PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
 
J Palliat Med. Aug 2011; 14(8): 929–939.
PMCID: PMC3146748
Hospice Care and Survival among Elderly Patients with Lung Cancer
Akiko M. Saito, M.D., Ph.D.,1 Mary Beth Landrum, Ph.D.,2 Bridget A. Neville, M.P.H.,3 John Z. Ayanian, M.D., M.P.P.,2,4 Jane C. Weeks, M.D., M.Sc.,3 and Craig C. Earle, M.D., M.Sc., FRCPCcorresponding author5
1Laboratory of Clinical, Epidemiological, and Health Services Research, Clinical Research Center, National Hospital Organization Nagoya Medical Center, Aichi, Japan.
2Department of Health Care Policy, Harvard Medical School, Boston, Massachusetts.
3Division of Population Sciences, Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, Massachusetts.
4Division of General Medicine, Brigham and Women's Hospital, Boston, Massachusetts.
5Cancer Care Ontario, Ontario Institute for Cancer Research, and the Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada.
corresponding authorCorresponding author.
Address correspondence to: Craig C. Earle, M.D., M.Sc., FRCPC, Institute for Clinical Evaluative Sciences, Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, Room G-106, Toronto, Ontario, M4N 3M5 Canada. E-mail:craig.earle/at/ices.on.ca
Accepted April 4, 2011.
Background
Recently observed trends toward increasingly aggressive end-of-life care may reflect providers' concerns that hospice may hasten death.
Methods
Using the Surveillance, Epidemiology, and End Results-Medicare linked database, we identified 7879 patients aged 65 years or older who died of advanced non-small–cell lung cancer from 1991 through 1999 after surviving for at least 3 months from their cancer diagnoses. Length of hospice admission post-cancer diagnosis and indicators of aggressive end-of-life care were ascertained based on claims data. We evaluated overall survival and care near death after controlling for baseline characteristics by using propensity score (PS) and instrumental variable analyses (IVA).
Results
Hospice patients were older, more likely to be non-Hispanic white and female, more likely to reside in urban areas with high hospice availability and higher socioeconomic status, more likely to be treated in a teaching hospital, and received less aggressive end-of-life care compared to nonhospice patients. Among hospice patients, those experiencing short-term hospice admissions within 3 days of death were more likely to be male, reside in urban areas, be treated in a teaching hospital, and receive more aggressive end-of-life care. PS analysis found that survival favored hospice patients slightly relative to nonhospice patients by 5.0 percentage points (25.7% versus 20.7%) at 1 year and 1.4 percentage points (6.9% versus 5.5%) at 2 years postdiagnosis (p < 0.001), while there was no significant difference between those with short- and longer duration hospice stays (p = 1.00). IVA confirmed these findings.
Conclusions
Hospice enrollment did not compromise length of survival following advanced lung cancer diagnosis.
The goal of hospice care is not to cure the illness, but to keep the pain and suffering of a terminally ill patient to a minimum. Although hospice and palliative care are now well established as appropriate13 and the use of hospice services in the United States has increased for over 30 years since the Medicare hospice benefit was established by Congress in 1982,4,5 hospice still remains underutilized.
In retrospective analyses, Connor et al.6 showed that hospice did not shorten survival among Medicare patients who died within 3 years with breast, colon, lung, prostate or pancreatic cancers, or congestive heart disease. The only randomized controlled trial, published in 1984 by Kane et al.,7 did not find a relationship between survival and hospice care for patients who were expected to die within 6 months of lung, prostate, ear, nose, throat, brain, or other cancers. Temel et al.8 concluded that patients with metastatic non-small–cell lung cancer (NSCLC) who received early palliative care, which provides care both in and outside the hospice, lived 2 months longer than those receiving standard care, although fewer patients in the palliative care group than in the standard care group received aggressive end-of-life care. Despite these results, recently observed trends toward increasingly aggressive care near death associated with late or nonhospice admission5,9,10 may reflect concerns by some practitioners or patients that hospice may hasten death.11 Thus, we compared survival and patterns of care near death in elderly patients with advanced NSCLC who received hospice care and those without hospice services, after controlling for baseline patient and disease characteristics with advanced statistical methods, to gain a better understanding of the clinical implications of hospice care.
Data sources for this study
The linked Surveillance, Epidemiology, and End Results (SEER)-Medicare database12 was used to identify the study cohort. SEER is a source of information on cancer incidence and survival. Eleven tumor registries participated in the SEER program during the study period and approximately 97% of incident cases for these regions were ascertained,13 covering a representative sample of approximately 14% of the U.S. population.12,14 Medicare claims for eligible patients have been linked to the SEER database, as have sociodemographic data from the 2000 Census.15,16
Identification of study cohort
Among the potentially eligible study cohort (N = 15,391) as patients who died from American Joint Committee on Cancer stage III/IV NSCLC between 1991 and 1999, we limited the study cohort as those after surviving at least 3 months with their cancer. Patients were continuously enrolled in Parts A and B of Medicare while simultaneously not in a health maintenance organization (HMO) in the last 3 months of life. They qualified for Medicare on the basis of age and diagnoses were not made from death certificate or autopsy, yielding a study cohort of 7879 patients. The 3-month survival criterion was to exclude patients with rapidly progressive disease in order to get more homogeneous population with respect to prognosis, as these patients were more likely to experience aggressive-approach chemotherapy near death and were less likely to receive longer term hospice care. Our sensitivity analyses confirmed that the inclusion of the excluded cohort did not change the results. In aggregate, we believe that the study cohort is valid. Patients in HMOs were excluded as complete data on explanatory variables is not available for them. This study was deemed exempt by the Dana-Farber Cancer Center/Partners Cancer Care Institutional Review Board because we used publicly available, deidentified data.
Variables used in this study
“Hospice patients” were defined as those with claims for hospice care at least once between diagnosis and death, while “nonhospice patients” were those without such claims. “Hospice patients” were further divided into two groups according to whether the length of hospice stay was 3 or fewer days (“short-term hospice patients”) or that of 4 or more days (“longer term hospice patients”). Length of stay was calculated from the first admission to hospice until death.
Explanatory variables included nine baseline characteristics (age at diagnosis, gender, race/ethnicity (non-Hispanic white or other), Charlson comorbidity index (0, 1, 2, or more), socioeconomic quintiles, urban/rural residence, year of diagnosis, provider characteristics (ever received care in a teaching hospital at any time after their cancer diagnosis or not and availability of hospice). Socioeconomic quintiles were developed using the following prioritization scheme: the median income from census tract, then zip code median income, and then per capita income according to zip code when median income was not available. A Charlson comorbidity index was calculated by examining ICD-9 Clinical Modification diagnosis and procedure codes recorded in the year prior to diagnosis, according to the Deyo method and applied to inpatient and outpatients claims as suggested by Klabunde et al.1719 Availability of hospice per million population in 1999 was calculated from the Area Resource file as the number of hospices or hospitals with hospice services within each Health Care Service Area (HCSA).20
Statistical analysis
Univariate comparisons for baseline characteristics and indicators of aggressive end-of-life care21,22 (a new chemotherapy regimen starting less than 30 days before death, the last dose of chemotherapy within 14 days of death, more than one emergency deparment visit in the last month of life, an intensive care unit admission in the last month of life, more than one hospital admission in the last month of life, and spending more than 14 days in hospital in the last month of life) were made with the χ2 test for categorical variables and the Wilcoxon rank-sum test for continuous variables. Overall survival (OS), defined as the length of time from the date of advanced cancer diagnosis to death from their disease was calculated by using the Kaplan-Meier method,23 and the log-rank test24 was used for group comparisons of hospice utilization. Cox proportional hazards regression models25 were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for comparisons of unadjusted hospice utilization and then adjusted for nine baseline characteristics. Whether a patient ever received chemotherapy at any time after their cancer diagnosis was also included in the Cox model, according to our previous findings.26
Because our prior work, as well as work by others9,21,22,2732 suggesting that the decision to refer patients to hospice care is affected by various patient and provider characteristics, we additionally analyzed patients closely matched for the likelihood that they would have hospice care by using PS methods.33 Since the PS approach is primarily designed to be applied to two-group comparisons, we developed two scenarios: (1) to evaluate the effect of hospice service versus (versus) nonhospice in all patients and (2) to assess the effect of longer-term versus short-term hospice stays among hospice patients. In each scenario, we included nine variables to estimate PSs for each patient as an adjustment for baseline characteristics. Each hospice patient was matched with a nonhospice patient (matched cohort A in Table 1) based on the “greedy” matching techniques described elsewhere, which matches the nearest neighbor within each matching caliper.34 Likewise, short-term hospice patients were matched with longer-term hospice patients (matched cohort B in Table 2) We first compared OS, practice patterns at the end of life and place of death (hospital, hospice, or other) by using the unmatched cohort, then did the same analyses by using the matched cohorts A and B. The impact of hospice care or length of hospice stay on survival, respectively, was also calculated across quintiles of PS. To see if the practice patterns near death affected survival via hospice utilization, we also evaluated these data after adding six indicators of aggressive end-of-life care to nine baseline characteristics in each model.
Table 1.
Table 1.
Characteristics of Patients by Hospice Utilization
Table 2.
Table 2.
Characteristics of Patients by Length of Hospice Stay
For sensitivity analyses, we performed an instrumental variable analysis (IVA)26,3539 to evaluate whether unobserved characteristics of patients might explain differences in survival associated with any hospice utilization or the length of hospice utilization. Based on our previous work demonstrating significant unexplained geographic variation in use of hospice but not of survival, we chose regional availability of hospice care as the instrument.2830,32 Under the assumption that unmeasured patient characteristics such as performance status do not differ according to where a patient receives care, comparison of survival for patients living in areas of higher availability of hospice with those living in areas of lower availability of hospice can indicate the effect on survival. The instrumental variable estimate (IVE) reflects the patients who received hospice care if they lived in an area of high availability of hospice but would not have received hospice care if they lived in an area of low availability of hospice, otherwise known as “marginal patients.”
All statistical tests were two-sided, and 95% CIs are presented for adjusted results. All analyses were conducted using SAS 9.1 software (SAS Institute, Cary, NC).
Characteristics of cohorts
The characteristics of 7879 patients before and after they were matched based on their propensity to receive care in a hospice are shown in Table 1. In the unmatched cohort, 3775 patients (47.9%) had ever received care in a hospice. Compared to the nonhospice group, the hospice group was older, more likely to be female and non-Hispanic white, more likely to reside in urban areas with higher socioeconomic status and greater availability of hospice, more likely to be treated in a teaching hospital, and was diagnosed in more recent years. All 6 indicators of aggressive end-of-life care were less likely to be provided to hospice patients compared to nonhospice patients. Of 3775 hospice patients, 3186 (84.4%) were matched based on the baseline characteristics to a similar nonhospice patient; no significant differences in baseline characteristics were noted among matched patients in these two groups (Table 1).
Ten percent (n = 379) of 3775 hospice patients experienced short-term hospice stays (Table 2). Short-term hospice patients were more likely to be male, reside in urban areas, and be treated in a teaching hospital. Indicators of aggressive end-of-life care were associated with short-term hospice stays. After controlling for the baseline characteristics, 100% of the 379 short-term hospice patients were matched with similar longer-term hospice patients. The availability of hospice differed significantly between the two groups in the unadjusted comparison, but not in the matched cohort B, and there were no significant differences in explanatory variables between the two groups in the matched cohort B.
Survival
OS at 1 and 2 years in the unmatched cohort were respectively 21.4% and 6.4% for nonhospice patients, and 24.6% and 6.2% for hospice patients (p < 0.01). After controlling for baseline variables, the absolute differences in survival between nonhospice and hospice patients of 3.2% at 1 year and 0.2% at 2 years in the unmatched cohort were increased to 5.0% (20.7% versus 25.7%) at 1 year and 1.4% (5.5% versus 6.9%) at 2 years in the matched cohort A, and remained statistically significant (p < 0.001). Using identical methods, among hospice patients, point estimates of survival between short-term and longer-term hospice patients were calculated (Table 2). The absolute difference in estimated survival between the two groups of 2.7% at 1 year and 1.7% at 2 years in the unmatched cohort (p = 0.39), and 1.5% at 1 year and 0.6% at 2 years in the matched cohort B (p = 1.00), respectively, did not differ significantly. Similar results were seen even after matching with practice patterns as well as baseline variables (data not shown). The survival curves for matched cohorts A and B are shown in Figure 1.
FIG. 1.
FIG. 1.
Kaplan-Meier Survival Curves after diagnosis of advanced non-small cell lung cancer by the hospice utilization. A: A matched cohort A of 3186 nonhospice patients and 3186 hospice patients after controlling for baseline characteristics. B: A matched cohort (more ...)
In the adjusted Cox regression analysis, longer-term hospice patients were found to have longer survival than nonhospice patients (HR 0.87, 95% CI 0.83 to 0.91, p < 0.001), and similar survival was observed between short-term and nonhospice patients (HR 0.94, 95% CI 0.85 to 1.05, p = 0.26). The effects were similar in all strata of propensity score. Worse survival was noted for men, those not treated in a teaching hospital, those with worse comorbidity, those diagnosed in more recent years, and those who never received chemotherapy (Table 3). Race/ethnicity, urban residence, hospice availability were not associated with survival. Inclusion of six indicators of aggressive end-of-life care in the model produced similar results, suggesting that the longer term hospice care is independently associated with survival.
Table 3.
Table 3.
Factors Significantly Associated with Survival in Multiple Cox Regression Analysis
We also estimated the impact of hospice use or length of hospice stay on survival by using IVA to test the robustness of the foregoing results from the two statistical tools. We separated the patients into two groups based on HCSA of higher or lower availability of hospice, using the median hospice availability (6.4 per million population in 1999) as the cut point. Patients who lived in an area of high hospice availability were more likely to experience care in a hospice compared to those who lived in an area of low hospice availability (53.1% versus 42.9%). Although the groups differed significantly in terms of some demographic variables, such as gender, race/ethnicity, urban/rural residence, year of diagnosis, and experiencing care in a teaching hospital, the difference was smaller than that between hospice patients vs. non-hospice patients in the unmatched cohort (Table 1). There were no significant differences between the two groups in the age at diagnosis, Charlson comorbidity index, socioeconomic status, and survival (21.8% versus 23.4% at 1 year; 5.5% versus 6.5% at 2 years, p = 0.08). These findings validated our assumption that the regional availability of hospice significantly influences whether a patient received care in a hospice, and is not associated with patients' severity of disease and outcomes, which supports the IV being a strong and valid instrument. We could not clearly identify a survival effect based on receipt of hospice care for marginal patients at 1 year (IVE [standard error, SE] 14.7% [10.2%], p = 0.15) or at 2 years (IVE [SE] 11.8% [5.9%], p = 0.04) even after controlling for the practice patterns at the end of life and/or baseline characteristics.
Place of death
In the unmatched cohort, 92.6% of the hospice patients died in hospice care and fewer hospice patients died in an acute care hospital compared to nonhospice patients (2.8% versus 39.7%, p < 0.001). After controlling for baseline characteristics, much the same pattern in the matched cohort A was observed. Among hospice patients, the proportion who died in a hospital was higher in the short-term hospice group both in the unmatched cohort and matched cohort B (both 5.3% versus 2.6%, p < 0.01).
Palliative, rather than life-prolonging or curative care, is the primary goal for hospice patients. As a result, despite the higher quality of life and better pain management that has been observed in hospice patients,40,41 there is a concern that hospice care might be associated with shorter survival compared to ongoing anticancer care outside of a hospice.
In accordance with the previous literature,9,2732,42 our data have shown that baseline characteristics, including age at diagnosis, gender, race/ethnicity, urban residence, experiencing care in a teaching hospital, and comorbid conditions were associated with choice of hospice care,30 although hospice availability did not affect length of hospice admission among those with hospice care. After controlling for these baseline characteristics by using the multiple Cox regression analysis and the propensity score analysis, we could not detect a detrimental effect on survival by entering hospice, as expected.68 However, these traditional logistic regression models could be biased because admissions to hospices are likely to be confounded by unobserved variables. Therefore, we attempted to remove residual selection bias by choosing the regional availability of hospices as an IV. Results from the IVA indicate that regional availability of hospice was associated with any hospice use, but not with length of hospice stay. Instead, experiencing aggressive end-of-life care was more predictive of (shorter) duration of hospice use. Despite a significant relationship between aggressive end-of-life care and no or only short-term hospice stay, hospice patients were found to have comparable or even longer survival compared to nonhospice patients based on three different statistical approaches.
Advances in medical technologies and a perception that patients favor receiving aggressive care even very near death for small expected benefits may reduce the number of patients referred to hospice,9,10,43,44 especially in the black patients.45 Short-term hospice use was not associated with longer survival, suggesting that continuing aggressive care close to death did not necessarily translate into better outcomes or inaccurate estimation of the prognoses. Together, these data suggest that provider or patient preferences and resource considerations, rather than clinical factors, may be driving end-of-life care for marginal patients.
Our study had several limitations. First, this study cohort did not include patients aged less than 65 years and those insured by an HMO in the Medicare program so the current results cannot necessarily be generalized to younger patients and those with managed care insurance. However, 60% of cancer cases occur in patients aged 65 or older46 and Medicare covers the majority of cancer patients in the United States.47 In addition, although a previous study showed that Medicare beneficiaries with managed care insurance were more likely than those with fee-for-service insurance to enroll in hospice and stay longer, the length of hospice stay for most hospice patients was less than one month regardless of whether they were enrolled in managed care or fee-for-service.48 We did not capture information if patients, who elected their hospice benefit, withdrew later. Our study was limited to patients who had died from lung cancer, which is the leading cause of cancer mortality in the United States, but the choice of hospice use, treatment options, and/or clinical outcomes for these patients may differ from those with other types of cancer.6,42 As with any study using administrative data, the accuracy of some of the variables used, including the identification of hospice use and the calculation of the comorbidity score, may be limited.49,50 Lastly, selection bias cannot be completely excluded without randomization. For example, physicians may be more likely to refer patients to hospice when they think they are going to live long enough to benefit from it.
In conclusion, use of hospice and length of hospice stay for Medicare patients with advanced NSCLC did not compromise survival. Appropriate timing of referral to hospice gives terminally ill cancer patients and their families more time and opportunity to benefit from palliative services and avoid futile interventions. Concern about hastening death should not be a barrier to hospice care.
Acknowledgments
This research was supported by a grant CA 91753-02 from the National Cancer Institute.
Author Disclosure Statement
No competing financial interests exist.
1. Doyle D, editor; Hanks G, editor; MacDonald N, editor. Oxford Textbook of Palliative Medicine. Oxford: Oxford University Press; 1998.
2. Jones AB. Moga DN. Davie KA. Transforming end-of-life care for the 21st century: The hospice vision. J Palliat Med. 1999;2:9–14. [PubMed]
3. Smith TJ. Schnipper LJ. The American Society of Clinical Oncology program to improve end-of-life care. J Palliat Med. 1998;1:221–230. [PubMed]
5. Han B. Remsburg RE. McAuley WJ. Keay TJ. Travis SS. National trends in adult hospice use: 1991–1992 to 1999–2000. Health Aff (Millwood) 2006;25:792–799. [PubMed]
6. Connor SR. Pyenson B. Fitch K. Spence C. Iwasaki K. Comparing hospice and nonhospice patient survival among patients who die within a three-year window. J Pain Symptom Manage. 2007;33:238–246. [PubMed]
7. Kane RL. Wales J. Bernstein L. Leibowitz A. Kaplan S. A randomised controlled trial of hospice care. Lancet. 1984;1:890–894. [PubMed]
8. Temel JS. Greer JA. Muzikansky A, et al. Early palliative care for patients with metastatic non-small-cell lung cancer. N Engl J Med. 2010;363:733–742. [PubMed]
9. Earle CC. Neville BA. Landrum MB. Ayanian JZ. Block SD. Weeks JC. Trends in the aggressiveness of cancer care near the end of life. J Clin Oncol. 2004;22:315–321. [PubMed]
10. Matsuyama R. Reddy S. Smith TJ. Why do patients choose chemotherapy near the end of life? A review of the perspective of those facing death from cancer. J Clin Oncol. 2006;24:3490–3496. [PubMed]
11. Gawande A. Letting Go. What should medicine do when it can't save your life? New Yorker, August 2, 2010. www.newyorker.com/reporting/2010/08/02/100802fa_fact_gawande. [Mar 30;2011 ]. www.newyorker.com/reporting/2010/08/02/100802fa_fact_gawande
12. The Surveillance Research Program, NCI: SEER Registries: National Cancer Institute. http://seer.cancer.gov/registries/data.html. [Mar 30;2011 ]. http://seer.cancer.gov/registries/data.html
13. Zippin C. Lum D. Hankey BF. Completeness of hospital cancer case reporting from the SEER Program of the National Cancer Institute. Cancer. 1995;76:2343–2350. [PubMed]
14. Nattinger AB. McAuliffe TL. Schapira MM. Generalizability of the surveillance, epidemiology, and end results registry population: Factors relevant to epidemiologic and health care research. J Clin Epidemiol. 1997;50:939–945. [PubMed]
15. Centers for Medicare and Medicaid Services: U.S. Department of Health & Human Services. www.cms.gov/ [Mar 30;2011 ]. www.cms.gov/
16. Potosky AL. Riley GF. Lubitz JD. Mentnech RM. Kessler LG. Potential for cancer related health services research using a linked Medicare-tumor registry database. Med Care. 1993;31:732–748. [PubMed]
17. Charlson ME. Pompei P. Ales KL. MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. J Chronic Dis. 1987;40:373–383. [PubMed]
18. Deyo RA. Cherkin DC. Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45:613–619. [PubMed]
19. Klabunde CN. Potosky AL. Legler JM. Warren JL. Development of a comorbidity index using physician claims data. J Clin Epidemiol. 2000;53:1258–1267. [PubMed]
20. Makuc DM. Haglund B. Ingram DD. Kleinman JC. Feldman JJ. Health service areas for the United States. Vital Health Stat 2. 1991:1–102. [PubMed]
21. Earle CC. Neville BA. Landrum MB. Souza JM. Weeks JC. Block SD. Grunfeld E. Ayanian JZ. Evaluating claims-based indicators of the intensity of end-of-life cancer care. Int J Qual Health Care. 2005;17:505–509. [PubMed]
22. Earle CC. Park ER. Lai B. Weeks JC. Ayanian JZ. Block S. Identifying potential indicators of the quality of end-of-life cancer care from administrative data. J Clin Oncol. 2003;21:1133–1138. [PubMed]
23. Kaplan EL. Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457–481.
24. Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep. 1966;50:163–170. [PubMed]
25. Cox DR. Regression models and life tables [with discussion] J R Stat Soc. 1972;Series B:187–220.
26. Earle CC. Tsai JS. Gelber RD. Weinstein MC. Neumann PJ. Weeks JC. Effectiveness of chemotherapy for advanced lung cancer in the elderly: Instrumental variable and propensity analysis. J Clin Oncol. 2001;19:1064–1070. [PubMed]
27. Christakis NA. Escarce JJ. Survival of Medicare patients after enrollment in hospice programs. N Engl J Med. 1996;335:172–178. [PubMed]
28. Christakis NA. Iwashyna TJ. Impact of individual and market factors on the timing of initiation of hospice terminal care. Med Care. 2000;38:528–541. [PubMed]
29. Connor SR. Elwert F. Spence C. Christakis NA. Geographic variation in hospice use in the United States in 2002. J Pain Symptom Manage. 2007;34:277–285. [PubMed]
30. Iwashyna TJ. Chang VW. Zhang JX. Christakis NA. The lack of effect of market structure on hospice use. Health Serv Res. 2002;37:1531–1551. [PMC free article] [PubMed]
31. Miller SC. Weitzen S. Kinzbrunner B. Factors associated with the high prevalence of short hospice stays. J Palliat Med. 2003;6:725–736. [PubMed]
32. Virnig BA. Kind S. McBean M. Fisher E. Geographic variation in hospice use prior to death. J Am Geriatr Soc. 2000;48:1117–1125. [PubMed]
33. Rosenbaum PR. Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55.
34. Parsons L. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. Proceedings of the 25th Annual SAS Users Group International Conference; 2001. pp. 214–126.
35. Bowden RJ. Turkington DA. Instrumental Variables. Cambridge University Press; Cambridge, MA: 1984.
36. Landrum MB. Ayanian JZ. Causal effect of ambulatory specialty care on mortality following myocardial infarction: A comparison of propensity score and instrumental variable analyses. Health Serv Outcomes Res Method. 2001;2:221–245.
37. McClellan M. McNeil BJ. Newhouse JP. Does more intensive treatment of acute myocardial infarction in the elderly reduce mortality? Analysis using instrumental variables. JAMA. 1994;272:859–866. [PubMed]
38. Newhouse JP. McClellan M. Econometrics in outcomes research: The use of instrumental variables. Annu Rev Public Health. 1998;19:17–34. [PubMed]
39. Stukel TA. Fisher ES. Wennberg DE. Alter DA. Gottlieb DJ. Vermeulen MJ. Analysis of observational studies in the presence of treatment selection bias: Effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods. JAMA. 2007;297:278–285. [PMC free article] [PubMed]
40. Greer DS. Mor V. Morris JN. Sherwood S. Kidder D. Birnbaum H. An alternative in terminal care: Results of the National Hospice Study. J Chronic Dis. 1986;39:9–26. [PubMed]
41. Wallston KA. Burger C. Smith RA. Baugher RJ. Comparing the quality of death for hospice and non-hospice cancer patients. Med Care. 1988;26:177–182. [PubMed]
42. Lackan NA. Ostir GV. Freeman JL. Mahnken JD. Goodwin JS. Decreasing variation in the use of hospice among older adults with breast, colorectal, lung, and prostate cancer. Med Care. 2004;42:116–122. [PubMed]
43. Earle CC. Do unto others. J Clin Oncol. 2006;24:3331–3332. [PubMed]
44. Emanuel EJ. Young-Xu Y. Levinsky NG. Gazelle G. Saynina O. Ash AS. Chemotherapy use among Medicare beneficiaries at the end of life. Ann Intern Med. 2003;138:639–643. [PubMed]
45. Johnson KS. Kuchibhatla M. Tanis D. Tulsky JA. Racial differences in hospice revocation to pursue aggressive care. Arch Intern Med. 2008;168:218–224. [PubMed]
46. Yancik R. Epidemiology of cancer in the elderly. Current status and projections for the future. Rays. 1997;22(1 Suppl):3–9. [PubMed]
47. Thorpe KE. Howard D. Health insurance and spending among cancer patients. Health Aff (Millwood) 2003;Suppl Web Exclusives:W3–189–198. [PubMed]
48. McCarthy EP. Burns RB. Ngo-Metzger Q. Davis RB. Phillips RS. Hospice use among Medicare managed care and fee-for-service patients dying with cancer. JAMA. 2003;289:2238–2245. [PubMed]
49. Riley GF. Potosky AL. Klabunde CN. Warren JL. Ballard-Barbash R. Stage at diagnosis and treatment patterns among older women with breast cancer: An HMO and fee-for-service comparison. JAMA. 1999;281:720–726. [PubMed]
50. Whittle J. Steinberg EP. Anderson GF. Herbert R. Accuracy of Medicare claims data for estimation of cancer incidence and resection rates among elderly Americans. Med Care. 1991;29:1226–1236. [PubMed]
Articles from Journal of Palliative Medicine are provided here courtesy of
Mary Ann Liebert, Inc.