Factors affecting survival of hospitalised patients with COPD

Chronic obstructive pulmonary disease (COPD) is a leading cause of death and disability worldwide. According to World Bank data, it is expected to move from its status in 2000 as the 4th and 12th most frequent cause of mortality and morbidity, respectively, to be the 3rd and 5th leading cause of mortality and morbidity, respectively, in 2020 1, 2. Moreover, about 10% of all hospitalisations are directly or indirectly attributable to COPD 3.

Despite being the only major disease showing increasing trends, factors that determine the short- and long-term outcomes of patients with COPD are not yet precisely understood. Identification of the factors that may influence survival in patients with COPD may enable clinicians to better assess life expectancy. This is extremely important, in that it may help offset the social and economic burden of COPD through the implementation of more individualised and effective treatment strategies, as well as better mobilising healthcare resources.

Investigations into the factors predicting outcome in COPD patients have included follow-up studies on stable COPD patients and on COPD patients admitted to the intensive care unit (ICU), as well as on patients admitted to hospital with hypercapneic acute attack. Among the parameters thought to be related to mortality are forced expiratory volume in one second (FEV₁) 4, carbon dioxide arterial tension (P_a,CO2), arterial oxygenation 5, cardiac status 6, body mass index (BMI) 7, serum albumin level, functional status 8 and the presence of other comorbid states 9. In a study that directly addressed the long-term outcome for COPD patients admitted to hospital, the in-hospital and 2-yr after-discharge mortality rates were found to be 11% and 49%, respectively 8. Connors et al. 8 even developed a multivariate model to predict the probability of survival of patients hospitalised with an acute exacerbation. There is only one prospective study that assessed the in-hospital and after-discharge outcome for all COPD patients admitted with acute exacerbation 9. Although the follow-up period was relatively short, it was found that 8% of patients died in hospital and 23% died within 1 yr after hospitalisation 9.

In the present study, the authors sought to prospectively identify factors that may influence in-hospital deaths and deaths after discharge in a large patient cohort. The current study differs from previous studies in that all COPD patients hospitalised for acute exacerbation were included, and the follow-up period was 3 yrs.

METHODS

RESULTS

The mean age of the patients was 65 yrs, and 60% were aged ≥65 yrs. Age did not seem to contribute significantly to in-hospital deaths or long-term survival. Only 25 of the 205 patients were female, in whom the main causes of exacerbation were: infections (n = 17; 68%), noncompliance with treatment (n = 4; 16%) and different combinations of the primary causes including air pollution (n = 3; 12%). These rates were similar to those observed in male patients, in whom the main causes of exacerbation were: infections (n = 115; 64%), noncompliance with treatment (n = 25; 14%) and the different combinations (n = 26; 14%). Due to the limited number of female patients in the study group and the standard treatment received after discharge by almost all patients, the analysis did not include detailing by sex and treatment after discharge.

Out of the 205 patients in the study group, 95 were current smokers (46%), 83 were ex-smokers (41%) and 27 had never smoked (13%). Smoking status and smoking load did not correlate with admission to the ICU, in-hospital mortality or long-term survival. Patients who had been receiving long-term oxygen treatment (n = 57) on admission did not differ from the rest of the population with regard to in-hospital mortality (n = 7) or admission to the ICU (n = 15; p>0.05; tables 1⇓ and 2⇓). After discharge, 63 patients received long-term oxygen treatment. Their mortality rates at 6 months, 1 yr, 2 yrs and 3 yrs were 29%, 37%, 44% and 56%, respectively. No significant difference in survival was observed between this group and the rest of the study population.

Medical insurance was available to 96% of the patients, and 93% were either married or living with relatives. According to Turkish standards, 84% of these patients belonged to the lower income group (<US$10,000 yearly income). Due to these very high rates, analysis based on these characteristics was not performed.

The general features of the 205 patients are shown in table 1⇑. Although arterial blood gas analysis could be performed in all patients, spirograms could not be obtained in 22 patients; 11 due to severity of the exacerbation, five due to technical problems and six due to lack of patient cooperation. Only the omission of patients on whom a spirometric examination could not be performed due to severe exacerbation may have created a selection bias, with the potential to change the results. However, since this number is small, the probability of their having a significant impact on the results should be low. Although the best effort was made to obtain stable period spirograms for all patients, an appropriate spirometric measurement could not be obtained for 36 patients. For the rest, the ERS/ATS stage of COPD was moderate in 34% of the patients (FEV₁ 50–80% predicted), severe in 44% of the patients (FEV₁ 30–50% pred) and very severe in 22% of the patients (FEV₁<30% pred). None of these patients had mild COPD (FEV₁≥80% pred).

In total, 17 patients (8.3%) died in the hospital, whereas 188 were discharged. Comparison of these two groups revealed that longer hospital stay (p = 0.042), lower P_a,O2 (p = 0.019), lower arterial oxygen saturation (S_a,O2; p = 0.01) and higher P_a,CO2 (p = 0.019) were significantly related to in-hospital death (table 1⇑). The curve analysis showed that P_a,CO2 at a cut-off of 6.916 kPa appeared to be the only factor with a relatively high sensitivity (0.71) and specificity (0.69) for predicting in-hospital mortality. The other parameters did not give any clinically revealing values for sensitivity or specificity. Whereas the rate of in-hospital death among the 45 patients (22%) admitted to the ICU was 27%, it was 3% for the non-ICU patients (five out of 160; p = 0.000). Clinical features of the ICU and non-ICU patients are shown in table 2⇑. Statistically significant differences between the ICU and non-ICU patients were FVC (p = 0.047), P_a,O2 (p = 0.001), P_a,CO2 (p = 0.025), and S_a,O2 (p = 0.000).

Long-term follow-up of all hospitalised patients revealed 6 month and 1-, 2- and 3-yr overall mortality rates of 24%, 33%, 39% and 49%, respectively (fig. 1⇓). Median survival for the patients who died during the 3-yr period was 154 days. Except for one patient, 6-month follow-up was completed for all discharged patients. In total, five, 9 and 15 patients were lost to follow-up at 1, 2 and 3 yrs, respectively. Table 3⇓ shows the results of the Cox proportional hazard analysis. This model revealed that mortality was related to longer duration of disease (relative risk (RR) 1.158; 95% CI 1.059–1.268; p = 0.001), lower albumin level (RR 0.411; 95% CI 0.205–0.824; p = 0.012), lower P_a,O2 (RR 0.871; 95% CI 0.784–0.969; p = 0.011), and lower BMI (RR 0.830; 95% CI 0.703–0.979; p = 0.027).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1—

Survival curve for all patients after hospitalistion.

Of the 205 patients, 156 had been hospitalised previously for acute exacerbation of COPD, with the first hospitalisation occurring ∼5 yrs after diagnosis of COPD. The correlation between duration of disease and time elapsed since first hospitalisation was 0.888 (p = 0.000). Since both parameters conveyed essentially the same information, they were not analysed simultaneously in the same model. Thus, this approach avoided the paradox of multicollinearity in the current regression model. As expected, substitution of duration of disease by time elapsed since first hospitalisation did not significantly change the statistical values of the other parameters in the model. Mortality related factors were found to be longer time since first hospitalisation (RR 1.195; 95% CI 1.061–1.346; p = 0.003), lower albumin level (RR 0.430; 95% CI 0.207–0.892; p = 0.023), lower P_a,O2 (RR 0.866; 95% CI 0.780–0.962; p = 0.007), and lower BMI (RR 0.795; 95% CI 0.666–0.950; p = 0.011) in this second run. Longer duration of disease and longer time elapsed since first hospitalisation were observed to be new determinants of mortality. To analyse the effects of age at disease onset on survival, the first model was run after excluding age from the model. While the age at disease onset did not appear to be a statistically significant parameter (p = 0.574), duration of disease was still highly significant (p = 0.011). Table 4⇓ describes the time related mortality rates for each of these statistically significant parameters after stratification according to clinically important classical thresholds. Among these parameters, poorest prognosis was encountered in patients with albumin level <2.5 g·dL⁻¹ at the time of admission.

DISCUSSION

The present study contributes some new data related to the survival of COPD patients hospitalised for an acute exacerbation. It also lends remarkable support to findings from studies assessing in-hospital and after-discharge survival of COPD patients. The current authors found that the in-hospital mortality rate for the patients was 8.3% and that the overall mortality rate at 3 yrs, one of the longest follow-up periods in the literature, was 49%. Longer duration of disease, longer time elapsed since first hospitalisation, lower P_a,O2, lower albumin level and lower BMI were the main factors related to long-term mortality after index hospitalisation with acute exacerbation of COPD.

The rate of in-hospital mortality for COPD patients hospitalised with acute exacerbation has been reported to be between 2.5% and 30%, depending on the methodology of data collection and the patient population 5, 16. Nonrespiratory organ system dysfunction and hospital days prior to ICU admission have been reported to be the most important predictors of in-hospital mortality, with the total in-hospital mortality rate being 24% 5. Others have reported an in-hospital mortality rate of 14.4%, with cardiac dysfunction being an important predictor of mortality 6. Recently, a cross-sectional nationwide study based on a national database reported a relatively low in-hospital mortality of 2.5% and identified older age, male sex, higher income, nonroutine admissions, and more comorbid conditions as independent risk factors 16. The present study found, however, that in-hospital mortality of COPD patients hospitalised for acute exacerbation was mainly influenced by lower P_a,O2 and S_a,O2, higher P_a,CO2, and longer length of hospital stay. Of these parameters, only P_a,CO2, at a cut-off of 6.916 kPa, demonstrated a relatively high sensitivity (0.71) and specificity (0.69). Admission to the ICU, which is virtually the end result of multiple clinical parameters, was associated with >70% of the in-hospital deaths in this study. When the ICU and non-ICU patients were categorised, the factors associated with ICU admission were found to be almost identical to the factors related to in-hospital mortality. Indeed, studies designed to identify the factors that direct a clinician to decide whether to admit a patient to the ICU may be another important step in determining the factors related to short-term survival.

A study on long-term survival of seriously ill, hospitalised COPD patients with P_a,CO2 ≥6.65 kPa showed that independent predictors for survival were severity of illness, age, prior functional status, BMI, P_a,O2/inspiratory oxygen fraction, congestive heart failure, serum albumin level and presence of cor pulmonale 8. In-hospital and 2-yr mortality rates after all-cause hospitalisations were 11% and 49%, respectively. The patient selection and flexible margins for cause of hospitalisation, however, make it difficult to generalise these results to all COPD patients. In another study, long-term mortality in patients admitted to the ICU with acute exacerbations was shown to be related to severity of respiratory organ dysfunction and the presence of nonrespiratory organ dysfunction 5. While they reported a 1-yr mortality rate of 48%, a major limitation of that study was the inability to obtain follow-up data from 54% of the patients 5. In contrast, the number of patients missing on follow-up in the current study was very small. A multi-centre study from Spain which analysed the association between re-admission of COPD patients and multiple modifiable factors reported a 1-yr after-discharge mortality rate of 29%, but there was no analysis of survival or mortality related to hospitalisation period 17.

As described above, only one study in the literature is comparable to the present study in directly addressing the long-term survival of all COPD patients hospitalised with acute exacerbations 9. This study reported that older age, longer stay in the hospital, comorbidity index, low FEV₁, higher P_a,CO2 and use of maintenance oral corticosteroids were factors related to mortality 9. The current findings regarding long-term mortality are somewhat different. This difference may be due to the different methodologies employed and the use of a longer follow-up period. As compared, the present model did not include analyses for sex and maintenance oral corticosteroid use. Sex was excluded because only 12% of the patients were females, due to the low smoking rate among older females in Turkey. Although the limited number of female patients in this study can be considered a problem for the generalisation of the data, varying degrees of male predominance have been accepted, for socio-economic reasons, worldwide.

As the Turgut Ozal Research Centre is the largest medical facility in the region, as well as being the only training and research centre, the current authors either directly treat the great majority of the COPD patients or consult in their diagnosis and treatment. For any patient classified as having COPD, the long-term treatment policy is not to prescribe maintenance oral corticosteroids. These patients are only allowed to use systemic corticosteroids during the treatment of acute exacerbations. Since this became a common practice in the authors region, no patients utilising maintenance systemic corticosteroids were seen during the study period. Thus, this parameter was excluded from consideration.

It was found that the statistically significant parameters related to mortality were lower P_a,O2, lower albumin level, lower BMI and longer duration of disease. Low P_a,O2 is direct evidence for limited pulmonary reserve and increased ventilation/perfusion mismatch, thus, emphasising the severity of the underlying disease. The critical level appeared to be 5.985 kPa in the present study (table 4⇑). These patients became less tolerant to alterations in their clinical condition, thus, showing a poorer prognosis.

Compatible with the current results, low BMI 7 and low serum albumin level 8 have been shown to be strong predictors of poor long-term survival. The suggested mechanisms for their roles in increased long-term mortality are respiratory muscle weakness, impaired gas exchange and impaired immune response. BMI is not generally thought to be influenced by acute events; rather, it correlates better with chronic health status. In agreement with this, it was found that BMI did not appear to be an important parameter for in-hospital mortality. For long-term mortality, the critical level appeared to be 20 kg·m⁻² (table 4⇑), below which mortality increased. In agreement with previous findings 18, the survival of overweight and obese patients was found to be increased, suggesting that increased BMI has a protective effect. When compared with patients having a BMI <20 kg·m⁻², those who were overweight or obese had a 33% decreased probability of death throughout the 3-yr period.

Serum albumin level has been considered to be part of the acute phase protein response. Low levels of this protein may, therefore, reflect a deterioration of clinical status or increased persistent inflammation during acute exacerbations of COPD. However, low serum albumin level is also a good indicator for long-term health status in chronically ill patients. Due to the low number of in-hospital deaths in the current study, present analysis could only show a relationship between low serum albumin level and higher long-term mortality. The clinically important threshold for mortality was 2.5 g·dL⁻¹ albumin, below which the mortality rate during follow-up increased sharply (table 4⇑). Approximately one-third and one-fifth of patients could survive beyond 1 and 3 yrs, respectively. When combined with disability and dependence on others, albumin values <2.5 g·dL⁻¹ may help clinicians discuss more conservative treatment options with patients and their relatives.

To the authors' knowledge, the present study is the first to show that longer duration of disease and time elapsed since first hospitalisation are predictors of mortality. These two parameters, however, are somewhat subjective, if directly asked to the patient. The date of first visit to a doctor for chronic cough, dyspnoea and/or sputum production to determine disease duration was used. This information, as well as the date of first hospitalisation, are usually recorded in personal and official medical documents and are relatively easy for patients to remember. Since all patients in the current study were handled individually and without time limitation by a team of investigators who were aware of the risks, the authors believe that the multiple questionings minimised these risks and lowered the chance of obtaining misleading information.

Table 4⇑ shows the relationship between disease duration/time elapsed since first hospitalisation and mortality at different cut-off levels. Patients with disease duration ≥20 yrs and >5 yrs since first hospitalisation had higher mortality rates. Since COPD is a progressive disease and the general status of patients unavoidably deteriorates over time, the impact of these two parameters on survival appears indisputable. The correlation between disease duration and time elapsed since the first hospitalisation was so high that these two parameters can be used interchangeably in future studies. To determine the real influence of these new parameters on the model, it was run without them. When the authors did this, previously significant parameters remained significant, while FEV₁/FVC (p = 0.023), P_a,CO2 (p = 0.007) and Systolic P_pa (p = 0.028) also appeared to be statistically significant. These latter parameters have been frequently linked to mortality. Thus, this analysis suggests that these parameters may be cofactors dependent on disease duration and time elapsed since first hospitalisation. Since the latter two parameters have never been included in previous survival models, this relationship may not have been revealed. In addition, the current analysis of age at disease onset using Cox regression analysis also failed to show any relationship to mortality. Inclusion of age at disease onset with duration of disease also did not significantly alter the statistical values of the other parameters. Thus, the present results suggest that neither age nor age at disease onset independently influences mortality. Rather, duration of disease is the only one of these three parameters that has a strong and independent relationship to mortality. However, these findings must be validated in future studies.

The National Health and Nutrition Examination Survey III study 19 showed that only 10–15% of all COPD patients are hospitalised, with the majority likely to be those with severe disease. Hence, the expected conclusion should be that the results of studies on hospitalised COPD patients may be specific to this group, and cannot be generalised for nonhospitalised COPD patients. Interestingly, the related data for this major group is extremely limited. Nevertheless, Soriano et al. 20 recently published a study directly addressing the survival of newly physician-diagnosed COPD patients identified in primary care. This large retrospective cohort study which was based on evaluation of national database revealed that the overall 3-yr mortality rate for these patients was around 33%. This high rate is quite surprising because the patients were nonhospitalised and virtually represented the general COPD outpatients quite well. Confirming the findings of Soriano et al. 20, almost identical mortality rates for outpatients with COPD were reported very recently by Celli et al. 21. They also designed an index for predicting the risk of death in outpatients with COPD. This index, called the BODE index, consists of four parameters: BMI, degree of airflow obstruction, level of dyspnoea and exercise capacity. It was found to be better than FEV₁ in predicting the risk of death 21. In designing the regression model, the authors also included both BMI and FEV₁, and found that BMI had a statistically significant influence on patient mortality. As in the current study, BMI has been widely recognised as having an influence on long-term mortality in both hospitalised and outpatients with COPD. In the present authors' opinion, these high morality rates strongly emphasise the poor long-term survival of COPD patients in general.

One limitation is that only overall death rates were reported, not the specific cause of death. Unfortunately, official records in some countries do not enable the investigators to identify the specific cause of death at all. To date, a few studies have been able to provide this specific data for patients with COPD, showing that >60% of outpatients and >80% of hospitalised patients die from COPD during follow-up 17, 20. Thus, the absence of exact cause of death is not considered to be a serious limitation in interpreting the data.

In conclusion, it has been shown that half of all chronic obstructive pulmonary disease patients hospitalised with acute exacerbations died within 3 yrs. These findings may help clinicians with important information about the probable short- and long-term survival of these patients. Thus, low levels of serum albumin strongly predict poorer long-term outcome, and longer duration of disease and longer time elapsed since first hospitalisation appear as new mortality-related independent factors. The authors believe that the current findings will provide clinicians with new insights, allowing them to implement more individualised treatment strategies by better predicting the life expectancy of chronic obstructive pulmonary disease patients.