Key words
diabetes mellitus - pre-diabetes - HbA1c measurement - in-hospital screening
Introduction
In Switzerland, one out of sixteen persons aged between 35 and 75 years has diabetes
mellitus (DM), and almost one third of subjects with diabetes is unaware of their
status [1]
[2]. Identification of patients with previously undiagnosed DM improves care and is
cost-effective [3]. Since 2009, measurement of haemoglobin A1c (HbA1c) is recommended both to screen and to diagnose DM [4]
[5].
Two studies conducted in the United States have suggested that in-hospital DM testing
might be more effective than community detection, because the prevalence of DM among
hospitalized patients is higher (approximately 35% in a general internal medicine
service) [6]
[7]. Such systematic screening might detect a previously unknown DM in 18% to 24% of
hospitalized patients [6]
[7].
As the prevalence of DM is lower in Europe than in the United States [8], it was expectable that DM detection among hospitalized patients might also be lower.
Interestingly, a recent study conducted in a German university hospital showed that
this hypothesis was partially wrong [9]. Actually, among patients hospitalized in internal medicine, the prevalence of DM
was 32% and up to 7% of them had a newly diagnosed DM [9]. Thus, among medical hospitalized patients, in-hospital DM screening might be as
effective in Europe as in the United States.
Because few information exists regarding current in-hospital DM screening, we aimed
to assess: 1) whether hospital doctors measure HbA1c in medical patients hospitalized
in a Swiss teaching hospital, 2) the prevalence of newly detected prediabetes and
DM and 3) whether newly detected prediabetes and DM are managed and reported at discharge.
Methods
Setting and sampling
The study was conducted in the Lausanne University Hospital (CHUV), one of the five
medical teaching hospitals in Switzerland. The CHUV has over 1400 beds and receives
over 45,000 hospitalizations per year (www.chuv.ch). We considered all adult (≥18 years) hospitalizations occurring in the division
of general internal medicine between January 2013 and December 2014 who gave their
informed consent to use their medical data (40% in 2014).
Data from hospital electronic files between 2002 and 2014 was collected. Patients
were categorized according to absence or presence of a history of DM. History of DM
was defined as: 1) the presence of ICD-10 codes E10.x to E14.x (x=any number) in the
medical files; 2) one glucose measurement>11.8 mmol/L (213 mg/dL); 3) HbA1c≥6.5% (47.5 mmol/mol), and 4) any anti-diabetic treatment at admission.
Data collection and outcomes
Socio-demographic data included age; gender; marital status (single/married or cohabitating/divorced/widowed);
nationality (Swiss/non-Swiss); housing (own house/nursing home/ other) and private
health insurance (yes/no). Hospital data included day of admission, total length of
stay, and time spent in the emergency ward.
Clinical data included treated hypertension (based on medical records); smoking (past
or current); obesity (based on ICD-10 codes E66.x); personal history of cardiovascular
disease (medication at admission or on medical records); dyslipidemia [medication
at admission, medical records, or fasting triglyceride>2.82 mmol/L (250 mg/dL) associated
to a HDL-cholesterol<0.9 mmol/L (35 mg/dL)] and kidney disease (estimated glomerular
filtration rate eGFR<30 ml/min/m2 according to Cockroft-Gault formula). Biological data included serum HbA1c, HDL-cholesterol, triglycerides, creatinin and glucose. Two glucose values measured
during the study period were considered: 1) first glucose assessment and 2) highest
value during hospitalization. Hba1c levels were assessed in EDTA whole blood using
Bio-Rad D-100TM HbA1c testing system (Bio-Rad Laboratories, Hercules, CA, USA) [10]. As only seven patients had two HbA1c measurements during their stay, only the first
HbA1c measurement was considered.
Outcomes included newly detected prediabetes or DM, the presence of an antidiabetic
drug treatment at discharge, the length of stay and in-hospital mortality. A newly
detected DM was defined as HbA1c≥6.5% (47.5 mmol/mol) within patients with no history of DM. We considered that patients
had a prediabetic state if HbA1c ranged between≥5.7 (38.8 mmol/mol) and<6.5% (47.5 mmol/mol) according to the American
Diabetes Association [4]. We also assessed whether newly detected prediabetes or DM were reported at discharge,
based on ICD-10 codes E10.x to E14.x in the electronic medical file.
Ethical considerations and data protection
Since January 2013, all patients hospitalized in the CHUV are asked for general consent
that allows future use of medical records and blood tests performed during their hospitalization.
The study was approved by the ethics committee and no further specific consent from
the patients was required as only already available administrative data was used.
Data was extracted by a dedicated team using the inclusion criteria defined previously
and was anonymized before being provided to the investigators. Thus, it was not possible
to assess how many patients were excluded.
Statistical analyses
Statistical analysis was conducted using Stata version 14.0 for windows (Stata corp,
College Station, TX, USA). The prevalence of diabetes testing was assessed overall
and according to presence/absence of previously known DM. Results were expressed as
number of patients (percentage), mean±standard deviation or median [interquartile
range] as appropriate. Bivariate comparisons were performed using chi-square or Fisher’s
exact test for categorical variables and Student’s t-test or nonparametric Kruskal-Wallis
test for continuous variables. Statistical significance was considered for a two-sided
test p-value<0.05.
Results
Overall, 2618 patients were included, 481 (18.4%) of whom had a history of DM. Patients
with DM were significantly older, less frequently Swiss nationals, presented with
more cardiovascular risk factors and stayed longer in the emergency department at
admission than patients without known DM ([Table 1]).
Table 1 Patients’ characteristics overall and according to previous history of diabetes mellitus.
|
All (n=2618)
|
History of diabetes mellitus *
|
p-value
|
Absent (n=2137)
|
Present (n=481)
|
Age, years
|
68.4±16.7
|
67.6±17.4
|
72±12.7
|
<0.001
|
Age ≥65 years (%)
|
1653 (63.1)
|
1303 (61.0)
|
350 (72.8)
|
<0.001
|
Female gender (%)
|
1187 (45.3)
|
1005 (47.0)
|
182 (37.8)
|
<0.001
|
Marital status (%)
|
|
|
|
0.074
|
Single
|
365 (14.4)
|
316 (15.2)
|
49 (10.8)
|
|
Married/cohabitating
|
1189 (46.9)
|
970 (46.7)
|
219 (48.1)
|
|
Divorced
|
453 (17.9)
|
372 (17.9)
|
81 (17.8)
|
|
Widowed
|
527 (20.8)
|
421 (20.3)
|
106 (23.3)
|
|
Coming from home (%)
|
2093 (80.0)
|
1685 (78.9)
|
408 (84.8)
|
0.003
|
Private health insurance (%)
|
128 (4.9)
|
111 (5.2)
|
17 (3.5)
|
0.127
|
Obesity (%)
|
58 (2.2)
|
5 (0.2)
|
53 (11.0)
|
<0.001
|
Hypertension (%)
|
199 (7.6)
|
56 (2.6)
|
143 (29.7)
|
<0.001
|
Past or current smoking (%)
|
23 (0.9)
|
6 (0.3)
|
17 (3.5)
|
<0.001
|
History of cardiovascular disease (%)
|
1352 (51.6)
|
958 (44.8)
|
394 (81.9)
|
<0.001
|
History of dyslipidemia (%)
|
657 (25.1)
|
395 (18.5)
|
262 (54.5)
|
<0.001
|
eGFR<30 ml/min/m2 (%)
|
242 (11.8) ‡
|
175 (10.6)
|
67 (16.8)
|
0.001
|
Glucose level at admission (mmol/l) §
|
6.3±3.0
|
6.3±3.0
|
7.9±3.8
|
<0.001
|
Time spent in emergency department (hours)
|
9 [5–16]
|
9 [5–16]
|
10 [7–18]
|
0.002 †
|
* History of diabetes mellitus was based on ICD-10 codes E10.x- E14.x, any anti-diabetic
medicines or a history of abnormal glucose homeostasis. §, n=108 (present) and n=390
(absent). eGFR, estimated glomerular filtration rate as assessed by the Cockroft-Gault
formula. Results are expressed as number of participants (column %), as average±SD
or as median and [interquartile range]. Statistical analysis by chi-square for categorical
data and by Student’s t-test of Kruskal-Wallis (†) test for continuous data.
‡ 573 observations were missing.
Do hospital doctors measure HbA1c in hospitalized patients?
Overall, HbA1c was measured in 298 (11.4%) patients, thereof 136 (45.6%) had no history of DM.
Median [interquartile range] HbA1c was 6.3 [5.6–7.6]% (45.4 [37.7-59.6] mmol/mol) ([Fig. 1]
, panel A). Of the 481 patients with known DM, 162 (33.6%) had HbA1c measured. Median [interquartile range] HbA1c among these patients was 7.2 [6.2–8.5]% (55.2 [44.3-69.4] mmol/mol) ([Fig. 1]
, panel B). Of the 2137 patients without known DM, 136 (6.4%) had HbA1c measured. Median [interquartile range] HbA1c among these patients was 5.8 [5.35–6.2]% (39.9 [35.0-44.3] mmol/mol) ([Fig. 1]
, panel C).
Fig. 1 Distribution of HbA1c values. Panel A: all patients; panel B: patients with known diabetes mellitus (based
on ICD-10 codes E10.x- E14.x, any anti-diabetic medicines or a history of abnormal
glucose homeostasis); panel C: patients without history of diabetes mellitus. Dashed
line indicates an HbA1c value of 6.5% (47.5 mmol/mol)
Restricting the analysis to patients without history of DM, significantly higher frequencies
of history of cardiovascular disease and hypertension, and a higher glucose levels
and a lower frequency of kidney disease were found in patients with HbA1c measurement, while no significant differences were found for gender, age, and history
of dyslipidemia ([Table 2]).
Table 2 Clinical characteristics of hospitalized patients without history of diabetes mellitus,
according to presence or absence of HbA1c measurement.
|
Not measured (n=2001)
|
Measured (n=136)
|
P-value
|
Age, years
|
67.6±17.5
|
66.6±16.0
|
0.514
|
Age ≥65 years (%)
|
1227 (61.3)
|
76 (55.9)
|
0.208
|
Female gender (%)
|
949 (47.4)
|
56 (41.2)
|
0.158
|
Hypertension (%)
|
25 (1.3)
|
31 (22.8)
|
<0.001
|
History of cardiovascular disease (%)
|
876 (43.8)
|
82 (60.3)
|
<0.001
|
Dyslipidemia (%)
|
363 (18.1)
|
32 (23.5)
|
0.117
|
eGFR<30 ml/min/m2 (%)
|
168 (11.1)
|
6 (4.6)
|
0.020
|
Glucose level at admission (mmol/l) ¦
|
6.1±1.7
|
7.6±6.7
|
<0.001
|
Results are expressed as number of patients (column %) or as average±standard deviation.
Between-group comparisons performed using chi-square for categorical data and by Student’s
t-test for continuous data. ¦ among 55 and 335 patients with and without measurement,
respectively.
Prevalence of newly detected prediabetes and diabetes mellitus
Of the 136 (6.4%) patients without history of DM and who had HbA1c measured, 51 (37.5%) had a prediabetic state and 23 (16.9%) had DM. Baseline characteristics
did not differ between patients with newly diagnosed prediabetes, newly diagnosed
DM and those without DM, with the exception that patients with newly detected DM had
a higher glucose level at admission ([Table 3]).
Table 3 Factors associated with abnormal glucose homeostasis in hospitalized patients without
history of diabetes at admission/discharge and with HbA1c measurement (n=136).
At admission
|
No diabetes (N=62)
|
Newly detected prediabetes* (n=51)
|
Newly detected diabetes† (N=23)
|
P-value
|
Age, years
|
64.6±17.3
|
68.5±14.8
|
68±14.7
|
0.388
|
Age ≥65 years (%)
|
32 (51.6)
|
33 (64.7)
|
11 (47.8)
|
0.263
|
Female gender (%)
|
29 (46.8)
|
21 (41.2)
|
6 (26.1)
|
0.226 ¦
|
Coming from home (%)
|
59 (95.2)
|
48 (94.1)
|
22 (95.7)
|
1.000 ¦
|
Obesity (%)
|
3 (4.8)
|
1 (2.0)
|
0 (0)
|
0.660 ¦
|
Hypertension (%)
|
12 (19.4)
|
11 (21.6)
|
8 (34.8)
|
0.326 ¦
|
Past or current smoking (%)
|
1 (1.6)
|
2 (3.9)
|
0 (0)
|
0.765 ¦
|
History of cardiovascular disease (%)
|
33 (53.2)
|
33 (64.7)
|
16 (69.6)
|
0.282
|
Dyslipidemia (%)
|
14 (22.6)
|
11 (21.6)
|
7 (30.4)
|
0.687
|
eGFR<30 ml/min/m2 (%)
|
1 (1.7)
|
2 (4.0)
|
3 (13.6)
|
0.098 ¦
|
Glucose level at admission (mmol/l) #
|
5.2±1.0
|
5.7±1.1
|
14.7±11.2
|
<0.001
|
Time spent in emergency department (hours)
|
21 [10–28]
|
20 [6–27]
|
11 [7–25]
|
0.320 §
|
At discharge
|
Length of stay (days)
|
13 [8–19]
|
8 [6–11]
|
16 [9–25]
|
0.002 §
|
In-hospital mortality (%)
|
0 (0)
|
1 (2.0)
|
1 (4.4)
|
0.155 ¦
|
Anti-diabetic treatment at discharge (%)
|
1 (1.6)§
|
0 (0)
|
11 (47.8)
|
<0.001
|
Reported in medical file
|
NA
|
3 (5.8)
|
15 (65.8)
|
NA
|
Abbreviations: eGFR: estimated glomerular filtration rate as assessed by the Cockroft-Gault
formula; NA: does not apply.
*Prediabetes defined as: HbA1c between ≥5.7 (38.8 mmol/mol) and<6.5% (47.5 mmol/mol).
†Newly diagnosed diabetes defined as HbA1c≥6.5%. § This patient received metformin
for a metabolic syndrome without diabetic state (Hba1c value 5.2%). # for 26, 16 and
13 patients with no diabetes, newly detected prediabetes and newly detected diabetes,
respectively
Results are expressed as number of participants (column %), as average±SD or as median
and [interquartile range]. Statistical analysis by chi-square or Fisher’s exact test
(¦) for categorical data and by Student’s t-test or Kruskal-Wallis (§) test for continuous
data
How are newly detected prediabetes and DM managed and reported at discharge?
Among the 23 patients with newly detected DM, 16 (69.6%) had an HbA1c≥7% (53.0 mmol/mol) ([Fig. 1]
, panel C). Patients with a newly detected DM had a longer length of stay, while no differences
were found regarding in-hospital mortality ([Table 3]). Eleven of them (48%) received anti-diabetic treatment at discharge. Also at discharge,
newly detected prediabetes or DM were reported in 5.8% (3/51) and 65.8% of cases (15/23),
respectively.
Discussion
In a Swiss teaching hospital, doctors seldom prescribed HbA1c measurement: slightly
over one tenth (11.4%) of medical hospitalized patients. This percentage was even
lower (6.4%) in medical patients without known DM. Interestingly, a prediabetic state
was found in 37.5% of patients and DM was detected in one of five. Indicating that
when a physician asks for HbA1c measurement, the likelihood of detecting (pre)DM is
high (roughly 50%). However, prescribers tended to underreport these newly detect
(pre)diabetic states.
To which hospitalized patients do hospital doctors prescribe HbA1c measurement?
According to the 2013 update of the Standards of Medical Care in Diabetes by the American Diabetes Association, testing for DM should be considered for all
adults with overweight (body mass index>25 kg/m2) and at least one of the known risk factors for DM such as physical inactivity; first-degree
relative with DM; high risk race/ethnicity; hypertension (≥140/90 mmHg or on therapy);
HDL<0.9 mmol/l and/or triglycerides>2.82 mmol/l; gestational diabetes or high weight
newborn; women with polycystic ovary syndrome; history of cardiovascular disease and
HbA1c≥5.7% (38.8 mmol/mol) on previous testing [4]. In our study, hospital doctors prescribed HbA1c measurement more often for patients with clinical factors associated with greater
risks of cardiovascular disease. The utilization of HbA1c among hospitalized patients seems to be in accordance with guidelines.
Interestingly, a recent study conducted in the ambulatory setting showed that it was
possible to half the number of patients needed to invite to screen by using electronical
medical data in a two-step procedure [11]. Patients without DM risk or with DM screening within the past 3 years were not
considered for screening. Further studies assessing the performance of opportunistic
in-hospital DM screening using electronical medical records and such are needed.
Hospital doctors prescribed HbA1c measurement to 11.4% of all medical patients, but only to 6.4% of patients with
no history of DM. In a German study, hospital doctors prescribed HbA1c measurement
to surgical and medical adults slightly more often (16% of all patients and 10% of
patients with no history of DM), but the prescription rates for medical patients was
not specified. A Canadian study reported that 14.2% of non-diabetic primary care patients
aged 45 or more had at least one HbA1c measurement during a three years period [12]. Indeed, most studies conducted on hospitalized patients used HbA1c as part of a systematic detection strategy [6]
[7]
[13] or focused on patients with a known DM [14]
[15]. Still, the Canadian study was performed prior to the release of the current guidelines
[4]
[5] and because the rates of HbA1c measurement tended to increase since 2009 [13], the frequency of HbA1c measurement in medical non-diabetic hospitalized patients might be at least twice
less frequent than in primary care patients without DM. Overall, our results suggest
that DM screening is seldom performed in a teaching hospital setting.
Prevalence of newly detected prediabetes and diabetes mellitus
Among hospitalized patients without history of DM, HbA1c measurement allowed detecting prediabetes in 37.5% and DM in 16.9%. Two studies
conducted in the United States have suggested that in-hospital DM testing might be
more effective than community detection, because the prevalence of DM among hospitalized
patients is higher. The first one showed that 35% of patients admitted to hospital
had DM, compared to 9.3% in the general population [6]. The second reported that 18% of hospitalized patients presented with previously
undetected DM, defined as HbA1c>6.1% (43.2 mmol/mol) [7]. In this study, the prevalence of DM among patients with HbA1c measurements was almost three fold higher than the one reported in the general population
using fasting plasma glucose levels [2].
How are newly detected prediabetes and DM managed and reported at discharge?
Hospital doctors underreported newly detected prediabetic state and DM at discharge,
precluding further HbA1c control as recommended by guidelines and good quality of care [4]. Still, the underreporting rate of DM in our study (40%) was lower than in a previous
Canadian study, which was approximately 80% [16]. Nowadays, at discharge, systematic report regarding whether an Hba1c was performed
and its corresponding result could be easily achieved using electronical medical records.
Similarly, antidiabetic drug treatment was prescribed to only half (48%) of new detected
DM cases, a value comparable to a study conducted in the Swiss general population
(53.3% in 2007) [17] but lower than in a study conducted in Lausanne (86%)[2].
In our study, hospital doctors were three times more likely to prescribe HbA1c in patients with a known DM, as if they were more concerned by the quality of DM
care than by hospital DM screening. This finding is in accordance with a study performed
in a German hospital, where 34% of patients with a known DM had a HbA1c measurement [9]. Hence, delivering DM prevention in asymptomatic patients requires skills that traditionally
trained clinicians may not possess [18]. Thus, hospital DM detection needs to be coupled with resources for physician education
and management in the outpatient setting.
Strengths and limitations
There is little information regarding in-hospital DM screening and management of newly
detected DM in hospitalized medical patients outside English-speaking countries. Hence,
our data provide important information for the public health management of DM in Europe.
Our study also has several limitations. Firstly, the detection of (pre)DM could only
be achieved in the sample of patients who benefited from HbA1c measurement. Thus, whether the current practice of HbA1c measurement in our teaching
hospital is sufficient and effective needs to be further studied. Although our results
cannot be generalized to the total population of hospitalized medical patients, still
they indicate that when a doctor asks for HbA1c measurement, the likelihood of detecting (pre)DM is high. Secondly, we could not
distinguish between DM type 1 and 2. However, because the prevalence of type 1 DM
is low, we assume that most of the cases of DM in our study were type 2 DM. Thirdly,
some patients with a history of DM could have been missed. Still, we used a broad
definition of DM history including ICD-10 codes in previous hospitalisations and glucose>11.8 mmol/L
(213 mg/dL), HbA1c≥6.5% (47.5 mmol/mol) or any antidiabetic treatment at admission. Fourthly, a possible
selection cannot be excluded, hospital doctors (un)consciously prescribing HbA1c measurements in patients who are more at risk of presenting with DM. Indeed, restricting
the analysis to patients without history of DM, significantly higher frequencies of
history of cardiovascular disease and hypertension and higher glucose levels and a
lower frequency of kidney disease were found in patients with HbA1c measurement, while no significant differences were found for gender, age, and history
of dyslipidemia ([Table 3]). Fifthly, no information regarding lifestyle intervention was available in the
electronic medical files. Hence, only drug interventions could be assessed. Finally,
the reporting of newly detected (pre)DM was based on ICD-10 codes at discharge, and
it is possible that results of HbA1c measurement might be present in the discharge letter but not coded in the medical
record. Still, it has been shown that diabetes is adequately reported using ICD-10
codes in Switzerland [19], so we believe that ICD-10 codes could adequately characterize the conditions indicated
in the discharge letter.
Conclusions
Hospital doctors do screen for diabetes, however they seldom prescribe HbA1c measurement
in medical hospitalized patients without known DM. Because prescription of HbA1c measurement leads to a high detection rate (53%) of (pre)DM among patients unaware
of their status, management and reporting of these conditions at discharge could be
further improved.
Author’s contribution
MM conducted the study and wrote most of the article; PMV made the statistical analyses
and wrote part of the article; GW revised the article for important intellectual content,
PMV had full access to the data and is the guarantor of the study. All authors have
read and approved this version of the article.