Abstract
The aims of this study were to characterize causes of death among veterans with COPD using multiple cause of death coding, and to examine whether causes of death differed according to timing of COPD diagnosis. Veterans with COPD who died during a five-year follow-up period were identified from national VA databases linked to National Death Index files. Primary, secondary, underlying, and all-coded causes of death were compared between recent and preexistent COPD cohorts using proportional mortality ratios (PMRs), which compares proportion dying from specific causes as opposed to absolute risk of death. Of 26,357 decedents, 7,729 were categorized preexistent and 18,628 were recent COPD cases. Unspecified COPD was listed as underlying cause of death in a significantly greater proportion of preexistent COPD cases compared to recent cases, 20% vs 10%, PMR = 2.0 (95% CI: 1.9–2.1). A relatively higher proportion of recently diagnosed cases died from lung/bronchus, prostate, and site-unspecified cancers. Respiratory failure (J969) was rarely coded as an underlying or primary cause (< 1%), but was a second-code cause of death in 9% of recent and 12% of preexistent cases. Differences in coded causes of death between patients with a recent diagnosis of COPD compared to a preexistent diagnosis of COPD suggests that there is either coded cause-related bias or true differences in cause of death related to length of time with diagnosis. Thus, methods used to identify cohorts of COPD patients, i.e., incidence versus prevalence-based approaches, and coded cause of death can affect estimates of cause-specific mortality.
| ABBREVIATIONS | ||
| BIRLS | = | beneficiary identification and record locator system |
| CHF | = | Congestive Heart Failure |
| CDC | = | Center for Disease Control |
| COPD | = | Chronic Obstructive Pulmonary Disease |
| ICD | = | International Classification of Diseases |
| NCHS | = | National Center for Health Statistics |
| NDI | = | National Death Index |
| PMR | = | proportional mortality ratio |
| VA | = | Veterans Affairs |
INTRODUCTION
Chronic obstructive pulmonary disease (COPD) is a major cause of morbidity, mortality, disability, and economic burden on a global level (Citation[1], Citation[2]). The prevalence and impact of COPD may be underestimated, as an estimated 10 million U.S. adults reported physician-diagnosed COPD while approximately 24 million adults have evidence of impaired lung function. A major known risk factor for COPD is cigarette smoking (Citation[3]), which has been also clearly identified as a direct cause of many types of cancers, ischemic heart disease, respiratory heart disease, stroke, and pneumonia (Citation[4]).
Current goals of COPD pharmacotherapy are to reduce exacerbations, improve health status and perhaps to prolong survival (Citation[5]). In order to understand the consequences of COPD and how existing and innovative new treatments may impact survival, it is important to understand causes of death in COPD patients. In mild to moderate COPD, it has been reported that most patients do not die from exacerbation due to respiratory failure, as cause of death is attributed to lung cancer, cardiovascular disease and coronary heart disease (Citation[6]). Patients with COPD often have multiple comorbidities, commonly co-occurring in combination with hypertension, ischemic heart disease, and diabetes (Citation[7]).
Thus, patients who have COPD may have their death attributed to a multiplicity of causes. Until the 1990s, the coding of routine mortality statistics in most countries included only underlying cause of death (Citation[8]). Furthermore, death registration data may not present a clear picture of mortality in COPD patients (Citation[9]). Examination of multiple cause mortality data in the United States indicated that mortality related to COPD is underestimated in studies examining only the underlying cause (Citation[10]).
Examination of primary, second, underlying and all-cause mortality in COPD patients can help to understand the extent to which coding can influence cause of death estimates. Another unknown factor of relevance from an epidemiological perspective is whether causes of death differ according to time since diagnosis of COPD. Thus, the aims of this study were to characterize causes of death among United States veterans with a diagnosis of COPD using data contained multiple coded causes of death, and to examine whether causes of death differed based on when COPD was diagnosed. The specific objectives of this study were: (Citation[1]) to describe the underlying, primary, secondary and all causes of death in patients with COPD in VA according to recent or longer term diagnosis of COPD, and (Citation[2]) to compare the proportion of deaths attributed to a specific underlying cause among patients with recent and longer-term COPD diagnoses to the general population.
METHODS AND MATERIALS
Data sources
The Veteran's Affairs (VA) Vital Status File mortality database was used to identify veterans who had died with a diagnosis of COPD. This file combines mortality information from several files that can be used to identify veterans that die, and includes the VA beneficiary identification and record locator system (BIRLS) file, Medicare, the Social Security Administration and VA hospital data.
Data on multiple causes of death was obtained from the multiple cause of death program managed by the National Center for Health Statistics (NCHS). Beginning in 1999, the NCDS switched to coding cause of death from ICD-9 to ICD-10 codes (Citation[11]). The record axis code files were used, which contained up to 20 coded causes of death for each individual.
Data on causes of death among the general population of the United States were obtained from the Centers for Disease Control and Prevention (Citation[12]). Data available from the website were limited to underlying cause of death, defined as the disease or injury which initiated the train of events leading directly to death or the circumstances of the accident or violence which produced the fatal injury. For greater comparability between VA and population-based estimates, statistics on underlying causes of death were restricted to males 45 years of age and older in the U.S. general population.
Study design and analysis
A retrospective cohort study design was used. IRB approval was obtained from the Edward Hines, Jr. VA Hospital/North Chicago VAMC Human Studies Subcommittee. Two cohorts were identified: (Citation[1]) recent diagnosis of COPD and (Citation[2]) preexistent diagnosis of COPD. Eligible ICD-9 codes for COPD were 490, 491.x and 492.x. Fiscal year 1999 (October 1998 to September 1999) was used as the index year for identification of incidence cases of COPD, i.e., recently diagnosed cases.
Recently diagnosed COPD was defined as an absence of any ICD-9 codes for COPD encounters relating to emergency room visits, hospitalizations, and outpatient visits prior to fiscal year 1999, but presence of 2 or more diagnostic codes during a twelve month period in any of the fiscal years from 1999 through 2004. This approach to case definition has been used in previous studies (Citation[12], Citation[13]). Patients were defined as preexistent COPD cases if they had any ICD-9 admission codes for COPD in the 12 months prior to fiscal years 1999-2004, i.e. in fiscal year 1998, which was the earliest data available. At the time of our request, data on multiple coded causes of death were available from NCHS to the end of 2003. Upon identification of patients with recent or preexisting COPD, the VA Vital Status File data were used to identify patients who died from 1999 to 2003. Subsequently, two random subsets of approximately 14,000 patients were selected in 2 separate data requests to NCHS.
Cases identified in the NDI data files were assigned a probability score and score-based class by the NHCS that represented the likelihood of a true (perfect) match based on personal identifiers (e.g., Social Security number, date of birth, first and last name) (Citation[11]). VA patient data was merged with the data compiled by NDI into analytic files based on probabilistic linkage. Decision rules for acceptable levels of thresholds of false positive and false negative were based upon NDI Plus guidelines (Citation[15]). Borderline cases were visually compared for similarities in identifiers between the datasets.
Differences in demographic characteristics between recent and preexistent diagnosis of COPD were identified using Chi-squares tests and t-tests for independent groups. Proportional Mortality Ratios (PMRs) with 95% confidence intervals were calculated for comparisons in the proportion of deaths attributed to an underlying cause between preexistent COPD cases, recent cases of COPD, and the US general population. The most common causes of death were listed and compared.
RESULTS
In total, multiple code causes of death records were positively linked for 26,357 COPD patients, including 18,628 recently diagnosed cases and 7,729 preexistent cases. Perfect matches were identified for 13.9% of the data request, i.e., perfectly fulfilled all NDI criteria for a match. Other than failure to meet the minimum criteria for a match, the only systematic exclusions related to small subset of patients from New York (n = 402) whose data was unavailable from the NCHS.
The mean age of patients identified with preexistent COPD was 73.9 (SD 10.2), which was slightly older than recent COPD patients who had a mean age of 73.1 (SD 9.1), (p < 0.001) (). The distribution of ethnicity was similar between recent and preexistent COPD cases except for African Americans. While approximately 70% of both recent and preexistent COPD cases were Caucasian, a higher proportion of recently diagnosed COPD cases were African American (11.6%) compared to preexistent cases of COPD (8.4%).
Table 1 Characteristics of patients who died with diagnosis of COPD (N = 26,357)
In comparing underlying causes of death between preexisting and recent cases of COPD, 20.0% of preexistent cases and 10.0% of recent cases died from unspecified COPD (PMR 2.00, 95% CI 1.88, 2.13) (). Similarly, 3.0% of preexistent cases and 1.4% of recent cases died from emphysema (PMR 2.20, 95% CI 1.85, 2.63). Compared to preexistent COPD cases, a higher proportion of recent COPD cases died from cancers of the lung/bronchus (13.9% vs 15.5%), prostate (1.5% vs1.9%), and site unspecified (1.0% vs 1.4%), respectively. As would be expected, proportionately fewer males 45 years and older from the general population died from lung cancer (8.4%) compared to the COPD cohorts.
Table 2 Proportional mortality ratios for underlying causes of death comparing preexistent COPD, recent COPD, and US general population
Congestive heart failure (I500) was an underlying cause of death in proportionately more preexistent COPD patients (2.4%) and recent COPD patients (2.3%) than among males 45 years and older in the general population (1.8%). However, PMRs were significantly higher for prostate cancer, colon cancer, atherosclerotic heart disease, AMI and stroke in the general population compared to veterans with recent and preexistent COPD ().
The 20 most common primary codes for cause of death were responsible for 65% of all recently diagnosed COPD cases and 69% of preexistent COPD cases (). Similar to underlying cause, primary coded cause of mortality was greater among preexistent COPD cases than recent cases for unspecified COPD (PMR 2.51, 95% CI 2.23, 2.82 and emphysema (PMR 2.54, 95% CI 1.85, 3.49). Septicemia and pneumonia were more commonly listed as a cause of death based on primary coded cause than using underlying cause for both preexistent and recent cases of COPD ( and ). Diabetes was a more common cause based on primary cause than underlying cause among recent COPD cases.
Table 3 Primary-coded cause of death, preexistent and recent COPD
Causes of death based on second code were more diffuse and the most commonly listed causes included 51% of longer term COPD cases and 45% of recent COPD cases. Interestingly, respiratory failure (J969) was among the 10 most common second-codes for causes of death for both preexisting and recent cases of COPD, but was rarely listed as primary code or underlying cause (). Lung/bronchial cancer (C349) and prostate cancer were seldom listed as second coded cancers (). Cause of death related to mental and behavioral disorder due to use of tobacco (F179), which occurred at an almost imperceptible rate among the general population, was coded as primary-code cause or second-code cause in 1–2% of preexisting and recently diagnosed COPD ( and ).
Table 4 Cause of death based on second code, pre-existing and recent COPD PMR
Pulmonary and cardiovascular disease-related causes dominate the top ten most common causes of death based on all codes (). Not surprisingly, unspecified COPD (J449) was the most frequently coded cause of death for both recent and preexistent diagnosis of COPD. Several causes of death that were common among all coded causes (> 5%) did not feature in the list of underlying causes, including respiratory failure, respiratory arrest, and cardiac arrest.
Table 5 All-coded causes of death, preexistent and recent COPD
DISCUSSION
Results of this study found differences in PMRs and coded causes of death between patients with a recent diagnosis of COPD compared to a preexistent diagnosis of COPD. These proportional rates of mortality, which relate to proportional risk of mortality due to specific causes among COPD patients and not absolute risk of death relative to the general population, suggest that there is either coded cause-related bias or true differences in cause of death related to length of time with COPD diagnosis. Thus, methods used to identify cohorts of COPD patients, i.e., incidence versus prevalence-based approaches, as well as level of coded cause of death (primary, secondary) can affect estimates of cause specific mortality. Such methods are typically employed in comparative effectiveness research, where large administrative claims databases are used to determine associations between medical treatments and outcomes of safety and effectiveness, such as risk of hospitalization or mortality.
Several previous studies have examined how multiple cause coding of death can illuminate bias in the reporting of mortality trends in COPD (Citation[8], Citation[10], Citation[16]) Our results were consistent with previous studies, finding that multiple coded causes of death provide additional information not conveyed by underlying cause. We also found that when patients were delineated according to timing of diagnosis, those diagnosed with COPD for a longer time were more likely to have a coded cause of death directly attributed to problems with respiration compared to those more recently diagnosed with COPD. Patients with a more recent diagnosis of COPD were more likely to die from cancer-related causes (lung/brochial, prostate, and unspecified neoplasms) compared to patients with a preexisting diagnosis of COPD.
These results are particularly relevant to clinical and epidemiological researchers concerned with cause of death as an outcome. Pharmacovigilance studies, for instance, focus on the evaluation of long-term outcomes associated with COPD medications that may include rare but fatal adverse events, and investigation into coded cause of death can be critical to understanding whether a specific mechanism is responsible. We found that respiratory failure was rarely coded as the underlying or primary cause (< 1% of cases), yet it was coded as one of the causes of death in 9% of recent cases of COPD and 12% of preexistent cases using second-coded cause. Among all coded causes, respiratory failure, respiratory arrest, and cardiac arrest were common codes but were not observed among underlying or primary coded cause. Thus, for investigations where specific causes of death such as respiratory failure are relevant, such as outcomes of pharmacotherapy in COPD, underlying cause may be insufficient.
COPD was associated with proportionately more death due to lung cancer and respiratory-related disease/complications, but proportionately more males 45 years and older in the US general population died from cardiovascular and cerebrovascular underlying causes. COPD patients were more likely to die from congestive heart failure (I500) as an underlying cause, perhaps evidence of the link between heart failure and respiratory disease (Citation[17]). The proportion of mentions of death due to AMI (8%) and heart failure (2%) among decedents in our study was similar to the characterization of mortality in patients with COPD who died in England and Wales using multiple cause coding (Citation[16]).
A major limitation of the study design is that it relies upon ICD-9 codes to identify patients with COPD. Diagnosis of COPD was not verified using lung function measurements and the direction of any misclassification bias is uncertain. Greater than 95% of the participants were male veterans, limiting generalizability. Similar distributions of age and self-reported ethnicity were generally noted between recent and preexistent COPD cases. However, a greater proportion of preexistent cases did not report ethnicity and proportionately more recently diagnosed cases were African American. We were limited in our ability to establish whether a larger proportion of African Americans declined to self-report their ethnicity among the preexistent cases.
In summary, we found that methods used to identify cohorts of COPD patients, i.e., incidence versus prevalence-based approaches, as well as level of coded cause of death (i.e., primary, secondary, underlying, all-cause) can affect estimates of cause specific mortality. Timing of diagnosis mattered, as patients with a preexistent diagnosis of COPD were more likely have COPD and emphysema as a coded cause of death compared to more recently diagnosed patients. In contrast, proportionately more patients with a recent diagnosis of COPD died from cancers related to the lung, prostate and unspecified neoplasm. Furthermore, when examining specific causes of death such as respiratory failure as an outcome, it appears necessary to obtain multiple coded causes of death, as underlying cause is insufficiently informative.
This study was supported by Veteran Affairs Health Services Research and Development (VA HS&R) Investigator Initiated Research (IRR) grant 03-307 (PI: Lee). The usual disclaimer applies, including that the views expressed are those of the authors do not necessarily represent the views of the VA. The authors have no actual or potential conflicts of interest to declare with respect to the results and interpretation of this manuscript.
REFERENCES
- Sullivan S D, Ramsey S D, Lee T A. The economic burden of COPD. Chest 2000; 117(2 Suppl)5S–9S
- Mannino D M, Homa D M, Akinbami L J, Ford E S, Redd S C. Chronic obstructive pulmonary disease surveillance—United States, 1971-2000. MMWR Surveill Summ 2002; 51(6)1–16
- Pauwels R A, Buist A S, Calverley P M, Jenkins C R, Hurd S S. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease. NHLBI/WHO Global Initiative for Chronic Obstructive Lung Disease (GOLD) Workshop summary. Am J Respir Crit Care Med 2001; 163(5)1256–1276
- Boyle P. Cancer, cigarette smoking and premature death in Europe: a review including the Recommendations of European Cancer Experts Consensus Meeting, Helsinki, October 1996. Lung Cancer 1997; 17(1)1–60
- Sin D D, Man S F. Pharmacotherapy for mortality reduction in chronic obstructive pulmonary disease. Proc Am Thorac Soc 2006; 3(7)624–629
- Anthonisen N R, Connett J E, Enright P L, Manfreda J. Hospitalizations and mortality in the Lung Health Study. Am J Respir Crit Care Med 2002; 166(3)333–339
- Lee T A, Pickard A S, Bartle B, Weiss K B. Osteoarthritis: A comorbid marker for longer life?. Ann Epidemiol 2007; 17(5)380–384
- Goldacre M J, Duncan M E, Cook-Mozaffari P, Griffith M. Trends in mortality rates comparing underlying-cause and multiple-cause coding in an English population 1979–1998. J Public Health Med 2003; 25(3)249–253
- Donaldson G C, Wedzicha J A. COPD exacerbations. 1: Epidemiology. Thorax 2006; 61(2)164–168
- Mannino D M, Brown C, Giovino G A. Obstructive lung disease deaths in the United States from 1979 through 1993. An analysis using multiple-cause mortality data. Am J Respir Crit Care Med 1997; 156: 814–818, (3 Pt 1)
- National Center for Health Statistics, Accessed at http://www.cdc.gov/nchs/ndi.htm National Death Index on July 11th, 2007
- Lee T A, Bartle B, Weiss K B. Spirometry use in clinical practice following diagnosis of COPD. Chest 2006; 129: 1509–1515
- Lee T A, Weiss K B. Fracture risk associated with inhaled corticosteroid use in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2004; 169: 855–859
- CDC Wonder. Compressed Mortality, Accessed at http://wonder.cdc.gov/controller/datarequest/D17;jsessionid=A12111BB7C4A36F6C04A8BF18BB4E5D7 on July 16, 2006
- Department of Health and Human Services, National Center for Health Statistics, Center for Disease Control and Prevention. National Death Index Plus: Coded Causes of Death. 1999
- Hansell A L, Walk J A, Soriano J B. What do chronic obstructive pulmonary disease patients die from? A multiple cause coding analysis. Eur Respir J 2003; 22(5)809–814
- Huiart L, Ernst P, Suissa S. Cardiovascular morbidity and mortality in COPD. Chest 2005; 128(4)2640–2646