This study aimed to assess the clinical severity and risk factors of diabetic ketoacidosis (DKA) at type 1 diabetes (T1D) diagnosis in children under 18 years in Greater Poland from 2006 to 2023, including temporal trends and the impact of COVID-19.

A retrospective cross-sectional study.

Greater Poland Province, Poland.

The study cohort comprised 2432 European Caucasian children (boys: 1335) aged 0–18 years with newly diagnosed T1D admitted to one hospital between 2006 and 2023.

DKA and its severity were classified according to the International Society for Pediatric and Adolescent Diabetes criteria. The multivariable analysis assessed the following risk factors for DKA at T1D diagnosis: age, sex, seasonality and the presence of T1D autoantibodies. Poisson regression models with a log link were used to assess the impact of the COVID-19 pandemic on monthly DKA cases at T1D onset, including time, pandemic period and their interaction as predictors.

DKA was diagnosed in 51.4% (1248) of newly diagnosed T1D patients, with 24.9% classified as mild, 14.4% as moderate and 12.1% as severe. Modest sex-related differences were observed, with DKA at T1D onset slightly more common in males than females (52.8% vs 47.2%). However, when comparing the DKA and non-DKA groups, a higher proportion of females presented with DKA (47.2%) than those without DKA (42.9%) (p=0.034). Children aged 0–2 years showed the highest DKA prevalence at T1D onset (76.4%), with a significant proportion experiencing severe DKA (33.6%). Factors like age, sex, season, glycaemia, glycated haemoglobin and autoantibodies did not independently predict DKA risk. The COVID-19 pandemic did not affect DKA rates at diagnosis.

The frequency of DKA is high, and its severity is substantial among children with newly diagnosed T1D in Greater Poland. Children aged 0–2 years are at the greatest risk of severe DKA at onset, underscoring the need for earlier recognition and intervention in this age group. Our findings emphasise the critical importance of increased awareness, education, point-of-care glucose testing, and targeted strategies such as T1D screening programmes to reduce the occurrence of DKA.

Marginalised populations—such as racialised groups, low-income individuals, newcomers and those in rural areas—disproportionately experience severe diabetes-related complications, including diabetic foot ulcers, retinopathy and amputations, due to systemic inequities and limited access to care. Although community-based programmes address cultural and accessibility barriers, their isolation from mainstream healthcare systems leads to fragmented care and missed opportunities for early intervention.

Artificial intelligence (AI)-powered technologies can enhance accessibility and personalisation, particularly for underserved populations. However, integrating AI into community settings remains underexplored, with socioethical concerns around inclusion, diversity, equity and accessibility requiring urgent attention.

This realist review aims to examine how, why and under what circumstances AI applications can be effectively integrated into community-based diabetic care for marginalised populations. The review will develop a programme theory to guide ethical, inclusive and effective AI implementation to ensure AI-driven innovations address health disparities and promote culturally sensitive, accessible care for all.

Using the Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA) extension for Reviews guidelines, this realist review will systematically search MEDLINE, Embase, CINAHL, Cochrane library, Google Scholar and Scopus, alongside grey literature. A two-stage screening process will identify eligible studies, and data extraction will use a developed tool. Synthesis will employ realist logic, analysing relationships between contexts (eg, organisational capacity), mechanisms (eg, AI functionalities) and outcomes (eg, reduced disparities).

Ethics approval is not required for conducting this realist review. Ethics approval will be obtained from the University of Toronto; however, following the completion of the realist review for patients and community members’ engagement to support knowledge mobilisation and dissemination to ensure practical application and reciprocity.

This protocol was registered at PROSPERO (CRD42025636284).