefore included four times in the tables of results. 16

Reminders in addition to usual invitation
Results are shown in Table 2. Eleven papers were identified, two in colorectal, six in breast, two in cervical and one in stomach cancer screening. 7,8,[18][19][20][21][22][23][24][25][26] Ten of the 11 reported an increase in participation with reminders (3/3 text reminder studies; 4/5 telephone reminder studies; 2/2 postal reminder studies; 1/1 telephone þ postal reminder study). The absolute magnitude of the effects was an increase in participation of the order of between 3 and 10 percentage points. Most studies were in urban areas, including substantial underserved and low socioeconomic status populations. All seven of the studies explicitly reporting results in such underserved populations found an increase in participation with the intervention. 7,8,18,[20][21][22][23] Primary care endorsement Results are shown in Table 3. Twelve studies were identified, six in colorectal, four in breast, and two in cervical cancer screening. 9,16,[70][71][72][73][74][75][76][77][78][79] Ten of the 12 reported positive results. One of the studies which did not find increased participation was in the context of a flexible sigmoidoscopy project in the UK before flexible sigmoidoscopy was included in the national programme, and rather than a letter with general practitioner (GP) endorsement, the intervention was the offer of a consultation with the GP to discuss the screening. 70 Typical increases in participation of the order of 2%-3% were observed, but some studies found increases of 10%-20%. All four studies explicitly reporting the effect in underserved populations found an increase in participation with the intervention. 16,72,75,76 One study of a multilingual intervention aimed at ethnic minorities in Cardiff found an increase in participation in people of south Asian origin, higher in Gujarati and Urdu speakers than in Bengali speakers. 72 Interventions targeted specifically on non-participants Results are shown in Table 4. Fifteen studies were identified, three in colorectal, eight in breast, and four in cervical cancer screening. 16,[27][28][29][30][31][32][33][34][35][36][37][38][39][40] Twelve of the fifteen studies found an increase in participation with the intervention; 3/6 telephone reminder studies had positive results, compared to 7/7 postal interventions (including two where the letter was from the subject's own GP), 1/1 study of combined telephone and postal reminders, and 1/1 which found increased participation with sending a second faecal occult blood test (FOBT) kit instead of a reminder letter. Typical effect sizes were of the order of a 10% increase in participation with reminder letters, and rather smaller effects with telephone reminders. Very large effects were noted for reminder letters compared with no further contact, 32 but as the former is standard practice in the UK, there is no policy change implied by this result. Substantially greater participation was observed for second timed appointment for breast screening non-attenders, compared with open invitation to call and make an appointment. 30 One study of an enhanced reminder letter,

Barcelona, Spain
Breast 233 women received usual invitation (control group), 470 received an additional text-message reminder (intervention group).
Effect significant in those with low educational status.
Taplin et al. 24 Telephone reminders and motivational telephone calls.

Seattle, USA
Breast 590 randomized to reminder postcard, 585 to reminder telephone call, and 590 to motivational telephone call addressing barriers, tailored to demographic features of the invitees.
Participation was 35% in the postcard group, 52% in the telephone reminder group, and 50% in the motivational call group.

Inequality dimension
Vidal et al. 25 Text-message reminders.

Catalonia, Spain
Breast Non-randomized study. 3719 sent text message in addition to invitation, compared to 9067 receiving invitation only.
Effective in areas where postal service less reliable.

Barcelona, Spain
Cervical Cluster-randomized trial in 3225 Coverage: Relatively low income and educational status of population.
Eaker et al. 20 Telephone and postal reminders.

Sweden
Cervical 12,240 women randomized to: (1) modified invitation vs. standard invitation (2) reminder letter vs. no reminder letter (3) telephone contact of nonattenders vs. no telephone contact.
Significant 9% increase in attendance with reminder letter. 31% increase with telephone reminder. Results: Effect stronger in less deprived groups.
Any intervention significantly improved attendance in neverscreened men. Attendance among never screened by group: Not clear. Table 3. Primary care endorsement studies.

Publication and intervention
Region Cancer Design

Inequality dimension
Wardle et al. 16 General practice endorsement as a banner across invitation letter.

England
Colorectal Cluster randomized trial. 134,011 invitees randomized to be sent usual invitation, 131,423 letter with GP endorsement banner.
Increased participation with GP endorsement (58% vs. 57%). Non-significant trend of increasing effect in more deprived populations.
Formally designed to investigate significant difference in effect among deprivation categories.
Hewitson et al. 9 Primary care endorsement letter and a patient leaflet.

South of England, UK
Colorectal Randomized controlled 2 Â 2 factorial trial of 1288 patients randomized to either a GP's endorsement letter and/or an enhanced information leaflet with their FOBT kit.
Including both an endorsement letter from each patient's GP and a more explicit procedural leaflet increased participation.
Conducted in an area of medium to high socioeconomic status.
Gray et al. 70 Letter offering opportunity to discuss flexible sigmoidoscopy screening with GP.

Colorectal
Patients aged between 50.5 and 60.5 randomly allocated to one of two groups. The first group was sent an invitation to have screening sigmoidoscopy along with an explanatory leaflet. The second group was sent the same invitation and leaflet but with an added option to discuss the test in the first instance with their GP.
The overall uptake rate was 24%. Significantly fewer people in the second group replied to the initial invitation.
Urban setting. Otherwise, none reported. Barthe et al. 71 Invitation signed by GP. Uptake: 15% in each group.
Urban setting. Otherwise, none reported.
Zajac et al. 78 Letter with primary care endorsement.

South Australia
Colorectal Randomized study. 1200 offered FIT without GP endorsement, 600 with GP endorsement mentioned in accompanying letter (GP2), and 600 with invitation explicitly from own practice (GP3).
Significant increase in participation in those with explicit invitation from own general practice. At first round, results were no endorsement, 33%; GP1, 39%; GP2, 42%.
None reported.
Cole et al. 79 Letter with primary care endorsement.

South Australia
Colorectal Similar to Zajac, above. 600 offered FOBT without GP endorsement, 600 with GP endorsement mentioned in accompanying letter (GP2), and 600 with invitation explicitly from own practice (GP3).
None reported.
(continued)   Turner et al. 31 Letter signed by own GP with second invitation.

Aberdeen, UK
Breast 234 non-responders randomized to GP letter with second invitation, 231 to usual second invitation.
None reported.

Dublin, Ireland
Breast Uncontrolled study of reminders for non-attenders.
None reported.
Hayes et al. 33 Second and third reminder letters.

Dublin, Ireland
Breast Non-responders to an invitation for screening were re-invited by computer-generated letter to attend for screening six weeks after issue of the first invitation and a final invitation was issued at 12 weeks.
Issue of second mailed invitations to women in the target age for breast screening increased uptake from 61% to 79%. Third invitations were not cost-effective. Women aged 55-64 were more likely to respond to first, second, or third invitations than those aged less than 55.
No difference in effect by whether or not the invitee had private health insurance.
Effect present in different educational status groups.

Belgium
Breast 3880 women who had not attended for screening randomized to written invitation or written invitation plus telephone call.
None reported.

Aarhus, Denmark
Cervical Cluster randomized trial. General practices were unit of randomization. 7527 nonattenders in intervention group, 7452 in control.
Significant (1%-2%) improvement in coverage in intervention group after nine months.
None reported.
Oscarsson et al. 36 Personal contact, including telephone contact, of non-attenders.

Kalmar County, Sweden
Cervical 400 randomized to intervention, 400 to control.
None reported.
(continued) one of additional reminder letters, and two studies of telephone reminders reported positive results in underserved populations. 16,28,31,34 Varying invitation materials or strategy Results are shown in Table 5. Eighteen intervention studies reported in 17 papers were identified, 11 in colorectal, four in breast, one in cervical, one in breast and cervical, one in melanoma screening. 10,16,17,[41][42][43][44][45][47][48][49][50][51][52][56][57][58] The 17 papers reported on evaluation of 20 interventions; some studies evaluated more than 1 intervention simultaneously. Thirteen interventions were observed to be associated with increased participation. Enclosing survey questionnaires with the invitation had little effect on participation, and one study found a significantly reduced participation with the intervention, with the reduction greatest in deprived areas. 42,56 Fixed screening appointment times compared with open invitations increased participation by around 20% (in absolute terms), and one study found a 3% increase when the invitation included the opportunity to switch to an evening or weekend appointment time for breast cancer screening. 10,51 The latter took place in Manchester and Bristol, both of which included areas of deprivation. 10 Out of five studies (in four publications) of varying the information with the invitation, only one found an increase in participation with the intervention. 16,17,43,52 Two studies of advance notice of the screening invitation found an increase in participation. 41,43 One study found an increased participation rate with the offer of a health check, 48 and another found no increase with the offer of counselling. 50 Two out of three evaluations of mass media campaigns found an increase in participation. 17,44,45 Two studies found that direct mailing of FOBT kits led to higher participation than the request to collect a kit at primary care. 57,58 One found increased participation with the mailing of a pack of equipment with the FOBT kit, including latex gloves and stool catchers to facilitate collection of stool samples. 17

Direct contact interventions
Results are shown in Table 6. Five studies were identified, one in colorectal, three in breast, and one in cervical cancer screening. 46,[53][54][55]59 Three studies were of home visits, two finding increased participation associated with the intervention. 53,55,59 One study of direct telephone contact by a health professional found an increase in participation with the intervention. 54 One study of opportunistic promotion of breast screening at clinic attendances for other reasons found an increase in participation, especially in women of low socioeconomic status. 46 Varying the screening test Results are shown in Table 7. Ten studies were identified, three in colorectal [60][61][62] and seven in cervical screening. [63][64][65][66][67][68][69] Faecal immunochemical testing (FIT) yielded 15%-20%   Study took place in a region of high mobility and low participation.
Offman et al. 10 Out of office hours screening appointments.

UK Breast
Four-armed randomized trial of women invited for routine breast screening randomized (3:1:1:1) to one of these screening invitations: standard office hour appointment, office hour appointment with the option to change to an out-of-hours appointment, weekday evening appointment, or weekend appointment.
The optimum strategy for improving attendance at breast screening was to offer a traditional office hour appointment and including in the letter of invitation an option to change to an evening or weekend appointment if wished (76% vs. 73%).
Trial took place in Manchester and Bristol, each centre containing urban areas of significant deprivation.
Banks et al. 49 Enclosure of a questionnaire with invitation to breast screening.

Breast
Randomized study of 6400 women invited for routine screening mammography were individually randomized to receive either the usual breast screening invitation alone, or to receive the usual invitation accompanied by a selfadministered questionnaire.
Screening uptake was not affected by the intervention.
Lower uptake in older women. Otherwise, none reported. Giordano

Results
Inequality dimension (4) invitation to contact the centre for information and arrangement of appointment.
Page et al. 44 Mass media campaign in Italian language.

New South Wales, Australia
Breast Italian language newspaper and radio promotion.
No change in uptake rates between pre-and postintervention.
Italian-speaking women in Australia, considered a hardto-reach group.
Nguyen et al. 45 Multiple outreach intervention.

Inequality dimension
Courtier et al. 59 Home visit.

Barcelona, Spain
Colorectal 1060 randomized to mailed kit and letter, 965 to home visit with kit.
Urban population.
Hoare et al. 53 Home visit.
Manchester, UK

Breast
The control group received no visits. The study population comprised all women with Asian names, from a batch of general practices where high proportions of patients were Asian, who were invited for screening.
No difference in attendance was found between the intervention and control groups. This study took place in an Asian population in an urban environment including areas of substantial deprivation.
Segura et al. 54 Direct contact by health professional.

Barcelona, Spain
Breast 564 women randomized to programme invitation letter, primary care invitation letter, or direct contact.
Increased participation was strongest in women of lower educational status.
Taylor et al. 46 Promotion at routine clinic visit for other medical reasons.

Seattle, USA
Breast 232 women randomized to promotion, 82 to usual care.
49% in intervention group participated, 22% in control. Inner city population. Significant effect in black women and in those with and without health insurance.

Thailand
Cervical Geographic zone randomization.
158 women in intervention group, 146 in control.
Took place in a developing country with corresponding levels of education and socioeconomic status. Significantly higher response to self-sampling (76% vs. 65%). Effect similar in different language groups. (continued) higher participation rates than either colonoscopy 61 or guaiac FOBT. 62 The improvements over both colonoscopy and guaiac FOBT did not vary substantially by age, sex, or ethnicity. One study in Germany comparing conventional with capsule colonoscopy found a small increase in participation with the latter. 60 All the cervical screening studies were of the offer of human papillomavirus (HPV) self-sampling, usually to women with a history of non-participation, [63][64][65][66][67][68][69] and all found increased participation with the offer of self-sampling, typically of the order of 10%. The one study reporting effects by socioeconomic status found the intervention equally effective in different socioeconomic groups. 66 The intervention is effective in previous non-participants, who are frequently characterized by lower socioeconomic status or specific ethnic profiles. 5

Discussion
A number of results seem to be observed consistently between studies and across different countries and health systems. Both pre-screening reminders and GP endorsement led to higher participation rates (albeit modest increases) and were observed to do so in deprived and otherwise underserved populations. [16][17][18]21,22,74,75,79 More personalized reminders for non-participants, whether by enhanced written materials or telephone contact (notably from primary care), were effective in increasing participation. These interventions too were successful in populations of low socioeconomic status. 16,30,33,40 Primary care endorsement and enhanced reminders for non-participants would incur almost no expense (other than the cost of screening larger numbers of people). In the UK, they might be expected to result in small, but arguably worthwhile, increases in participation. Larger increases might be expected in more deprived populations with lower current participation rates. 16 The choice of screening test itself is associated with participation. FIT is clearly more popular than other bowel screening modalities and was observed to increase participation by 15%-20%. HPV self-sampling raised participation rates, notably in previous nonparticipants, by around 10%. This is likely to be effective in socioeconomic or ethnic populations traditionally less easy to reach with cervical screening. 62 Less consistent results were observed for different invitation strategies and home visits. Inclusion of questionnaires for research along with the invitations seemed to have a negative effect. 38 It is accepted that interventions to promote cancer screening should be non-coercive and should respect the principle of informed choice. 80 However, it is also the case that there is a strong socioeconomic gradient in participation in screening, with lower participation being associated with lower socioeconomic status, 3 and there is evidence that non-participants often report not having read the information provided. 81 The search for interventions to remedy this would, therefore, seem to be ethically justified.
Significant difference in effect among centres.
included in Table 7 as interventions varying the screening test could have been included in Table 4 as interventions targeting non-participants. Similarly, some of the latter could have been included as primary care endorsement studies. However, the results are generally clear. It is also worth noting that, in classifying the studies, a degree of oversimplification was inevitable, in that some multi-component interventions have been classified into one category or another. One example is the study by Bell et al., 72 which is classified as primary care endorsement, but which also included multilingual approaches and offers of transport to screening centre. The magnitude of effects varied considerably, even within intervention types. The effects tended to be larger in environments where participation rates were relatively low. For example, reminder studies showing particularly large effects on participation were that of Baker et al. 21 in Chicago, where opportunistic rather than organized screening was taking place, and one in inner north-east London, where there are large deprived populations, high levels of ethnic diversity, and usually low screening participation rates. 8,22 Similarly, in the primary care endorsement studies, greatest effects were seen in populations with previously low participation. 72,75 In the studies identified, the patient navigation approach, 82 whereby a 'navigator' guides the patient/invitee through the complexities of a screening, diagnostic, or therapeutic process, was largely absent. However, some of the telephone interventions involved detailed scripts and briefing of the staff, so that they were able to answer questions; indeed, one such study in the context of bowel cancer screening showed good results. 22 The concept of patient navigation is already established in the USA and may well spread to Europe in the immediate future.
We did not sub-classify the studies by design or quality but have noted in the tables whether the studies were randomized trials or observational studies. Of the 71 intervention studies, 52 (73%) were randomized, either by individual or cluster. The majority of positive results were seen in randomized studies.
As noted in the Methods section, due to time and resource considerations, we restricted our search to peerreviewed publications listed in PubMed and adopted an unconventional expanding search strategy rather than a comprehensive search followed by successive narrowing by abstract and paper review. Although we only ceased the search expansion when it yielded relatively small numbers of potentially eligible publications and built in a safeguard by specifying that the search had to include a number of key publications, it is possible that some eligible material has been missed. 83 Watt et al. 84 advocate not a standard methodology for rapid review, but clear reporting and transparency with respect to the methods used. We have tried to adhere to this in terms of the description in the Methods section and the information in Table 1. It would be interesting if another group with the time and resources carried out a traditional systematic review, to see what, if anything, we have missed.

Conclusion
Interventions which were found most consistently to improve participation in cancer screening, including in underserved populations, were pre-screening reminders, general practice endorsement, more personalized reminders for non-participants, and offering a more acceptable screening test in cervical and bowel screening, both of which may suffer from social and cultural taboos.