Hospital Pharmacy - September 2017 - 549

549

Hemenway and Terry
Table 4. Binary Logistic Regression Predicting Elevated Blood
Glucose From Covariables and Dose Groups.
Predictor
Age
Charlson score
Dose groups
Medium
High

Wald χ2

−0.015
0.165

2.930
5.728

.087
.017

0.870
1.179

6.902
14.327

.009
<.001

Table 5. Linear Regression Predicting Length of Stay (Hours)
From Covariables and Dose Groups.
Predictor
Age
LACE
Charlson score
Confirmed ADHF
Confirmed pneumonia
Confirmed smoker
Dose groups
Medium
High

0.419
6.580
−5.977
19.456
10.560
−9.752

2.321
6.225
−3.822
2.599
1.821
−2.097

.021
<.001
<.001
.01
.07
.037

−2.158
21.313

−0.348
3.675

.728
<.001

Note. LACE = Length of stay, Acuity of admission, Comorbid conditions,
number of Emergency department visits; ADHF = acute decompensated
heart failure.

Discussion
All data were collected from our community, regional medical center. The overall severity of our population, as determined by the average Charlson score, is typical for a
community medical center.11 In addition, our readmission
rate for AECOPD is similar to other published data; however, one limitation of our study is that we were only able to
determine readmission to our own facility.12 The focus of our
study was patients admitted to our medical, cardiac, or surgical floors. Patients admitted to a higher level of care were
excluded, and the potential need for higher doses of corticosteroids in this population was not determined in this study.
Our medical center does not currently provide a suggested
corticosteroid dose for AECOPD. This leads to a wide range
of doses used in the study. This variety helped make our
study possible, and these data will help our hospital make an
informed decision on the optimal dose. However, the dose
range skewed toward the high-dose (>500 mg) group, which
could have had implications on our sample size. We did not
find a statistically significant difference in readmission rates.
However, it is unknown whether this is due solely to there
being no difference, or whether data availability contributed.
We were also limited in the amount of data available, which
was determined by the date we converted to an electronic
health record.
This study was performed as a retrospective, single-center
trial. Readmission is a multifactorial outcome, with facility

differences that can affect readmission rates.13,14 We decided
to limit our study to just 1 site to attempt to control for facility factors related to readmission. We also used the LACE
tool to help compare readmission risk for the different
groups. However, there are data suggesting that LACE does
not determine readmission risk as well in certain populations, such as heart failure.15 It is unknown how this variability may have affected our readmission outcome data. As the
data were collected retrospectively, a list of potential covariables and confounders were collected. Although several of
the baseline characteristics were similar, we did find that the
high-dose group had a higher percentage of patients who had
COPD verified through spirometry, and more use of home
oxygen. In addition, the low-dose group had a higher percentage of patients who had a concomitant diagnosis of
ADHF. Taken together, it is possible that the higher dose
group had a portion of patients with a more reliable diagnosis
of COPD. We attempted to control for these group differences by using multivariable regression analysis.
Our study found that the low-dose group was associated
with a lower amount of patients who experienced a blood
glucose increase of >30% from baseline as compared with
both the medium- and high-dose groups. This finding is consistent to those found in some studies, but not others.5,8,16,17
This potential difference may be due to the way we determined an increase in blood glucose, which differed compared with other studies.5,16 The use of a 30% increase from
each patient's baseline allowed patients to act as their own
reference point, and assess impact of the corticosteroids
instead of overall control of diabetes.8 This potentially provided a more even comparison while including both patients
with and without diabetes. We also chose to use blood glucose concentrations from their daily metabolic panel which
may have provided more consistency between concentrations. Our results may have been affected by additional blood
glucose data points in patients with longer length of stays
which could have led to a larger chance for a higher blood
glucose ratio.
We did not find a statistically significant association
between readmission rate and dose group, even when holding possible covariables constant. This is similar to the findings in the REDUCE trial, which did not directly evaluate
readmission, but assessed re-exacerbation.5 It is also consistent with the results of 2 large meta-analyses.2,3 Our evaluation of differences in length of therapy found that the highest
dose group was associated with the longest length of stay,
while holding other covariables constant. As our study is a
retrospective cohort, we are only able to ascertain an association between lengths of stay and dose group, not determine a
direct cause and effect. However, as this was not seen with
the low- or medium-dose group, and similar results were
seen in the REDUCE trial, we feel confident that using lower
amounts of corticosteroids should not lead to an increase in
length of stay.5 Future evaluation of length of stay differences after implementation of a lower dose protocol could

Table of Contents for the Digital Edition of Hospital Pharmacy - September 2017