Comparison of statistical methods for calculating life expectancy after spinal cord injury

Study design Retrospective observational. Objectives To compare results of several different methods for calculating life expectancy in the same sample of people with spinal cord injury (SCI), and critically assess their advantages and disadvantages. Setting Two spinal centres in Great Britain. Methods Survival status of persons with traumatic SCI injured between 1943 and 2010 with follow-up to 2015 was determined. Standardised mortality ratios (SMRs) were calculated using age at injury and current (attained) age, and compared. Life expectancy was then estimated using the SMR methods and compared with the results of a method based on multivariate logistic regression of a person-year dataset. Life expectancy estimates calculated by applying SMRs based on current age to general population period (current) and cohort (projected) life tables were also compared. Results The estimated life expectancies were significantly higher when the SMRs were based on age at injury. They were also higher when a general population cohort life table was used, particularly for younger ages. With the exception of the ventilator-dependent group, the life expectancy estimates derived from logistic regression were slightly lower than those derived from SMRs based on current age and a general population period life table. Conclusions The multivariate logistic regression of person-years method offers several advantages compared to the SMR method for calculating life expectancy after SCI, the main ones being: greater statistical power and precision with smaller sample sizes, the ability to include more predictive factors and to distinguish the otherwise confounded effects of current age, time post-injury, and calendar time.