(165) Length of Lookback Period for The Charlson Comorbidity Index and Prediction of Mortality Among Hip-Fracture Patients in Denmark

Sunday, August 28, 2022

8:00 AM – 1:45 PM CEST

Location: Hall C

Publication Number: 1210

Presenting Author(s)

Morten Madsen, MSc
– Biostatistician, Department of Clinical Epidemiology, Aarhus University, Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus N, Midtjylland, Denmark

Co-Author(s)

Morten Schmidt
– Department of Clinical Epidemiology, Aarhus University and Aarhus University Hospital, Aarhus, Denmark, Denmark
Alma Becic Pedersen, ABP
– Aarhus University, Department of Clinical Epidemiology, Aarhus University, Aarhus N, Midtjylland, Denmark
Vera Ehrenstein, MPH, DSc
– Aarhus University, Aarhus, Denmark
Johnny Kahlert
– Department of Clinical Epidemiology, Aarhus University Hospital and Aarhus University, Denmark

Background: In pharmacoepidemiology comorbidity burden is commonly measured for confounder adjustment or subgroup stratifications. The Charlson Comorbidity Index (CCI) is commonly used for that purpose and can be calculated using registry data. However, it is unclear what is the most appropriate lookback period for calculating CCI.

Objectives: To assess the association between the length of CCI lookback period with its ability to predict one-year mortality.

Methods: We included patients with hip fracture in Denmark (2015-2017) and on the date of admission, computed CCIs with 1 -30 years’ lookback periods based on hospital diagnoses in the Danish National Patient Registry. Deaths were ascertained in the Danish Civil Registration System. The predictive ability of 1-year mortality of different CCI lookback periods was measured with the C-index for a model with CCI added to a base model of age and sex using logistic regression. This C-index was used as a proxy for the model’s ability to adjust for comorbidity as a confounder.
In addition, we compared mortality of the subgroup with high CCI (patients with 3 or more points on the weighted CCI-scale) as identified with 1- and 30-years lookback periods.

Results: The 11,391 hip fracture patients had an overall 1-year mortality of 23%.
The base model had a C-index of 0.711. In all the models, including CCI with any choice of lookback-period between 1 and 30 years C-indices ranged between 0.745 and 0.748.
Based on 1-year and 30-years of lookback period we identified 723 and 2796 patients respectively with a high CCI. The 1-year mortality in these respective groups was 45% and 35%.

Conclusions: While CCI added additional predictive ability to the base model, the choice of length of lookback period between 1 and 30 years had minimal impact on the model’s ability to predict 1-year mortality in hip fracture patients.
Lookback period plays a role if the comorbidity is used for subgroup stratification. A long lookback period identified more patients with high CCI, compared with short lookback period, but the mortality among high CCI patients based on short lookback period was clearly higher, indicating more severe cases of the comorbidities.