Effect of comorbid diabetes on length of stay and risk of death in patients admitted with acute exacerbations of COPD

INTRODUCTION

Comorbid diabetes mellitus (DM) and hyperglycaemia without diagnosed DM are both associated with poor outcomes, including increased length of stay and increased mortality, among patients hospitalized^1,2 with acute myocardial infarction,³ other acute coronary syndromes, heart failure, stroke,⁴ pneumonia,⁵ trauma and surgery.^3,6–10 There is limited published evidence about the effect of comorbid DM on clinical outcomes in patients hospitalized with an acute exacerbation of COPD (AECOPD).^2,11,12

However, other evidence points to the possibility that DM and AECOPD may have additive adverse effects on outcome. For example, DM and poor glycaemic control are associated with altered lung volumes and reduced flow rates^13–16 as well as a reduction in gas transfer due to microangiopathy.¹⁷ Treatment with insulin improves the defect in gas transfer.¹⁸ Furthermore, hyperglycaemia is a risk factor for adverse outcomes of lung infections, which are a key feature of AECOPD. DM is associated with an increased risk for the development of pulmonary infection (both community-acquired and nosocomial) and for increased mortality¹⁹ and length of stay, even after adjusting for other prognostic factors.²⁰ Finally, treatment for AECOPD may worsen the course of DM. Systemic glucocorticoid steroids, which are commonly administered to patients admitted with AECOPD, increase the risk of hyperglycaemia.^21,22 These observations suggest that the presence of DM may have an adverse impact on clinical outcomes in patients admitted with AECOPD.

The objective of this study was to assess the impact of comorbid DM on length of hospital stay and risk ofdeath, as indicators of the outcome of hospital care, among patients admitted to hospital with AECOPD.

METHODS

This is a cross-sectional analytical study based on the analysis of records of patients admitted to one large tertiary hospital during the calendar year 2007 with a diagnosis of AECOPD. Data were collected from the case notes by a single reviewer (A.P.) and entered into an Excel spreadsheet.

The study was approved by the Sydney South West Area Health Service (Western Zone) Human Research Ethics Committee.

The target population comprised all patients aged 45 years and over who were admitted with an AECOPD under the care of the respiratory medicine team at Liverpool Hospital between 1 January 2007 and 31 December 2007. These patients were identified by retrieving and reviewing available medical records for all admitted patients for whom the attending physician was a respiratory physician, who were within the required age range, whose admission date was in the designated study period and who had been allocated an International Classification of Diseases (ICD) code of J41–J44. In some cases review of the case notes indicated that the main reason for admission was another diagnosis, commonly pneumonia. These cases were excluded from further consideration.

The diagnosis of diabetes was based on history reported by the patient as documented in the medical record. Hyperglycaemia was defined as any random blood glucose ≥10 mmol/L during the admission. In most cases (200/246, 81%) this was a bedside capillary blood glucose measurement, but in the remainder it was a laboratory plasma measurement. This threshold was based on the recommended upper limit random blood glucose levels for non-critically ill inpatients.²³ Length of stay was ascertained from the record of admission and separation date. All admissions for each patient within the study year were included in the analysis.

The distribution of length of stay was skewed and hence was log-transformed for analysis. Analysis was by generalized estimating equations to adjust for correlation between multiple admissions in some individuals. Log-transformed length of stay and death were the dependent variables. A linear link and normal error structure was used for the length of stay and a logit link and binomial error structure was used for death. The following potential confounders were included as covariates: use of home oxygen, initial blood gas pH, requirement of invasive or non-invasive ventilation support, and presence of pneumonia, cancer, dementia or disabling arthritis as other comorbidities. However, for the analysis of death as the dependent variable some of these covariates had to be removed to enable the model to converge. Data on use of oral corticosteroids were also collected. However, as all patients with AECOPD were treated with oral corticosteroids, we did not adjust for this in the analysis.

RESULTS

During the study period there were 432 separate episodes for patients with a final diagnosis of AECOPD, of which 404 had charts available for review (Fig. 1). We excluded 158 episodes where the chart revealed that the principal reason for admission was not an AECOPD. The majority of these patients had pneumonia. Finally, we collected data for 246 (57% of the original sample) of these episodes. These individual episodes involved 172 individual patients who had between one and six admissions during this year. Of patients 56% were male (Table 1) with a mean age at the time of their first admission of 71.3 years. Fifty-three episodes (22%) occurred in patients with diagnosed DM (Table 2) and in 47 of these cases (89%) the patient was on therapy for DM at the time of presentation. Hyperglycaemia was identified in 42% of admissions including 36 admissions in which DM was not recorded as a diagnosis. Cancer, disabling arthritis and dementia were relatively less commonly recorded as comorbidities (Table 1).

Most patients were ex-smokers (Table 1) and 21% were using long-term home oxygen therapy. A substantial number had severe episodes manifest as acute respiratory acidosis, need for non-invasive ventilation or referral to the intensive care unit at the time of admission (Table 2).

The geometric mean length of hospital stay among the patients with diabetes was 7.8 days and the length of stay among patients without diagnosed diabetes was 6.5 days. After adjusting for confounders the length of stay was 10.3% (95% CI: −12% to +38%) longer in patients with diabetes (Table 3). There were few deaths in the cohort. However, the risk of death in people with diabetes tended to be greater than the risk of death in people without diabetes (odds ratio 1.93, 95% CI: 0.43–8.64, after adjustment for confounders) (Table 4). The effects of hyperglycaemia observed during the hospital admission were of a similar magnitude.

DISCUSSION

This study shows a trend towards increased length of stay and increased deaths among patients admitted with AECOPD who have comorbid DM as compared with those not diagnosed DM. After adjustment for potentially confounding factors, such as disease severity, age, gender and other comorbidities, the differences were not statistically significant. Adjustment for confounders tended to reduce the magnitude of the effect on length of stay.

The strength of this study is the inclusion of data on other comorbidities and markers of disease severity. These were not measured in a previous study on this subject¹¹ and, as shown here, may have confounded the apparent associations observed in that study. The main limitation of this study is the relatively small sample size, which has resulted in CI around the estimates effect that are relatively broad, indicating that the study may be underpowered when considered alone. Also, as this is a retrospective study we were not able to define the covariates in all the subjects. In particular, only 153 of 246 patients had spirometry recorded during the admission, and therefore we could not define disease severity using this objective measure. The blood glucose measures used in the study were random (non-fasting) samples and 26% of episodes did not have a record of blood glucose measurement.

The previous study on this subject also found a non-significant trend for increasing length of stay with higher quartiles of blood glucose measured during the hospital admission.¹¹ The presentation of data from that study is different from the current one and it is not possible to formally combine the data. Both studies are probably underpowered to detect a clinically relevant effect on length of stay. Nevertheless, taken together, these two independent investigations do lend weight to the view that comorbid DM or hyperglycaemia increase length of stay in patients admitted with AECOPD.

A number of factors have previously been associated with worse clinical outcomes for patients admitted with AECOPD: lower arterial pH²⁴, older age,^24,25 male gender, ICU admission, need for non-invasive ventilation, requirement for home oxygen and the presence of comorbidities. The important predictors of increased length of stay are: age 65 years or more, poor performance status, lowest FEV₁, intervention with assisted ventilation, oxygen saturation at admission of <86% and admission PEF of <150 L/min and increased blood glucose level.²⁶ Although we have not directly measured performance status, we have adjusted for the comorbidity due to arthritis and dementia, both of which are important contributors to poor performance status in older people.^26–28 Most of these factors are not amenable to intervention during a hospital stay. However, improved glycaemic control is achievable with appropriate intervention and our data imply that this may alter the course of the disease. However, this is yet to be established with certainty.

One possible explanation for prolonged length of stay and increased risk of death in patients with DM is that hyperglycaemia impairs response to treatment in patients with AECOPD. Hyperglycaemia causes impairment in immune response and depresses neutrophil chemotaxis, phagocytosis, intracellular bacterial activity, opsonization and cell-mediated immunity.²⁹ These abnormalities tend to develop when glucose levels exceed 11 mmol/L and do improve with glycaemic control. Hyperglycaemia is also associated with increased production of superoxide, which is associated with damage to lipid, protein and DNA. Its reaction with nitric oxide can lead to nitration of proteins, which can be reversed if euglycaemia is re-established¹² Hence, it is plausible that patients with infective exacerbations of COPD respond less well to antibiotic treatment than other patients.

Another plausible explanation for increased length of stay in patients with comorbid DM is that these patients require longer inpatient care in order to attain euglycaemia prior to discharge. As noted above, most patients with AECOPD are treated with systemic glucocorticoid steroids, which exacerbate hyperglycaemia in patients with DM. Adjustment of treatment with insulin and/or oral hypoglycaemic agents in response to this iatrogenic hyperglycaemia and then readjustment when systemic steroids are withdrawn can take several days and may, itself, be a reason for prolonged length of stay.

Our study supports the view that DM may be an adverse prognostic factor for patients admitted with an AECOPD. Further studies are now required to elucidate the reasons for this prolonged length of stay in particular whether premorbid glycaemic control or inpatient control is responsible, and if so whether more intensive diabetes management, either in the community or in association with their hospital admission, can influence the clinical course of patients with AECOPD and DM.