Substance use disorder and its effects on outcomes in men with advanced-stage prostate cancer

Authors

  • Sumedha Chhatre PhD,

    Corresponding author
    1. Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania
    • Corresponding author: Sumedha Chhatre, PhD, 3535 Market Street, Suite 4051, Philadelphia PA 19104; Tel: (215) 746-7348; rasu@mail.med.upenn.edu

    Search for more papers by this author
  • David S. Metzger PhD,

    1. Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for more papers by this author
  • S. Bruce Malkowicz MD,

    1. Department of Surgery, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for more papers by this author
  • George Woody MD,

    1. Department of Psychiatry, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Ravishankar Jayadevappa PhD

    1. Department of Surgery, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania
    2. Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania
    Search for more papers by this author

Abstract

BACKGROUND

Substance use disorder in patients with cancer has implications for outcomes. The objective of this study was to analyze the effects of the type and timing of substance use on outcomes in elderly Medicare recipients with advanced prostate cancer.

METHODS

This was an observational cohort study using Surveillance, Epidemiology, and End Results (SEER)-Medicare linked data from 2000 to 2009. Among men who were diagnosed with advanced prostate cancer between 2001 and 2004, we identified those who had a claim for substance use disorder in the year before cancer diagnosis, 1 year after cancer diagnosis, and an additional 4 years after diagnosis. The outcomes investigated were use of health services, costs, and mortality.

RESULTS

The prevalence of substance use disorder was 10.6%. The category drug psychoses and related had greater odds of inpatient hospitalizations (odds ratio [OR], 2.3; 95% confidence interval [CI], 1.9-2.8), outpatient hospital visits (OR, 2.6; 95% CI, 1.9-3.6), and emergency room visits (OR, 1.7; 95% CI, 1.2-2.4). Substance use disorder in the follow-up phase was associated with greater odds of inpatient hospitalizations (OR, 2.0; 95% CI, 1.8-2.2), outpatient hospital visits (OR, 2.0; 95% CI, 1.7-2.4), and emergency room visits (OR, 1.7; 95% CI, 1.5-2.1). Compared with men who did not have substance use disorder, those in the category drug psychoses and related had 70% higher costs, and those who had substance use disorder during the follow-up phase had 60% higher costs. The hazard of all-cause mortality was highest for patients in the drug psychoses and related category (hazard ratio, 1.3; 95% CI, 1.1-1.7) and the substance use disorder in treatment phase category (hazard ratio, 1.5; 95% CI, 1.3-1.7).

CONCLUSIONS

The intersection of advanced prostate cancer and substance use disorder may adversely affect outcomes. Incorporating substance use screening and treatments into prostate cancer care guidelines and coordination of care is desirable. Cancer 2014;120:3338–3345. © 2014 American Cancer Society.

INTRODUCTION

The number of Americans aged >65 years is expected to double between 2000 and 2050. Because prostate cancer incidence increases exponentially with advancing age, there may be a surge in the number of older patients with prostate cancer,[1] which will pose a challenge to our health care system. Like most elderly patients, those with prostate cancer likely will acquire a higher number of maladies, both physical and psychosocial, with advancing age. Substance use disorder is an important comorbidity. An estimated 23.9 million individuals in the United States were current drug users in 2012.2 Prescription drug use is a recent trend in substance use in the United States and is now the second most common form of illegal substance use.[2] The exact prevalence of substance use disorder in the elderly is uncertain. However, estimates indicate that, by 2020, there will be approximately 4.5 million older adults with substance use disorder.[3, 4] Health problems related to substance use disorder can reach unprecedented levels in the Baby Boomer generation as it reaches Medicare eligibility. For the first time, compared with younger adults, the proportion of older adults seeking treatment for substance use disorder is on the rise.[5]

Despite these trends, the issue of substance use disorder remains understudied in cancer treatment,[6] and patients' alcohol and drug use assessments continue to be sketchy.[7] Advanced prostate cancer care involves combinations of drugs, surgery, radiation therapy, and palliative care.[8] Many patients with advanced prostate cancer experience cancer-related pain and impaired outcomes.[9, 10] The intersection of aging and advanced disease stage may exacerbate the potential for substance use, leading to adverse outcomes.[3, 6, 11-14] The objective of this study was to analyze the prevalence and modifying effects of substance use disorder on the use of health services, the cost of care, and mortality in elderly, fee-for-service Medicare patients with advanced prostate cancer. We hypothesized that the timing and the type of substance use disorder would have varying effects on outcomes.

MATERIALS AND METHODS

Data

The Surveillance, Epidemiology, and End Results (SEER)-Medicare data are the linkage of 2 large, population-based data sources and provide detailed information about Medicare beneficiaries with cancer. The SEER Program of the National Cancer Institute collects data on cancer incidence, treatment, and mortality from 16 SEER sites and encompasses 26% of the US population. Of individuals aged ≥65 years diagnosed with cancer who are enrolled in SEER registries, 93% have a match in Medicare enrollment records.[15]

Study Cohort

For this retrospective case-control study, we used SEER-Medicare–linked data to create a cohort of men aged ≥66 years who were diagnosed with advanced prostate cancer between 2001 and 2004. The local institutional review board approved this study. Patients with advanced prostate cancer were identified from the SEER Patient Entitlement and Diagnosis Summary (PEDSF) file by selecting regional or distant codes for the variable summary stage 2000 (summ2k1). The summary stage variable is derived from collaborative stages for 2004 and later and, before that, from extent of disease and is used in most SEER publications.[16] We excluded men who were aged <66 years at the time of diagnosis to ensure there were sufficient claims for medical care before the diagnosis of prostate cancer, because these claims were necessary to adjust for prediagnosis comorbidity. With the date of prostate cancer diagnosis as the anchor, we defined the following phases of care: the prediagnosis phase is defined as the year before prostate cancer diagnosis, the cancer treatment phase is the 1-year period after prostate cancer diagnosis, and the follow-up phase is the period up to 4 years beyond the treatment phase.

Type of Substance Use Disorder

The key independent variable in our analyses was substance use disorder. Substance use disorder was defined using International Classification of Diseases, 9th edition (ICD-9) codes 291.xx (alcoholic psychosis and related), 292.xx (drug psychoses and related), 303.xx (alcohol dependence syndrome), 304.xx (drug dependence), and 305.xx (nondependent use of drugs). Among our cohort of patients with advanced prostate cancer, we identified those who had at least 1 Medicare inpatient or outpatient claim for any of these codes.

Timing of Substance Use Disorder

We created 3 exclusive categories based on the time that substance use disorder was identified: the prediagnosis phase, the treatment phase, and the follow-up phase.

Outcomes

The main outcomes of our study were the use of health services, the cost of care, and mortality. The SEER-Medicare–linked data yield information on inpatient hospitalization, outpatient hospital visits (including emergency room [ER] visits), physician or provider services, durable medical equipment, home health services, skilled nursing facility use, and hospice care. For this study, we defined health service use as the number of inpatient hospitalizations, outpatient hospital visits, and ER visits. We operationalized the cost of care as reimbursements received from Medicare.[17-19] Total costs were calculated as the sum of reimbursements from inpatient hospitalizations, outpatient hospital visits, physician or provider services, durable medical equipment, home health services, and hospice care. Using the SEER PEDSF file, we obtained all-cause mortality data reported by both SEER and Medicare. For SEER-reported mortality, we obtained the date of death using the SEER month of death and the SEER year of death. Because SEER does not report the day of death, we assigned the middle of the month (ie, the 15th) as the day of death. For Medicare-reported mortality, the Medicare day, month, and year of death yielded the date of death. A patient was coded as dead if SEER and/or Medicare reported that the patient had died. The time to death was calculated as the time between prostate cancer diagnosis and death. We censored those patients who were alive at the end of 5-year follow-up. The variable SEER cause-specific death classification was used to establish whether the death was prostate cancer-specific.

Covariates

From the PEDSF file, we used sociodemographic characteristics, disease severity, medical comorbidity, and type of prostate cancer treatment to adjust our measures of association. The Elixhauser comorbidity index was derived using Medicare inpatient claims for the 1-year period before prostate cancer diagnosis.[20] The procedure codes helped identify prostate cancer treatments from Medicare claims and lead to following the exclusive treatment categories: surgery, radiation therapy, multimodal therapy, and no treatment/watchful waiting.

Statistical Analysis

First, we tested for differences in the demographic and clinical characteristics of patients with and without substance use disorder from our cohort of elderly patients with advanced prostate cancer using t tests and chi-square tests, as appropriate. For assessing health service use, negative binomial models were used.[21] The dependent variables were the number of total inpatient hospitalizations, outpatient hospital visits, and ER visits. To analyze the association of substance use disorder with the cost of care, we used a 2-part model. The first part consisted of logistic regression to model the odds of incurring any cost. For the second part, we used nonzero costs to model the association between substance use and cost using a generalized linear model with log-link and gamma distribution variance function.[21] For analyzing the association between substance use disorder and mortality, we used Cox regression models. We used 3 sequential sets of models to study the effect of substance use disorder on health service use, cost of care, and mortality. In Model 1 we estimated the unadjusted association of substance use disorder with outcomes. In Model 2, we adjusted for sociodemographic attributes and Elixhauser comorbidity score.[20, 22] The type of treatment for prostate cancer may affect outcomes; however, in our assessment of the relation between substance use disorder and outcome, treatment assignments were nonrandom. To minimize the bias caused by type of prostate cancer treatment, we used propensity score analysis. Using multinomial logistic regression, we estimated the propensity of receiving a particular prostate cancer treatment for each patient as a function of demographic and clinical characteristics.[23] We compared the t statistics for these covariates before and after adjusting with the propensity score to observe the extent to which the different prostate cancer treatment groups matched. In Model 3, we used type of treatment and the propensity score as additional covariates. In all our analyses, the reference category was ‘those without a substance use disorder.’ We used the software package Statistical Analysis System (SAS), version 9.3 (SAS Institute Inc., Cary, NC) for data analysis.

RESULTS

We identified 14,277 fee-for-service Medicare patients aged ≥66 years who were diagnosed with advanced-stage prostate cancer between 2001 and 2004. From this cohort, we identified 1509 patients (10.6%) who had a diagnosis of substance use disorder. The prevalence of substance use disorder was 1.8% in the pretreatment phase, 4.2% in the cancer treatment phase, and 4.6% in the follow-up phase. Because the frequencies of 2 categories of substance use disorder—alcoholic psychosis and related and drug dependence—were very small, we excluded them from analysis. Therefore, the 3 major categories of substance use disorder that were used in our analysis were: drug psychoses and related (n=136), alcohol dependence syndrome (n=142), and nondependent use of drugs (n=1201).

In Table 1, we present the comparison between patients with and without substance use disorder in our cohort of men with advanced prostate cancer. Compared with patients who did not have a substance use disorder, those with a substance use disorder were younger (mean age±standard deviation: 74.5±5.6 years vs 72.4±5.6 years) and included a greater proportion of African Americans (11.3% vs 17.6%). Those with substance use disorder were less likely to be married and to be from a metropolitan area. The comparison of clinical characteristics indicated that those with substance use disorder had more medical comorbidities and were more likely to have a higher grade of prostate cancer compared with those who did not have a substance use disorder. Finally, those with a substance use disorder were more likely to have received multimodal treatment and were less likely to have undergone surgery alone compared with those who did not have a substance use disorder.

Table 1. Comparison of Demographic and Clinical Characteristics, n = 14,277
 No. of Patients (%) 
CharacteristicWithout a Substance Use Disorder, n = 12,768With a Substance Use Disorder, n = 1509P
  1. Abbreviations: SD, standard deviation.

Age at diagnosis: Mean ± SD, y74.5 ± 6.972.4 ± 5.6<.0001
Census track median income: Mean ± SD, $43,026 ± 16,62337,458 ± 15,923<.0001
Ethnicity   
White10,208 (79.9)1127 (74.7)<.0001
African American1430 (11.2)266 (17.6) 
Hispanic1130 (8.9)116 (7.7) 
Marital status   
Married9038 (70.8)900 (59.6)<.0001
Single/separated/divorced3185 (24.9)537 (35.7) 
Unknown545 (4.3)71 (4.7) 
Geographic area   
Metropolitan11,190 (87.7)1237 (79.8)<.0001
Nonmetropolitan1570 (12.3)304 (20.2) 
Elixhauser comorbidity index   
011,446 (89.7)1228 (81.4)<.0001
≥11327 (10.4)282 (18.7) 
Grade   
Moderately differentiated4242 (33.2)475 (31.5)<.0001
Poorly differentiated6232 (48.8)752 (49.8) 
Other2294 (17.9)282 (18.7) 
Treatment   
Surgery2902 (22.7)273 (18.1)<.0001
Radiation1252 (9.8)144 (9.5) 
Multimodal5803 (45.5)893 (59.2) 
No treatment/watchful waiting2811 (22)199 (13.2) 

Table 2 details the unadjusted outcomes across different categories of substance use disorder. The highest proportion of patients in the drug psychoses and related category had used both inpatient services and outpatient services. This group also had more inpatient and outpatient episodes, higher costs of care, higher all-cause mortality, and higher prostate cancer-specific mortality. Conversely, those in the nondependent use of drug category had higher ER use.

Table 2. Unadjusted Health Service Use, Costs, and Mortality Outcomes, n = 14,247
 Percentage of Patients
VariableWithout a Substance Use Disorder, n = 12,768Nondependent Use of Drugs, n = 1201Drug Psychoses and Related, n = 136Alcohol Dependence Syndrome, n = 142
  1. Abbreviations: ER, emergency room; SD, standard deviation.

Any health service use    
Inpatient hospitalizations    
040.215.813.214.8
1-347.257.648.553.5
≥412.626.638.231.7
Outpatient hospital visits    
055.932.823.528.2
1-324.834.125.735.9
≥419.333.150.735.9
ER visits    
048.528.638.240.9
1-316.620.526.523.2
≥434.950.935.335.9
Total costs: Mean ± SD (median), $44,345 ± 86,192 (18,141)73,383 ± 110,766 (39,973)76,654 ± 112,827 (34,381)65,420 ± 79,492 (39,527)
Mortality    
All-cause41.741.465.454.9
Prostate cancer-specific28.525.939.728.9

Association Between Substance Use Disorder and Health Service Use

Inpatient hospitalizations

Compared with patients who did not have a substance use disorder, all types of substance use categories were associated with higher inpatient use, as indicated in Table 3. In particular, patients in the drug psychoses and related category had the most inpatient hospitalizations (odds ratio [OR], 2.3; 95% confidence interval [CI], 1.9-2.8). Those who had substance use disorder during the follow-up-phase had the most inpatient hospitalizations (OR, 2.0; 95% CI, 1.8-2.2).

Table 3. Association Between Health Service Use and Substance Use
 OR (95% CI)a
VariableHospitalizationsOutpatient VisitsER Visits
  1. Abbreviations: CI, confidence interval; ER, emergency room; OR, odds ratio.

  2. a

    ORs and 95% CIs were calculated using negative binomial models.

  3. b

    For all models, the reference category is those without a substance use disorder.

Type of substance use disorderb   
Model 1: Unadjusted   
Alcohol dependence syndrome2.0 (1.6-2.4)1.6 (1.2-2.3)1.1 (0.70-1.5)
Drug psychoses and related2.4 (2.0-2.9)2.7 (1.9-3.8)1.3 (0.89-1.8)
Nondependent use of drugs1.8 (1.7-1.9)1.7 (1.5-1.9)1.4 (1.5-1.8)
Model 2: Adjusted for sociodemographic characteristics and geographic area   
Alcohol dependence syndrome1.9 (1.6-2.3)1.8 (1.3-2.5)1.1 (0.76-1.5)
Drug psychoses and related2.3 (1.9-2.8)2.7 (1.9-3.8)1.3 (0.90-1.8)
Nondependent use of drugs1.7 (1.6-1.9)1.8 (1.6-2.1)1.5 (1.3-1.7)
Model 3: Adjusted for treatment, clinical characteristics, and propensity score   
Alcohol dependence syndrome1.9 (1.6-2.3)1.8 (1.3-2.4)1.2 (0.85-1.7)
Drug psychoses and related2.3 (1.9-2.8)2.6 (1.9-3.6)1.7 (1.2-2.4)
Nondependent use of drugs1.7 (1.6-1.8)1.8 (1.6-2.0)1.5 (1.3-1.7)
Period in which substance use disorder was identifiedb   
Model 1: Unadjusted   
Preprostate cancer diagnosis1.7 (1.5-1.9)1.8 (1.4-2.2)1.1 (0.86-1.4)
Cancer treatment phase1.8 (1.6-1.9)1.8 (1.5-2.1)1.5 (1.2-1.7)
Follow-up phase2.0 (1.8-2.2)1.7 (1.4-2.1)1.8 (1.5-2.2)
Model 2: Adjusted for sociodemographic characteristics and geographic area   
Preprostate cancer diagnosis1.4 (1.2-1.6)1.6 (1.3-2.0)1.1 (0.90-1.4)
Cancer treatment phase1.8 (1.6-1.9)2.0 (1.7-2.3)1.4 (1.1-1.6)
Follow-up phase2.1 (1.9-2.3)2.0 (1.7-2.4)1.7 (1.4-2.0)
Model 3: Adjusted for treatment, clinical characteristics, and propensity score   
Preprostate cancer diagnosis1.4 (1.2-1.6)1.6 (1.3-1.9)1.1 (1.0-1.4)
Cancer treatment phase1.8 (1.5-1.9)2.0 (1.7-2.3)1.4 (1.2-1.6)
Follow-up phase2.0 (1.8-2.2)2.0 (1.7-2.4)1.7 (1.5-2.1)

Outpatient hospital visits

For outpatient hospital visits, we observed a pattern similar that for inpatient hospitalizations. Patients in the drug psychoses and related category had the most outpatient visits (OR, 2.6; 95% CI, 1.9-3.6). Substance use disorder identified in the treatment phase (OR, 2.0; 95% CI, 1.7-2.3) and in the follow-up phase (OR, 2.0; 95% CI, 1.7-2.4) had a statistically significant association with more outpatient visits.

ER visits

Compared with patients who did not have a substance use disorder, those in the drug psychoses and related category had more ER visits (OR, 1.7; 95% CI, 1.2-2.4). Patients who had substance use disorder in the follow-up phase had the most ER visits (OR, 1.7; 95% CI, 1.5-2.1).

Association Between Substance Use Disorder and Cost of Care

Table 4 presents results from the 2-part models for cost of care. Part 1 consists of logistic models in which the binary dependent variable was any cost. Compared with the reference category of those without a substance disorder, those with all types of substance use disorder had higher odds of incurring any cost. In particular, the categories alcohol dependence syndrome and nondependent use of drugs had 4 times higher odds of incurring any cost. With regard to timing, substance use disorder in the cancer treatment phase had very high odds of incurring any cost (OR, 7.6; 95% CI, 4.7-12.0).

Table 4. Association Between, Cost, Mortality, and Substance Use
 Two-Part Model: Total Costa 
VariablePart 1: OR (95% CI)Part 2: OR (95% CI)All-Cause Mortality: HR (95% CI)b
  1. Abbreviations: CI, confidence interval; HR, hazard ratio; OR, odds ratio.

  2. a

    For the 2-part model, part 1 is a logistic model, and part 2 is a generalized linear model (gamma distribution with log-link).

  3. b

    All-cause mortality was calculated using a Cox proportional hazards model.

  4. c

    For all models, the reference category is those without a substance use disorder.

Type of substance use disorderc   
Model 1: Unadjusted   
Alcohol dependence syndrome4.6 (2.3-9.5)1.2 (0.99-1.5)0.97 (0.66-1.4)
Drug psychoses and related3.9 (2.0-7.8)1.4 (1.2-1.8)1.8 (1.5-2.7)
Nondependent use of drugs4.9 (3.8-6.4)1.4 (1.3-1.5)0.68 (0.58-0.80)
Model 2: Adjusted for sociodemographic characteristics and geographic area   
Alcohol dependence syndrome4.6 (2.3-9.5)1.2 (1.0-1.5)0.89 (0.61-1.3)
Drug psychoses and related3.6 (1.8-7.2)1.5 (1.2-1.9)1.3 (1.0-1.6)
Nondependent use of drugs4.7 (3.7-6.1)1.3 (1.2-1.4)0.80 (0.68-0.94)
Model 3: Adjusted for treatment, clinical characteristics, and propensity score   
Alcohol dependence syndrome4.1 (1.9-8.5)1.2 (1.0-1.5)0.84 (0.57-1.2)
Drug psychoses and related3.4 (1.7-6.8)1.7 (1.4-2.1)1.3 (1.1-1.7)
Nondependent use of drugs4.3 (3.3-5.6)1.3 (1.2-1.4)0.82 (0.69-0.96)
Period in which substance use disorder was identifiedc   
Model 1: Unadjusted   
Preprostate cancer diagnosis3.0 (2.1-4.4)1.1 (0.95-1.3)1.8 (1.6-2.1)
Cancer treatment phase8.3 (5.5-13.2)1.2 (1.1-14)1.1 (1.0-1.3)
Follow-up phase4.4 (3.1-6.2)1.7 (1.5-1.9)0.14 (0.09-0.19)
Model 2: Adjusted for sociodemographic characteristics and geographic area   
Preprostate cancer diagnosis2.5 (1.7-3.8)1.1 (0.94-1.2)1.4 (1.2-1.6)
Cancer treatment phase8.1 (5.1-12.8)1.2 (1.1-1.3)1.3 (1.1-1.5)
Follow-up phase4.4 (3.1-6.3)1.6 (1.4-1.8)0.18 (0.12-0.25)
Model 3: Adjusted for treatment, clinical characteristics, and propensity score   
Preprostate cancer diagnosis2.3 (1.6-3.4)1.1 (0.95-1.2)1.2 (1.1-1.4)
Cancer treatment phase7.6 (4.7-12.0)1.2 (1.1-1.3)1.5 (1.3-1.7)
Follow-up phase3.9 (2.7-5.6)1.6 (1.4-1.7)0.18 (0.13-0.26)

Part 2 consists of generalized linear model models with log-link and gamma distribution for those with nonzero costs. Costs were higher for all types of substance use. Patients in the drug psychoses and related category had 70% higher costs that those without a substance use disorder. In addition, patients who had substance use disorder in the follow-up phase had 60% higher costs compared with those in the reference category.

Association Between Substance Use Disorder and Mortality

All-cause mortality

Kaplan-Meier survival curves are provided in Figures 1 and 2. The log-rank statistic (P<.0001) indicates a difference in survival probabilities between various substance use categories. Similarly, log-rank statistics indicate differences in the probability of survival according to the timing of substance use disorder (P<.0001). The category of drug psychoses and related had the highest hazard of all-cause mortality (HR, 1.3; 95% CI, 1.1-1.7). The hazard of mortality was lower for the category nondependent use of drugs (HR, 0.82; 95% CI, 0.69-0.96). The hazard of mortality also was high for substance use disorder in the pretreatment phase and in the treatment phase (hazard ratio: 1.2 [95% CI, 1.1-1.4] and 1.5 [95% CI, 1.3-1.7], respectively). We observed comparable results from an analysis that focused on prostate cancer-specific mortality (data not shown).

Figure 1.

This Kaplan-Meier curve illustrates survival according to the type of substance use.

Figure 2.

This Kaplan-Meier curve illustrates survival according to duration of substance use. TX indicates treatment.

DISCUSSION

Our results provide important evidence regarding the intersection of age, advanced prostate cancer, and substance use disorders. The prevalence of substance use disorder in our cohort of elderly fee-for-service Medicare patients with advanced prostate cancer was 10.6%. Another important observation was that the type and timing of substance use disorder had a strong association with outcomes. Although it was not observed most frequently, the category drug psychoses and related had the strongest association with health service use, cost of care, and all-cause mortality. In addition, substance use disorder identified after the diagnosis of prostate cancer had the highest impact on health service use and cost of care. One possible explanation is that pain and psychiatric comorbidities, especially anxiety and depression, may influence substance use during this period, and the interaction of these factors ultimately may affect outcomes.[24-26] Past and current substance use disorder in cancer patients can affect their treatment and pain management and, thus, complicate the management of their disease.[14, 27] A study of substance use disorder in patients with chronic myelogenous leukemia or myelodysplastic syndrome indicated that lifetime cocaine use was associated with a 6-fold increased risk of death.[6]

We note some limitations of our study. SEER-Medicare–linked data have been used to study cancer-related health services and costs, including for prostate cancer.[15, 17, 18] In our study, the sample consisted only of men aged ≥66 years who lived in a SEER region and were fee-for-service Medicare enrollees. The SEER-Medicare–linked database does not include patients with Medicare Advantage or Part C coverage. Furthermore, although the age and sex distribution for individuals aged ≥66 years is comparable to that of older adults in the United States, the SEER regions have a higher proportion of nonwhite individuals. Mortality rates derived from SEER data may not be representative of national data on cancer mortality rates.[15] Administrative data have become an important source of information for public health and health services research but are subject to error.[28] Our sample did not capture men from the group ages 50 to 64 years, who also are at risk for prostate cancer. We also did not include surrogate alcohol disorders as part of our substance use disorder group. Finally, substance use disorder may be underreported in the Medicare claims, leading to conservative estimates of the association of substance use disorder with health services use, costs of care and mortality.

In conclusion, substance use disorder among elderly Medicare patients with prostate cancer can pose unique challenges to the delivery of effective and efficient care. Substance use disorder can be multifaceted and chronic; therefore, it demands intense management similar to that for cancer. There is an urgent need to identify the missed opportunities for identifying and treating substance use disorder among elderly men with prostate cancer. As a first step, guidelines for prostate cancer care can incorporate screening for substance use disorder as a recommendation. Enhanced screening and coordinated care may help with the early identification of elderly prostate cancer patients who have substance use disorder, facilitate appropriate treatments and clinical management, and, thus lead to improved outcomes.

FUNDING SUPPORT

This work was supported by a grant from the Department of Defense (W81XWH-12-1-0089 PC110707) and by a grant from the National Institute on Aging, National Institutes of Health (R21AG034870-01A1).

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary