Effects of cancer history and comorbid conditions on mortality and healthcare use among older cancer survivors^†‡

Abstract

BACKGROUND

Older cancer survivors use healthcare services to an increased extent relative to their counterparts who have no history of malignant disease. In the current study, the authors set out to assess the effects of cancer history and comorbid conditions on healthcare use and mortality.

METHODS

Using information from the 1992 North Carolina Established Populations for Epidemiologic Study of the Elderly database, study participants were classified as having no history of malignant disease or as having a recent (cancer diagnosed < 1 year earlier), intermediate (cancer diagnosed 1–6 years earlier), or remote (cancer diagnosed > 6 years earlier) history of malignancy. Overall, 15 different comorbid conditions were ascertained. Logistic regression models adjusted for sociodemographic factors, tobacco and alcohol use, and functional measures were used to determine the risk of emergency room, hospital, and nursing home (NH) admission in 1992 and also in 1996 according to history of malignancy and presence of comorbid conditions. Using data from the National Death Registry, a similar controlled analysis of 7-year mortality also was performed.

RESULTS

There were 2567 participants in the current study (mean age, 79 years; range, 71–102 years); 69% of all participants were women, 55% were African American, and 14% reported having a history of malignancy. Participants with a history of malignancy had an average of 3 comorbid conditions, and differences across groups in terms of cardiovascular and lung disease incidence were noted. Controlled analyses revealed that recent cancer history (odds ratio [OR], 15.5; 95% confidence interval [CI], 7.0–34.2) and intermediate cancer history (OR, 2.1; 95% CI, 1.4–3.3) were associated with same-year hospital admission. In addition, having a recent history of malignancy in 1992 was found to be correlated with NH admission 4 years later (OR, 3.1; 95% CI, 1.1–9.1). History of malignancy was not associated with mortality.

CONCLUSIONS

Cancer history had limited influence on healthcare use and mortality. Efforts aimed at improving health-related outcomes in older cancer survivors should continue to focus on attenuating the impact of comorbid conditions. Cancer 2004. Published 2004 by the American Cancer Society.

The National Cancer Institute has determined that cancer-related mortality rates stabilized in the late 1990s, leaving a group of individuals whose malignancies have been cured or who are living with cancer.1 It is estimated that in 1997, 16% of all persons age > 65 years were cancer survivors, and the number of older cancer survivors will increase as the general population ages.2 In cohort studies, older cancer survivors have been found to experience changes in emotional health and quality of life as well as persistent treatment-related problems.3, 4 In addition, cancer survivors are more likely to seek the services of physician specialists, therapists, and mental health professionals than are individuals who do not have a history of malignancy.2 Nonetheless, in general, the incidence of comorbid conditions, which also can affect health and quality of life, increases with increasing age. Thus, it may be the case that cancer's effects on health must be characterized in the context of other comorbid conditions. The current study compared a group of older cancer survivors with a community-based cohort of individuals who had no history of malignant disease; study participants were assessed in 1992 and then again 4 years later to evaluate the impact of various types of cancer history and comorbidity on healthcare use and mortality.

MATERIALS AND METHODS

Study Participants

Participants in the multisite National Institute of Aging Established Populations for Epidemiologic Study of the Elderly (EPESE) project who were recruited at Duke University (Durham, NC) were included in the current study. The EPESE project was designed to evaluate health and well-being and identify predictors of mortality and morbidity in older community-dwelling individuals. A probability sample of such individuals was obtained using a 4-stage stratified household sampling design that selected persons age ≥ 65 years who were living within a 5-county area (containing both urban and rural components) in the Piedmont region of North Carolina. African-American individuals were oversampled, accounting for 54% of the study population; the actual proportion of African Americans living in the community of interest was 35%.5-8 EPESE was constructed as a 6-year longitudinal study (1986–1992) with annual interviews. An additional interview was performed at 10 years from baseline, in 1996. In all cases, informed consent was obtained from participants or their legal representatives.

Reference Time Points and Data on Patient Demographics and Habits

The 1992 interview (n = 2567) was chosen as the starting time point for the current analysis. This time point provided the most complete set of data on comorbid conditions and made it possible for us to examine cancer histories of various durations. The 1996 interview was used as the 4-year time point.

The sociodemographic factors and habits on which data were obtained included race, gender, education level (with a ceiling at 17 years), marital status, employment status, smoking status (in pack-years), and amount of alcohol consumed per month. For the purposes of the analysis, the 24 participants for whom educational information was not available were assigned the cohort's average level of education (8.76 years), and the 140 participants for whom data on smoking status were unavailable were treated as having 0 pack-years of smoking history.

Cancer History

In 1986, participants were asked whether a doctor had ever told them that they had a malignancy of any type. In subsequent annual interviews, participants were asked to report any cancer diagnoses made within the past year. If a diagnosis was reported, further inquiry was made into the tumor type (lung, breast, colon/bowel/rectum, lymphoma, leukemia, melanoma, nonmelanoma skin malignancy, or other malignancy).

Participants evaluated in 1992 were divided into four groups according to history of malignancy. Reports of ‘other skin cancers’ were excluded from the analysis. The recent cancer history group consisted of participants who had had malignancies diagnosed within the past year, the intermediate cancer history group consisted of participants who had had malignancies diagnosed between 1986 and 1991, and the remote cancer history group consisted of participants who had had malignancies diagnosed before 1986; participants without a diagnosis of malignant disease at any time were included in the no cancer history group.

Comorbid Conditions

Participants were asked annually between 1986 and 1992 whether they had experienced certain health problems—coronary atherosclerotic disease (CAD; heart attack, coronary/myocardial infarction, coronary thrombosis, or coronary occlusion), cerebrovascular disease (CVD; stroke or brain hemorrhage), hip fracture, or fracture of a bone other than the hip—or had been diagnosed with diabetes (as indicated by sugar in urine or high blood sugar) or hypertension.

Depression was assessed using the 20-item Center for Epidemiological Studies—Depression (CES-D) instrument9; participants reporting 9 or more affirmative answers on the CES-D were categorized as being at risk for depression. In addition, cognitive functioning was measured using the 10-item Short Portable Mental Status Questionnaire10; participants answering 3 or more questions incorrectly on the questionnaire were classified as cognitively impaired.

In 1992, participants were asked whether they were experiencing additional health conditions, including arthritis or rheumatism; osteoporosis; glaucoma; Parkinson disease; anemia or tired blood; emphysema, chronic bronchitis, asthma, or other lung disease; and cirrhosis or liver disease. For each illness that was present, participants were asked to rate the extent to which that illness interfered with their typical activities using a 3-point Likert scale (‘not at all’, ‘a little [some]’, or ‘a great deal’). Histories of these 15 comorbid conditions were included in the current analysis.

Functional Status Measures

Functional status was evaluated in 1992 by assessing participants' ability to perform activities included in the Activities of Daily Living (ADL) and Independent Activities of Daily Living (IADL) indexes. The ADL index measures an individual's ability to perform seven simple self-care tasks: walking, bathing, personal grooming, dressing, eating, toileting, and moving from bed to chair.11 The 6 participants for whom ADL scores were unavailable were assigned the cohort's average score (0.64). The IADL index evaluates an individual's ability to complete 5 more complex self-care tasks: telephone use, shopping, taking of medication, finance management, and community mobility (i.e., use of an automobile/other modes of transportation). The 145 participants for whom IADL scores were unavailable were assigned the cohort's average score (1.32).

Mortality and Healthcare Use Outcome Measures

Mortality was ascertained using data obtained by the National Death Registry through October 1999. At each annual interview, participants were asked to report all visits to the emergency room (ER) of a hospital, all hospitalizations extending at least overnight, and any time spent as a patient in a nursing home (NH) or rest home in the preceding year. Events occurring in 1992 were assessed for cross-sectional associations. The same questions were also asked in 1996, with ER visits, hospital, and NH admissions being treated as prospective outcomes.

Statistical Analysis

Bivariate analyses of sociodemographic factors, participant habits, comorbid conditions, and outcome measures were performed according to cancer history group. Categoric variables were evaluated using the chi-square test with Fisher exact analysis. Analysis of variance was used to compare continuous variables across the four cancer history groups. Two-sided P values < 0.05 were considered indicative of statistical significance.

Multivariate Models

With regard to healthcare use, logistic regression models for 1992 ER visitation and hospital admission were controlled for age, gender, race, education, marital status, employment status, pack-years of smoking, alcohol consumption, and ADL and IADL functional measures. The logistic regression model for 1992 NH admission was controlled for the same variables as above, excluding employment status and functional measures.

No participants were lost to follow-up in the prospective analyses of 1996 healthcare use and mortality. The models for ER visitation, hospital admission, and mortality were controlled for age, gender, race, education, marital status, employment status, pack-years of smoking, alcohol consumption, and ADL and IADL functional measures. The model for NH admission was controlled for the same variables as above, excluding employment status and functional measures.

Because moderate alcohol consumption is associated with decreased coronary and congestive heart failure events, a squared alcohol consumption term was included in all models.12

The final models for the analyses of 1992 healthcare use and mortality included 11 conditions: CAD, CVD, diabetes mellitus, hypertension, arthritis, Parkinson disease, lung disease, cirrhosis/liver disease, hip fracture, depression, and cognitive impairment. Glaucoma, anemia, osteoporosis, and bone fractures not involving the hip were removed from all models because of their limited impact on the analyses. Cirrhosis/liver disease was removed from the models for 1996 healthcare use and mortality due to the scarcity of participants with these conditions. Logistic regression–derived odds ratios (ORs) with 95% confidence intervals are presented. All analyses were performed using SAS statistical software (Version 8.0; SAS Institute, Cary, NC).

RESULTS

In 1992, the age of the 2567 participants ranged from 71 years to 102 years (mean, 79 years). Sixty-nine percent of all participants were women, and 55% of all participants were African American. Three-hundred sixty-nine participants reported having a history of malignancy; 42 had a recent history of malignancy, 127 had an intermediate history, and 200 had a remote history. The majority of participants (n = 2198) did not report having any history of malignancy.

Age distributions were similar across cancer history groups (Table 1), with unmarried patients and unemployed patients both accounting for more than half of each group. The relative proportion of women was smaller in the recent cancer history group than in the other three groups, whereas the relative proportion of African Americans was smaller in the no cancer history group than in the remaining three groups. The average educational level was lowest among patients without a history of malignancy (group mean, 8.6 years); this group also had the shortest average duration of tobacco use. Alcohol consumption was highest among patients with an intermediate history of malignancy.

Table 1. Participant Characteristics in 1992

Characteristic	History of malignancya				P valueb
	None (n = 2198)	Remote (n = 200)	Intermediate (n = 127)	Recent (n = 42)
Age (yrs)					0.303
Mean	78.7	78.9	77.8	78.5
Range	71–102	71–96	71–91	71–91
Female (%)	69.0	73.5	59.8	54.8	0.015c
African American (%)	42.4	61.0	64.6	54.8	< 0.0001c
Married (%)	31.7	32.5	33.9	45.2	0.298c
Employed (%)	8.2	4.5	3.9	11.9	0.052c
Mean education level (yrs) ± SD	8.6 ± 4.0	9.6 ± 4.1	9.5 ± 4.1	10.0 ± 4.6	0.0001
Mean tobacco use (pack-yrs) ± SD	12.6 ± 24.2	18.4 ± 34.8	20.7 ± 32.2	18.5 ± 25.9	< 0.0001
Mean alcohol consumption (drinks/mo) ± SD	2.5 ± 11.9	2.0 ± 10.7	6.1 ± 21.8	4.1 ± 10.8	0.0101
Mean ADL score ± SD	0.64 ± 1.40	0.60 ± 1.33	0.60 ± 1.29	1.00 ± 1.79	0.170
Mean IADL score ± SD	1.31 ± 1.81	1.35 ± 1.82	1.36 ± 1.87	1.32 ± 1.84	0.950

• SD: standard deviation; ADL: Activities of Daily Living index; IADL: Instrumental Activities of Daily Living index.
• a Remote history: diagnosis > 6 years before baseline; intermediate history: diagnosis 1–6 years before baseline; recent history: diagnosis < 1 year before baseline.
• b Omnibus test for group differences. P values calculated using analysis of variance unless otherwise noted.
• c P value calculated using chi-square Fisher exact test.

Because it was possible for participants to have more than 1 malignancy, a total of 475 malignancies were reported, with breast and colon malignancies being the ones most commonly observed (Table 2). The other cancers category included genitourinary (prostate, bladder, and kidney), gynecologic (uterine, cervical, and ovarian), gastrointestinal (stomach and intestinal), and head and neck malignancies.

Table 2. Distribution of Cancer Diagnosesa

Malignancy type	History of malignancyb
	Remote (%)	Intermediate (%)	Recent (%)
Lung	8 (4.0)	8 (6.3)	1 (2.4)
Breast	55 (27.5)	30 (23.6)	7 (16.7)
Colon	32 (16.0)	21 (16.5)	7 (16.7)
Leukemia	6 (3.0)	3 (2.4)	0
Lymphoma	4 (2.0)	2 (1.6)	1 (2.4)
Melanoma	16 (8.0)	9 (7.1)	3 (7.1)
Other	119 (59.5)	72 (56.7)	24 (57.1)

• a Due to the presence of patients with multiple malignancies, column percentages may sum to > 100%.
• b Remote history: diagnosis > 6 years before baseline; intermediate history: diagnosis 1–6 years before baseline; recent history: diagnosis < 1 year before baseline.

Participants experienced an average of three comorbid conditions. Differences with respect to only two conditions, CAD and lung disease, were noted across cancer history groups (Table 3). Five hundred fifty-three participants had CAD, with fewer in the no cancer history group than in any of the other groups. In addition, emphysema/lung disease, which was reported in 11% of all participants (n = 290), occurred with greater frequency in the intermediate cancer history group than in any of the other groups. Aside from these conditions, the most common comorbidities were hypertension and arthritis. One-hundred ninety-eight participants (7.7%) screened positively for depression, and 512 (20%) were found to be cognitively impaired; these conditions were similarly prevalent across cancer history groups.

Table 3. Prevalence of Comorbid Conditionsa

	History of malignancya				P valueb
	None (%)	Remote (%)	Intermediate (%)	Recent (%)
Hypertension	70.4	72.0	67.7	66.7	0.795
Arthritis	68.1	69.5	72.4	59.5	0.444
Fracture of bone other than hip	29.9	35.0	26.8	42.9	0.109
Diabetes mellitus	22.8	28.5	25.2	31.0	0.161
CAD	20.5	27.5	28.4	26.2	0.022
Cognitive impairment	20.7	16.5	16.5	9.5	0.121
CVD	13.6	15.0	19.7	11.9	0.258
Emphysema/lung disease	10.7	12.5	18.9	14.3	0.036
Glaucoma	9.4	11.5	10.2	14.3	0.484
Depression	7.3	10.5	8.7	11.9	0.231
Osteoporosis	6.4	9.0	7.9	7.1	0.423
Hip fracture	6.5	9.5	3.9	4.8	0.247
Anemia	5.0	5.0	6.3	9.5	0.428
Parkinson disease	1.4	1.5	3.2	0	0.376
Cirrhosis/liver disease	0.3	0.5	0.8	0	0.422

• CAD: coronary atherosclerotic disease; CVD: cerebrovascular disease.
• a Remote history: diagnosis > 6 years before baseline; intermediate history: diagnosis 1–6 years before baseline; recent history: diagnosis < 1 year before baseline.
• b Omnibus test for group differences. P values calculated using the chi-square Fisher exact test.

To investigate the effects of tobacco use on comorbid conditions, bivariate analysis of cancer history groups was performed after removing participants with a history of lung carcinoma. The prevalence of CAD continued to differ across groups (20.6%, 28.1%, 26.9%, and 24.4% in the no cancer history, remote cancer history, intermediate cancer history, and recent cancer history groups, respectively [P = 0.041]). In contrast, emphysema/lung disease became equally prevalent across cancer history groups (11.0%, 11.5%, 16.0%, and 12.2% in the no cancer history, remote cancer history, intermediate cancer history, and recent cancer history groups, respectively [P = 0.396]). Furthermore, differences across groups in terms of the prevalence of cognitive impairment (20.7%, 16.7%, 15.1%, and 7.3% in the no cancer history, remote cancer history, intermediate cancer history, and recent cancer history groups, respectively [P = 0.044]) arose.

There was a 96–100% rate of response to the question asking patients to report the degree to which comorbid conditions interfered with their typical activity. The majority of participants perceived at least some amount of interference caused by these conditions. With regard to conditions that had at least 10% prevalence, at least one-third of all participants felt that their arthritis, lung disease, glaucoma, and osteoporosis caused a great deal of interference, with little difference across cancer history groups (Fig. 1).

Figure 1

Degree of limitation caused by various comorbid conditions according to cancer history (Hx) group.

In 1992, there were 627 hospital admissions, with the largest proportion of these hospitalizations occurring in the recent cancer history group, whereas the 563 ER visits and 263 NH admissions documented occurred in equivalent proportions across cancer history groups (Table 4). In 1996, the 442 ER visits, 266 NH admissions, and 394 hospital admissions documented occurred in similar proportions across cancer history groups.

Table 4. 1992 and 1996 Healthcare Use Data

	History of malignancya				P valueb
	None (%)	Remote (%)	Intermediate (%)	Recent (%)
1992
ER visitation	21.3	24.5	28.4	23.8	0.207
Hospital admission	22.3	27.5	38.6	78.6	< 0.0001
NH admission	10.1	11.0	11.8	11.9	0.804
1996
ER visitation	22.7	26.8	25.0	34.8	0.334
Hospital admission	27.5	31.3	30.3	42.9	0.359
NH admission	15.2	21.3	15.0	21.7	0.243

• ER: emergency room; NH: nursing home.
• a Remote history: diagnosis > 6 years before baseline; intermediate history: diagnosis 1–6 years before baseline; recent history: diagnosis < 1 year before baseline.
• b Omnibus test for group differences. P values calculated using chi-square Fisher exact test.

The results yielded by the adjusted logistic regression models for 1992 healthcare use are summarized in Table 5. Conditions significantly associated with 1992 hospital admission included recent history of malignancy (OR, 15.5; 95% CI, 7.0–34.2), intermediate history of malignancy (OR, 2.1; 95% CI, 1.4–3.3), hypertension, CAD, and hip fracture. Arthritis and cognitive impairment were negatively associated with hospital admission. CAD, CVD, lung disease, and depression were significantly associated with ER visitation. NH admission was positively correlated with cognitive impairment, CVD, and hip fracture and negatively correlated with arthritis.

Table 5. 1992 Healthcare Use Odds Ratios According to History of Malignancy and Comorbidity

	Odds ratio (95% confidence interval)
	ER visitationa	Hospital admissiona	NH admissionb
Recent history of malignancyc	1.0 (0.5–2.3)	15.5 (7.0–34.2)	2.2 (0.7–7.0)
Intermediate history of malignancyc	1.4 (0.9–2.2)	2.1 (1.4–3.3)	1.5 (0.8–3.0)
Remote history of malignancyc	1.2 (0.8–1.7)	1.1 (0.8–1.6)	1.0 (0.6–1.7)
Hypertension	1.3 (1.0–1.6)	1.7 (1.4–2.2)	0.8 (0.6–1.1)
Arthritis	1.2 (0.9–1.5)	0.8 (0.6–0.9)	0.7 (0.5–0.9)
Diabetes mellitus	1.1 (0.8–1.4)	1.0 (0.8–1.3)	1.1 (0.8–1.6)
CAD	2.3 (1.8–2.9)	2.4 (1.9–3.0)	1.0 (0.7–1.4)
Cognitive impairment	1.1 (0.9–1.5)	0.6 (0.4–0.8)	7.7 (5.6–10.6)
CVD	1.4 (1.1–1.9)	1.2 (0.9–1.5)	1.8 (1.2–2.6)
Emphysema/lung disease	1.6 (1.2–2.2)	1.3 (0.9–1.8)	1.3 (0.8–2.1)
Depression	1.5 (1.1–2.1)	0.9 (0.7–1.4)	1.2 (0.7–2.1)
Hip fracture	1.1 (0.7–1.5)	2.0 (1.4–3.0)	3.3 (2.2–5.1)
Parkinson disease	1.0 (0.4–2.0)	0.7 (0.3–1.7)	1.9 (0.7–5.0)
Cirrhosis/liver disease	0.5 (0.1–2.9)	1.0 (0.2–5.1)	0.8 (0.7–8.4)

• ER: emergency room; NH: nursing home; CAD: coronary atherosclerotic disease; CVD: cerebrovascular disease.
• a Logistic regression models were controlled for age, gender, race, education, marital status, employment status, tobacco use, alcohol use, and functional status.
• b Logistic regression model was controlled for age, gender, race, education, marital status, tobacco use, and alcohol use.
• c Remote history: diagnosis > 6 years before baseline; intermediate history: diagnosis 1–6 years before baseline; recent history: diagnosis < 1 year before baseline.

The results of the adjusted logistic regression analysis of 1996 healthcare use are summarized in Table 6. NH admission was correlated with recent history of malignancy (OR, 3.1; 95% CI, 1.1–9.1), cognitive impairment, and CVD. ER visitation was significantly associated with arthritis, CAD, lung disease, hip fracture, and Parkinson disease. Finally, participants with arthritis, diabetes mellitus, CAD, or CVD had an elevated risk of hospital admission.

Table 6. 1996 Healthcare Use Odds Ratios According to History of Malignancy and Comorbidity

	Odds ratio (95% confidence interval)
	ER visitationa	Hospital admissiona	NH admissionb
Recent history of malignancyc	2.5 (1.0–6.1)	2.0 (0.8–5.0)	3.1 (1.1–9.1)
Intermediate history of malignancyc	1.0 (0.6–1.8)	1.0 (0.6–1.8)	1.0 (0.5–2.0)
Remote history of malignancyc	1.1 (0.7–1.6)	1.1 (0.7–1.7)	1.5 (0.9–2.5)
Hypertension	1.2 (0.9–1.6)	1.3 (1.0–1.7)	1.2 (0.8–1.7)
Arthritis	1.4 (1.1–1.8)	1.7 (1.3–2.2)	0.8 (0.6–1.1)
Diabetes mellitus	1.2 (0.9–1.5)	1.5 (1.2–2.0)	1.1 (0.8–1.6)
CAD	1.5 (1.1–2.0)	1.5 (1.1–2.0)	0.8 (0.5–1.2)
Cognitive impairment	1.1 (0.8–1.6)	1.0 (0.7–1.5)	5.3 (3.8–7.5)
CVD	1.4 (1.0–2.1)	1.5 (1.1–2.1)	1.7 (1.1–2.7)
Emphysema/lung disease	1.7 (1.2–2.4)	1.3 (0.9–1.9)	1.3 (0.8–2.2)
Depression	1.2 (0.8–1.8)	1.4 (1.0–2.0)	1.5 (0.9–2.4)
Hip fracture	1.8 (1.1–2.9)	0.9 (0.6–1.6)	1.3 (0.8–2.2)
Parkinson disease	2.6 (1.1–6.3)	0.7 (0.2–1.8)	1.7 (0.6–4.6)

• ER: emergency room; NH: nursing home; CAD: coronary atherosclerotic disease; CVD: cerebrovascular disease.
• a Logistic regression models were controlled for age, gender, race, education, marital status, employment status, tobacco use, alcohol use, and functional status.
• b Logistic regression model was controlled for age, gender, race, education, marital status, tobacco use, and alcohol use.
• c Remote history: diagnosis > 6 years before baseline; intermediate history: diagnosis 1–6 years before baseline; recent history: diagnosis < 1 year before baseline.

There were a total of 878 deaths in the current cohort. Bivariate analysis revealed equal mortality rates across cancer history groups during the 1992–1996 follow-up period. After 1996, fewer deaths were documented in the no cancer history group compared with the other three groups. Having a history of malignancy did not increase mortality risk in the adjusted model, however (Table 7).

Table 7. Mortality Data

	History of malignancya				P valueb
	None	Remote	Intermediate	Recent
No. of deaths (%)	729 (33.2)	81 (40.5)	48 (37.8)	20 (47.6)	0.034
1992–1996	539 (24.5)	54 (27.0)	34 (26.7)	14 (33.3)	0.482
1996–1999	190 (11.8)	27 (18.5)	14 (15.1)	6 (21.4)	0.028
Mortality riskc	1.0	1.4 (1.0–2.0)	1.1 (0.7–1.7)	2.0 (1.0–4.0)

• a Remote history: diagnosis > 6 years before baseline; intermediate history: diagnosis 1–6 years before baseline; recent history: diagnosis < 1 year before baseline.
• b Omnibus test for group differences. P values calculated using the chi-square Fisher exact test.
• c Logistic regression model was controlled for age, gender, race, education, marital status, employment status, tobacco use, alcohol use, and functional status. Values are odds ratios with 95% confidence intervals in parentheses.

DISCUSSION

The current study indicates that older individuals tend to experience a similar range of comorbid conditions regardless of whether they have a history of malignancy. Other studies assessing comorbid conditions at the time of cancer diagnosis in older patients have yielded results similar to those observed in the recent cancer history group in the current study. For example, Yancik et al.13 found that patients age 70–79 years who had breast carcinoma had an average of 3–4 comorbidities, with the most common conditions being hypertension, arthritis, and heart disease. In addition, in a study of women in Detroit, MI, who had breast carcinoma, Satariano et al.14 reported that the subset of patients with ≥ 3 comorbidities had a mean age of 70 years; yet again, the most common comorbid conditions in that subpopulation were hypertension, arthritis, and heart disease.

In the current study, the observed comorbidity distributions were similar for cancer survivors and patients with no history of malignancy, and a previous analysis also indicated that cancer diagnoses were not associated with an increased risk of developing other major comorbid conditions.7 Two other studies examining the presence of comorbid conditions in older cancer survivors, however, yielded different results. Schultz et al.15 investigated a cohort of cancer survivors (mean age, 48 years; mean time from diagnosis, 18 years) treated at the University of Texas M. D. Anderson Cancer Center (Houston, TX); in comparing cancer survivors age > 65 years with a gender- and age-matched group of 1997 National Health Interview Survey (NHIS) respondents who had no history of malignancy, those investigators found that the cancer survivors had a lower incidence of 6 comorbid conditions (arthritis, diabetes, migraine headache, heart disease, cataracts, and hearing loss). In a larger study, Hewitt et al.2 separated 1998–2000 NHIS respondents into those who had a history of malignancy and those who did not and found that patients age > 65 years who did have a history of malignancy had chronic conditions (cardiovascular disease, diabetes, emphysema, asthma, ulcer, weak kidneys, and liver disease) more frequently compared with their age-matched counterparts who had no history of malignancy.

When all reported lung malignancies were disregarded, the prevalence of lung disease equalized across the four cancer history groups, whereas significant differences in the prevalence of CAD and cognitive impairment persisted; these findings suggest that tobacco use may be associated with CAD and cognitive impairment. Only 17 lung malignancies were reported, however, and as a consequence, these results may be underpowered.

For participants in the current study who had more recent histories of malignancy, the presence of malignant disease was associated with hospital admission around the time of diagnosis. One could postulate that this association is consistent with the need to manage the symptoms of, diagnose, or treat the malignancy in question. Stafford et al.,16 examining healthcare use in older patients with cancer, surveyed 9745 Medicare enrollees in 1991. In that study, cancer diagnoses were found to be associated with the largest annual Medicare reimbursements; however, patients with other chronic conditions received reimbursements that were similarly large. Patients with cancer diagnoses also had higher rates of physician visitation and more days of inpatient hospitalization compared with patients who did not have cancer, although these differences became less pronounced when patients with cancer diagnoses were compared with patients who had other chronic illnesses. Those investigators did not specify when each cancer diagnosis was made, and consequently, it is difficult to separate recently diagnosed patients from survivors in that study. The NHIS-based study conducted by Hewitt et al.2 found that cancer survivors age > 65 years, compared with their counterparts who had no history of malignancy, were more likely to visit physician specialists and therapists but similarly likely to visit general practitioners.

When cancer survivors were analyzed at 4 years from baseline, NH admission emerged as the only type of healthcare use that varied with varying cancer history. This suggests that among older cancer survivors, healthcare use is driven by other comorbid conditions and not by history of malignancy, despite the finding that cancer survivors may use healthcare services more regularly compared with their counterparts who have no history of malignancy.2, 17 Nonetheless, with approximately 25% of all participants in each cancer history group dying between 1992 and 1996, the diminished size of the recent cancer history group may have made it impossible to detect differences in healthcare use over that 4-year period. It is possible that an annual analysis of healthcare use would have revealed different usage patterns, especially near the time of death. Thus, the association between NH admission and history of malignancy may simply reflect a cancer survivorship effect.

More deaths were documented among cancer survivors compared with individuals who had no history of malignancy. Nonetheless, in the adjusted analysis, time of cancer diagnosis was not correlated with overall mortality, whereas the comorbid conditions CAD, lung disease, and cognitive impairment were associated with increased mortality risk.

Aside from the data obtained from death registries, self-reported data represent the basis of the current study, and thus there is a risk that diagnoses and dates were reported erroneously; however, other studies suggest a reasonably good correlation between self-reported data and data found in medical records, especially when the self-reported data are obtained via interview and pertain to cancer history.18, 19 Another potential limitation of the current study is the lack of additional malignancy-related information (e.g., stage at diagnosis, treatment received, and presence or absence of recurrent disease) that could influence morbidity and mortality; future studies using Medicare payment data or other prospectively collected data may address this limitation.

In conclusion, the current study, which involved a community-based population of older individuals, set out to assess whether older persons with various types of cancer history had comorbid conditions and healthcare use patterns that differed from those of their counterparts who had no history of malignancy. We ultimately found that associations between comorbid conditions and healthcare use and mortality were independent of cancer history. Geriatric assessment and management of older patients with cancer may help to identify comorbidities and direct treatment so as to minimize the effects of these comorbidities and reduce healthcare use. Thus, for older individuals with a history of malignancy extending back by > 1 year, healthcare-related decisions should continue to be made with a focus on attenuating comorbid conditions to improve functional status and health-related outcomes.