Over the next 20 yr, the burden of cancer for older adults (age ≥65 yr) will increase (Smith, Smith, Hurria, Hortobagyi, & Buchholz, 2009). By 2030, almost 20% of the U.S. population, or approximately 72 million people, will be age ≥65 yr (Administration on Aging, 2011), and 70% of all cancers will be diagnosed within this age group (Smith et al., 2009). Older adults are at greater risk of suffering adverse consequences of cancer and its treatments (Parry, Kent, Mariotto, Alfano, & Rowland, 2011). For example, they are more likely to report having fair or poor health during and after cancer treatment, and their quality of life declines after a diagnosis of cancer, regardless of cancer type (Mohile et al., 2009; Reeve et al., 2009). After treatment, many older adults are unable to return to their previous levels of activity, a situation that decreases their quality of life (Courneya & Friedenreich, 1997; Courneya et al., 2003) and increases mortality and morbidity (Extermann & Hurria, 2007). Moreover, older adults who report daily fatigue, a common symptom of cancer and cancer treatment, are more likely to report depression and experience pain and least likely to report this symptom to their practitioner (Curt et al., 2000).

One possible explanation for older cancer survivors experiencing decreased quality of life is that this population has more limitations in both activities of daily living (ADLs) and instrumental activities of daily living (IADLs) than its younger counterparts (Mohile et al., 2009; Stafford & Cyr, 1997). With the growth in Medicare beneficiaries who have cancer, as well as the advent of health care reform, the need to identify services that effectively improve older adults’ quality of care and quality of life will become increasingly important. Occupational therapy has the potential to increase participation in daily activities, improve quality of care and, ultimately, enhance quality of life for adults with cancer (Campbell, Pergolotti, & Blaskowitz, 2009; Clark et al., 1997; Lloyd & Coggles, 1988; Lyons et al., 2011; Palmadottir, 2010). However, little is known about the use of occupational therapy services among the growing number of older adults with cancer.

Research on health services use should begin by examining patterns—how services are used, under what conditions, and by whom (Andersen & Newman, 2005). To date, health services research examining the patterns of use of occupational therapy is scant and is typically bundled with other rehabilitative services such as physical therapy (Cook, Stickley, Ramey, & Knotts, 2005; Freburger & Konrad, 2002). Instead, research is needed to understand large-scale utilization of occupational therapy and the effectiveness of these services for older adults with cancer (Bass-Haugen, 2009; Braveman & Bass-Haugen, 2009; Morello, Giordano, Falci, & Monfardini, 2009).

The Behavioral Model of Health Services Use (Andersen, 1968) is the most commonly used model for predicting health service use (Babitsch, Gohl, & von Lengerke, 2012). The Andersen model considers both the individual and contextual levels by examining three types of factor: (1) predisposing (propensity of individuals to use services), (2) enabling (resources to access services), and (3) need (illness level; Andersen, 1995). Inequitable access (disparity) occurs when a predisposing factor (e.g., race) and enabling resources (e.g., income) determine who gets health care instead of need variables (Babitsch et al., 2012).

The Andersen model conceptualizes the complex nature of utilization and has been widely used to shape related inquiry. We used this model to examine differences between users and nonusers of occupational therapy in older adults with cancer and considerable variations in patterns of occupational therapy service use between groups. In particular, we focused on variations in patterns of occupational therapy service use by age, sex, race, cancer type, and stage of cancer (see Figure 1). Because of differences between sex, race, geographic location, and functional abilities in use of health care services and postacute rehabilitation for other diseases, we hypothesized there would be similar differences within the population of people with cancer (Fisher et al., 2003; Ottenbacher et al., 2008). Specifically, we hypothesized that Medicare beneficiaries with cancer who used occupational therapy would be more likely to be White women living in large urban counties, where access to an academic center is more likely to occur; to have breast cancer (the most common cancer type in North Carolina; Carpenter, Yeh, Wobker, & Godley, 2011); and to be diagnosed at Stage III or IV, when adults may have more obvious functional deficits leading to a referral to occupational therapy.

In this retrospective cohort study, we used secondary data from the Integrated Cancer Information and Surveillance System (ICISS), which links multiple data sources including the North Carolina Central Cancer Registry (NCCCR) and administrative claims from both public and private insurance payers. ICISS includes about 80% of the North Carolina population with cancer (UNC Lineberger Comprehensive Cancer Center, 2010). The other 20% of people with cancer were either uninsured or had insurance plans not captured within ICISS. The University of North Carolina at Chapel Hill institutional review board approved this study.

The study sample was limited to individuals enrolled in Medicare, aged >65 with incident cases of breast, prostate, lung, colorectal, or melanoma (skin) cancers between 2004 and 2007. These cancer diagnoses represent the five highest incidence rates within North Carolina. Cancer cases in the NCCCR were identified by International Classification of Diseases for Oncology, 3rd Edition, diagnosis codes and were subsequently linked to the Medicare insurance claims files (UNC Lineberger Comprehensive Cancer Center, 2010). These cases were further linked to the area resource file for county-level data.

We excluded adults who (1) qualified for Medicare because of end-stage renal disease or disability, (2) were diagnosed at death or during an autopsy, (3) were diagnosed before their 66th birthday, (4) had a previous diagnoses of cancer, or (5) were not enrolled in Medicare Part A or Part B (and thus would lack claims data). See Figure 2 for a participant flowchart.

Occupational therapy users were defined as beneficiaries who had submitted a billing claim for occupational therapy service using Current Procedural Terminology; the International Classification of Diseases, Ninth Revision, clinical modification section; and Healthcare Common Procedure Coding System codes (see UNC Lineberger Comprehensive Cancer Center, 2010). We identified 28 codes that best defined use of occupational therapy services from inpatient, outpatient, home health, hospice, and skilled nursing facilities. This coding includes evaluations and treatments for rehabilitation as well as palliative and end-of-life care. The final sample consisted of 27,131 older adults with various forms of cancer, of whom 8,720 used occupational therapy services during the 2 yr postdiagnosis.

Our primary dependent variable was occupational therapy use within 2 yr of the date of the cancer diagnosis. To check the basis of this decision, we examined the relationship of time and therapy utilization related to cancer, using histograms and frequency tables, to see whether there was a specific pattern or signal for when occupational therapy use spiked. Frequency of occupational therapy visits appeared stable throughout the time frames initially chosen (1 yr, 18 mo, and 2 yr). Within oncology research, Sehl, Satariano, Ragland, Reuben, and Naeim (2009)  found that limitations in ADLs and IADLs persisted beyond 1 yr for older women with breast cancer. In addition, Reeve et al. (2009)  examined adults with cancer pre- and postdiagnosis and found that although some older adults were able to improve within the first year, others did not recover compared with the general health scores of adult control participants without cancer more than 19 mo after the cancer sample’s diagnoses. Thus, the 2-yr time period was chosen based on clinical experience of the first author (Pergolotti) and the literature describing functional deficits from a cancer diagnosis as still present after 1 yr or longer (Deimling, Sterns, Bowman, & Kahana, 2005; Reeve et al., 2009; Sehl et al., 2009; Sehl, Lu, Silliman, & Ganz, 2013).

Independent variables were chosen on the basis of the conceptual model, literature review, clinical experience of the first author (Pergolotti), and available data (Andersen, 1995). Predisposing variables included age, sex, race, and county-level percentage of adults with less than a high school degree. Enabling variables included eligibility marker for low socioeconomic status (measured as Medicaid supplement to Medicare), county classification as defined by the U.S. Department of Agriculture Economic Research Service (2013)  continuum coding scheme (rural [<19,999], urban [>20,000], and metropolitan [>250,000]), county-level average household income, and previous use of occupational therapy (defined as at least one claim for an occupational therapy visit in the year before the date of cancer diagnosis, ending the month before diagnosis). Need variables included cancer type, cancer stage, and comorbidity status.

Comorbidities were measured with the Charlson Comorbidity Index (CCI; Klabunde, Potosky, Legler, & Warren, 2000), which uses inpatient, outpatient, and physician claims from 12 mo before cancer diagnosis until the month preceding diagnosis. This index was categorized into none, 1, 2, 3, and ≥4 comorbidities, with higher scores associated with increased risk of mortality and morbidity (Klabunde et al., 2000). Tumors were staged as 0–IV, with IV representing the most progressed (Greene et al., 2002).

In bivariate analyses, we compared occupational therapy users and nonusers using likelihood ratio chi-square tests for categorical variables (race, county classification, cancer type) and dichotomous variables (sex, dual eligibility for Medicare and Medicaid) and t tests for continuous variables (age, education, household income, stage, CCI). The multivariable analyses used a hierarchical regression approach to assess the contribution of the different types of care utilization determinants, as outlined in the Andersen model (Nathans, Oswald, & Nimon, 2012; Quick, 2010). A binomial distribution was chosen for this analysis because of the dichotomous dependent variable (yes or no—use of occupational therapy within 2 yr of diagnosis of cancer).

Each generalized linear model was analyzed with a log link to obtain relative risk ratios (RR) of occupational therapy use and the corresponding 95% confidence intervals (CIs). We chose RR because of the outcome event having >10% incidence (McNutt, Wu, Xue, & Hafner, 2003). The first model included only predisposing variables, and the second model added the enabling variables. In the third model, need variables were added to the second model.

We found missing data within three of the variables (rural–urban character, household income, and cancer stage). Less than 0.01% of the variables were missing. Cases with missing variables were excluded. Because of the large sample size, we used a significance level of p < .001 for all tests. The software used for this analysis included RStudio for Unix (v.0.96.122; RStudio, Boston) and SAS/STAT software, Version 8 of the SAS System for Unix (SAS Institute, Cary, NC).

Of the 27,131 North Carolina Medicare beneficiaries who were diagnosed with breast, prostate, lung, colorectal, and melanoma cancers in 2004–2007, only 32% (8,720) used occupational therapy within the first 2 yr of their cancer diagnosis. In the bivariate analyses (Table 1), older adults who used occupational therapy were significantly older (77 yr vs. 75 yr) and disproportionately female (55% vs. 43%). As for differences between groups within the enabling variables, occupational therapy users were more likely to be dually eligible for Medicare and Medicaid (17% vs. 12%), to use occupational therapy within 1 yr before cancer diagnosis (28% vs. 15%), and to be from metropolitan areas (64% vs. 60%). Occupational therapy users were more likely to be diagnosed with breast (25% vs. 19%) and colorectal cancers (21% vs. 16%); to be Stage I (22% vs. 18%) and Stage III (14% vs. 12%); and to have one (28% vs. 26%), two (13% vs. 10%), three (6% vs. 4%), or more than four comorbid conditions (6% vs. 4%).

Hierarchical linear regression identified variables associated with the use of occupational therapy services in three different models in sequential fashion (Table 2). When only considering predisposing variables (Model 1), occupational therapy users’ age, sex, and education were the strongest predictors of occupational therapy use. The strength of the relationships between predisposing variables and occupational therapy use was attenuated when adding enabling variables (Model 2). Including the need variables (Model 3) lessened the predictive ability of age, sex, race, dual eligibility, and previous occupational therapy use.

According to Andersen’s model, need variables would predict utilization. Within the final model, however, predisposing, enabling, and need variables all predict use. As we hypothesized, for every 5-yr increase in age, adults were 11% more likely to use occupational therapy. Women were 16% more likely, and those who were diagnosed with breast cancer were 14%–23% more likely to use occupational therapy than were adults with prostate and lung cancer. As for adults with different stages of cancer, adults with Stage I, II, or III cancers were more likely to use occupational therapy than those with Stage 0 or Stage IV. Last, adults with a score ≥1 on the CCI were more likely to use occupational therapy.

In terms of race in the fully adjusted model, African-Americans were 4% more likely to use occupational therapy. This finding was marginally significant in the intermediate model, and the CI included 1.00 in the final model. Also, household income and urban location had no relationship to use of occupational therapy services (RR = 1.02, 95% CI [0.99, 1.06]). Previous use of occupational therapy remained the strongest predictor over and above all other predictors within the model, and adults who used occupational therapy within 1 yr before their diagnosis were 35% more likely to use occupational therapy after diagnosis.

Some disparities in care were suggested by the findings. Only 32% of the sample used occupational therapy within the first 2 yr of their cancer diagnosis, a rate lower than the estimated ≤87% of adults who are in need of such services (Holm et al., 2012; Lehmann et al., 1978). The occupational therapy users were significantly older than nonusers, and women were the majority. The literature substantiates this finding (Evashwick, Rowe, Diehr, & Branch, 1984; Holmes, Freburger, & Ku, 2012; Stoddart, Whitley, Harvey, & Sharp, 2002). Although we hypothesized that occupational therapy users would differ by race and that race would predict use of occupational therapy, the difference appears to be small based on percentage of users. African-Americans appeared more likely to use the service; however, the magnitude of relative risk is small and minimally significant. This finding could be considered encouraging because it suggests only a minimal difference based on race, and that difference gives the advantages to African-Americans.

Freburger et al. (2011) reported that sociodemographics predicted increased use of higher institutional rehabilitation. However, as Freburger et al.  described, even relatively small increases in use by minority groups may be concerning when considering the differences of outcomes and quality of survivorship for minorities overall (National Cancer Institute [NCI], 2011). Therefore, a small difference in use of services by minority compared with White patients may actually be more likely because of delayed, unmet health care needs and delayed use of services (Freburger et al., 2011; Moon & Shin, 2005).

Surprisingly, household income and the rural or urban character of the county of residence did not predict use of occupational therapy. Unlike in previous studies (Freburger et al., 2011; Harada, Chun, Chiu, & Pakalniskis, 2000), geographic location did not seem to be related to disparities in utilization. Harada et al. (2000)  examined the geographic location of the hospitals where adults with hip fractures received physical therapy and found location to be highly important. Their findings may speak to the differences between health service use for urban and rural hospitals, not necessarily the county in which the adult lives as was examined in this study. Also, our finding could be related specifically to North Carolina; Freburger et al. (2011)  did not examine adults from North Carolina. It could also be specific to the use of health care after a diagnosis of cancer, which may be different than for other conditions (Au, Udris, Fihn, McDonell, & Curtis, 2006). These results are different from previous research, even though the designation of rural and urban character was similar to other studies examining health service use (Jacobs, Kelley, Rosson, Detrani, & Chang, 2008; O’Malley, Forrest, Feng, & Mandelblatt, 2005). This curious finding suggests a need for additional investigation of the spatial distribution of access to occupational therapy relative to residence and cancer care sites and the need for more detailed individual level variables for analysis.

Beneficiaries with breast, colorectal, and melanoma skin cancer were more likely to be seen by an occupational therapist when compared with adults with prostate and lung cancer. Although those with a lung cancer diagnosis were the largest group, they were the least likely to be seen by an occupational therapist. This finding is disconcerting because the literature shows that older adults with lung cancer are most likely to experience a decline in ADLs, specifically, bathing, dressing, getting in and out of a chair, and using the toilet, after their diagnosis (Reeve et al., 2009). Compared with breast cancer, adults with lung cancer were more likely to report poorer health status (Hewitt, Rowland, & Yancik, 2003). Baker, Denniston, Smith, and West (2005)  identified similar results and stated that adults with lung cancer report the most problems, including feeling helpless and dependent. Moreover, Esbensen, Østerlind, Roer, and Hallberg (2004)  reported that having a diagnosis of lung cancer alone predicted poor quality of life and called for targeted interventions for this group. Adults with lung cancer are typically diagnosed at a later stage and have poorer survival rates than adults with the other cancer types represented in this study (NCI, 2011). However, considering their poorer survival rates and quality-of-life status, older adults with lung cancer may need special attention and intervention.

Adults with Stage IV cancers were least likely to be treated with occupational therapy, although recent literature suggests that occupational therapy would be beneficial for this population (Kasven-Gonzalez, Souverain, & Miale, 2010; Schleinich, Warren, Nekolaichuk, Kaasa, & Watanabe, 2008). Similar to what Cheville (2005)  reported, a considerable number of adults with late-stage cancer do not have access to occupational therapy services, although they may benefit from such services. According to Cheville, cancer rehabilitation (understood as making specific gains toward restoring previous levels of independence and functional ability) is commonly “dismissed as an oxymoron” (p. 219), particularly within the later stages. This stereotype could explain why older adults with later stage cancers were least likely to be seen by an occupational therapist in this study. Future research is warranted to examine whether other predictors of use may determine use at this stage, including attitudes, values toward health care, or availability of occupational therapists to provide care.

Previous use of occupational therapy remained the strongest predictor in the final model. Once adults are aware of the services available, they become more likely to use them again. The literature on cancer rehabilitation commonly reports physician unawareness of occupational therapy and poor communication among fields as barriers to use because a referral is needed for access to care (Cheville, 2005; McCartney, Butler, & Acreman, 2011). Possibly, physicians (or nurse practitioners) who are aware of occupational therapy are more likely to refer. Future research could focus on awareness of occupational therapy as a potential way to expand access to care.

To our knowledge, this is the first study to examine the patterns of use of occupational therapy alone in a population-based study. These results suggest underuse of occupational therapy by older adults with cancer, a population with considerable functional needs. Given the known relationships between functional status and overall well-being in cancer care, further research exploring both barriers to occupational therapy use and opportunities for intervention will be critical in strengthening cancer survivorship care in North Carolina and beyond.

Our study had several limitations. First, because the types of occupational therapy provided are tailored to clients’ specific needs, types of occupational therapy intervention and evaluation are likely to differ among adults. Moreover, occupational therapy billing codes do not include diagnosis codes to verify the reason for therapy. Some of the adults in our study may have been receiving occupational therapy for other reasons. Second, data were lacking on important predictors of occupational therapy use not found in claims data, such as personal beliefs and individual functional status. Third, although billing codes for occupational therapy could be used to represent other services such as physical therapy, a conservative approach to the codes was used to decrease that possibility. Fourth, income and education level were represented at the county level. Fifth, the study was conducted only in North Carolina, which may limit its generalizability.

To examine for the first time the patterns of use of occupational therapy by older adults with cancer, we identified several predictors of occupational therapy use in this population, including sex, age, previous use of occupational therapy, cancer type, and stage. Our results suggest possible underuse of occupational therapy by older adults with cancer. Future research could narrow the focus to one cancer type because cancers differ by type and stage. Moreover, research could include other large surveys linked to Medicare claims, which would include both functional status and billing claims and provide a more thorough understanding of the appropriateness, effectiveness, and possible disparity of occupational therapy services.

These analyses addressed an important problem that has received little attention. We identified several sociodemographic variations and lower usage than reported need in the patterns of occupational therapy use of older adults with cancer. Although cancer rehabilitation, defined to include occupational therapy and physical therapy, has been recommended, we noted large numbers of older adults not receiving services and considerable differences between those who did and did not use occupational therapy (Holm et al., 2012; Lehmann et al., 1978; Movsas et al., 2003;Ross, Petersen, Johnsen, Lundstrøm, & Groenvold, 2012; Stafford & Cyr, 1997). Because the burden of cancer and its treatments is greater for older adults, we stress that future researchers continue to understand the utilization of occupational therapy services and the appropriateness of the services for this population; it is especially critical for adults with lung cancer, who demonstrate the highest need and are least likely to use occupational therapy. Although evidence for occupational therapy is growing in other fields, we reiterate the need for future research within this population.

This study is the first description and analysis of the use of occupational therapy services by older adults with cancer. Examination of use of occupational therapy is the first step to understanding the quality of care provided to older adults. The findings from this study suggest the following implications for occupational therapy research and practice:

We thank Seth Tyree and Huan Liu, Statistics and Data Management for the Integrated Cancer Information and Surveillance System, and Chris Weisen of the Odum Institute, University of North Carolina at Chapel Hill, for their helpful suggestions and support in preparation of this study.

This article is part of the research completed for a doctoral dissertation. This research was partially supported by the National Cancer Institute (R25CA116339); a National Research Service Award Predoctoral Traineeship from the Agency for Health Care Research and Quality, sponsored by the Cecil G. Sheps Center for Health Services Research, University of North Carolina (UNC) at Chapel Hill (T32-HS000032); a North Carolina Translational and Clinical Science Institute Pilot Award (2KR401206); and the Integrated Cancer Information and Surveillance System, UNC Lineberger Comprehensive Cancer Center, with funding provided by the University Cancer Research Fund through the state of North Carolina. Weinberger was supported by a Research Career Scientist Award (RCS 91-408) from the U.S. Department of Veterans Affairs, Veterans Health Services Research and Development Service. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funders.

Mackenzi Pergolotti presented this research at the American Occupational Therapy Association Annual Conference & Expo in San Diego, California, in April 2013 and in preliminary form at poster sessions for the 18th Annual National Service Research Award Trainees Research Conference and AcademyHealth’s Annual Research Meeting in Orlando, Florida, in June 2012.