Derivation and external validation of risk algorithms for cerebrovascular (re)hospitalisation in patients with type 2 diabetes: Two cohorts study

Aims

Cerebrovascular disease is one of more typical reasons for hospitalisation and re-hospitalisation in people with type 2 diabetes. We aimed to derive and externally validate two risk prediction algorithms for cerebrovascular hospitalisation and re-hospitalisation.

Methods

Two independent cohorts were used to derive and externally validate the two risk scores. The development cohort comprises 4704 patients with type 2 diabetes registered in 18 general practices across Cambridgeshire. The validation cohort includes 1121 type 2 patients from a post-trial cohort data. Outcomes were cerebrovascular hospitalisation within two years and cerebrovascular re-hospitalisation within ninety days of the previous cerebrovascular hospitalisation. Logistic regression was applied to derive the two risk scores for cerebrovascular hospitalisation and re-hospitalisation from development cohort, which were externally validated in the validation cohort.

Results

The incidence of cerebrovascular hospitalisation and re-hospitalisation was 3.76% and 1.46% in the development cohort, and 4.99% and 1.87% in the external validation cohort. Age, gender, body mass index, blood pressures, and lipid profiles were included in the final model. Model discrimination was similar in both cohorts, with all C-statistics > 0.70, and very good calibration of observed and predicted individual risks.

Conclusion

Two new risk scores that quantify individual risks of cerebrovascular hospitalisation and re-hospitalisation have been well derived and externally validated. Both scores are on the basis of a few of clinical measurements that are commonly available for patients with type 2 diabetes in primary care settings and could work as tools to identify individuals at high risk of cerebrovascular hospitalisation and re-hospitalisation.