Associations between chronic comorbidity and exacerbation risk in primary care patients with COPD

The study used routine data from a general practice database from the Department of Primary and Community Care at the Radboud University Medical Center, Nijmegen, the Netherlands. De-identified electronic medical records from primary care patients diagnosed with COPD from 179 general practices in the eastern part of the Netherlands were available in the database.

For each registered subject, the following data were extracted: age, sex, all diagnoses using the International Classification of Primary Care (ICPC), extended with Dutch ICPC sub-codes, [12] and all prescribed medication. ICPC-2 or ICD10 coding data were recoded into ICPC-1. Medication prescriptions (i.e., prescription start and end dates, dosage, frequency, and duration) were extracted and categorized using the Anatomical Therapeutic Chemical (ATC) classification system [13]. For the current study only the data on prescriptions for oral corticosteroids were used.

Subjects aged ≥40 years were included in the study population when they had physician-diagnosed COPD (as labeled with ICPC code R95 in the electronic medical record) before or during the study period. Asthma (ICPC R96) in addition to the COPD code was not an exclusion criterion. The follow-up period covered the years 2012 and 2013. The observation period for patients terminated either at the end of the study period (31 December 2013), or when a subject died or deregistered from the practice.

The selection of chronic comorbid diseases studied was based on existing literature [1, 14], the authors’ clinical expertise and expert opinions (Nielen MM, Spronk I, Davids R, Korevaar JC, Poos MJ, Hoeymans N, Opstelten W, van der Sande MAB, Biermans MCJ, Schellevis FG, RA V: A new method for estimating morbidity rates based on routine electronic medical records in primary care, submitted). We considered all chronic diseases as comorbidities, regardless whether the disease had been diagnosed before the COPD diagnosis or thereafter. Apart from all ‘obligatory’ chronic diseases we also included several recurrent diseases (i.e., depression, anxiety, anemia, dyspepsia, urinary tract infection) which could potentially influence COPD outcomes. After reaching consensus about these recurrent comorbidities within the research team, ICPC (sub)codes were linked (see Appendix 1). Selection of the recurrent comorbidities in our population was based on the patient’s history in terms of these particular ICPC codes. To define whether a history of ICPC codes was relevant or irrelevant for the aim of the study, we added specific selection criteria based on published clinical guidelines for the respective diseases (see Appendix 1).

Finally, a total of 82 chronic comorbid conditions were selected and included in the analyses. The comorbidities were clustered and analyzed based on their ICPC codes into the following 14 categories: respiratory; cardiovascular; digestive; endocrine; metabolic/nutrition; musculoskeletal; neurologic; psychiatric; urogenital; blood (−forming organs)/lymphatics; infectious; eye/ear/skin; non-pulmonary cancer; and pulmonary cancer. Low prevalence categories were merged (see Appendix 2). To restrict ourselves, we focused on conditions with a high prevalence and cardiopulmonary comorbidities (other than COPD) with a lower prevalence (7 conditions, see Table 2). High-prevalent comorbidities (19 conditions), further referred to as ‘frequent comorbidities’, were defined as being present in ≥5% of the study population. This resulted in a total of 26 comorbidities remaining for further analyses.

The outcomes for the study were (i) prevalence of comorbidities in the study population, (ii) annual rate of exacerbations (dichotomized as <2 versus ≥2 exacerbations/year based on the cumulated 2012/13 data), and (iii) time (in days) until first exacerbation. An exacerbation was defined as a prescription of oral corticosteroids (i.e., prednisolone (ATC H02AB06) or prednisone (ATC H02AB07)) with a minimum daily dose of 20 mg for a minimum duration of 5 days and a maximum duration of 15 days (based on Dutch GP guidelines for treatment of COPD exacerbations [15]). As there is no consensus in the literature regarding a cut-off to differentiate between relapse of an earlier exacerbation and a new exacerbation, [16] we considered a subsequent predniso(lo)ne prescription after an oral corticosteroid-free interval of ≥14 days since the end-date of the previous prescription as a new exacerbation.

We explored associations between comorbidities and exacerbation risk using univariable analyses. Hazard ratios for comorbidities were calculated using Cox regression, in which the time variable consisted of time to the first exacerbation. Data from patients who died or were otherwise lost to follow up were right-censored. Subsequently, all frequent and cardiopulmonary comorbidities (Table 2), age, and gender were included as covariates in multivariate Cox regression analyses. The model was reduced through backward exclusion to produce a final model that consisted of only non-collinear, independently associated, statistically significant covariates. The same modeling approach was used for comorbidity categories using all other categories, with age and gender as covariates.

Overall, data of 16,427 subjects diagnosed with COPD were available for analyses. Of these patients, 1824 (11 · 1%) were lost to follow-up during the 2-year study period. Reason for loss to follow-up was known for 800 (44 · 5%) of these patients, with death being the predominant reason. Table 1 shows baseline characteristics of the patients with complete follow-up (i.e., the final study population, n = 14,603). Mean (SD) age was 66 · 5 (11 · 5) years and 53% were males. At baseline, 89 · 1% of patients suffered from ≥1 chronic comorbid conditions, while 23 · 1% had ≥5 comorbidities. Most prevalent comorbid conditions were hypertension (35 · 2%), coronary heart disease (19 · 2%), osteoarthritis (17 · 6%), diabetes (17 · 3%), and peripheral vascular disease (14 · 3%). Table 2 shows the prevalence rates of the frequent and cardiopulmonary comorbidities. Table 3 shows the prevalence of ICPC-categorized comorbidities.

During the 2-year study period the mean number of exacerbations per patient was 0.72 (SD 1 · 5). 68% of patients had no exacerbation and 5 · 7% had ≥4 exacerbations during the study period.

Tables 2 and 3 show the univariable associations between comorbidities and comorbidity categories and the exacerbation frequency subgroups, respectively. Overall, patients with one or more comorbid conditions more often had ≥2 exacerbations/year compared to patients without any comorbidity (5 · 9% vs 4 · 0%, p = 0 · 001). Patients with any other chronic respiratory disease next to their COPD, (n = 2,294, 15 · 7%) more often had ≥2 exacerbations per year compared to patients without respiratory comorbidity (8 · 2% vs 5 · 7%, p < 0 · 001).

Univariable logistic regression analysis showed that COPD patients with pulmonary cancer had 1.81 higher odds for ≥2 exacerbations per year compared to patients without pulmonary cancer (Fig. 1, p = 0.002). Patients who, next to their COPD, also suffered from asthma, blindness or low vision, coronary heart disease, depression, dyspepsia, heart failure, osteoporosis or osteopenia, peripheral vascular disease, or prostate disorders, had a higher risk of having frequent exacerbations compared to those who did not suffer from these comorbid conditions (Fig. 1).

Table 4 lists the comorbidities and comorbidity categories significantly associated with having ≥2 exacerbation per year. In the multivariable logistic regression analysis, among the statistically significant associations, the highest ORs for having ≥2 exacerbations per year were observed for pulmonary cancer (OR 1 · 85; 95% CI 1 · 28–2 · 67), heart failure (OR 1 · 72; 1 · 38–2 · 14), prostate disorders (OR 1 · 50; 1 · 13–1 · 98) and blindness/low vision (OR 1 · 46; 1 · 21–1 · 75) as comorbid conditions (Table 4). Dislipidemia was not statistically significant, but did show a trend, with an OR of 0 · 81 (95% CI 0 · 65–1 · 01, p = 0 · 071). When looking at comorbidity categories, patients with other chronic respiratory conditions (OR 1 · 37; 1 · 15–1 · 64) and psychiatric comorbidities (OR 1 · 35; 1 · 13–1 · 60) were at highest risk for frequent exacerbations.

Table 5 summarizes the results from the Cox regression analyses. Among the statistically significant associations, the comorbid conditions with the highest risk of developing a first exacerbation were recurrent sinusitis (Cox hazard ratio 1 · 53; 95% CI, 1 · 05–2 · 24), bronchiectasis/chronic bronchitis (HR = 1.50; 1.31–1.73) and heart failure (1 · 41; 1 · 29–1 · 55). For dislipidemia a non-statistically HR of 0 · 92 was observed (p = 0 · 067, 95% CI 0 · 85–1 · 00).

Having another chronic respiratory disease beside COPD was also associated with risk of developing a first exacerbation (Cox hazard ratio 1 · 26; 1 · 17–1 · 36), see Fig. 2.

Previous research has shown that cardiovascular, psychiatric, and metabolic comorbidity are highly prevalent in COPD patients, [8, 17] and our results confirm these findings. In addition to the finding by Rutten et al. [18] that unrecognized heart failure is rather common in elderly patients with stable COPD, our data also indicate that heart failure may increase the risk of having frequent exacerbations. Recent clinical trial data have shown correlations between several comorbidities and mortality risk if a COPD patient is admitted to hospital with an acute exacerbation [19, 20]. Our observations support the association between chronic comorbidity and exacerbation risk in a primary care study population, i.e., the COPD population without selection of any kind, which is unprecedented and impossible to derive from clinical trial populations [21].

We observed a trend towards statistical significance that COPD patients with dislipidemia had less frequent exacerbations compared to patients without dislipidemia (HR 0.92; p = 0.067). This observation seems to be in line with findings by Ingebrigtsen et al., who recently reported that statin use for treatment of dislipidemia was associated with reduced odds of exacerbations in individuals with COPD [22] and findings by Chan et al. that hyperlipidemia in COPD was associated with decreased incidence of pneumonia and mortality in retrospective analyses of health insurance data [23]. Intuitively, the observed lower risk of frequent exacerbations in COPD patients with comorbid diabetes might be sought in GPs’ reluctance to prescribe oral corticosteroids in these patients because the impact this may have on glucose levels, but a survey among Dutch GPs showed that most of them do not adjust treatment of exacerbations to the presence of diabetic comorbidity [24]. Gastroesophageal reflux disease (OR = 1.25 (95% CI 1.03–1.50) in our analyses) was recognized as a significant predictor of acute exacerbations of COPD in a recent review by Lee et al [25]. A relationship between prostate disorders and exacerbations has not been described in the literature, but might be related to use of inhaled anticholinergics.

A strength of this study is the inclusion of >14 thousand COPD patients from a real-life, unbiased primary care setting. However, the main strength is not so much the uniqueness or even the size of our dataset. Other existing general practice databases essentially contain the same, or even more detailed data regarding diagnoses and medication prescriptions, [26‐29] but the meticulousness with which we have looked at ALL chronic comorbidity, including recurrent episodes of conditions that are not necessarily chronic in all patients, seems unprecedented. Moreover, other existing databases with real-life general practice COPD data mainly stem from the UK and Denmark, and now there is also one available from the Netherlands. We intentionally applied minimal exclusion criteria in order to maximize generalizability of the results. Another strength is the wide range of chronic comorbidities investigated, summing up to a total of 82 conditions. Apart from all commonly known chronic comorbid diseases, we also included several recurrent diseases (i.e., depression, anxiety, anemia, dyspepsia, urinary tract infection) and applied criteria to define their chronicity based on disease specific guidelines (see Appendix 1). Inclusion of patients with recurrent diseases seems relevant when studying risk factors for COPD exacerbations, but has not been done in previous studies.

Our study was based on patients’ medical records in general practice. Limited agreement between medical record-based and objectively identified comorbidities of COPD [30] and undiagnosed comorbidity in COPD patients is common [18, 31]. This may have resulted in underestimation of the presence of comorbidity in our study population. The use of real-life data presents limitations, for instance the fact that patients’ smoking history and lung function could not be included because this information is not consistently and uniformly documented in general practice medical records. We chose to limit the analyses to comorbidities with a relatively high (i.e., ≥5%) prevalence. This may mean that comorbidities that are related to increased exacerbation risk but have a low prevalence rate in the COPD patient population were missed.

We defined an exacerbation as an oral corticosteroid prescription, which is the recommended treatment for acute exacerbations in Dutch COPD guidelines [15]. Consequently, mild exacerbations treated with bronchodilators only are not included in our analyses. Oral steroid prescriptions during GP out-of-office hours, emergency department visits and hospitalizations, and prescriptions by pulmonary specialists may not always have been included for all patients, as these are not automatically added to patients’ medical records in all electronic patient record systems. Because there is no international consensus about a definition that discriminates relapse of an earlier exacerbation from a new one, our (arbitrary) choice to use an interval of ≥14 days since the end date of the previous oral steroid prescription may have led to under- or overestimation of the number of exacerbations. Unfortunately, the rather crude prescription information did not allow us to look at the impact of comorbidities on the duration or progression of exacerbations. Although observational studies such as ours lack the rigorous internal validity that is typical for randomized controlled trials, they provide valuable insight into comorbidity prevalence in COPD and its relation with an important outcome, i.e., exacerbations. As such, our findings should be considered in conjunction with those arising from other study designs, including randomized trials.

Better knowledge about the role that comorbidity plays in COPD exacerbation risk may contribute to lower exacerbation rates in COPD patients through patient-tailored and systems medicine approaches. In turn, reduction of exacerbations may improve patients’ quality of life and prevent disability, hospitalizations, and mortality. A challenge for researchers is to find ways to enable physicians to take comorbidity into account when assessing COPD patients’ exacerbation risk. Putcha et al. developed a simple score that includes 14 comorbidities, where one point increase in comorbidity count was associated with 21% higher exacerbation risk [9]. However, their comorbidity score does not include comorbidities such as asthma, lung cancer and depression, while our results indicate that these comorbidities are also related to exacerbation risk. Neither does Putcha’s score take differences in exacerbation risk for different comorbidities into account. This highlights the importance of including a wide range of comorbid chronic conditions like we did in our study.

Beside Putcha’s comorbidity score, several prognostic indices to support COPD patient care have been developed, [32] most of them predicting prognosis in terms of mortality or hospitalization. Only few indices predict exacerbation risk and only one (the DOSE index [33]) has been developed and validated in primary care [34]. Comorbidity is not included in the existing prognostic indices, with the exception of the COTE index, which assesses mortality and not exacerbation risk [10, 11]. Our results may contribute to the development of a prognostic index that connects comorbidities with exacerbation risk to identify patients at highest risk, thereby potentially reducing disease progression.