Vol. 0 No.

0 Month 2016 1

Blood Glucose Control in Noncritically Ill

Patients Is Associated With a Decreased
Length of Stay, Readmission Rate, and
Hospital Mortality
Adrienne Z. Ables, Patricia J. Bouknight, Heather Bendyk, Rebecca Beagle, Rebecca Alsip, Jill Williams

Purpose Purpose: Multiple studies have shown that hyperglycemia

Our primary objective was to determine correlates with mortality and morbidity in critically ill patients.
whether keeping glucose values below 180 This has not been demonstrated in noncritically hospitalized
mg/dL decreased the length of hospital patients. The primary objective of this study was to determine
stay in hospitalized non-intensive care unit whether glycemic control shortens the length of stay (LOS).
(ICU) patients. Secondary objectives Secondary objectives included assessing readmissions, in-
included assessing readmission within 30 hospital mortality, and rates of hypoglycemia.
days, in-hospital mortality, and rates of Methods: A retrospective review of hospitalized patients
hypoglycemia. admitted between 2008 and 2012 with fingerstick blood sugar
(FSBS) was performed. Patients were divided into two groups:
“controlled” FSBS ($80% of FSBS were ,180 mg/dL) and
Review of the Literature “uncontrolled” FSBS (,80% of FSBS were ,180 mg/dL). The
Diabetes mellitus is a chronic illness that is average LOS (ALOS) in days, in-hospital mortality, readmission
quickly becoming more prevalent in the rates, and rates of hypoglycemia was compared.
United States. According to the Centers for Results: A total of 32,851 patient records were reviewed. ALOS
Disease Control (CDC), diabetes mellitus for patients with controlled and uncontrolled FSBS was 5.86
affects 29.1 million people greater than 20 and 6.17 days, respectively (p , .0001). Readmission within 30
years of age nationwide and is the country’s days and hospital mortality were significantly lower in patients
seventh leading cause of death (Centers for with controlled FSBS (p = .0000, .00001), whereas rates of
Disease Control and Prevention, 2014). In hypoglycemia were significantly higher in the uncontrolled
2009, there were 688,000 hospital dis- group (p = .00000).
charges with diabetes as the first-listed Conclusions: Glycemic control was associated with decreased
diagnosis with an average length of stay LOS, hospital mortality, and 30-day readmission rate in non-
(ALOS) of five days (Centers for Disease critically ill patients regardless of the presence or absence of
Control and Prevention, 2014). Hypergly- diabetes.
cemia during hospitalization is associated
with worse outcomes such as higher
in-hospital mortality and LOS, even in non-
diabetic patients (Bruno et al., 2008; Um- therapy, defined as blood glucose targets of
pierrez et al., 2002). Furthermore, higher less than 140 mg/dL for all hospitalized
mortality rates have been demonstrated in patients. This recommendation was based
patients hospitalized with community- on a systematic review, which showed that
acquired pneumonia and chronic obstruc- intensive insulin therapy did not clearly KeyWords:
tive pulmonary disease (COPD) who pre- show beneficial health outcomes but diabetes mellitus
sented with hyperglycemia at the time of increased the risk of severe hypoglycemia glycemic control
admission (Baker et al., 2006; Lepper et al., (Kansagara et al., 2011; Qaseem et al., length of stay
2012; McAlister et al., 2005). 2011). This study did not include trials in the
In 2011, the American College of Physi- general medical ward setting. Despite the Journal for Healthcare Quality
Vol. 0, No. 0, pp. 1–8
cians (ACP) published clinical guidelines paucity of evidence for management of © 2016 National Association for
recommending against intensive insulin patients in the noncritically ill setting, the Healthcare Quality
2 Journal for Healthcare Quality

American College of Endocrinology (ACE) study if they were pregnant, required any
and the American Diabetes Association form of critical care or ICU admission
(ADA) advocated for more intensive therapy during the hospital stay, or were admitted
in the general inpatient setting (Moghissi to hospice.
et al., 2009). The Joint Commission on
Accreditation of Healthcare Organizations
(JCAHO) followed suit, launching an ini- Data Collection
tiative on inpatient glycemic control (Wexler Data of eligible patients were collected
and Cagliero, 2007). using the Horizon Lab integrated Labo-
Most healthcare systems are graded ratory Information Systems (LIS) software
based on the recommendations given by in conjunction with Integrated Medical
the American Diabetic Association, which Software Solution (MIDAS). These data
recommended target blood sugar values were de-identified by randomly assigning
less than 180 mg/dL for inpatient man- a number specific for the individual
agement in noncritically ill patients. This patient and removing all protected health
recommendation was adopted by the information.
Diabetes Initiative of South Carolina as the We collected the following data: age,
guideline that hospitals and other health- sex, diagnosis, FSBS, LOS, hypoglycemic
care organizations should follow (Diabe- episode(s), and outcome (discharge,
tes Initiative of South Carolina, 2015). Our death, and/or readmission within 30 days).
institution adopted a diabetes manage- The LIS was used to extract FSBS levels,
ment program to track point-of-care glu- whereas the MIDAS software was used to
cose values measured for all inpatients collect patient demographics and out-
starting at 48 hours after admission comes. Most frequent diagnoses and
as a quality measure for physician severity of illness were determined using
performance. the 3M All Patient Refined Diagnosis
Related Groups (APR-DRG) severity of ill-
ness scoring system. Severity of illness
Study Design and Methods scores range from one to four, one
being minor, and four being extreme, and
Study Design take into account the patient’s primary
We conducted a retrospective review of diagnosis and underlying conditions. (APR-
medical records from hospitalized pa- DRGs, 2014) These data were exported into
tients in whom blood glucose monitoring an Excel spreadsheet for analysis. All data
was performed. were stored on a password-protected file
belonging to the investigators.

Sample
Our study was conducted at Spartanburg Data Analysis
Medical Center (SMC), a 588-bed tertiary Blood sugar readings were defined as
care hospital in upstate South Carolina. controlled if 80% or more of FSBS were
Data were abstracted from all patients between 71 and 180 mg/dL, and uncon-
admitted to general medical, surgical, or trolled if less than 80% FSBS fell within
psychiatric units between January 1, 2008, this range, as per hospital policy. Hypo-
and December 31, 2012. Patients’ charts glycemia was defined as FSBS as less than
were reviewed if they were 18 years of age 71 mg/dL. These values are aligned with
and older, and had a fingerstick blood targets set by the ADA and ACP. Demo-
sugar (FSBS) checked on the nursing unit graphics of the two samples were analyzed
after 48 hours of admission. Patients who using frequency statistics for categorical
were admitted for diabetic ketoacidosis or variables and measures of center/spread
hyperosmolar coma but did not require for continuous variables. We compared
care in the ICU were also eligible for the the difference in LOS between the groups
study. Patients were excluded from the using the Wilcoxon Rank-Sum Test.
 Vol. 0 No. 0 Month 2016 3

Mortality, readmission rates and rates of nificantly lower in patients whose FSBS
hypoglycemia were analyzed using a bino- were 71 to 180 mg/dL, 80% or more of the
mial proportion test. Diagnoses and time (Table 4). Of the most common APR-
severity of illness as they related to LOS, DRGs, mortality rates were only signifi-
hospital mortality, and readmission rates cantly higher in uncontrolled patients with
were compared by Chi-Square analysis. All COPD, and 30-day readmission rates were
calculations were performed with a = 0.05 significantly higher in uncontrolled pa-
as the level of significance. tients with pneumonia only (Table 5).
Rates of hypoglycemia were signifi-
cantly higher in the group with uncon-
Institutional Review Board trolled FSBS regardless of age (Table 6).
This study was approved by the Spartan-
burg Regional Healthcare System Institu-
tional Review Board as exempt research. Limitations
We recognize the limitations of our study.
First, it was a retrospective noninterven-
Results tional investigation as opposed to the gold
Between January 1, 2008, and December standard prospective, randomized, placebo-
31, 2012, there were 32,851 inpatients with controlled study. Therefore, definitive
FSBS monitoring. The average age of cause and effect conclusions cannot be
these patients was 65 years; 56% were drawn from the results. Additionally, we did
female. Patient demographics were not not examine treatment of hyperglycemia
significantly different in patents with con- with insulin or other medication adminis-
trolled and uncontrolled FSBS readings tration such as corticosteroids that might
(Table 1). There were significantly more have affected the results. We did not docu-
patients with controlled FSBS in both the ment patients with preexisting diabetes.
male and female populations. The higher However, we did not set out to study the
the severity of illness score, the greater the diabetic population but all patients with
likelihood of uncontrolled blood glucose hyperglycemia. Finally, because we enrolled
levels (Table 2). patients after 48 hours of admission, we may
ALOS, our primary outcome, was sig- have selected a sicker population than pre-
nificantly shorter in patients with con- vious studies.
trolled blood glucose levels compared
with patients whose FSBS were uncon-
trolled (Figure 1). In patients within one of Directions for Future Research
our most common APR-DRGs, LOS was A prospective, randomized controlled
significantly shorter in the controlled study evaluating routine testing of FSBS in
group (Table 3). noncritically ill hospitalized patients
Overall, hospital mortality and read- would help to definitively answer the
mission rates within 30 days were also sig- research question.

Table 1. Patient Demographics

Characteristics Controlled FSBS Uncontrolled FSBS p*
Age (years) 64.49 64.63 .97
Gender (n)
Male 7,508 7,034 .022
Female 9,685 8,624 .022
*
a = 0.05.
FSBS, fingerstick blood sugar.
4 Journal for Healthcare Quality

Table 2. APR-DRG Severity of Illness

1 (N) 2 (N) 3 (N) 4 (N)
Controlled FSBS 1,330; 8.31% 5,746; 35.91% 7,061; 44.13% 1,861; 11.63%
Uncontrolled FSBS 714; 4.80% 4,319; 29.03% 7,363; 49.50% 2,480; 16.67%
1 = minor, 2 = moderate, 3 = major, 4 = extreme.
p , .0001.
FSBS, fingerstick blood sugar.

Discussion blood sugars were controlled during their

hospital stay than in those who were
Length of stay uncontrolled. This is consistent with pre-
The difference in LOS between the two viously published data regarding hypergly-
patient groups, although statistically signifi- cemia and in-hospital survival in patients
cant, may seem trivial (.31 days, P , .0001). and independent of age, sex, underlying
However, inpatient bed capacity and occu- severity of illness, and previous history of
pancy are significant factors in providing diabetes (Baker et al., 2006; Bruno et al.,
quality care. LOS is an indicator of effective 2008; Evans and Dhatariya, 2012; Lepper
capacity management and throughput in et al., 2012; McAlister et al., 2005;
healthcare organizations. A .31-day reduction Umpierrez et al., 2002). These studies were
means that the hospital earns an additional either retrospective in nature or pro-
patient day every 3 days. Thus, the hospital spective cohort studies without an inter-
has effectively earned an additional 113 ventional component. In a systematic
patient days for a year with the LOS reduction review of 19 studies assessing the effect of
as noted. Optimizing inpatient throughput intensive therapy to achieve tight glycemic
enhances patient access, reduces unit cost, control in hospitalized patients, no studies
and improves service levels (Thomas, 2006). used death as an outcome in nonsurgical,
Additionally, average LOS is used as a hospi- noncardiac patients on a general medical
tal efficiency measurement by payors in the floor (Murad et al., 2012). Although ret-
current pay for performance healthcare rospective, our study was specifically de-
environment (Cromwell et al., 2011). signed to assess mortality rates in patients
on general medical wards with FSBS
checked after 48 hours of admission rather
Mortality than on admission. This indicates that not
Overall hospital mortality rate in this study only admission hyperglycemia is a risk fac-
was significantly lower in patients whose tor, as demonstrated by previous studies
(Baker et al., 2006; Lepper et al., 2012;
McAlister et al., 2005; Umpierrez et al.,
Figure 1. Length of Stay in Days. 2002), but ongoing hyperglycemia also
FSBS, Fingerstick contributes to in-hospital mortality.
Blood Sugar.

Readmission Rates
In 2010, hospital readmissions cost the
Medicare program $17.5 billion (Centers
for Medicare and Medicaid Services,
2014). Centers for Medicare and Medicaid
Services are attempting to identify bene-
ficiaries at risk of readmission and target-
ing resources to prevent these patients
from returning to the hospital. Few studies
 Vol. 0 No. 0 Month 2016 5

Table 3. Length of Stay in Days by Top 5 APR-DRG Diagnoses

Controlled Uncontrolled
APR-DRG FSBS FSBS p*
COPD 4.78 5.53 ,.0001
Heart failure 5.28 6.04 ,.0001
Pneumonia 5.43 6.09 ,.0001
Pulmonary edema and respiratory failure 5.45 6.29 ,.0001
Renal failure 5.92 7.06 ,.0001
*
a = 0.05.
COPD, chronic obstructive pulmonary disease; FSBS, fingerstick blood sugar.

have examined the relationship between Hypoglycemia

hyperglycemia and readmission rates. In 1 Intensive insulin therapy has been associ-
analysis of 1,502 patients admitted to ated with higher rates of hypoglycemia
a general acute medical unit, 28-day re- (Kansagara et al., 2011; Murad et al., 2012).
admission rates were significantly higher Although our study did not examine
when admission blood glucose values were treatment of hyperglycemia, patients with
greater than 162 mg/dL (Evans and uncontrolled blood sugars had a signifi-
Dhatariya, 2012). Rates were higher in cantly higher rate of hypoglycemia. This
persons with values greater than 360 mg/ result is consistent with previous studies
dL, 15% versus 6% in patients with values and may have been due to the aggressive
less than 126 mg/dL. Godar and col- treatment for high blood sugar readings
leagues demonstrated that in patients (Murad et al., 2012).
admitted to the hospital with community-
acquired pneumonia, serum glucose levels
on admission were not associated with 30- Discharge Diagnosis/Severity of Illness
day readmission (Godar et al., 2011). In Patients with a discharge diagnosis of
our study, 30-day readmission rates in pa- COPD, heart failure, pneumonia, pulmo-
tients with uncontrolled blood sugars were nary edema/respiratory failure, and renal
significantly higher than in persons with failure were more likely to have uncon-
FSBS less than 180 mg/dL, 16.6 % versus trolled FSBS. Although not documented
14.1%, respectively. Routinely monitoring in our study, we can postulate a few rea-
and controlling FSBS in noncritically ill sons for some of these glucose excursions.
patients would target at least 1 risk factor It is probable that patients with COPD
for hospital readmission. Whether poorly were treated with systemic corticosteroid
controlled blood sugars after hospital dis- therapy during their hospital stay, making
charge impact readmissions remains to be blood sugar control challenging. Addi-
determined. tionally, in the case of any acute illness,

Table 4. Mortality and readmission rates

Controlled Uncontrolled
Endpoint FSBS FSBS p*
Hospital mortality Rate (%) 0.721 1.19 .00001
Readmission rate within 30 days (%) 14.1 16.57 .0000
*
a = 0.05.
FSBS, fingerstick blood sugar.
6 Journal for Healthcare Quality

Table 5. Mortality and Readmission Rates by Top 5 APR-DRG

Diagnoses
Controlled, Uncontrolled,
% % p
COPD (n = 2,388)
Mortality rate 0.0 0.5 .0255*
Readmission rate 17.68 19.22 .3629
Heart failure (n = 1,773)
Mortality rate 1.11 1.98 .1808
Readmission rate 20.89 21.59 .7271
Pneumonia (n = 1,275)
Mortality rate 0.68 0.73 1.0000
Readmission rate 11.53 18.39 .0004*
Pulmonary edema and respiratory failure
(n = 1,538)
Mortality rate 0.49 1.2 .1834
Readmission rate 22.00 19.7 .2742
Renal failure (n = 1,209)
Mortality rate 1.04 1.6 .4602
Readmission rate 17.04 19.7 .2359
*
Indicates statistically significant difference at the a = 0.05 level.
COPD, chronic obstructive pulmonary disease.

stress hormones such as catecholamines, These outcomes are meaningful not only
glucagon, growth hormone, and cortisol to the patient but to the healthcare insti-
rise, and either oppose the action of tution regarding utilization. Finally, pre-
insulin or independently elevate blood vious studies investigated blood glucose
glucose levels (Peters-Harmel and Mathur, levels on admission but not values during
2004). the hospitalization. We studied blood glu-
cose levels 48 hours after admission, indi-
cating that continued hyperglycemia is
Strengths a risk factor for increased LOS, mortality,
Although a retrospective study, the large and readmission.
number of patient records included pro-
vides a robust sample for data analysis.
Another strength of this study is the Implications for Practice
measurement of patient-oriented outcomes, This study provides further evidence that
i.e., LOS, readmission rate, and mortality. blood glucose levels over 180 mg/dL

Table 6. Rates of Hypoglycemia

Age Range, years Controlled, % Uncontrolled, % p
n = 1,804 ,60 27.8 72.2 .0001*
n = 1,485 60–69 24.6 75.3 .0001*
n = 1,293 70–79 24.3 75.6 .0001*
n = 907 $80 26.4 73.6 .0001*
*
Indicates statistically significant difference at the a = 0.05 level.
 Vol. 0 No. 0 Month 2016 7

impact length of hospital stay, in-hospital clinical outcomes in patients hospitalized

mortality, and 30-day readmission rates in with community-acquired pneumonia. WMJ
2011;110:14–20.
noncritically ill patients regardless of the
Kansagara, D., Fu, R., & Freeman, M., et al.
presence or absence of diabetes. A pro- Intensive insulin therapy in hospitalized
spective, randomized, controlled study is patients: a systematic review. Ann Intern Med
needed to definitively answer the question 2011;154:268–282.
of whether routine testing of FSBS in all Lepper, P.M., Ott, S., & Nuesch, E., et al. Serum
glucose levels for predicting death in
hospitalized patients is warranted. In
patients admitted to the hospital for
addition, based on our findings and the community-acquired pneumonia: prospective
recommendations of the ACP advising cohort study. BMJ 2012;344:e3397.
against blood glucose targets of less than McAlister, F.A., Majumdar, S.R., & Blitz, S.,
140 mg/dL, the optimal blood glucose et al. The relation between hyperglycemia
and outcomes in 2,471 patients admit-
range must be determined.
ted to the hospital with community-
acquired pneumonia. Diabetes Care 2005;
28:810–815.
References Moghissi, E.S., Korytkowski, M.T., &
All Patient Refined DRGs (APR-DRGs): An Over- DiNardo, M., et al. American Association of
view. Available at: https://www.bcbst.com/ Clinical Endocrinologists and American
providers/webinar/APRDRG.pdf. Accessed Diabetes Association consensus statement
July 8, 2014. on inpatient glycemic control. Endocr Pract
Baker, E.H., Janaway, C.H., & Philips, B.J., et al. 2009;15:1–17.
Hyperglycemia is associated with poor out- Murad, M.H., Coburn, J.A., & Coto-Yglesias, F.,
comes in patients admitted to the hospital et al. Glycemic control in non-critically ill
with acute exacerbations of chronic hospitalized patients: a systematic review
obstructive pulmonary disease. Thorax 2006; and meta-analysis. J Clin Endocrinol Metab
61:284–289. 2012;97:49–58.
Bruno, A., Gregori, D., & Caropreso, A., et al. Peters-Harmel, A., & Mathur, R. Davidson’s
Normal glucose values are associated diabetes mellitus. Philadelphia, PA: Saunders;
with a lower risk of mortality in hospi- 2004.
talized patients. Diabetes Care 2008;31: Qaseem, A., Humphrey, L.L., & Chou, R., et al.
2209–2210. Use of intensive insulin therapy for the
Centers for Disease Control and Prevention. management of glycemic control in hospi-
National Diabetes Statistics Report: Estimates of talized patients: a clinical practice guideline
Diabetes and its Burden in the United States. from the American College of Physicians.
Atlanta, GA: US Department of Health and Ann Intern Med 2011;154:260–267.
Human Services; 2014. Thomas, J.W. Hospital cost efficiency measure-
National Medicare Readmission Findings: Recent ment: methodological approaches. 2006.
Data and Trends. Centers for Medicare and Available at: http://www.pbgh.org/programs/
Medicaid Services. Available at: http:// documents/PBGHHospEfficiencyMeas_01-
www.academyhealth.org/files/2012/sunday/ 2006_22p.pdf. Accessed September 22, 2014.
brennan.pdf. Accessed September 22, 2014. Umpierrez, G.E., Isaacs, S.D., & Bazargan, N.,
Cromwell, J., Trisolini, M.G., & Pope, G.C., et al, et al. Hyperglycemia: an independent
eds. Pay for Performance in Health Care: Methods marker in in-hospital mortality in patients
and Approaches. Research Triangle Park, NC: with undiagnosed diabetes. J Clin Endocrinol
RTI Press. RTI Press publication No. BK- Metab 2002;87:978–982.
0002-1103. Available at: http://www.rti.org/ Wexler, D.J., & Cagliero, E. Inpateint diabetes
rtipress. Accessed September 22, 2014. management in non-ICU setting: evidence
Diabetes Initiative of South Carolina. South and satrategies. Curr Diabetes Rev 2007;3:
Carolina adult guidelines for diabetes 239–243.
care in the hospital. Available at: http://
clinicaldepartments.musc.edu/medicine/
divisions/endocrinology/dsc. Accessed May Authors’ Biography
1, 2015.
Adrienne Z. Ables, Edward Via College of Osteopathic
Evans, N.R., & Dhatariya, K.K. Assessing the
Medicine, Carolinas Campus.
relationship between admission glucose
levels, subsequent length of hospital stay,
readmission and mortality. Clin Med 2012;12: Patricia J. Bouknight, Spartanburg Family Medicine
137–139. Residency Program.
Godar, D.A., Kumar, D.R., & Schmelzer, K.M.,
et al. The impact of serum glucose on Heather Bendyk, Spartanburg Medical Center.
8 Journal for Healthcare Quality

Rebecca Beagle, Spartanburg Family Medicine Resi- This study was presented in part at the North
dency Program. American Primary Care Research Group Annual
Meeting. New Orleans, LA: 2012; December 5.
Rebecca Alsip, Edward Via College of Osteopathic
Medicine, Carolinas Campus. The authors declare no conflict of interest.

Jill Williams, University Family Medicine, For more information on this article, contact
Columbia, S.C. Adrienne Z. Ables at aables@carolinas.vcom.edu.