 |
|
Controlling for Patient Case Mix at the End of Life: Issues in Identifying Cause of Death
To the Editor: Researchers have used various methods to identify cause of death for case-mix classification at the end of life, but there has been no systematic evaluation of the consistency between these methods. The most common approach is to use the death certificate underlying cause of death, but the limitations of this method include problematic data quality,1 prohibitive cost of obtaining National Death Index data, and identification of the underlying cause of death (the condition that started the train of fatal events) may not be consistent with a particular research objective. Some researchers have used health care utilization data from their study data sets to identify cause of death for case-mix classification. The methods have included identifying the last major diagnosis (last-diagnosis method)2-3 and using the diagnosis responsible for the most resource use near death (cost method).4-5 We assessed the consistency among death certificate data and these 2 other methods.
Methods
Using the California Merged Death Files, the Veterans Affairs Benefits Identification and Records Locator Subsystem Death File, and the Department of Veterans Affairs (VA) and Medicare utilization files, we extracted utilization and cost data for 12 808 VA users who died in California between October 1, 1999, and September 30, 2001, and had used VA or Medicare in the final year of life. We compared patient classification from 3 methods: the death certificate underlying cause of death, the last major diagnosis, and the health care cost during the final year of life. We identified the 16 cause-of-death categories other than accidents that accounted for at least 0.25% of elderly deaths nationwide in 2002.6 For the last-diagnosis method, we found the last principal diagnosis within these categories. For the cost method, we identified the category responsible for the plurality of health care costs in the final year of life. We measured the percentage of patients who were classified into each of the 10 most common cause-of-death categories, analyzed agreement among the methods at the patient level, and assessed consistency by death certificate cause-of-death category. Analyses were performed using SAS version 9.13 (SAS Institute Inc, Cary, NC).
Results
Table 1 shows the percentage of patients by the different methods placed into each of the 10 most common cause-of-death categories. Table 2 presents the patient-level agreement between methods when classifying patients into the top 2, 4, and 10 cause-of-death categories. Although the overall distributions were relatively similar when placing patients into the top 10 cause-of-death categories, the patient-level agreement between the death certificate data and the other 2 methods was only slightly higher than 50%. The last-diagnosis and cost methods had relatively high agreement with each other. Considering the actual death certificate cause-of-death category, each method agreed with the death certificate more than 75% of the time when the underlying cause of death was cancer, but less than 50% for heart disease, chronic obstructive pulmonary disease, and cerebrovascular diseases.
|
|
|
|
Table 1. Distribution of Patients Into 10 Most Common Cause-of-Death Categories by Method*
|
|
|
|
|
|
|
Table 2. Patient-Level Agreement Between Methods for the Top 2, 4, and 10 Cause-of-Death Categories
|
|
|
Comment
The method to identify cause of death most used by public health and health services researchers is classification based on the death certificate underlying cause of death.1 Although this is often appropriate, patients typically have multiple diseases near the end of life,4 and the underlying cause of death may have little relation to the condition that necessitates the majority of health care utilization near death.1 The last-diagnosis method may be optimal in studies analyzing care very close to the end of life, because it best represents the reason for health care utilization closest to death. The cost method could result in more reliable classification than the last-diagnosis method, because it takes utilization directly into account. However, the cost method is sensitive to differences in practice patterns and insurance coverage and could be affected heavily by expensive procedures. When using the last-diagnosis and cost methods, there is the potential for errors if the validity of the diagnosis or cost data has not been established.
Although the methods classified similar percentages of patients into top cause-of-death categories, they did not place the same patients into each category. None of these methods is adequate in capturing the complete story of a decedent’s cause of death, especially given the increasingly high proportion of deaths that have multifactorial causes.1 However, each method provides potentially useful information on case mix at the end of life and researchers should consider the objectives of their study, the feasibility of applying each method, and the reasons they need to identify cause of death when deciding which method to use. When critically interpreting these studies, readers should also consider these issues.
Access to Data: Mr Richardson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosures: None reported.
Funding/Support: This study was funded by grant IIR-02-189 from the US Department of Veterans Affairs Health Services Research and Development Service.
Role of the Sponsor: The US Department of Veterans Affairs Health Services Research and Development Service was not directly involved in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Acknowledgment: We thank Shirley Kim, MHSA, and Pon Su, MS, Health Economics Resource Center of Health Services Research and Development Service, for their acquisition and management of the VA and Medicare utilization data sets used for this study. The contributions of Ms Kim and Mr Su were funded by the US Department of Veterans Affairs Health Services Research and Development Service.
Samuel S. Richardson, BA
samuel.richardson3{at}med.va.gov
Wei Yu, PhD
Health Economics Resource Center of Health Services Research and Development Service
US Department of Veterans Affairs
Menlo Park, Calif
1. Rosenberg HM. Cause of death as a contemporary problem. J Hist Med Allied Sci. 1999;54:133-153.
FREE FULL TEXT
2. Temkin-Greener H, Meiners MR, Petty EA, Szydlowski JS. The use and cost of health services prior to death: a comparison of the Medicare-only and the Medicare-Medicaid elderly populations. Milbank Q. 1992;70:679-701.
FULL TEXT
| PUBMED
3. Bhattacharya J, Garber AM, MaCurdy T. Cause-Specific Mortality Among Medicare Enrollees. Cambridge, Mass: National Bureau of Economic Research; 1996. Working Paper 5409.
4. Hogan C, Lynn J, Gabel J, Lunney J, O'Mara A, Wilkinson A. Medicare Beneficiaries' Costs and Use of Care in the Last Year of Life. Washington, DC: Medicare Payment Advisory Commission; 2000. No. 00-1.
5. Garber AM, McCurdy TE, McClellan MC. Medical care at the end of life: diseases, treatment patterns, and costs. In: Garber A, ed. Frontiers in Health Policy Research. Vol W6748. Cambridge: Massachusetts Institute of Technology Press; 1999:77-98.
6. Kochanek KD, Murphy SL, Anderson RN, Scott C. Deaths: Final Data for 2002: National Vital Statistics Reports; Vol 53 No. 5. Hyattsville, Md: National Center for Health Statistics; 2004.
Letters Section Editor: Robert M. Golub, MD, Senior Editor.
JAMA. 2005;294:793-794.
 CiteULike  Connotea  Del.icio.us  Digg  Reddit  Technorati  Twitter
What's this?
|
