Background Screening for diabetic retinopathy is a key prevention programme in the UK. However, inequity in participation remains a persistent problem. Certain aspects of this, such as ethnicity gaps or non-attendance, are not well-studied. We sought to address this with a quantitative analysis of non-attendance in the Diabetic Eye Screening Wales (DESW) programme, including associations with demographic factors.

Methods Available data were extracted from the DESW database from 2008 onwards (excluding the Covid-19 suspension of screening). This included appointment attendance, appointment non-attendance with no prior notice (’Did Not Attend’), and key characteristics, including age, gender, deprivation quintile, local area, ethnicity, travel time amongst others.

First, a descriptive analysis was undertaken to show the distribution of attendance across different demographics. Second, chi² and regression analyses were used to assess associations between individual factors and non-attendance. Finally, we used multi-variable logistic regression to create a combined model for non-attendance.

Results 1048574 invitations were included covering January 2008 to January 2023. The median age was 66–70. The female: male ratio was 44:56. Ethnicity was 95% white British/Welsh. Roughly 92% of invitees had type II diabetes. Non-attendance was 16%.

We found that, individual associations existed between non-attendance and age group, deprivation, ethnicity, local authority, travel time, appointment month, appointment time, gender, diabetes type, and health board. In our combined analysis, several factors were associated with non-attendance. The largest associations were with age group (Odds Ratio 1.55 working-age vs. retired/school-age [95% confidence intervals 1.52–1.58], p-value <0.001), Ethnicity (Odds ratio 1.35 white British/Welsh vs. non-white [95% confidence intervals 1.34–1.36], p-value <0.001), and area-leveldeprivation (Odds Ratio 1.13 per quintile [95% confidence intervals 1.12–1.14], p-value <0.001).

Other associated factors included travel time (Odds ratio 1.16 20+ minutes travel vs. less [95% confidence intervals 1.12–1.20], p-value <0.001), diabetes type (Odds Ratio 1.12 non-type II vs. type II [95% confidence intervals 1.07–1.15], p-value <0.001), and gender (Odds ratio 1.11 female vs. male [95% confidence intervals 1.08–1.14], p-value <0.001). Finally, appointment Month (Odds ratio 1.05 winter months vs. rest [95% confidence intervals 1.03–1.07]; p-value <0.001) and local authority (Odds ratio 1.04 South Wales Cities/Valleys & Northeast Wales vs. rest [95% confidence intervals 1.02–1.06]; p-value <0.001) displayed small associations with non-attendance.

Conclusion A number of socio-economic and logistical factors as well as age, gender, and ethnicity are associated with non-attendance. Future service developments should investigate such factors in further depth, using mixed methods, with an ambition to reduce unwarranted variation.