trends in recorded capillary blood glucose and hypoglycaemia in hospitalised patients with diabetes
TRANSCRIPT
Trends in recorded capillary blood glucose andhypoglycaemia in hospitalised patients withdiabetes
G.C. Jones a,*, H. Casey a, C.G. Perry a, B. Kennon b, C.A.R. Sainsbury a
aDepartment of Diabetes, Gartnavel General Hospital, Glasgow G11 0YN, United KingdombDepartment of Diabetes, Southern General Hospital, Glasgow G51 4TF, United Kingdom
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e 1 0 4 ( 2 0 1 4 ) 7 9 – 8 3
a r t i c l e i n f o
Article history:
Received 29 August 2013
Received in revised form
15 October 2013
Accepted 17 January 2014
Available online 25 January 2014
Keywords:
Capillary blood glucose
Near patient testing
Hypoglycaemia
Inpatient
Diabetes
a b s t r a c t
Aims: To utilise whole-system analysis of capillary glucose measurement results to exam-
ine trends in timing of glucose monitoring, and to investigate whether these timings are
appropriate based on observed patterns of hypoglycaemia.
Methods: Near-patient capillary blood glucose results from eight acute hospitals collected
over 57 months were analysed. Analysis of frequency of measurement, and measurements
in the hypoglycaemic (<4 mmol/l) and severe hypoglycaemic (<2.5 mol/l) range per time of
day was made.
Results: 3 345 241 capillary glucose measurements were analysed. 1 657 594 capillary blood
glucose values were associated with 106 624 admissions in those categorised as having
diabetes. Large peaks in frequency of glucose measurements occurred before meals, with
the highest frequency of capillary glucose measurement activity being seen pre-breakfast.
Overnight, an increase in measurement activity was seen each hour. This pattern was
mirrored by frequency of measured hypoglycaemia. 27 968 admissions (26.2%) were asso-
ciated with at least one hypoglycaemic measurement. A greater proportion of measure-
ments were within the hypoglycaemic range overnight with 61.7% of all hypoglycaemia
between 2100 and 0900 h, with peak risk of measured capillary glucose being hypoglycaemic
between 0300 and 0400 h.
Conclusions: Hypoglycaemic is common with the greatest risk of hypoglycaemia overnight
and a peak percentage of all readings taken being in the hypoglycaemic range between 0300
and 0400 h. Measurement activity overnight was driven by routine, with patterns of
proportion of measurements in the hypoglycaemic range indicating that there may be a
significant burden of undiscovered hypoglycaemia in the patients not routinely checked
overnight.
# 2014 Elsevier Ireland Ltd. All rights reserved.
Contents available at ScienceDirect
Diabetes Researchand Clinical Practice
journal homepage: www.elsevier.com/locate/diabres
1. Introduction
Capillary blood glucose monitoring is performed routinely in
hospital to manage patients at risk of hyperglycaemia and
hypoglycaemia and point of care capillary blood glucose
* Corresponding author. Tel.: +44 141 211 3259.E-mail address: [email protected] (G.C. Jones).
0168-8227/$ – see front matter # 2014 Elsevier Ireland Ltd. All rights
http://dx.doi.org/10.1016/j.diabres.2014.01.021
measurement can be considered as analogous to an additional
‘‘vital sign’’ for hospitalised patients with diabetes.
Diabetes is common amongst hospital inpatients in the
United Kingdom, Europe and North America being thought to
be present in between 1 in 5 and 1 in 10 hospital stays [1–3].
reserved.
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e 1 0 4 ( 2 0 1 4 ) 7 9 – 8 380
Diabetes has been unequivocally shown to negatively impact
both mortality and length of stay in hospital in a variety of
medical and surgical settings prompting an increasing focus
on improving inpatient glycaemic control [2,4–7].
In hospitalised patients with diabetes, hypoglycaemia is
common, and in patients both with and without diabetes its
frequency and severity has been associated with an increase
in pre and post discharge mortality and length of stay [8–10].
The ACE/ADA Task Force on Inpatient Diabetes in 2006
stressed the importance of glycaemic control and raised
concerns regarding under and over treatment of glycaemia
citing insufficient frequency of glucose monitoring as a
potential safety issue [11]. The 2013 ADA standards of medical
care for patients with diabetes suggest glucose monitoring in
all people with, and at risk of, diabetes citing the importance of
glycaemic control and hypoglycaemia avoidance. Guidance on
frequency of glucose monitoring by point of care testing is only
given for patients treated with insulin with between 4 and 6
hourly to ½ hourly to 2 hourly for those on intravenous insulin
suggested [12]. In the UK the National Diabetes Inpatient Audit
has recommended that glucose is checked weekly for stable
diet controlled patients and 4 times a day for unwell patients,
unstable diabetes or those treated by basal bolus insulin
therapy [13].
The Greater Glasgow and Clyde Health Board in Scotland
performs around 700 000 near patient capillary blood glucose
tests per year. The purpose of this study is to utilise whole
system analysis of these results to examine current trends in
timing of monitoring and assess if these timings are appro-
priate based on observed patterns of hypoglycaemia.
2. Materials and methods
Capillary blood glucose data were extracted from the Abbott
Precision Webb system [Abbott, UK]. Whole-hospital data
were included from eight acute hospitals and associated units
within Greater Glasgow and Clyde Health Board (Gartnavel
General Hospital, Western Infirmary, Glasgow Royal Infirmary,
Victoria Infirmary, Southern General Hospital, Royal Alexan-
dria Hospital, Inverclyde Hospital and Vale of Leven Hospital)
between dates 01/01/2008 and 01/05/2013. Data collected from
intensive care environments were excluded as it was felt likely
that the capillary glucose monitoring in these areas would be
very different from general medical and surgical wards. Data
available from the system included patient identifier, patient
location, test date and time and capillary glucose value.
All data manipulation and analysis was performed using
bespoke code written in statistical language R [14].
Prior to analysis of the complete data set it became
apparent that a number of capillary blood glucose measure-
ment values were not attributable to an identifiable patient
and were removed.
We wished to limit our analysis as much as possible to data
from patients with diabetes and we were aware that a
proportion of capillary blood glucose measurements are
performed as a routine admission test in people without
diabetes. Based on the World Health Organisation diagnostic
criteria for diabetes we assumed that either morning (between
0500 and 0700 h) capillary blood glucose of �7 mmol/l or 2
random capillary blood glucose of �11.1 mmol/l indicated
diabetes. This was felt to give the best possible dataset for
further analysis, consisting of subjects likely to have diabetes
without unduly excluding individuals with diabetes from
analysis.
Dates of admission and discharge from hospital were
implied following examination of the comparison with case
note review dataset. Any gap of five or more days between
capillary blood glucose measurements was taken as the
threshold for determining a new admission episode.
Analysis of frequency of glucose measurement per time of
day was made. A similar frequency analysis was made of
hypoglycaemic measurements (<4 mmol/l), and severe hypo-
glycaemic capillary glucose measurements (<2.5 mmol/l). We
used the National Health Service Diabetes guideline treatment
cut-off value (blood glucose values <4 mmol/l) to categorise
hypoglycaemia. Severe hypoglycaemia is best categorised by
the need for third-party assistance in treating the episode. As
this information was not available to us a value of <2.5 mmol/l
was used to describe severe hypoglycaemia [15]. Comparison
was made between the frequency of hypoglycaemic measure-
ments during daytime (0900–2100) and night time (2100–0900)
hours.
The number of capillary glucose measurement in the
hypoglycaemic range (<4 mmol/l) as a proportion of all
capillary glucose measurements was calculated for the
dataset as a whole and for each hour of the day.
We observed peaks and troughs of glucose measurement
activity overnight with peaks occurring at 0100, 0200, 0300, and
0400 h. We calculated the proportion of capillary blood glucose
measurements in the hypoglycaemic range for a 30 min
window centred on these peaks in activity, and also in the
corresponding 30 min activity troughs bordering the peaks.
We repeated the analysis of the four peaks in measurement
activity at 0100, 0200, 0300 and 0400 h excluding any measured
capillary blood glucose occurring within 1 h of any initial
measurement <4 mmol/l.
3. Results
The initial raw dataset consisted of 3 345 241 capillary glucose
measurements. 1 836 258 capillary glucose values were asso-
ciated with 60 256 valid patient identifiers, corresponding to
172 771 admission episodes over 57 months of data acquisi-
tion. 1 657 594 capillary blood glucose values were associated
with 106 624 admissions during which a diagnosis of diabetes
was made using glycaemic parameters. The remainder of the
discussion will consider those subjects in whom a glycaemic
diagnosis of diabetes was made.
4. Timing of capillary blood glucosemeasurements
Increases in measurement activity were identified with large
peaks in frequency of capillary blood glucose measurements
at pre-meal times. The highest frequency of capillary blood
glucose measurement activity was seen pre-breakfast, with
less pronounced peaks identified before lunchtime and
010
0020
0030
0040
0050
0060
00
time (hour)
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0
num
ber
of C
BG m
easu
rem
ents
per
min
ute
Fig. 1 – Frequency of capillary blood glucose measurement
per minute over 24 h day. Data from 28 697 individuals
determined to have diabetes (1 657 594 capillary blood
glucose measurements).
010
020
030
040
0
time (hour)0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 0nu
mbe
r of
CBG
val
ues
<4
mm
ol/L
per
min
ute
Fig. 2 – Frequency of capillary blood glucose measurements
<4 mmol/l per minute. Data from 11 174 individuals
(75 529 capillary blood glucose measurements).
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e 1 0 4 ( 2 0 1 4 ) 7 9 – 8 3 81
evening meals. A diffuse increase in activity was observed
during the late evening. Overnight, frequency of testing fell.
Peaks of activity were observed at 0000, 0100, 0200, 0300, 0400
and 0500 h (Fig. 1).
5. Hypoglycaemia
Peaks of ascertainment of hypoglycaemiccapillary blood glucose
measurement (<4 mmol/l) were observed prior to mealtimes,
with the pre-breakfast peak being dominant, mirroring the
patterns identified in capillary blood glucose measurement
activity. Overnight, increases in hypoglycaemia identification
were observed at 0000, 0100, 0200, 0300, 0400 and 0500 h (Fig. 2).
27 968 admissions were associated with at least one
hypoglycaemic capillary blood glucose measurement, repre-
senting a proportion of 26.2%. A majority (46 654, 61.7% of all
0000
-010
001
00-0
200
0200
-030
003
00-0
400
0400
-050
005
00-0
600
0600
-070
007
00-0
800
0800
-090
009
00-1
000
1000
-110
011
00-1
200
1200
-130
013
00-1
400
1400
-150
015
00-1
600
1600
-170
017
00-1
800
1800
-190
019
00-2
000
2000
-210
021
00-2
200
2200
-230
023
00-2
400
time (hour)
0.00
0.02
0.04
0.06
0.08
prop
ortio
n of
CBG
val
ues
<4
mm
ol/L
Fig. 3 – Proportion of capillary blood glucose measurements <4
day.
hypoglycaemic measurements) were obtained nocturnally
between the hours of 2100 and 0900 h. The large majority of
these were measured between the hours of 0600 and 0800
(23 423, 31.0% of all hypoglycaemic measurements). Of 11 792
severe hypoglycaemic measurements (<2.5 mmol/l), 6492
(55.1%) occurred between 2100 and 0900, again with the large
majority occurring in the early morning. 75 529 capillary blood
glucose measurements were in the hypoglycaemic range
(<4 mmol/l) representing 4.1% of the total number of capillary
blood glucose values measured. This proportion varied
throughout the day, with a greater proportion of capillary
blood glucose values being hypoglycaemic between 0000 and
0700 than between 0700 and 0000 h.
11 792 capillary blood glucose measurements were in the
severe hypoglycaemic range (<2.5 mmol/l) representing 0.6%
of the total number of capillary blood glucose values
measured. Again a greater proportion of measurements were
0.00
00.
005
0.01
00.
015
0000
-010
001
00-0
200
0200
-030
003
00-0
400
0400
-050
005
00-0
600
0600
-070
007
00-0
800
0800
-090
009
00-1
000
1000
-110
011
00-1
200
1200
-130
013
00-1
400
1400
-150
015
00-1
600
1600
-170
017
00-1
800
1800
-190
019
00-2
000
2000
-210
021
00-2
200
2200
-230
023
00-2
400
prop
ortio
n of
CBG
val
ues
<2.
5 m
mol
/L
time (hour)
mmol/l (left panel) and <2.5 mmol/l (right panel) by hour of
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e 1 0 4 ( 2 0 1 4 ) 7 9 – 8 382
in the severe hypoglycaemic range between 0000 and 0700 h
(Fig. 3).
Overnight capillary blood glucose encapsulating the four
peaks in measurement activity at 0100, 0200, 0300 and 0400 h
were examined. 7.1% of measurements obtained during 30 min
windows centred on the peaks were in the hypoglycaemic range
(<4 mmol/l) compared to 8.6% of the capillary glucose results
during the intervening trough periods. 1.4% of measurements
obtained during 30 min windows centred on the peaks were in
the severe hypoglycaemic range (<2.5 mmol/l) against 1.7% of
the capillary glucose results during the intervening trough
periods.
In the repeated analysis of the four peaks in measurement
activity at 0100, 0200, 0300 and 0400 h excluding any measured
capillary blood glucose occurring within 1 h of any initial
measurement <4 mmol/l. 5.6% of measurements obtained
during 30 min windows centred on the peaks were in the
hypoglycaemic range (<4 mmol/l) compared to 6.5% of the
capillary glucose results during the intervening trough periods.
6. Conclusions
Hypoglycaemia is a common problem in diabetic inpatients
and confers a significant risk in terms of morbidity and
mortality [2,4–7]. Monitoring of capillary blood glucose plays a
key role in identifying and preventing hypoglycaemia. Our
data set is the largest to be examined to describe trends in
timing of capillary blood glucose monitoring and frequency of
ascertainment of hypoglycaemia. Our data show that mon-
itoring appears to be done mainly before breakfast with
increased testing pre-meals and testing on the hour.
Our study shows that of all inpatient measured capillary
glucose 5.1% is in the hypoglycaemia range (<4 mmol/l). This
is consistent with previously described rates of 3.5–5.7% for
glucose <70 mg/dl (3.9 mmol/l) in the US for non-intensive
care patients [16,17] and similar to that described in the United
Kingdom by Kerry et al. of 4.9% [18].
It has been shown previously using continuous glucose
monitoring in type 1 and 2 diabetes that glucose is more often
low in the night time than daytime (7% compared to 11% of
readings) [19] and in the Diabetes Control and Complications
Trial severe hypoglycaemia was also more common at night
[20]. Our data confirm this overnight increased risk of
hypoglycaemia in an inpatient setting and we found 61% of
all hypoglycaemia occurred between 2100 and 0900 h which
compares with the findings of Kerry et al. who found that 70%
of all hypoglycaemia occurs in this time period [18]. In 206
diabetic inpatients Bailon et al. showed that 78% more
hypoglycaemic events occurred during night shift with most
occurring between 0400 and 0500 and 0600 and 0700 h [21]. It is
possible that the increased recorded hypoglycaemia in these
studies is at least partly explicable by ascertainment bias as
large proportion of all measured glucose is done at the
beginning of early nursing shift between 0600 and 0900 h in
the morning. In our own study we were able to confirm the
high risk of overnight hypoglycaemia with the absolute
number of hypoglycaemic episodes also highest between
0600 and 0700 h. Importantly we were also able to show a
higher proportion of all measured capillary blood glucose
reading to be hypoglycaemic between 0000 and 0700 h with the
peak proportion of recorded glucose levels in the hypogly-
caemic range occurring between 0300 and 0400 h.
In an attempt to further reduce ascertainment bias we
compared the hourly peaks of monitoring (which are likely to be
predominantly driven by routine and planned monitoring
rather than by acute clinical concerns or patient symptoms)
with troughs of monitoring activity. The difference in rate of
hypoglycaemia between these peaks and troughs of measure-
ment frequency are minor suggesting that overnight increased
hypoglycaemic frequency is unlikely to be caused solely by an
increased proportion of the tests being performed because of
symptoms reported by patients. Indeed the difference in rate of
capillary blood glucose found to be in hypoglycaemic range
reduced when removing tests performed within the hour
following uncovered hypoglycaemia. Even in the severe
hypoglycaemia group (<2.5 mmol/l) who would likely be most
symptomatic there was still minimal difference between peak
and trough rates. It appears that overnight routine capillary
blood glucose testing of patients will uncover a high rate of
hypoglycaemia albeit in a group of patients who have been
deemed high risk enough to require an hourly monitoring
regime and may be on intensive intravenous or subcutaneous
insulin therapy. These data suggest that there may be a
significant burden of undiscovered hypoglycaemia in the
patients not routinely checked overnight and this should be
considered in patients at high risk of hypoglycaemia. It would
appear that the optimal time to look for hypoglycaemia
overnight would be between 0300 and 0400 h. As pointed out
by Kerry et al. this increased risk of overnight hypoglycaemia
may be worsened by early hospital meals and lack of night time
carbohydrate snack availability which should be borne in mind
when planning hospital mealtimes [18].
Our data confirm that hypoglycaemia is common in
inpatients and suggests times where monitoring might be
focused. Future research is needed to evaluate patterns of
capillary blood glucose testing to better detect and prevent
hypoglycaemia.
Conflict of interest statement
None.
r e f e r e n c e s
[1] Sampson MJ, Crowle T, Dhatariya K, Dozio N, GreenwoodRH, Heyburn PJ, et al. Trends in bed occupancy forinpatients with diabetes before and after the introductionof a diabetes inpatient specialist nurse service. Diabet Med2006;23:1008–15.
[2] Olveira-Fuster G, Olvera-Marquez P, Carral-Sanlaureano F,Gonzalez-Romero S, Aguilar-Diosdado M, Soriguer-EscofetF. Excess hospitalizations, hospital days, and inpatientcosts among People with diabetes in Andalusia, Spain.Diabetes Care 2004;27:1904–9.
[3] American Diabetes Association. Economic costs of diabetesin the U.S. in 2007. Diabetes Care 2008;31:596–615.
[4] Umpierrez GE, Isaacs SD, Bazargan N, You X, Thaler LM,Kitabchi AE. Hyperglycaemia: an independent marker of
d i a b e t e s r e s e a r c h a n d c l i n i c a l p r a c t i c e 1 0 4 ( 2 0 1 4 ) 7 9 – 8 3 83
inpatient mortality in patients with undiagnosed diabetes. JClin Endocrinol Metab 2002;87:978–82.
[5] Krinsley JS. Association between hyperglycaemia andincreased hospital mortality in a heterogeneouspopulation of critically ill patients. Mayo Clin Proc2003;78:1471–8.
[6] Frisch A, Chandra P, Smiley D, Peng L, Rizzo M, Gatcliffe C,et al. Prevalence and clinical outcome of hyperglycaemia inthe perioperative period in noncardiac surgery. DiabetesCare 2010;33:1783–8.
[7] Evans NR, Dhatariya KK. Assessing the relationshipbetween admission glucose levels, subsequent length ofhospital stay, readmission and mortality. Clin Med2012;12:137–9.
[8] Turchin A, Matheny ME, Shubina M, Scanlon J, GreenwoodB, Pedergrass M. Hypoglycemia and clinical outcomes inpatients with diabetes hospitalised in the general ward.Diabetes Care 2009;32:1153–7.
[9] Nirantharakumar K, Marshall T, Kennedy A, Narendran P,Hemming K, Coleman JJ. Hypoglycaemia is associated withincreased length of stay and mortality in people withdiabetes who are hospitalised. Diabet Med 2012;29:445–8.
[10] Tan HK, Flannagan D. The impact of hypoglycaemia onpatients admitted to hospital with medical emergencies.Diabet Med 2013;30:574–80.
[11] American College of Endocrinology and American DiabetesAssociation. Consensus statement on inpatient diabetesand glycemic control: a call to action. Diabetes Care2006;29:1955–62.
[12] American Diabetes Association. Standards of medical carein diabetes-2013. Diabetes Care 2013;36(Suppl. 1):S11–66.
[13] National Diabetes Inpatient Audit 2012. Published March2013; 2013, Available at http://www.hscic.gov.uk/ [lastaccessed July 2013].
[14] R Core Team. R: a language and environment for statisticalcomputing. Vienna, Austria: R Foundation for StatisticalComputing; 2012, ISBN: 3-900051-07-0, Available at http://www.R-project.org/ [last accessed July 2013].
[15] Joint British Diabetes Society. The hospital management ofhypoglycaemia in adults with diabetes mellitus. NewcastleUpon Tyne: Joint British Diabetes Society; 2010.
[16] Swanson CM, Potter DJ, Kongable GL, Cook CB. Update oninpatient glycaemic control in hospital in the United States.Endo Pract 2011;17:853–61.
[17] Cook CB, Konglable GL, Potter DJ, Abad VJ, Leija DE,Anderson M. Inpatient glycaemic control: a survey of 126US hospitals. J Hosp Med 2009;9:E7–14.
[18] Kerry C, Mitchell S, Sharma S, Scott A, Rayman G. Dailytemporal patterns of hypoglycaemia in hospitalised peoplemay reveal potentially correctable factors. Diabet Med 2014.http://dx.doi.org/10.1111/dme.12256.
[19] Bode BW, Schwartz S, Stubbs HA, Block JE. Glycemiccharacteristics in continuously monitored patients withtype 1 and type 2 diabetes: normative values. Diabetes Care2005;28:2361–6.
[20] The DCCT research Group. Epidemiology of severehypoglycaemia in the diabetes control and complicationstrial. Am J Med 1991;90:450–9.
[21] Bailon RM, Cook CB, Hovan MJ, Hull BP, Seifert KM, Miller-Cage V. Temporal and geographical patterns ofhypoglycaemia among hospitalised patients with diabetesmellitus. J Diab Sci Technol 2009;3:261–8.