Objective
T
o apply group-based trajectory modelling
(GBTM) to the hospital administrative data to evaluate,
model and visualise trends and changes in the frequency
of long-term hospital care use of the subgroups of patients
with cerebrovascular conditions.
Design
A retrospective cohort stud
y of patients with
cerebrovascular conditions.
Settings
Secondar
y care of all patients with
cerebrovascular conditions admitted to English National
Hospital Service hospitals.
Participants
All pa
tients with cerebrovascular conditions
identified through national administrative data (Hospital
Episode Statistics) and subsequent emergency hospital
admissions followed up for 4 years.
Main outcome measure
Annual number of emergenc
y
hospital readmissions.
Results
GBTM model c
lassified patients with
intracranial haemorrhage (n=2605) into five subgroups,
whereas ischaemic stroke (n=34
208) and transient
ischaemic attack (TIA) (n=20
549) pa
tients were shown
to have two conventional groups, low and high impact.
The covariates with significant association with high-
impact users (17.1%) among ischaemic stroke were
epilepsy (OR 2.29), previous stroke (OR 2.18), anxiety/
depression (OR 1.63), procedural complication (OR
1.43), admission to intensive therapy unit (ITU) or high
dependency unit (HDU) (OR 1.42), comorbidity score
(OR 1.36), urinary tract infections (OR 1.32), vision loss
(OR 1.32), chest infections (OR 1.25), living alone (OR
1.25), diabetes (OR 1.23), socioeconomic index (OR
1.20), older age (OR 1.03) and prolonged length of stay
(OR 1.00). The covariates associated with high-impact
users among TIA (20.0%) were thromboembolic event
(OR 3.67), previous stroke (OR 2.51), epilepsy (OR
2.25), hypotension (OR 1.86), anxiety/depression (OR
1.63), amnesia (OR 1.62), diabetes (OR 1.58), anaemia
(OR 1.55), comorbidity score (OR 1.39), atrial fibrillation
(OR 1.27), living alone (OR 1.25), socioeconomic index
(OR 1.13), older age (OR 1.04) and prolonged length of
stay (OR 1.02). The high-impact users (0.5%) among
intracranial haemorrhage were strongly associated
with thromboembolic event (OR 20.3) and inversely
related to older age (OR 0.58).
Conclusion
GBTM effectively assessed trends in the
use of hospital care by the subgroups of pa
tients with
cerebrovascular conditions. High-impact users persistently
had higher annual readmission during the follow-up
period.