Outcomes for general medical inpatients with diabetes mellitus and new hyperglycaemia

Scott T Baker, Cherie Y Chiang, Jeffrey D Zajac, Leon A Bach, George Jerums and Richard J MacIsaac
Med J Aust 2008; 188 (6): 340-343. || doi: 10.5694/j.1326-5377.2008.tb01650.x
Published online: 17 March 2008

A 2002 study of hospital inpatients in Atlanta, USA, reported prevalences for diabetes mellitus and new hyperglycaemia of 26% and 12%, respectively.1 New hyperglycaemia was associated with adverse outcomes, including increased mortality, increased hospital length of stay, and increased rates of high-level residential care placement. Inpatient hyperglycaemia in the non-diabetic patient is a risk factor for in-hospital mortality and heart failure in the setting of acute myocardial infarction.2-6 Hyperglycaemia at the time of acute ischaemic stroke is also linked to worse functional outcome and increased mortality.7-9 A more recent study has linked hyperglycaemia with adverse outcomes in patients with community-acquired pneumonia.10

It has been unclear whether hyperglycaemia directly affects the outcome of disease, or whether hyperglycaemia merely reflects a stress response to the severity of the intercurrent illness. It is also difficult to differentiate true “stress hyperglycaemia” from existing but previously undiagnosed diabetes mellitus. However, one study has suggested that a glycated haemoglobin (HbA1c) level > 6.0% in acutely ill inpatients with hyperglycaemia on admission can reliably diagnose diabetes.11

The prevalence of diabetes in Australian tertiary referral hospitals has been estimated to be 23%.12 The prevalence of newly diagnosed hyperglycaemia and its associations has not been documented in Australian hospitals. The aim of this prospective study was to determine the prevalence of known diabetes mellitus and new hyperglycaemia in patients admitted to a general medical ward in a tertiary referral hospital, and to investigate their relationship with in-hospital mortality and length of stay.


We studied 903 consecutive admissions in three recruitment periods spread across the year (25 Feb 2003 to 24 Apr 2003; 18 Nov 2003 to 8 Apr 2004; 5 Jul 2004 to 12 Jul 2004). A total of 139 patients were excluded because of readmissions (only the first admission during the recruitment period was included), death or discharge during the first 24 hours, or missed pathology testing. Background demographics (age, sex), admission diagnosis, diabetes status (non-diabetic, type 1 or type 2 diabetes), and information regarding intravenous fluid therapy administered (none, normal saline, 4% dextrose + 1/5 normal saline, or 5% dextrose) were recorded, as well as the outcome of each patient’s admission (discharge or death), cause of death, and length of stay.

Ethics approval

The study was approved by the Human Research Ethics Committee of Austin Health.


Fasting plasma glucose (FPG) and HbA1c levels were measured the morning after admission. The fasting sample was drawn after at least 6 hours of overnight fasting. The American Diabetes Association definition of normoglycaemia, FPG < 5.6 mmol/L, was used to define the euglycaemic threshold, and hyperglycaemia was defined as FPG ≥ 7.0 mmol/L in subjects not known to have pre-existing diabetes.13 Patients with known diabetes were classified as “Group 1”. The remaining patients with no documented history of diabetes were subdivided according to their FPG as either patients with new hyperglycaemia (FPG, ≥ 7.0 mmol/L, “Group 2”), patients with indeterminate or uncertain glucose status (FPG, 5.6–6.9 mmol/L, “Group 3”), or patients with normoglycaemia (FPG, < 5.6 mmol/L, “Group 4”).

Plasma glucose levels were measured using a hexokinase method (Beckman Coulter, Fullerton, Calif, USA). HbA1c was measured using Primus CLC 330 (Primus Diagnostics, Kansas City, Mo, USA) affinity high performance liquid chromatography, which was calibrated against the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) approved reference measurement procedure.14

Statistical analysis

Categorical data were analysed by χ2 test. Continuous variables were not normally distributed, so characteristics of Groups 1–4 were analysed using the Kruskal–Wallis test and characteristics of high and normal HbA1c groups were analysed by the Mann–Whitney test. Logistic regression was used to determine (i) the relationship between mortality and the following variables: age either as a categorical (≤ 75 years or > 75 years) or continuous variable, length of stay, sex, and category of glycaemia; (ii) the relationship between mortality and the following variables in non-diabetic patients: (a) age, HbA1c level, length of stay, sex, and FPG; (b) age, length of stay, sex, and category of FPG and/or HbA1c level (≤ 6.0% or > 6.0%). Analyses were performed using SPSS, version 11 (SPSS Inc, Chicago, Ill, USA).


The characteristics and outcomes for subjects with and without a history of diabetes are shown in Box 1. Of the total patient population, 49 (5.4%) died during their inpatient stay, with age > 75 years (odds ratio [OR], 4.6; 95% CI, 1.8–11.8; P = 0.001) and admission FPG ≥ 5.6 mmol/L (OR, 2.0; 95% CI, 1.1–3.7; P = 0.02) being the only independent predictors of mortality. In patients without a known history of diabetes (Groups 2 to 4), mortality increased with increasing admission FPG levels, with a mortality of 3.8% in Group 4, compared with 8.0% in Group 3 and 10.3% in Group 2 (P = 0.04). Compared with those with FPG < 5.6 mmol/L (Group 4), uncertain glycaemic status and new hyperglycaemia (Groups 3 and 2) were associated with significant increases in the risk of inpatient death (Box 2). In those without known diabetes, there was a 33% increase in mortality for each 1 mmol/L rise in admission FPG (P = 0.02). There was a linear association between mortality and admission FPG in these patients (P = 0.005). Patients in Group 2 represented 16% of those who died and 8% of those who survived (P = 0.04); conversely, patients in Group 4 represented 35% of those who died and 51% of survivors (P = 0.04, Group 2 v Group 4). Comorbidities did not vary significantly across the groups.

HbA1c measurements were used to attempt to identify undiagnosed diabetes in patients without a documented history of diabetes at admission. Patients without a prior diagnosis of diabetes and an HbA1c level > 6.0% (elevated), suggestive of unrecognised pre-existing hyperglycaemia, had a mortality of 11.3%, whereas non-diabetic patients with an HbA1c level ≤ 6.0% (normal) had a mortality of 4.4% (P = 0.006). FPG levels were significantly higher in patients with an elevated HbA1c level (Box 3). In a multivariate analysis that included age, sex, length of stay, HbA1c level (normal or elevated) and FPG category, only age and HbA1c level showed a significant relationship with mortality, with odds ratios of 1.06 (95% CI, 1.03–1.10; P = 0.001) and 2.47 (95% CI, 1.16–5.26; P = 0.005), respectively.

There was no significant difference in mortality between those with normoglycaemia and those with known diabetes. In patients with known diabetes, there was no association between admission FPG levels, HbA1c level and mortality. There was no variation in FPG according to the type of intravenous therapy administered.

The most common causes of death, ascertained from the medical records, were infections (51%), including urinary and respiratory sepsis and septicaemia. There was no significant relationship in the total study population between glycaemic status and cause of death, perhaps due to the small number of deaths. No differences in length of stay were found among the four groups. Patients who died had a longer length of stay than survivors (median, 9 v 6 days; P = 0.01), but this relationship was of borderline significance after accounting for age, sex and glycaemic group using multivariate logistic regression (Box 2).


The background prevalence of diabetes mellitus in the Australian adult population is 7.4%, with an undiagnosed case for every known case of diabetes.15 The prevalence rises to 23% for people older than 75 years.15 We found that only half of a general medical inpatient population had a normal FPG level, with a prevalence of known diabetes of 22% and a further 28% having an FPG level ≥ 5.6 mmol/L. Mortality and length of stay for patients with known diabetes were similar to those of the group with FPG < 5.6 mmol/L. This suggests that identification of patients with diabetes before hospital admission may be beneficial in reducing mortality in this older population. Nearly half (41%) of those with admission fasting hyperglycaemia had an HbA1c level > 6.0%. One study found that an HbA1c level > 6.0% in hospital inpatients with an admission random glucose level > 7 mmol/L and no prior history of diabetes was 100% specific and 57% sensitive for diagnosing diabetes.11 However, the utility of HbA1c level in diagnosing and stratifying patients in the acute setting may be limited by results not being available within the first 24 hours of admission. This raises the importance of outpatient screening for diabetes, especially in at-risk populations, as preadmission diagnosis and treatment of patients with known diabetes may lead to better inpatient outcomes.

It has been shown that elevated HbA1c level has prognostic significance in non-diabetic patients undergoing percutaneous coronary intervention.16 In a study of patients admitted for coronary artery bypass surgery, admission HbA1c level > 7.0%, suggestive of undiagnosed diabetes, predicted increased length of stay; however, patients with HbA1c levels of 6.0%–6.9% had no increased length of stay.17 In our study, we found no significant difference in length of stay for patients with or without diabetes, regardless of HbA1c level. However, an elevated HbA1c level in patients without known diabetes at admission was associated with a significantly higher mortality than for patients with a normal HbA1c level (11.3% v 4.4%). Interestingly, when FPG and HbA1c categories were both entered as independent variables in the logistic regression analysis, FPG appeared to be a less powerful predictor of mortality than HbA1c. It should be noted that these two variables are correlated (Spearman ρ = 0.27; P < 0.001). The very high accuracy and precision of the HbA1c level method at our centre (bias of < 0.1% compared with the recently endorsed IFCC standardisation system;14 coefficient of variation, 1.9% [Mr Ian Goodall, Senior Scientist, Special Chemistry, Austin Health, personal communication]) allows confidence in discriminating between normal and abnormal HbA1c values; this may require further validation in centres using other methods.

Improved inpatient glycaemic control has been shown to be beneficial in specific hospital populations, including cardiac surgery,18 myocardial infarction,19 and intensive care20 settings. More recently, studies have been published which failed to confirm these findings,21-23 suggesting that more studies need to be performed to establish the benefit of tight glycaemic control, and to identify glycaemic targets for therapy and which inpatient populations are most likely to benefit from intensive glucose-lowering interventions. A recently published study of intensive care patients found a twofold higher mortality in non-diabetic patients compared with patients with diabetes, despite lower blood glucose levels in the non-diabetic group.24 This suggests that target levels for glycaemic control may differ between patients with and without diabetes. Further studies are required to evaluate whether improvements in glycaemic control in the older general medical inpatient population, especially in people with likely undiagnosed diabetes, results in reduced mortality.

Our study has some limitations. First, as the patients admitted to the general medical ward were older (median age, 79 years), survival bias could potentially explain the similar mortality rates between the diabetic and normoglycaemic groups. This may limit the extent to which these results can be generalised to the population at large, but the study group is representative of patients in the general medical wards of Australian tertiary referral hospitals. Admission criteria for diabetic and non-diabetic patients did not differ, so this cannot explain the similar mortality rates in the diabetic and normoglycaemic groups.

Second, patients who were discharged home or died within the first 24 hours after admission were excluded from the study, as they were unable to have the fasting blood test. Therefore, extremes of the clinical spectrum — the very well and the very sick — were excluded from the study. This could have resulted in a lower detection rate of new hyperglycaemic patients. One study found that patients with new hyperglycaemia had a longer hospital stay and an increased rate of intensive care unit (ICU) admission compared with patients with a history of diabetes or normoglycaemia.1 We did not study the ICU admission or transfer rate, as considerations such as advanced age result in proportionally fewer critically unwell patients being considered candidates for intensive care management. Thus, ICU transfer does not accurately reflect deterioration in or severity of this cohort’s condition, unlike in some other inpatient populations.

Third, post-discharge follow-up to confirm or reject a diagnosis of impaired glucose tolerance or diabetes proved difficult in the group with indeterminate plasma glucose levels. An outpatient post-discharge oral glucose tolerance test was performed in a small number of patients; however, testing a high percentage of patients with indeterminate FPG was not feasible because of their advanced age, with many requiring supported accommodation and assistance with travel. The abnormal FPG result was documented on patients’ discharge summaries for primary health care provider follow-up. People in this study with an HbA1c level > 6% and an FPG level ≥ 7.0 mmol/L were labelled as having diabetes, and appropriate therapy, including dietary and lifestyle advice, was instituted in hospital. Those with HbA1c level ≤ 6% and an FPG level ≥ 7.0 mmol/L were referred for repeat testing by their primary health care provider.

In conclusion, in patients without known diabetes, the risk of death was increased for admission FPG levels ≥ 5.6 mmol/L. However, pre-existing abnormal glucose metabolism, reflected by elevated HbA1c level, appeared a more important predictor of inpatient mortality than glucose levels in patients without known diabetes. Further studies are needed to determine if identification and early treatment of these patients leads to reduced inpatient mortality rates.

1 Characteristics of 903 patients admitted to a general medical ward

Group 1
n = 202 (22%)

Group 2
n = 78 (9%)

Group 3
n = 174 (19%)

Group 4
n = 449 (50%)


Age (years)

78 (70–84)

79 (72–84)

80 (71–86)

79 (71–86)


Sex (males)






HbA1c level

7.0% (6.2%–8.0%)

5.9% (5.6%–6.3%)

5.7% (5.5%–6.0%)

5.6% (5.4%–5.9%)

< 0.001

Prevalence of HbA1c level > 6%





< 0.001

FPG (mmol/L)

7.1 (5.7–8.9)

7.6 (7.2–8.4)

6.0 (5.8–6.3)

4.9 (4.6–5.2)

< 0.001







Age of death (years)

82 (74–91)

81 (78–93)

85 (83–90)

82 (77–88)


Length of stay (days)

7 (4–12)

8 (5–14)

6 (4–11)

6 (4–11)


FPG = fasting plasma glucose level. Group 1 = known diabetes. Group 2 = new hyperglycaemia (FPG ≥ 7 mmol/L). Group 3 = uncertain glycaemic status (FPG, 5.6–6.9 mmol/L). Group 4 = normoglycaemia (FPG, < 5.6 mmol/L). HbA1c = glycated haemoglobin. Results are shown as median (interquartile range) for continuous variables. P values were determined by Kruskal–Wallis test for continuous variables and χ2 test for categorical variables.

2 Odds ratios for mortality

Odds ratio (95% CI)


1.06 (1.03–1.10)

Length of stay

1.01 (0.996–1.03)

Female sex

0.80 (0.44–1.45)

Glycaemic group:

Known diabetes (Group 1)

1.41 (0.63–3.16)

FPG 5.6–6.9 mmol/L (Group 3)

2.20 (1.05–4.62)

FPG ≥ 7.0 mmol/L (Group 2)

2.87 (1.17–7.00)

FPG = fasting plasma glucose. Odds ratios were estimated using logistic regression. Female sex relative to male; glycaemic group relative to normoglycaemia (Group 4, FPG < 5.6 mmol/L).

3 Characteristics of patients without a history of diabetes (Groups 2, 3 and 4) according to glycated haemoglobin (HbA1c) level

HbA1c ≤ 6.0% (n = 569)

HbA1c > 6.0% (= 115)


Age (years) (median [IQR])

79 (71–84)

80 (72–86)


Sex (male)




Fasting plasma glucose (mmol/L) (median [IQR])

5.2 (4.7–5.8)

5.9 (5.2–7.1)

< 0.001

Length of stay (days) (median [IQR])

6 (4–12)

7 (4–12)






IQR = interquartile range. Continuous data were analysed by Mann–Whitney test and categorical data analysed by χ2 test.

Received 15 June 2007, accepted 19 November 2007

  • Scott T Baker1
  • Cherie Y Chiang1
  • Jeffrey D Zajac1
  • Leon A Bach2
  • George Jerums1
  • Richard J MacIsaac1

  • 1 Austin Health and University of Melbourne, Melbourne, VIC.
  • 2 Department of Diabetes and Endocrinology, Alfred Hospital and Monash University, Melbourne, VIC.


This study was supported by a Novo Nordisk Diabetes Regional Support Grant 2004.

Competing interests:

None identified.

  • 1. Umpierrez GE, Isaacs SD, Bazargan N, et al. Hyperglycemia: an independent marker of in-hospital mortality in patients with undiagnosed diabetes. J Clin Endocrinol Metab 2002; 87: 978-982.
  • 2. Capes SE, Hunt D, Malmberg K, Gerstein HC. Stress hyperglycaemia and increased risk of death after myocardial infarction in patients with and without diabetes: a systematic overview. Lancet 2000; 355: 773-778.
  • 3. Yudkin JS, Oswald GA. Hyperglycaemia, diabetes and myocardial infarction. Diabet Med 1987; 4: 13-18.
  • 4. Norhammar AM, Ryden L, Malmberg K. Admission plasma glucose: independent risk factor for long-term prognosis after myocardial infarction. Diabetes Care 1999; 22: 1827-1831.
  • 5. Oswald GA, Smith CC, Betteridge DJ, Yudkin JS. Determinants and importance of stress hyperglycaemia in nondiabetic patients with myocardial infarction. BMJ 1986; 293: 917-922.
  • 6. Fava S, Aquilina O, Azzopardi J, et al. The prognostic value of blood glucose in diabetic patients with acute myocardial infarction. Diabet Med 1996; 13: 80-83.
  • 7. Capes SE, Hunt D, Malmberg K, et al. Stress hyperglycemia and prognosis of stroke in nondiabetic and diabetic patients: a systematic overview. Stroke 2001; 32: 2426-2432.
  • 8. Weir CJ, Murray GD, Dyker AG, Lees KR. Is hyperglycaemia an independent predictor of poor outcome after acute stroke? Results of a long term follow up study. BMJ 1997; 314: 1303-1306.
  • 9. Williams LS, Rotich J, Qi R, et al. Effects of admission hyperglycemia on mortality and costs in acute ischemic stroke. Neurology 2002; 59: 67-71.
  • 10. McAlister FA, Majumdar SR, Blitz S, et al. The relation between hyperglycemia and outcomes in 2,471 patients admitted to the hospital with community-acquired pneumonia. Diabetes Care 2005; 28: 810-815.
  • 11. Greci LS, Kailasam M, Malkani S, et al. Utility of HbA1c levels for diabetes case finding in hospitalized patients with hyperglycemia. Diabetes Care 2003; 26: 1064-1068.
  • 12. Rando L, Keith C, Sardjono DA, et al. Diabetes ward management — room for improvement. J Pharm Pract Res 2004; 34: 95-99.
  • 13. Genuth S, Alberti KG, Bennett P, et al. Expert Committee on the Diagnosis and Classification of Diabetes Mellitus. Follow-up report on the diagnosis of diabetes mellitus. Diabetes Care 2003; 26: 3160-3167.
  • 14. Mosca IF, Goodall I, Hoshino T, et al. Global standardization of glycated hemoglobin measurement: the position of the IFCC Working Group. Clin Chem Lab Med 2007; 45: 1077-1080.
  • 15. Dunstan DW, Zimmet PZ, Welborn TA, et al. The rising prevalence of diabetes and impaired glucose tolerance: the Australian Diabetes, Obesity and Lifestyle Study. Diabetes Care 2002; 25: 829-834.
  • 16. Corpus RA, O’Neill WW, Dixon SR, et al. Relation of hemoglobin A1c to rate of major adverse cardiac events in nondiabetic patients undergoing percutaneous coronary intervention. Am J Cardiol 2003; 92: 1282-1286.
  • 17. Medhi M, Marshall MC Jr, Burke HB, et al. HbA1c predicts length of stay in patients admitted for coronary artery bypass surgery. Heart Dis 2001; 3: 77-79.
  • 18. Furnary AP, Wu Y, Bookin SO. Effect of hyperglycemia and continuous insulin infusions on outcomes of cardiac surgical procedures: the Portland Diabetic Project. Endocr Pract 2004; 10 Suppl 2: 21-33.
  • 19. Malmberg K, Rydén L, Efendic S, et al. Randomized trial of insulin-glucose infusion followed by subcutaneous insulin treatment in diabetic patients with acute myocardial infarction (DIGAMI study): effects on mortality at 1 year. J Am Coll Cardiol 1995; 26: 57-65.
  • 20. Van den Berghe G, Wouters P, Weekers F, et al. Intensive insulin therapy in critically ill patients. N Engl J Med 2001; 345: 1359-1367.
  • 21. Gandhi GY, Nuttall GA, Abel MD, et al. Intensive intraoperative insulin therapy versus conventional glucose management during cardiac surgery: a randomised trial. Ann Intern Med 2007; 146: 233-243.
  • 22. Mehta SR, Yusuf S, Diaz R, et al. Effect of glucose-insulin-potassium infusion on mortality in patients with acute ST-segment elevation myocardial infarction: the CREATE-ECLA randomized controlled trial. JAMA 2005; 293: 437-446.
  • 23. Collier B, Diaz J, Forbes R, et al. The impact of a normoglycemic management protocol on clinical outcomes in the trauma intensive care unit. JPEN J Parenter Enteral Nutr 2005; 29: 353-359.
  • 24. Rady MY, Johnson DJ, Patel BM, et al. Influence of individual characteristics on outcome of glycaemic control in intensive care unit patients with or without diabetes mellitus. Mayo Clin Proc 2005; 80: 1558-1567.


remove_circle_outline Delete Author
add_circle_outline Add Author

Do you have any competing interests to declare? *

I/we agree to assign copyright to the Medical Journal of Australia and agree to the Conditions of publication *
I/we agree to the Terms of use of the Medical Journal of Australia *
Email me when people comment on this article

Online responses are no longer available. Please refer to our instructions for authors page for more information.