Using survival modelling as a composite measure of outcome to demonstrate insulin pump therapy effectiveness in type 1 diabetes

Abstract

Glycated haemoglobin (HbA1c) is a time‐series measure subject to variability; quantifying the difference between the first and last readings over a defined period may have little value in measuring success in service evaluation models. Identifying a composite measure for outcome would be an invaluable tool for service providers. We demonstrate the efficacy of survival modelling as a composite measure of outcome. This was a retrospective observational cohort analysis of patients with type 1 diabetes on insulin pump therapy. From these data, a service evaluation project was undertaken. Survival (time‐to‐event) analysis was undertaken for two HbA1c values – 58 and 64mmol/mol (7.5% and 8.0%); analyses were based on time taken to reach the index HbA1c value (58 or 64mmol/mol) for the first time from start of pump therapy. A total of 113 patients were treated with the insulin pump. The median time to reach an HbA1c of 58 and 64mmol/mol was 25 and 13.9 months, respectively. Cox regression using gender, HbA1c, weight and age prior to pump therapy as factor variables identified HbA1c as the only predictor that significantly influenced the time‐to‐event (HR 0.94, p=0.001 for HbA1c 58mmol/mol, and HR 0.97, p=0.008 for HbA1c 64mmol/mol). Survival modelling appears to be a robust statistical tool to evaluate performance. Furthermore, adjusting for covariates identifies factors that may influence outcome. Pump therapy is considered a high cost area by commissioners, therefore having auditable outcome measures would be useful to inform authorities on the effectiveness of the service. Copyright © 2020 John Wiley & Sons.