Hospital Pharmacy - September 2017 - 547

547

Hemenway and Terry
trial, additional corticosteroids were allowed, which resulted
in a cumulative mean of 379 mg in the short-course group and
793 mg in the long-course group.5 And finally, in the United
States, there is a concern regarding rates of 30-day hospital
readmission for patients with AECOPD, which was not
assessed by the REDUCE trial.7 Our study was performed to
determine whether, in a US community hospital, low doses of
corticosteroids provide the lowest risk of adverse effects
without increasing length of stay or readmission rate.

Methods
Patients and Setting
Institutional review board approval was obtained on August
28, 2015. Patients were selected from a report generated
from our electronic health record using the International
Classification of Diseases, Ninth Revision, code of 491.21
between May 1, 2013, and July 31, 2015. Patients were
included in the study if they met clinical criteria for AECOPD
defined as 2 of the following: increase in dyspnea, increase
in cough frequency, increase in cough strength, or increase in
amount or purulence of sputum. They also needed to receive
at least 1 dose of a systemic corticosteroid. Patients were
excluded if they were admitted to an intensive care unit or
the step-down intensive care unit. They were also excluded if
they were discharged with hospice, or died during the index
admission. Data were collected using a retrospective chart
review. A total of 665 charts were reviewed, with 369
included in the final analysis. Data abstraction was performed by the 2 authors using a detailed protocol.
The data were split into 3 groups: low: ≤250-mg prednisone equivalents, medium: 251- to 500-mg prednisone
equivalents, and high: ≥501-mg prednisone equivalents.
These groups were chosen using the REDUCE trial's planned
short-course total of 200-mg prednisone, actual short-course
mean total of 379 mg due to the allowance of extra steroids,
and the planned long-course total of 560 mg.5 Less than or
equal to 250 mg was chosen as our low-dose group in place
of 200 mg because providers at our hospital often give a first
dose of methylprednisolone and this inclusion would put our
lowest dose over the 200-mg cutoff. Inpatient doses that
were documented on the medication administration record
were included.

Endpoints and Measurements
This study was undertaken to determine whether a lower
dose of corticosteroid was as effective as higher doses, while
providing a lower risk of adverse effects. The primary effectiveness outcomes were rates of 30-day hospital readmission
and length of stay. The adverse effect measured was an elevation of blood glucose greater than 30% above the patient's
baseline. To determine the 30-day readmission rate, we
reviewed the electronic health record for any inpatient

readmissions to our facility within 30 days. Length of stay in
hours was included on the initial report generated from our
electronic health record. An increase of blood glucose greater
than 30% above baseline was determined by dividing the
patient's maximum blood glucose during their hospitalization by the first recorded for the index admission. We chose
to use a 30% increase from each patient's baseline based
upon data from Islam et al, who found an average increase of
~30% in blood glucose even with low doses of corticosteroids, and higher increases seen with larger doses.8 Only
blood glucose concentrations drawn as part of the patient's
metabolic panel were included; point-of-care glucose tests
were not used.
Baseline characteristics of age, sex, diagnosis of chronic
obstructive pulmonary disease (COPD) verified with spirometry, use of home oxygen, oxygen use (L) at admission,
LACE readmission risk score (Length of stay, Acuity of
admission, Comorbid conditions, number of Emergency
department visits), and Charlson comorbidity score were collected.9,10 Potential confounders were collected, which
included if the patient had concomitant diabetes mellitus,
notation of hemoglobin A1c within 6 months of admission to
help assess control of diabetes, if antibiotics were given during admission, concomitant diagnosis of pneumonia verified
with a chest radiograph, concomitant diagnosis of acute
decompensated heart failure (ADHF) verified with a
N-terminal pro-B-type natriuretic peptide level greater than
their age-adjusted range, and if they were a current smoker.

Statistical Analysis
All statistical tests were run using IBM SPSS version 24
(IBM, Inc, Armonk, New York). The significance level was
determined a priori for the univariate and multivariable tests
using a Bonferroni adjusted alpha level of P < .0167 per outcome. Patient baseline characteristics and potential confounders were described using descriptive and univariate
statistical tests (1-way analysis of variance for continuous
data and chi-square for categorical data). Binary logistic
regression was performed for the rate of readmission within
30 days, which included covariates that could be considered
impactful in addition to the dose groups. From those potential covariates, a backward Wald χ2 process was used to
determine the variables included in the final model. A similar
process was performed for the rate of blood glucose elevation greater than 30%. For length of stay, multivariable linear
regression was performed. A backward process was utilized
to determine the variables included in the final model, and
collinearity diagnostics were determined. For readmission
and length of stay, the covariables included age, sex, confirmed diagnosis of COPD, home oxygen use, amount of
oxygen used at admission, LACE, Charlson comorbidity
score, antibiotic use during admission, confirmed ADHF,
confirmed pneumonia, and whether they were a current
smoker. For increase in blood glucose, the covariables

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