Latent class analysis identifies three subtypes of aggressive end-of-life care: a population-based study in Taiwan.

The aggressiveness of end-of-life (EOL) cancer care has often been analysed by the occurrence of several indicators, separately or aggregately. Whether aggressive EOL cancer care has different subtypes is unknown. This study sought to identify distinct subtypes of aggressive EOL care based on usage patterns of aggressive EOL-care indicators and to explore demographic, disease and treatment factors associated with the identified subtypes. This retrospective study linked data from 2001 to 2006 from three Taiwanese databases: National Registration of Death Database, Cancer Registration System and National Health Insurance claims database. Adult cancer patients (N=203,642) who died in 2001-2006 were selected. For these cancer patients' last month of life, we analysed eight indicators of aggressive EOL care: receiving chemotherapy, >1 emergency room visit, >1 hospitalisation, hospitalisation for >14 days, intensive care unit admission, received cardiopulmonary resuscitation, received intubation and received mechanical ventilation. Subtypes of aggressive EOL care were identified by latent class analysis. Among the study population, only 22.3% were treated by medical oncologists. Based on their profiles of EOL care, deceased cancer patients were classified into three subgroups: 'not aggressive' (45%), 'intent to sustain life' (33%) and 'symptom crisis' group (22%). Patients assigned to the 'intent to sustain life' group were less likely to have metastatic disease and to receive hospice care in the last year of life, but more likely to be cared for by non-medical oncologists, to die within 2 months after diagnosis and to die in hospital. EOL cancer care may be improved by understanding factors related to different subtypes of aggressive EOL care.