Aim Poor glycaemic control in type 1 diabetes mellitus increases risk of poor health in adulthood. It is not known if available interventions in specialist centres improve diabetes control.

Method retrospective data analysis from a single hospital. We used a threshold of 9.5% (80mmol/mol) to define poor glycaemic control and analysed outcomes of those newly referred with poor control.

First, HbA_1c change associated with individual treatments (insulin pump, inpatient rehabilitation programme, and psychological support) over 24 months were measured using paired ttest. Where patients dropped out before 24 months, we used their last recorded HbA_1c as the final outcome measure.

Second, a binomial logistic regression model including clinic attendance, interventions and demographic data was used to analyse this complex intervention. Clinic appointments were grouped into >10, 6–10 and compared with <6 over 24 months.

Results 197 new patients aged 12.0–18.0 years (51.8% female) were seen between January 2008 and March 2014, of which 85 (43.1%) had HbA_1c 9.5%. No difference was seen in demographics between the groups.

Of those with poor control, an overall mean (SD) improvement in HbA_1c of 0.82% (2.1) was observed over 24 months. 78 (92%) remained for the full 24 months. Uptake of interventions was 29 (34%) for an insulin pump, 8 (9%) for inpatient rehabilitation and 42 (49%) for psychological support.

Mean (SD) change over 24 months were 0.6% (1.6) for pump therapy, 0.5% (1.6) for inpatient rehabilitation, and 0.8% (2.0) for psychological support. Each additional clinic attendance was associated with an 0.1% improvement in HbA_1c, (F(1,83) = 3.026, p=0.09).

Regression analysis showed improved control with an insulin pump (Odds Ratio=5.6; 95% CI, 1.2–27.3), and clinic attendance >10 appointments (OR=10.0; 1.1–90.8) but not inpatient rehabilitation (OR=0.4; 0.0–2.6), psychological support (OR=0.4; 0.1–1.6) or demographic variables including ethnicity, gender, deprivation and age over 24 months.

Conclusion Insulin pumps were the treatment modality most likely to improve diabetes control yet only a minority fulfilled eligibility criteria. Further work is needed to understand low uptake of interventions, and if improved access to them would benefit overall glycaemic control.