If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Admission glucose level was associated with increased short-term mortality and length-of-stay irrespective of diagnosis, treating medical specialty or concomitant laboratory values
Department of Clinical Sciences, Endocrinology and Diabetes, Lund University, Lund, SwedenDepartment of Internal Medicine, Endocrinology and Diabetes, Central Hospital, Region Kronoberg, Växjö, Sweden
Department of Clinical Sciences, Endocrinology and Diabetes, Lund University, Lund, SwedenDepartment of Research and Development, Region Kronoberg, Växjö, Sweden
Corresponding author at: Department of Clinical Sciences, Endocrinology and Diabetes, Lund University and Department of Research and Development, FoU, Region Kronoberg, Box 1223, S 351 12 Växjö, Sweden.
Department of Clinical Sciences, Endocrinology and Diabetes, Lund University, Lund, SwedenDepartment of Internal Medicine, Endocrinology and Diabetes, Central Hospital, Region Kronoberg, Växjö, SwedenDepartment of Research and Development, Region Kronoberg, Växjö, Sweden
Influence of admission glucose is less known in heterogeneous, non-intensive care populations.
•
Admission hyper- and hypoglycemia increased mortality in all patients, with or without diabetes.
•
Length-of-stay was 1–2 days longer with higher admission glucose if no known diabetes.
•
Both above irrespective of diagnosis, medical specialty or simultaneous laboratory analyses.
•
The first Scandinavian study of emergency admission glucose and mortality and length-of-stay.
Abstract
Background
Glucose is a routine emergency sample. General guidelines for inpatient hyperglycemia are scarce, except in myocardial infarction, stroke, and perioperative/ICU. Previous studies found admission glucose associated with increased mortality in specific conditions. Scandinavian data, and for general patients, are scarcer. We investigated admission glucose levels, 30-day mortality, and length-of-stay (LoS), in a Swedish hospital.
Methods
From 8146 emergency visits data regarding age, gender, dates of admission, discharge and death, diagnoses, admission p-glucose, s-sodium, s-potassium, b-hemoglobin, b-WBC and s-CRP, was collected, and for 6283 information regarding diagnosis of diabetes the previous 5 years. Visits were grouped in hypoglycemia (≤4.0), normoglycemia (>4.0–≤7.0), modest (>7.0–≤11.1) and severe hyperglycemia (>11.1) mmol/l.
Results
Short-term mortality was 1.5% in the normoglycemic, 2.6% in the hypoglycemic, 4.0–4.5% in modest and severe hyperglycemia, p < 0.001; Cox proportional hazard ratios (HR) for groups of patients without/with diabetes were 6.8; 1; 3.4; 4.4/7.3; 3.9; 4.0; 2.1 compared to the normoglycemic without diabetes (p 0.0001–0.05); adjusted for age, and concurrent levels of sodium, potassium, Hb, WBC and CRP 1.51 (1.07–2.1, p 0.02) with modest hyperglycemia, and 1.08 (0.60–1.95, p 0.80) in severe hyperglycemia. Mean LoS was 1.2 and 1.7 days longer with modest and severe hyperglycemia.
Conclusions
Short-term mortality increased substantially with admission hypo- and hyperglycemia for patients both with and without diabetes, irrespective of treating medical specialty, main discharge diagnosis, or concurrent laboratory values. Patients with diabetes (16%) were older, with higher glucose levels at admission, and with a different pattern of the association of admission glucose and mortality.
Hyperglycemia without a history of diabetes is often encountered in acutely ill patients, and might be an indication of stress, although there are also patients with undiagnosed diabetes [
]. Hyperglycemia is shown to negatively impact morbidity and mortality during and after the acute phase of conditions like myocardial infarction, stroke, trauma and pneumonia [
Intensified insulin-based glycaemic control after myocardial infarction: mortality during 20 year follow-up of the randomised diabetes mellitus insulin glucose infusion in acute myocardial infarction (DIGAMI 1) trial.
Elevated admission glucose is common and associated with high short-term complication burden after acute myocardial infarction: insights from the Validate-Swedeheart study.
]. It is still unclear whether abnormal admission glucose is associated with increased mortality in a heterogeneous group of medical and surgical emergency patients with different diagnoses, and if so as independent risk factor, or just indicator of increased neuroendocrine, inflammatory and/or metabolic activity in response to critical illness [
Intensified insulin-based glycaemic control after myocardial infarction: mortality during 20 year follow-up of the randomised diabetes mellitus insulin glucose infusion in acute myocardial infarction (DIGAMI 1) trial.
Elevated admission glucose is common and associated with high short-term complication burden after acute myocardial infarction: insights from the Validate-Swedeheart study.
], and some newer European studies of cardiac events, most early studies of admission glucose were performed in North America or Australia/New Zealand [
Intensified insulin-based glycaemic control after myocardial infarction: mortality during 20 year follow-up of the randomised diabetes mellitus insulin glucose infusion in acute myocardial infarction (DIGAMI 1) trial.
Prevalence and relevance of abnormal glucose metabolism in acute coronary syndromes: insights from the PLATelet inhibition and patient Outcomes (PLATO) trial.
P-glucose is a routine sample for most patients in emergency departments (ED). Due to unclarities regarding association of admission hyperglycemia with mortality and morbidity, and LoS, and lack of consensus regarding glucose lowering treatments in acute situations, in Sweden and elsewhere there are no established general guidelines for routine treatment of hyperglycemia in acute conditions, outside of ICU and surgical areas, except for stroke, sepsis and myocardial infarction [
Intensified insulin-based glycaemic control after myocardial infarction: mortality during 20 year follow-up of the randomised diabetes mellitus insulin glucose infusion in acute myocardial infarction (DIGAMI 1) trial.
]. Few, if any, studies investigated glucose and mortality in a heterogeneous emergency population of all adult ages, irrespective of diagnoses or treating specialty, or in secondary care. The aim of this study was to investigate associations between admission glucose level and mortality, and LoS, in a heterogeneous group of emergency patients at a secondary hospital in Sweden, and to explore differences in the patterns of that relation between those without and with diabetes.
2. Methods
Data were collected retrospectively via Cambio Cosmic® – the electronic records system in Health Care Region Kronoberg, in southern Sweden, that provided the data deidentified to the researchers. Only patients seeking acute care at the Emergency Department (ED) at Växjö Central Hospital were included (no scheduled admissions). It is the larger of the only two in Kronoberg, serving the eastern catchment area (136 000 inhabitants), and the only ED available to the entire population, regardless of socioeconomic status or other parameters. Inclusion criteria were visits by patients aged ≥18 years, where plasma glucose was sampled. All consecutive emergency visits to the ED during six months, January, 1st, 2014 to June 29th, 2014 were screened. For exclusions see Fig. 1, Flow chart. Information regarding age, gender, dates of admission and discharge, visited/admitting service/specialty within the ED, which serves internal medicine, infectious diseases, surgery, orthopedic, and geriatric specialties; laboratory values, dates of death during 30 days of follow-up, and discharge diagnoses according to ICD-10, grouped by affected organ systems or other likeness (ie malignancies), were collected. Previously known diabetes was defined as diagnosis of any type of diabetes mellitus (E10–11 or E14), except gestational diabetes, recorded at least once by hospital or primary care any time during the preceding five years. All Region Kronoberg, the two hospitals and primary care, including private practices, use the same electronic records system. From analyses comparing groups with/without previously known diabetes 1863 patients were excluded due to incomplete information regarding potential diagnosis of diabetes the preceding 5 years. Visits were grouped based on admission plasma glucose levels, defined as hypoglycemia ≤4.0 mmol/l, normoglycemia >4.0–≤7.0 mmol/l, modest hyperglycemia >7.0–≤11.1 mmol/l and severe hyperglycemia >11.1 mmol/l [
], the limit of 4.0 relating to common definitions of hypoglycemia in clinical trials. Venous p-glucose and other venous samples were routinely collected at triage, and analyzed within 60–90 min at the Department of Clinical Chemistry. P-glucose was analyzed with the Olympus AU2700 instrument (Olympus, Melville, NY); the other analyses with ADVIA 2120 routine hematology analyzer (Siemens AG, Erlangen, Germany). Levels of b-Hemoglobin, s-Sodium, s-Potassium, s-CRP and b-WBC were documented if available. Blood samples had been collected from 95% of internal medicine, 75% surgical and most infectious patients. The study was approved by the Ethical Review Board of Linköping University.
Fig. 1Flow chart of sample selection and distribution of patients. Patients excluded from the study are indicated with dashed arrows.
Normally distributed descriptive data, presented as mean±SD, were analyzed with One-Way ANOVA for comparisons between groups, when applicable with Bonferroni post-hoc test for multiple comparisons. Non-parametric data, median and quartiles (q1, q3, max-min), were analyzed with Mann–Whitney U test or the Kruskal–Wallis test. All confidence intervals are 95%, all R2 are Nagelkerke; p values ≤0.05 were considered significant; all tests were two-sided. Univariate analysis was applied to all variables, those with p < 0.25 were included in multiple regression models. Logistic regression (Backward: Wald) was used to explore possible predictors of mortality and LoS, the goodness-of-fit was confirmed with Hosmer–Lemeshow, p = 0.93 (ns), so the model used was a good fit. For odds ratios ordinal regression was applied. Hazards were screened with Kaplan-Meier method for proportionality allowing analysis with the Cox proportional hazards method. SPSS (Statistical Package for the Social Sciences) versions 22 + 23 (IBM, Chicago, Illinois) were used.
3. Results
3.1 Demographic data
There were 14 981 visits to the ED during the 6-month period. For exclusions see Fig. 1. Internal Medicine, Surgery and Infectious diseases emergency services had 10 823 visits by adult patients; p-glucose was collected in 75.3% (8146) of those; 63.8% (5201) to medicine, 23.7% (1931) surgery and 7.9% (610) infectious services. Data regarding specialty service was missing for 404 (5.0%) patients who were admitted to the common emergency ward, for maximum 36 h, then discharged or admitted to any of the three clinics. For study population see flow chart, Fig. 1, and Table 1. For glucose levels in patients with and without known diabetes see Table 3. Median (q1, q3; min-max) age of all the 8146 patients was 64 (43, 78; 18–103), mean 59.9 ± 21.5, years (Table 1). The patients with modest or severe hyperglycemia were significantly older, median age 72 (59, 82; 18–101) years and 74 (62, 82; 18–101) years respectively, p < 0.001 for both compared to 60 (40, 76; 18–103) years for the normoglycemic and 47 (30,67; 19–85) for the hypoglycemic (without/with diabetes 40.6 ± 21.5, 19–83/61.2±, 28–85). Of the 8146 patients, 49% were women and 51% were men (ns). There were no significant differences between the sexes neither between the different glucose groups, nor with/without diabetes, nor in any other parts of the study.
Table 1Distributions of age (years), gender and rate of admittance by admission glucose level (mmol/l), by treating specialty, for 7742 patients with this information (the 404 admitted to the common emergency ward excluded).
The majority of the 8146 emergency patients were normoglycemic, 65.5%, few were hypoglycemic, 0.5%, 26.7% had modest hyperglycemia and 7.3% severe hyperglycemia (Table 2a). The majority of the patients visited the service of internal medicine (Table 1). Distribution of admission p-glucose among patients examined by the different specialties was equal to the total group (Table 1).
Table 2aMortality, admission and length-of-stay (LoS) in relation to admission glucose level.
P-Glucose mmol/l
n (%)
Admitted n (%)
Median LoS (q1, q3; range)
Mean LoS ±SD
Mortality 30-day n (%)
In-hospital mortality n (%)
≤ 4.0
38 (0.5)
20 (52.6)
2 (1, 4; 0–26)
4.5 ± 5.8
1 (2.6)
1 (2.6)
4.0–≤ 7.0
5334 (65.5)
2414 (45.3)
2 (1, 5; 0–61)
4.0 ± 5.4
78 (1.5)
39 (0.7)
7.0–≤ 11.1
2179 (26.7)
1372 (63.0)
3 (1, 7; 0–110)
5.2 ± 6.5
88 (4.0)
48 (2.2)
> 11.1
595 (7.3)
401 (67.4)
4 (2, 8; 0–24)
5.6 ± 5.3
27 (4.5)
20 (3.4)
Total
8146 (100)
4207 (51.6)
2 (1, 6; 0–110)
4.5 ± 5.8
194 (2.4)
108 (1.3)
Mortality, admission rates and Length-of Stay (LoS). Crude (COR) and Adjusted (AOR) Odds Ratios for different levels of admission glucose.
Table 2bDifferences between mean Length-of Stay, LoS, and median LoS, per admission glucose level group, compared to the normoglycemic group (> 4.0–≤ 7.0).
P-Glucose mmol/l
Median LoS (days) (q1, q3; range)
P-value
Mean LoS (days) (95% CI)
P-value
≤ 4.0
2 (1, 4; 0–26)
0.95
4.4 (1.3–7.4)
0.99
> 4.0–≤ 7.0
2 (1, 5; 0–61)
–
4.0 (3.7–4.2)
–
>7.0–≤ 11.1
3 (1, 7; 0–110)
< 0.0001
5.2 (4.8–5.5)
< 0.001
> 11.1
4 (2, 8; 0–24)
< 0.0001
5.6 (5.1–6.1)
< 0.001
Mortality, admission rates and Length-of Stay (LoS). Crude (COR) and Adjusted (AOR) Odds Ratios for different levels of admission glucose.
The total 30-day mortality was 2.4%, for women 2.3%, men 2.5% (ns). Mortality of the normoglycemic was 1.5% and increased to 2.6% with admission hypoglycemia, 4.0% if modest hyperglycemia and 4.5% if severe hyperglycemia, p < 0.001, (Table 2a). The 30-day mortality was significantly lower for the surgical, 1.3%, compared to the medical, 2.7%, patients, p < 0.001, both differed significantly from the total, 2.4%, p < 0.001, but that for the infectious patients did not, 2% (Table 2a). The in-hospital mortality was 1.3% overall and increased with higher admission glucose, but was significantly higher only for the patients with severe hyperglycemia, p < 0.02 (Table 2a). The in-hospital mortality for internal medicine was 1.6%, surgery 0.9%, and infectious diseases 1.3%, not significantly different between the groups or from the total. Admission glucose explained 3.2% (R2 = 0.032, p < 0.01) of the variance in mortality, and admission glucose, age, sodium, potassium, hemoglobin, WBC and CRP levels together explained 16% (R2 = 0.159, p < 0.001). Crude hazard ratio (HR) for all was 3.01 (2.19–4.15. p < 0.001) for modest and 2.50 (1.47–4.25, p 0.001) if severe hyperglycemia, compared to normoglycemia. Patients with hypoglycemia had the highest HR, 5.87 (1.85–18.7, p 0.003). Age had a significant influence on risk of 30-day mortality, HR 1.09 (1.07–1.11, p < 0.0001), so if age was increased with 10 years the HR was increased to 2.32 (1.99–2.27).
Adjusted hazard ratio (AHR) for age and the mentioned laboratory variables was 8.11 (1.95–33.6, p < 0.004) for the hypoglycemic; 1.51 (1.07–2.1, p 0.02) with modest hyperglycemia, and lost its significance in those with severe hyperglycemia, 1.08 (0.60–1.95, p 0.80). For crude HR for the 6200 patients without/with diabetes and complete information regarding both mortality and diabetes status see Fig. 2 and Table 5. The differences found in patterns of mortality in relation to admission glucose level between those with a previous versus no diagnosis of diabetes are visualized in Figs. 2 and 3 and Table 5.
Fig. 2Cox regression survival curve for patients without and with diabetes, with different levels of admission glucose.
Admission rates varied between medical specialties. Of the patients triaged to infectious diseases 66% were admitted, but only 48% of the surgical and 55% of the internal medicine patients (Table 1). The admission rate for the three clinics medicine, surgery and infectious diseases was mean 56.4%, and increased if the patient had modest or severe hyperglycemia, both p < 0.001, as did the LoS. LoS was longer with higher admission glucose (Table 2a + b + c). Between specialties LoS only significantly differed for infectious diseases where the patients were admitted mean 5.8 ± 6.0 days compared to internal medicine 4.5 ± 5.6 days and surgery 4.0 ± 6.4 days (ns). Length-of-stay increased with degree of admission hyperglycemia (Table 2a + c + d), and was one day longer for patients with modest hyperglycemia, mean 1.2 days, and two days, mean 1.7 days, if severe admission hyperglycemia compared to normoglycemic patients. Mean LoS for patients with admission hypoglycemia was longer than in the normoglycemic but without significance (Table 2c). For further analysis LoS was categorized into five groups which were considered clinically relevant: 0–2, 3–4, 5–7, 8–14 and >15 days, with COR of 1.70 (1.50–1.93) for modest and 2.27 (1.87–2.75) for severe hyperglycemia, for risk of having a LoS in the next higher category than if normoglycemic at admission. After adjustments for age, electrolytes, hemoglobin, WBC and CRP, the AORs were 1.16 (1.01–1.34), and 1.46 (1.17–1.81) for patients with modest or severe hyperglycemia (Table 2c).
Table 2cOdds ratios regarding the risk to be in the adjacent higher LoS category (0–2, 3–4, 5–7, 8–14 and >15 days) in comparison to the normoglycemic group.
The most common groups of discharge diagnoses for the medical patients were Diseases of the Circulatory system I00-I99, 39.9%, Diseases of the Respiratory system J00-J99, 16.6% and Diseases of the Nervous system G00-G99, 8.3%; for surgical patients Diseases of the Digestive system K00-K93, 54.3%, Diseases of the Urogenital system N00-N99, 16.9% and Neoplasms C00-D48, 12.9%; for infectious patients Certain Infections and Parasitic diseases A00-B99, 34.8%, Diseases of the Respiratory system J00-J99, 31.6% and Diseases of the Urogenital system N00-N99, 15.1%. No significant differences were found in mortality or LoS between the groups of diagnoses. The OR for being admitted from the ER, was increased if glucose was increased, modestly 2.0 (1.8–2.2) and severely 2.4 (2.0–2.9), in line with those with a diagnosis of COPD (Chronic Obstructive Pulmonary Disease) 2.6 (1.3–5.4) but not as high as with myocardial infarction 7.2 (2.6–20.5), stroke 5.4 (2.3–13.1) or heart failure 3.7 (1.8–7.3), all p < 0.001.
3.5 Other laboratory values
There were no significant differences between the glucose groups in levels of electrolytes or hemoglobin on arrival to the ED. The levels of WBC and CRP, however, were significantly higher in the groups with modest or severe hyperglycemia compared to the normoglycemic patients (Table 4). The HR of mortality within 30 days of admission was 1.5 (1.1–2.2), if s-potassium increased with 0.4 mmol/l (p < 0.02), and 1.03 (1.02–1.05) if LPK increased with 0.03 × 103/l, (p < 0.03). Small increases in S-Na, or CRP, carried very small, but significant, increases in risk of 30-day-mortality, and small decreases of B-Hb were protective (data not shown).
3.6 Diabetes status
Information regarding previously known diagnosis of diabetes was available for 77% (6283/8146). A previously known diagnosis of diabetes was present in 15.7% (987/6283) of the patients with information available for the five years preceding the emergency visit. The associations of admission glucose with mortality were significant when analyzed with the Cox proportional hazards method (Fig. 2 and Table 3). Admission hypoglycemia carried the highest HR for death within 30 days regardless of diabetes status. Apart from that the patterns of association differed between groups without/with diabetes (Figs. 2 and 3). Regarding LoS, the association was only significant for ED patients without a known diagnosis of diabetes (Tables 2a–c). Admission hyperglycemia was more frequent in ED patients with known diabetes, only 20.8% of them were normoglycemic (Table 5).
Table 3Hazard Ratios (HR) for mortality within 30 days of admission by level of admission glucose, and diabetes status, in medical, surgical and infectious patients visiting the emergency department at Växjö Central Hospital, Kronoberg; compared to the group without diabetes and with normal level of glucose (> 4.0–≤ 7.0).
Cox regression model. **AHR = Adjusted Hazard Ratio; adjusted for age, and levels of p sodium, potassium, Hemoglobin, White Blood Cell count and CRP (C-reactive protein): modest hyperglycemia 1.51 (1.07–2.01), p<0.02; severe hyperglycemia 1.08 (0.60–1.95), p 0.8.
Cox regression model. **AHR = Adjusted Hazard Ratio; adjusted for age, and levels of p sodium, potassium, Hemoglobin, White Blood Cell count and CRP (C-reactive protein): modest hyperglycemia 1.51 (1.07–2.01), p<0.02; severe hyperglycemia 1.08 (0.60–1.95), p 0.8.
CI (95%)
p
≤ 4.0
6.8
1.68–28.1
0.007
7.3
1.01–52.8
0.05
> 4.0–≤ 7.0
1
–
–
3.9
2.02–7.36
0.0001
>7.0–≤ 11.1
3.4
2.35–4.92
0.0001
4.0
2.45–6.49
0.0001
> 11.1
4.4
1.58–12.0
0.004
2.1
1.06–4.10
0.03
Cox regression model. **AHR = Adjusted Hazard Ratio; adjusted for age, and levels of p sodium, potassium, Hemoglobin, White Blood Cell count and CRP (C-reactive protein): modest hyperglycemia 1.51 (1.07–2.01), p<0.02; severe hyperglycemia 1.08 (0.60–1.95), p 0.8.
Table 4Laboratory values collected simultaneously with p-glucose upon arrival to the Emergency Department. Values were available for 96–99% of the included patients, except for s-potassium (93%).
Both in patients without and with previously diagnosed diabetes 30-day mortality was increased if admission plasma glucose level was abnormal in this sample of mixed medical, surgical and infectious emergency patients, irrespective of main discharge diagnosis group, or treating medical specialty. The highest significant risk was found in emergency patients arriving with hypoglycemia, despite a quite low number of episodes, and irrespective of diabetes status. This is a new finding, and the reasons for hypoglycemia may be quite diverse in patients with, and without diabetes, and its treatment. The HR was highly significant for the large groups with glucose levels above 7.0 mmol/l, and also after correction for age, and simultaneous levels of electrolytes, hemoglobin, WBC and CRP (except for diabetes patients with glucose >11.1 mmol/l). The patterns of association between admission glucose level and mortality were different between those with and without diabetes. Apart from both groups displaying the highest mortality risk with admission hypoglycemia, the association increased linearly with increasing glucose level in those without diabetes. With known diabetes and admission normoglycemia or modest hyperglycemia HR for 30-day mortality was of the same magnitude as with no diabetes and modest or severe hyperglycemia (3.5–4.5), while with diabetes and severe hyperglycemia the HR was only half of that (2.1), compared to no diabetes and normoglycemia. These detailed findings regarding the patterns of glycemia are new, and the most likely explanation is assumed to be that hyperglycemia is taken more seriously, and being more promptly treated, in patients with known diabetes. Our general findings regarding admission hyperglycemia were in agreement with findings from a university hospital in Dublin, Ireland, and an Australian tertiary hospital [
]. The study populations were comparable regarding distribution of age and gender. Their definitions of glucose groups were close to ours. The levels of mortality, however, were lower in our population, suggesting differences, as our population based catchment, and including all ED patients, not only the admitted, compared to those of larger city tertiary referral EDs, meaning other populations were more selected with accumulation of more severely ill patients. This will account both for higher mortality rates, and higher prevalence of diabetes in those cohorts. Another possible explanation might be true differences in mortality due to differences in care performance. Differences in care settings have been frequently discussed regarding intensive insulin treatment in the ICU setting [
We found LoS with modest or severe admission hyperglycemia 1 and 2 days longer compared to normoglycemia. The hyperglycemic patients had higher OR for longer LoS, partly consistent with findings from Ireland, and upstate New York, US [
]. Both studies found LoS one day longer, irrespective of glucose level, compared to this study, raising the question if LoS times are shorter in Sweden/Kronoberg in general, or is it an effect of different study populations?
4.3 Hypoglycemia
Hypoglycemia upon admission was the most dangerous level of admission glucose regarding short-term mortality and reached significant HR both without and with diabetes, albeit wide confidence intervals, despite the low number of episodes in that group. One explanation may be that most hypoglycemic events are treated immediately outside the hospital, by patients themselves or bystander assistance. Severe hypoglycemia in need of intravenous treatment is treated immediately on-site by arriving paramedics. Either those patients chose to stay home, or were treated, resulting in higher glucose on ED arrival. Mean LoS with admission hypoglycemia tended to be longer than with normoglycemia. The proportion with hypoglycemia corresponded to findings from Ireland, but not from the US [
], why higher admission p-glucose might indicate a more severe condition, also indicated by the higher age in patients with hyperglycemia. Diabetes is increasingly prevalent with increasing age [
]. Alternatively hyperglycemia might induce a more severe condition, rather than being marker of severity. Support for this assumption might be studies where surgical ICU and perisurgical patients were treated with intensive insulin therapy to lower glucose [
]. The risk of hypoglycemia was found, as elsewhere, to be the main obstacle to better glucose control also in the ICU, why advice for treatment have been modified [
For the laboratory values taken simultaneously with p-glucose levels were quite robust, even when abnormal glucose levels, but then 65% were normoglycemic. In the hyperglycemic for instance a tendency towards hyponatremia may have been counteracted by dehydration. Our findings are congruent with sodium levels at two large Swedish EDs where prevalence of hyponatremia was rare, only 3%, and with no differences in mortality between patients with different sodium levels [
]. In our study only WBC and CRP differed significantly between different groups of glycemia. We are not aware of any other articles regarding mortality, LoS and admission glucose that investigated relations with other simultaneous laboratory values. A study of admission laboratory results and 30-day mortality for 8000 patients with acute heart failure found that albumin, sodium, urea, uric acid and WBC influenced risk of 30-day mortality [
]. Their groups all had admission p-glucose ca 9.0 mmol/l. Of those patients, however, 38% had known diabetes, as compared to 16% in our sample. We found higher risk of 30-day mortality with higher WBC, in line with the heart failure study, where increased WBC indicated 60% increased risk of 30-day mortality in that high risk group, where also sodium level influenced mortality, but for levels <136 mmol/l, lower than we found [
Strengths of our study are first that we have explored and further described the differences in patterns of associations between admission glucose and mortality in patients without/with diabetes, and secondly the cases without diabetes were verified so for five years before the admission. Third, the large number of episodes investigated, with possibility to detect differences between groups not only of age, gender, and admission glucose levels but also different categories of diagnoses, and, forth, relations to other laboratory values. Fifth, our study is one of few that include results from EDs of different medical specialties, not just specific conditions [
Intensified insulin-based glycaemic control after myocardial infarction: mortality during 20 year follow-up of the randomised diabetes mellitus insulin glucose infusion in acute myocardial infarction (DIGAMI 1) trial.
]. Sixth, it is population based, being the only ED in the area. Seventh, often settings are critical/intensive/coronary care units, leaving out the large numbers of general patients [
Intensified insulin-based glycaemic control after myocardial infarction: mortality during 20 year follow-up of the randomised diabetes mellitus insulin glucose infusion in acute myocardial infarction (DIGAMI 1) trial.
]. In this respect our results confirm and expand the knowledge of the influence of admission glucose level also in more ordinary secondary care hospital settings. Eighth, to our knowledge this is the only study from Sweden or Scandinavia that explores admission glucose with 30-day mortality and LoS for a heterogeneous group of emergency patients.
Limitations of this study are retrospectivity, using deidentified records data, why controlling for fasting was not possible. Determining whether a patient without known diabetes had new/unknown diabetes was only possible for the 64 patients, 1.2%, without diabetes, that presented with a venous p-glucose of >11.1 mmol/l [
]. Ensuring that controls were free of diabetes five years prior to admission ought to minimize this risk. Another limitation was that previous diabetes status was unclear for 1863 patients that were excluded from comparisons of diabetes/no diabetes. Reasons may have been migration, or attending a physician outside Kronoberg.
With the Cox proportional hazards method there were significant associations of admission glucose levels and 30-day mortality, for all groups of glucose values in patients with diabetes, and all with abnormal glucose levels and no diabetes, compared to normoglycemic without diabetes, in disagreement with many previous studies [
], as did an Israeli study of atrial fibrillation, where increased levels of admission glucose increased HR for mortality both in patients with and without diabetes [
]. Level of admission glucose per se, and not diabetes diagnosis, was associated with increased frequency of inhospital complications, and mortality, in two recent studies of stroke patients, from Israel and Spain [
We found lower short-term mortality, and LoS, in all the departments investigated, than studies both from other European countries, and elsewhere, which is a matter for further exploration per se. A further step would be treating patients with insulin more intensively irrespective of cause for the emergency visit, if an elevated glucose is detected. A few studies have explored this outside of intensive or surgical care settings [
Early intervention for diabetes in medical and surgical inpatients decreases hyperglycemia and hospital-acquired infections: a cluster randomized trial.
]. They found tendencies to improvement for the insulin treated, but with no significant differences in morbidity or mortality, perhaps due to small numbers. A recent Canadian study found that patients who received guideline-based care were hyperglycemic only 43%, compared to 64%, of the time [
], and other hospital settings, has probably contributed to the paucity of guidelines for treatment of general inpatients with hyperglycemia, as in Sweden, where only few specified conditions have national guidelines for treatment of inpatient hyperglycemia, as stroke, myocardial infarction, and ICU [
]. This might change soon, however, considering the technical devices for glucose monitoring now quickly entering the diabetes scene, already used to prevent inpatient hypoglycemia [
]. There is a need for further investigations for more definite answers regarding intervention to impact the associations of glucose levels with mortality and LoS [
]. This might also affect comorbidity, and amount of complications, as was found in the ICU studies, and have been found in several surgical settings, especially cardiothoracic surgery [
]. Potential applications of the study apply not only to the hospital stay. A follow-up should be performed after discharge if the patient had a pathological admission p-glucose. This would allow for early prevention and diagnosis of diabetes and prevention of future complications and morbidity [
In this first Scandinavian study of admission glucose in the emergency department of a Swedish secondary hospital we found that admission glucose level, especially hypoglycemia, was associated with in-hospital and 30-day mortality, both in patients without, and with, diabetes, irrespective of main discharge diagnosis, treating medical specialty or levels of simultaneously collected routine laboratory analyzes. The mortality rates and length-of-stay found were lower than those in studies from other countries and settings. Since admission glucose levels influence the prognosis for hospitalized patients, and the need for resources, further research regarding the best way to treat this, also outside of ICU and tertiary settings, is still warranted.
Funding
This work was supported by the Health Care Region Kronoberg; The Medical Faculty of Lund University; and the Kamprad Family Foundation, all Sweden. The funding sources had no influence on the conception or design of the study, the collection, analysis or interpretation of data, the writing of the manuscript or the decision to publish it.
Declaration of Competing Interest
The authors declare they have no conflict of interest.
Acknowledgements
The authors are grateful to PhD Anna Lindgren, center for Mathematical Sciences, Lund University, Lund, Sweden, for statistical advice.
References
Cho N.H.
Shaw J.E.
Karuranga S.
Huang Y.
da Rocha Fernandes J.D.
Ohlrogge A.W.
et al.
IDF diabetes atlas: global estimates of diabetes prevalence for 2017 and projections for 2045.
Intensified insulin-based glycaemic control after myocardial infarction: mortality during 20 year follow-up of the randomised diabetes mellitus insulin glucose infusion in acute myocardial infarction (DIGAMI 1) trial.
Elevated admission glucose is common and associated with high short-term complication burden after acute myocardial infarction: insights from the Validate-Swedeheart study.
Prevalence and relevance of abnormal glucose metabolism in acute coronary syndromes: insights from the PLATelet inhibition and patient Outcomes (PLATO) trial.
Early intervention for diabetes in medical and surgical inpatients decreases hyperglycemia and hospital-acquired infections: a cluster randomized trial.
AOR: Odds Ratio Adjusted for age and levels of sodium, potassium, Hb, WBC and CRP.
30010-8&id=gr1.jpg){kind=link}
30010-8&id=gr2.jpg){kind=link}
30010-8&id=gr3.jpg){kind=link}