Shaun Zhai, MBChB,* Fergus Gardiner, BMSc, MBA, PhD(c),†,‡ Teresa Neeman, PhD,§
Brett Jones, BN,‖ and Yash Gawarikar, MBBS, MD†,§
Journal of Stroke and Cerebrovascular Diseases, Vol. ■■, No. ■■ (■■), 2017: pp ■■–■■ 1
ARTICLE IN PRESS
2 S. ZHAI ET AL.
outcomes, and, more importantly, to concurrently analyze or intern doctor. The neurologist may have been con-
the financial cost of the operation of an SU in real life, sulted for the patient’s care during the admission; however,
thereby assessing its cost-effectiveness. It was hypoth- this practice was highly variable. A general ward phys-
esized that the SU will provide a cost-effective improvement iotherapist, occupational therapist, or a speech therapist
in patient outcomes. It was hoped that this study may was involved in the patient care, upon referral from the
provide further justification on the implementation of SUs treating team.
in the Australian health system. Patients were eligible for inclusion in the study if they
had any type of stroke, including hemorrhagic or isch-
Methods emic, as well as transient ischemic attacks (TIAs) (defined
as <24 hours neurologic deficit caused by a focal lesion
The researchers conducted an observational study,
in areas including the brain, the retina, and the spine).
comparing stroke patients who have been admitted to
In the post-SU phase, as data were prospectively re-
Calvary Public Hospital, Canberra, before and after the
corded, any presentation that resembled a stroke with
establishment of the local SU from October 2013 to October
subsequent admission to the SU was included. For some
2014 (52 weeks). The hospital ethics committee approv-
of the stroke-like presentations, an alternative diagnosis
al was received for data collection, analysis, and publication
than stroke was identified eventually. These were clas-
(reference number 24-2014).
sified as stroke-mimics and were included in the analysis.
The pre-SU patients were identified retrospectively
If the patient was transferred to another facility after pre-
by matching International Classification of Diseases-10
senting briefly to Calvary Hospital (i.e., less than 24 hours)
(ICD-10) codes for cerebrovascular diseases (ICD-10:
and was not admitted to the local SU, they were ex-
I60-I69). On reviewing physical and electronic clinical
cluded from the study. Patients who were transferred to
records, patient files were retrieved and their diagnoses
Calvary Hospital from another health institution for re-
reviewed. Specifically, investigators evaluated all pre-
habilitation or placement purposes after a stroke were
and post-SU clinical notes to ensure the accuracy of the
also excluded.
diagnosis. The investigators adjudicated cases where the
ICD-10 coding was incorrect based on clinical documen-
tation, imaging, and other investigation findings. The Study Endpoints
disease characteristics, investigations, treatment, and The primary endpoint for the study was to compare
outcomes were recorded. The data of the post-SU pa- the modified Rankin Scale (mRS) at 90 days after the stroke
tients were collected prospectively by the stroke neurologists in pre- and post-SU patients. Patients were identified ret-
and stroke nurse specialist. rospectively by ICD-10 code, and as such, the 90-day mRS
outcomes were obtained via the following measures:
Study Population
• Neurology/stroke follow-up clinic at 90 days;
The SU is on a dedicated ward in the hospital and has • If patients were inpatients at day 90 either for the
4 beds equipped with continuous cardiac monitoring. same admission, for rehabilitation purposes, or for
There is a computed tomography (CT) scanner and a readmission purposes, their mRS scores were evalu-
magnetic resonance imaging (MRI) scanner on site for ated based on clinical notes;
neuroimaging support. The SU is staffed by a stroke neu- • If patients had died within a 90-day period;
rologist, a stroke nurse specialist, and specially trained • If patients came for a clinic appointment and the
nurses. The SU also receives input from the allied health clinical documentation was adequate for an mRS
team for speech therapy, dietary therapy, occupational evaluation;
therapy, and physiotherapy. The SU was established on • Delayed follow-up visits, with patient recollection
April 28, 2014; therefore patients presented to Calvary of their neurologic status at the 90-day mark after
Hospital in the 6-month period between October 28, 2013 a stroke.
and April 27, 2014 were referred to as the pre-SU group, The main secondary endpoint was to compare the
whereas for the 6 months between April 28, 2014 and length of stay (LOS) and the in-hospital cost of each stroke
October 27, 2014, it was referred to as the post-SU group. patient during the pre- and post-SU periods. Other sec-
In the post-SU period, all stroke patients were admitted ondary endpoints included the timeliness and completeness
to the SU in the first instance. Once stabilized the patient of investigations, adherence to best medical therapy,
could be discharged directly from the SU or stepped down thrombolysis rates, and allied health involvement. Out-
to a conventional ward. The stroke service provided the comes at discharge from the stroke service (stroke service
entire acute care. is different from SU, it includes whole period patient in
In the pre-SU period, stroke patients were treated by hospital for SU and subsequent conventional ward care)
various general physicians (internists). These physicians and nosocomial infections were also analyzed. The dis-
often had various specialty interests (but not neurology). charge destination involved the patient’s transfer location
Under these physicians, there was a registrar and a resident upon departure from the stroke service. In regard to the
ARTICLE IN PRESS
COST-EFFECTIVENESS OF A STROKE UNIT 3
pre-SU patients, the physical discharge of the patients Table 1. Baseline age characteristics of all patients under
from their treating team was regarded as the equivalent stroke service (stroke-mimic patients included)
data point.
Nosocomial infections were grouped into 3 categories: Pre-SU Post-SU (n = 186)
pneumonia, urinary tract infections (UTIs), and others. (n = 103) (all types)
Others often included cannula-associated infections,
Mean (SD) 75.5 (12.8) 68.3 (15.7)
cellulitis, etc. All infective events were reviewed and ad- Median 77 70
judicated by the investigator. If the patient had a new Min 31 18
lung infiltrate plus clinical evidence that the infiltrate is Max 98 95
of an infectious origin, which fulfills at least 2 of the 4 IQR 25/75 69/84 57/80
criteria, including the new onset of fever, purulent sputum,
leukocytosis, and decline in oxygenation, then a diag- Abbreviations: IQR, interquartile range; Max, maximum; Min,
nosis of pneumonia was made. If the clinical criteria were minimum; SD, standard deviation; SU, stroke unit.
not fully met, then the diagnosis was at the discretion
of the treating physician.12 For the diagnosis of a UTI, Table 2. Baseline age characteristics of ischemic stroke
the patient had to have a positive urinalysis or a urine patients only
culture, with the presence of symptoms.13 If the treating
physician initiated a course of antibiotic treatment based Pre-SU (n = 54) Post-SU (n = 87)
on clinical evidence or suspicion, which did not fulfill (ischemic stroke) (ischemic stroke)
the above criteria, then these events were grouped as
“others” in the nosocomial infection analysis. Mean (SD) 75.5 (10.9) 72.3 (15.7)
Median 76 75
Statistical Analysis Min 45 24
Max 98 95
EPI Info 7.0 software (Centers for Disease Control and IQR 25/75 68/84 64/84
Prevention, Atlanta, GA) was used for the data acquisi-
tion of all patients in the study. Prespecified characteristics Abbreviations: IQR, interquartile range; Max, maximum; Min,
minimum; SD, standard deviation; SU, stroke unit.
of each patient were recorded in our Stroke Study da-
tabase. Ordinal regression was used to compare mRS
outcomes in ischemic stroke patients between the 2 periods.
Average hospital length of stay between the 2 periods 76 and 75, respectively; interquartile range (IQR 25/75)
was compared using a linear model, adjusting for type. was 68/84 and 64/84. Of the pre-SU patients, 56.3% were
Subgroup analyses by type were performed post hoc. In- male, compared to 56.6% in the post-SU group.
vestigation rates, treatments, infection rates, and destination As shown in Table 3, in the pre-SU group, types of
following hospital discharge were compared between events were divided into TIA 38.8%, ischemic stroke
periods using Pearson chi-squared tests. SPSS version 23.0 52.4%, hemorrhagic stroke 6.8%, and stroke mimic 1%.
software (IBM SPSS Statistics for Windows, IBM Corp., In the post-SU phase, the types of events were TIA
Armonk, NY) was for all statistical modeling. Observed 28.0%, ischemic stroke 46.8%, hemorrhagic stroke 2.7%,
difference were considered significant if P < .05. and stroke mimic 22.0%.
Results
There were 112 patients identified in the pre-SU period, Table 3. Percentage and frequency of types of events for all
and subsequently 9 were excluded. In the post-SU period, patients under stroke service before and after the stroke
197 patients were recorded and 11 were excluded. The unit establishment
majority (90%) of excluded cases were transfers from
another facility for rehabilitation or placement pur- Post-SU
poses. The remaining exclusions included missing data Pre-SU (n = 186)
(N = 1) and duplicated record (N = 1). There were 103 (n = 103) (all types)
patients included in the pre-SU group compared to 186 % (number) % (number)
in the post-SU group. The mean age (and the standard
Type of TIA 38.8 (40) 28.0 (52)
deviation) was 75.5 (12.8) in pre-SU patients and 68.3 (15.7)
events Ischemic stroke 52.4 (54) 46.8 (87)
in the post-SU group (Table 1). In the pre- and post-SU
Hemorrhagic stroke 6.8 (7) 2.7 (5)
groups, 54 and 87 patients had ischemic stroke, respec- Stroke mimic 1.0 (1) 22.0 (41)
tively. When comparing ischemic stroke patients only, the Undetermined 1.0 (1) .5 (1)
mean age was 75.5 in the pre-SU group and 72.3 in the
post-SU group, as shown in Table 2. The median age was Abbreviations: SU, stroke unit; TIA, transient ischemic attack.
ARTICLE IN PRESS
4 S. ZHAI ET AL.
Figure 1. Modified Rankin Scale at 90 days of pre- and post-SU ischemic stroke patients.
Study Outcomes days in the SU, and a further 1.9 days on the conven-
tional ward. The total cost per stroke patient amounted
The ordinal analysis of the functional status of stroke
to $6061 (AUD) for the post-SU period. This is in com-
patients, in terms of mRS, at 90 days, demonstrated a
parison to the total amount of $6382 (AUD) for the cost
significant trend favoring a better outcome in the post-
per stroke patient with conventional medical ward care
SU period (P = .01) (Fig 1). Ninety days after initial stoke,
considering the average LOS of 9.7 days (Table 5).
70% of all patients were followed-up to determine their
Other secondary endpoint measurement can be divided
functional status and mortality data. There were no sig-
into 3 categories: investigations, therapeutics, and imme-
nificant differences between the follow-up rate between
diate outcomes. There were significantly more MRI
the pre- and post-SU groups.
and CT angiogram scans of the brain carried out in the
There was a significant improvement in the LOS of
post-SU group; 74.7% versus 48.5% for the former (P < .001)
stroke patients from 9.7 days in the pre-SU group com-
and 71.5% versus 4.9% for the latter (P < .001) (Table 6).
pared with 4.6 days in the post-SU group (P = .001). During
There was a nonsignificant reduction in CT head acqui-
the post-SU phase, stroke patients spent on average 2.7
sition in the same period. The door to CT acquisition time
days in the SU (median 2, range 1-13, standard devia-
(from the patient presenting to the emergency depart-
tion 2.4). Subgroup analysis of the LOS data showed that
ment doorstep to obtaining a CT scan) had been reduced
the improvement was predominant in TIA (3.8 days
from 135 to 89 minutes (P = .004). When comparing the
pre-SU and days 1.7 post-SU, P = .16) and ischemic stroke
completion rate of secondary preventative investigations,
patients (13.8 days pre-SU and 5.9 days post-SU, P < .001)
there was a significant improvement in multiple areas. These
(Table 4). The average cost of patient care on a normal
included the increased carotid imaging from 75.7% to 89.3%
medical ward amounted to approximately $660 (AUD)
(P = .007), cardiac monitoring from 43.7% to 100% (P < .001),
per patient day, whereas the cost averaged $1317 (AUD)
per day on the SU mainly due to higher staffing ratio
and resource distribution. Despite the significantly higher Table 5. Breakdown of total cost per patient admission with
cost of care in the SU, the duration of stroke patients’ conventional ward care versus stroke unit care
admission was substantially shorter—an average of 2.7
Conventional Stroke
ward care unit care
Table 4. Average length of stay (LOS) for pre- and $ (AUD) $ (AUD)
post-SU patients
Nursing $1611.5 $1824.9
Days—mean (SE) Pre-SU Post-SU P value Medical staff $808.7 $895.5
Allied health $721.6 $601.8
TIA 3.8 (.93) 1.7 (.13) .18 Imaging $737.6 $1041.5
Ischemic stroke 13.8 (2.24) 5.9 (.57) <.001 Pathology $978.9 $366.2
Hemorrhagic stroke 11.7 (4.89) 13.2 (5.76) 1.0 Pharmacy $386.5 $531.1
Stroke mimic 2.4 (.55) Miscellaneous $1137.5 $799.7
All strokes 9.7 (1.35) 4.6 (.37) .001 Total cost per admission $6382 $6061
Cost per admission day $660 $1317
Abbreviations: SE, standard error; SU, stroke unit; TIA, transient
ischemic attack. Abbreviation: AUD, Australian dollar.
ARTICLE IN PRESS
COST-EFFECTIVENESS OF A STROKE UNIT 5
Table 6. Completion rate and timeliness of CT and MRI Table 8. Treatment (pharmacologic and nonpharmacologic)
imaging for all strokes/TIAs for ischemic stroke patients
Pre-SU (%) Post-SU (%) Table 10. Nosocomial infection rates for stroke patients
(n = 94) (n = 139) P value
Pre-SU Post-SU
Carotid imaging 75.7 89.3 .004 (%) (%) P value
Cardiac monitoring 43.7 100 <.001
Fasting blood sugar 39.8 94.1 <.001 Nosocomial infection 23.8 9.2 .03
and lipids Pneumonia 11.5 2.2 .03
Transthoracic 65.1 64.0 .9 (including aspiration)
echocardiogram UTI 4.9 6.5 .9
Abbreviations: SU, stroke unit; TIA, transient ischemic attack. Abbreviations: SU, stroke unit; UTI, urinary tract infection.
ARTICLE IN PRESS
6 S. ZHAI ET AL.