Factors influencing attendance at Abdominal Aortic Aneurysm (AAA) Screening and interventions to reduce inequalities Emma Davies1, Mike Crilly2, Paul Manson3 on behalf of the AAA Inequalities Project Group 1 Specialty Trainee in Public Health Medicine, NHS Grampian 2 Consultant in Public Health Medicine, University of Aberdeen 3 Information Specialist, NHS Grampian November 2017 Background An abdominal aortic aneurysm (AAA) is a swelling of the aorta which can ultimately rupture and lead to death. Risk factors include being male, older age, smoking and high blood pressure. Men aged 65 years in the UK are invited to attend a one off ultrasound scan as part of the national population based screening programme. Screening is intended to reduce the number of deaths due to AAA through the early detection, monitoring and surgical repair of AAAs. The effectiveness of AAA-screening in men is supported by evidence from four randomised controlled trials which have demonstrated that screening reduces AAA-related deaths (Guirguis-Blake 2014). These results are based on uptake rates of over 80 per 100 men invited and AAA detection rates of around 5 per 100 men screened. Eligible men receive a postal invitation to attend for screening. Typically men who fail to attend after one or two further reminders are removed from the programme and not followed up. Given the national coverage of the programme, the absolute number of non-responders is substantial and will continue to grow. The human right to health includes universal physical and economic access to healthcare and appropriate accessible information (WHO 2015). Knowing the reasons for non- attendance would allow programme managers to evaluate whether men are making an informed decision or whether there are barriers to uptake which could be addressed. Further to this, those from more deprived backgrounds are more likely to have an AAA (Kim 2004). It is important for these men to attend for screening to allow them to benefit from early detection and to ensure the cost- effectiveness of the programme. Lack of attendance by this group of men may inadvertently increase health inequalities. In fact, if any one group of men is less likely to attend than another, this could lead to differences in health status between the different population groups. It is in the interest of all to avoid this and the responsibility of those leading the screening programme to ensure the implementation of interventions to minimise inequalities in attendance. 1 The aims of this systematic review are to assess what factors influence attendance at AAA screening and what interventions are effective at reducing inequalities in attendance. Methods Criteria for considering studies for this review Types of studies Systematic reviews, randomised controlled trials, observational studies, analysis of routinely collected data Types of participants Adult men or men and women invited for AAA screening Types of screening programme and intervention AAA screening Any intervention to reduce inequalities in attendance Types of outcome measures Factors associated with attendance excluding sex and age, since these cannot influence attendance in the National AAA Screening Programme. Effectiveness of the intervention in reducing inequalities Search methods for identification of studies Ovid Medline (1946-2017), Embase (1974-2017), and Proquest Public Health (1963-2017) databases were searched on 30th June 2017. The search strategy was deliberately broad. Medline and Embase were searched using exp(Abdominal aortic aneurysm) AND exp(Mass screening) AND (uptake or non- respon* or barriers or facilitat* or non-particip* or disadvantage* or depriv* or socioeconomic* or inequality*or factors).ti,ab. Proquest Public Health database was searched using (SU.exact(Aortic aneurysm abdominal) and SU.exact(Mass screening)) OR ((disadvantage* or depriv* or socioeconomic* or inequalit* or uptake or non-respon* or barriers or facilitate* or non-particip*) and SU.exact(Aortic aneurysm abdominal)) OR (ti(abdominal aortic aneurysm or AAA) and ab(screening) and (disadvantage* or depriv* or socioeconomic* or inequalit* or uptake or non-respon* or barriers 2 or facilitate* or non-particip*). The Cochrane Database of Systematic Reviews was searched. Google Scholar was used to look for grey literature as well as the following sources: National Institute for Health Research (NIHR), Agency for Healthcare Research and Quality (AHRQ), Canada Agency for Drugs and Technologies in Health (CADTH). No language or date restrictions were used. Additional search An initial scope of results suggested that there is some evidence available for factors influencing attendance but very little on interventions. Based on the lack of papers looking at interventions specifically for AAA screening, it was decided that an additional search would be helpful. Types of studies Systematic reviews Types of participants Adult men or women Types of screening programme and intervention Breast, bowel or cervical screening Any intervention to reduce inequalities in attendance excluding: i) interventions that simplify or change the test ii) interventions that reduce or eliminate the fee for screening Comparator Standard invitation alone Types of outcome measures Effectiveness of the intervention in increasing uptake Effectiveness of the intervention in reducing inequalities We searched for systematic reviews assessing interventions to reduce inequalities in uptake of bowel, breast and cervical screening using the Cochrane Database of Systematic Reviews, Ovid Medline, Embase and Proquest Public Health databases on 31st July 2017. Medline and Embase were searched with the following strategy: exp(Mass screening) AND (uptake or non-respon* or barriers or facilitat* 3 or non-particip* or disadvantage* or depriv* or socioeconomic* or inequalit* or rates).ti,ab. AND systematic review.ti AND interven*.mp. Proquest Public Health was searched using SU.exact(Mass screening) AND ti(systematic review) AND interven*. Data collection and analysis Selection of studies For the main search, two researchers (PM and ED) independently undertook an initial selection of studies based on titles and abstracts. One of these researchers (ED) than read the full text papers of the selected studies and decided if they met the inclusion criteria. For the additional search, one researcher (ED) did both the initial and final selections. The decision to include or reject full text papers was checked by a further researcher (MC). Those excluded at this stage are listed in Appendix A. Data extraction and management One researcher (ED) extracted the data using a predesigned data extraction form (Appendix B). For the main search, information collected included country of origin, setting and duration of study, details of the screening programme, ‘did not attend’ definition, inclusion and exclusion criteria, total (and percentage) who attended screening and author recommendations for reducing inequalities. Factors investigated for possible influence on attendance were also extracted and the associated reported outcomes. For the additional search, information collected included inclusion and exclusion criteria for the review, type of intervention, comparator and reported effectiveness of the intervention. Assessment of quality in included studies The CASP (Critical Appraisal Skills Programme) tools were used to generate a basic framework for quality assessment, which was incorporated into the data extraction forms. In the main search, the following were considered: missing data, other possible data inaccuracies, risk of bias, potential confounders and role of chance. The risk of selection, detection, performance and attrition bias were assessed, as recommended by the Cochrane Collaboration and using their risk of bias tool. In the additional search, quality assessment of the systematic reviews was achieved through critique of the 4 selection and quality appraisal processes and assessment of the overall risk of bias. The results are presented alongside the list of included studies in Appendix C. Measures of effect Percentage attendance and/or adjusted odds ratios with 95% confidence intervals (main search) Absolute percentage change in overall uptake/uptake between subgroups and/or adjusted odds ratios with 95% confidence intervals (additional search) 5 Results Description of studies See Appendices A and C. Results of the main search (see Figure 1) The search produced 217 papers. Of these, 191 were excluded after reading titles and abstracts, leaving 26 papers which were retrieved in full text form. Out of the 26, 9 met the inclusion criteria and were used in the narrative synthesis. Figure 1: Study flow diagram 214 papers 3 papers identified identified from from other sources database searching 191 excluded as Titles and not relevant abstracts read =217 papers Full text read = 17 excluded: 26 papers 9 duplicates 8 did not meet inclusion criteria 9 included in narrative synthesis Included studies All included studies were published in English. Five were analyses of routinely collected data, three were observational studies (case-control) and one was a retrospective analysis of interventional group data collected as part of a randomised controlled trial. Six were from the UK (three Scottish, two English and one from Northern Ireland), two from Sweden and one from the USA. Four used data from 2010 onwards, three between 2000 to 2009 and one prior to 2000. 6 Participants The nine studies included 724,654 participants. The number in each studied varied from 3652 to 593,032. Eight studies had included only men and one included men and women. Four had invited only those aged 65 years, three had invited 65 to 74 year olds and one had invited a random selection of individuals aged 65 years and older. The final study had altered the age criteria half way through so that initially 65 to 74 year olds had been invited and then this changed to only 65 year olds. Participants Study IDs Number of studies Sex Men only Badger 2008; Crilly 2015; 8 Jacomelli 2017; Kim 2004; Lindsay 2006; Linne 2014; Ross 2013; Zarrouk 2013 Men and Schermerhorn 2008 1 women Age 65 years Crilly 2015; Jacomelli 2017; 4 Linne 2014; Zarrouk 2013 65-74 years Badger 2008; Kim 2004; 3 Lindsay 2006 65-74 years Ross 2013 1 (50%) 65 years (50%) 65 years + Schermerhorn 2008 1 Screening programme The methods of invitation and screening were similar in all studies; they invited participants by postal letters and screening was by ultrasound scan. Four were part of a national programme, three were prior to a national programme being implemented and recruitment was through GP practices, one was a mix of these two options and one invited a selection of Medicare beneficiaries. In seven studies the screening was free, in one there was a fee and in the final one the fee was scrapped three-quarters of the way through the study. Seven of the studies used the same ‘did not attend’ definition which was no attendance after an invitation and one reminder or formally opted out if part of a national screening programme. One study used a similar definition which was no attendance after an invitation and two reminders or formally opted out. The final study analysed those who opted out separately and those who did not attend were only those who were not present for their appointment after an invitation and two reminders. 7 Screening programme Study IDs Number of studies Method of Postal letter Badger 2008; Crilly 2015; 9 invitation Jacomelli 2017; Kim 2004; Lindsay 2006; Linne 2014; Ross 2013; Schermerhorn 2008; Zarrouk 2013 Method of Ultrasound Badger 2008; Crilly 2015; 9 screening scan Jacomelli 2017; Kim 2004; Lindsay 2006; Linne 2014; Ross 2013; Schermerhorn 2008; Zarrouk 2013 Recruitment National Crilly 2015; Jacomelli 2017; 4 screening Linne 2014; Zarrouk 2013 programme GP practices Badger 2008; Kim 2004; 3 Lindsay 2006 GP practices Ross 2013 1 then national programme once implemented Random Schermerhorn 2008 1 selection of Medicare beneficiaries Cost Free Badger 2008; Crilly 2015; 7 Jacomelli 2017; Kim 2004; Lindsay 2006; Ross 2013; Schermerhorn 2008 Fee Zarrouk 2013 1 Fee then free Linne 2014 1 Did not attend After one Badger 2008; Kim 2004; 7 definition invitation and Lindsay 2006; Linne 2014; Ross one reminder 2013; Schermerhorn 2008; or opted out Zarrouk 2013 After one Crilly 2015 1 invitation and two reminders or opted out After one Jacomelli 2017 1 invitation and two reminders Intervention to reduce inequalities in attendance There were no studies which looked at interventions to reduce inequalities in attendance at AAA screening. 8 Outcomes We extracted data on the following outcomes: total number and percentage attended; number and percentage attendance by deprivation score; median deprivation score and interquartile range for those who attended and did not attend; percentage uptake and median private cost in pounds sterling by rural classification; percentage who attended and did not attend who had more than two hospital admissions in the last ten years, more than two outpatient visits, ischaemic heart disease, chronic obstructive pulmonary disease, diabetes, stroke, renal failure and malignancy; coefficient for correlation between attendance and mean income, unemployment levels, percentage of subjects on welfare support, distance to clinic, proportion of immigrants and smoking rates; adjusted odds ratios and 95% confidence intervals for attendance by deprivation score, rural classification, clinic type, distance to clinic and season, income quintile, education level, marital status and immigration status. Excluded studies There were eight papers excluded (see Appendix A). The primary reason for exclusion was that there was no reporting of the outcomes of interest (six papers). The other reason for exclusion was that there was not sufficient evidence provided to determine effectiveness of interventions implemented aiming to reduce inequalities in attendance (two papers). Risk of bias in included studies As described in Appendix C, each study was assessed for the likelihood of selection, performance, detection and attrition biases. Figure 2 summarises the findings. All studies had low risk of performance and detection biases. Only one study had a high risk of any type of bias (selection). Despite this, only one study had a low risk of any bias. This was largely due to a lack of reporting of how participants were identified, specific reasons for excluding participants and how this was achieved, characteristics of those who declined to participate and how missing data was dealt with. This meant that the risk of selection and attrition biases had to be categorised as ‘uncertain’ in 67% and 44% of studies, respectively. 9 Figure 2: Risk of bias graph Selection as Performance bi f o e p y Detection T Attrition 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Percentage of papers Low Uncertain High Factors influencing AAA screening attendance See Appendix D: Data used in review Socioeconomic status Six studies assessed how socioeconomic status influenced AAA screening attendance by using an index of multiple deprivation. For all five studies (Badger 2008; Crilly 2015; Jacomelli 2017; Kim 2004; Ross 2013) which broke deprivation score into categories, as deprivation level increased, the less likely it was for men to attend. Linne (2014) produced similar results for income level and education level (as decrease, less likely to attend). Lindsay (2006) compared median Scottish Index of Multiple Deprivation (SIMD) score between those who attended and did not attend. A higher value indicated higher deprivation. The results were a score of 17.27 for those who attended and 19.83 for those who did not. This was statistically significant (p<0.001). Zarrouk (2013) looked at correlations between attendance and mean income, unemployment rates and percentage of subjects on welfare support at datazone level. There was no adjustment for age or sex and no ability to match data on a smaller or more specific level. Areas with higher mean income had higher uptake rates. However, the correlation between lower unemployment and higher uptake was only found in the city and the correlation between lower percentage on welfare support and higher uptake was only found in the suburbs. 10