Chronic obstructive pulmonary disease (COPD) is one of the most common respiratory diseases. It is associated with high morbidity and mortality, and a significant social and economic burden that is expected to worsen in coming decades1–3, and characterized by chronicity and exacerbations that contribute to a significant deterioration of health, affect the progression of the disease, and increase the risk of death2–4. In patients with advanced disease, the course of COPD is often complex, heterogeneous, and difficult to predict in some cases, making it difficult to establish an accurate prognosis and to determine the point at which palliative care should progressively begin to replace conventional treatment4,5.

Many different studies have focused on identifying reliable predictors of mortality, although the available evidence varies widely, having been obtained from studies with different methodologies, follow-up times, and criteria for patient inclusion (stable phase or after hospitalization)4,5. After an exacerbation, prognostic variables described so far can be grouped into patient characteristics before admission (age, comorbidities, dyspnea, functional capacity, malnutrition, and previous hospitalizations), or clinical severity at admission or during hospitalization5,6–13. However, as the predictive capacity of these variables is limited, the prognostic validity of the combination of several multidimensional indices formed by combinations of these factors has been investigated5,13. The most widely accepted are BODE, BODEx, ADO, DOSE, and CODEx, which include variables such as age, the degree of bronchial obstruction and dyspnea, nutritional status, exercise tolerance, the frequency of previous severe exacerbations, the number of significant comorbidities, or persistent smoking habit4,13–18. Although their value is limited5, the use of these indices is recommended in most COPD guidelines2,3. Recently, a consensus document from the European and American respiratory medicine societies suggested that comparative studies between the various prognostic indices were now priority area for research19.

Activities of daily living (ADL) are among the main elements considered when measuring the health and quality of life of the population20–25. The capacity of patients to live independently within the community must also be determined in order to adapt their care situation and establish priorities for intervention. In contrast to other diseases, surprisingly little attention has been paid to the predictive capacity of dependency for performing ADL in COPD, and the results of the few studies analyzing this factor have been heterogeneous6,7,12. There is as yet no evidence to show the predictive power of functional capacity determined by ADL, which include basic tasks (BADL) of self-care (eating, bathing, dressing, mobility, etc.) and instrumental tasks (IADL) in the immediate environment (shopping, preparing food, cleaning the house, etc.). Moreover, these variables have not been included in any of the prognostic indices developed to date5.

The aim of this study was to evaluate the predictive capacity of dependency for BADL and IADL for mortality after hospitalization for COPD exacerbation, and to analyze if the inclusion of these indices in a multidimensional score would improve this predictive power, in comparison with accepted indices, such as BODEx, ADO, DOSE, and CODEx.

This study provides new information by demonstrating that limitation for performing ADLs, especially the most basic activities, is a powerful predictor of mortality in the months following hospitalization for COPD exacerbation that is equivalent to age and superior to clinical variables, such as obstruction or dyspnea grade, low weight, and number of comorbidities. It is also the first study to demonstrate that the inclusion of this variable in a mortality score following a severe exacerbation can significantly improve the prognostic classification of these patients.

In COPD, existing evidence on the capacity of functional dependency to predict mortality is very limited33. Almagro et al.7 and Sanjaume et al.12 described 2 small cohorts, in which they found that limitation in BADL measured by the Katz scale, which reflects dependency for 7 sociobiological functions also included in the Barthel index, is a predictor of mortality following a severe exacerbation in bivariate analyses only, but not when adjusted for other variables included in the BODEx index or factors such as physical activity and co-morbidities. Connors et al.6, in a study published over 20 years ago, found that BADL dependency measured by this same index was an independent predictor, especially in the case of limitation for at least 2 activities. Similarly, in this study we found that the percentage of deaths in the follow-up period after an exacerbation was 9.6%, 25%, 36%, and 47% (p < 0.01), according to zero, mild, moderate, or severe dependency on the Barthel index, respectively (data not shown). This would mean that the greater the degree of dependency, the higher the likelihood of death. However, in order to simplify scoring, we chose not to stratify for the degree of BADL dependency, despite the fact that this variable has shown great predictive power, both individually and when included in a multidimensional score.

To our knowledge, no previous studies have analyzed the capacity of limitations in IADL for predicting mortality in COPD patients, an association that was revealed in our study. However, although the individual association of IADL with dependency appears similar to that detected for BADL, when it is combined in a multidimensional score its predictive capacity is less than for BADL, although it is superior to some of the previously validated indices, such as BODEx or DOSE. IADL are more dependent on sociocultural factors or gender32, and these factors may have influenced these findings. In addition, when both indices were included in the multivariate predictive models (Barthel and Lawton and Brody), the results are very similar to those described in model 1 (data not shown), which suggests that limitation for BADL is a better predictor, and no added value is gained from the evaluation of both BADL and IADL.

The other predictive variables detected in the different models (age, comorbidities, lung function, dyspnea, low BMI, previous severe exacerbations) were also identified in other studies and are included in many known indexes, including those used in this study4,5,13–18. It is striking that in our study, the score that does not include dependency for ADL (score 3) seems to be a better predictor than those already described, although this may be because it is based on the analysis of the factors that predict mortality in the sample itself.

Our study has some limitations. The first is that it was performed in a single center and with a limited number of cases. This may restrict its external validity, and our findings must be confirmed in multicenter studies with larger sample sizes. However, our study population does not appear to differ much from those found in other centers or countries, as it reflects clinical and demographic characteristics similar to those described in large national and international audits34. Moreover, the number of deaths adjusted for follow-up time and cause was similar to figures reported in other Spanish series7,12,35. In the absence of questionnaires to determine COPD-specific functional capacity validated for the aims of our study33, we used those recommended by the health authorities36,37. This also allowed us to make comparisons with other chronic diseases and to demonstrate the utility of questionnaires that do not focus solely on dyspnea as an essential component of ADL limitation, and may therefore provide a more comprehensive assessment of the general status of the patient, because their results are also related to changes in other areas, such as cognitive, sensory, or musculoskeletal factors, which can be key in the prediction of events in chronic patients5,13,21. Furthermore, collecting information during hospitalization on the stable phase situation of the patient a few weeks previously is prone to inaccuracies.

We believe, however, that our study has some strengths, including its design, the consecutive sampling for 1 year, the assessment of dependency by expert personnel, and the absence of patients lost to follow-up. Our results indicate that evaluating the patient's functional capacity on the basis of a single question instead of measuring all indices, and eliminating the need for intermediate values by dichotimising the remaining variables, make this score very easy to use. Furthermore, the score does not require other functional tests, such as those required by BODE or other recent proposals14,38 that may be unavailable or unevaluable at the time of admission of many patients, and generates evidence on the usefulness of variables of daily life that differ from commonly used clinical parameters. Stratification in tertiles is useful to differentiate very clearly from the start of follow-up 3 groups with different probabilities of death, which can be of help in decision-making at the time of recommending discharge of a patient after a severe exacerbation.

In conclusion, ADL limitation, especially for the most basic activities, is an independent predictor of mortality following a severe COPD exacerbation, and its inclusion in multidimensional scores improves their predictive ability. These results must be validated in multicenter studies and other populations of COPD patients.

The authors state that they have no conflict of interests.