Oncologist recommendations for adjuvant chemotherapy for patients with breast cancer are based on numerous factors, including patient clinical factors and tumor characteristics. In breast cancer patients with a high risk of recurrence by virtue of stage or tumor biology, guidelines recommend treatment with adjuvant chemotherapy in the absence of substantial patient comorbidity [1]. For patients with a lower risk of recurrence and a corresponding lower marginal benefit of adjuvant chemotherapy [2,3], factors such as patient preferences and physician beliefs about the benefit of chemotherapy enter into the treatment decision [4–6].

It is not known to what extent non-clinical patient factors play a role in medical oncologists' recommendations regarding chemotherapy. In other medical conditions, including diabetes and cardiovascular disease, factors such as presence of social support, heavy career and family demands, a patient’s desire for an active lifestyle, and perceived patient education level have been shown to influence physician treatment recommendations [7–9]. Primary care providers have reported that patient socioeconomic status affects their clinical management decisions including diagnostic testing, medication prescription, and referrals to subspecialists [5,10]. In other work, social, behavioral, and psychological patient characteristics influence physician decision to recommend initiation of protease inhibitor triple combination regimens for the treatment of HIV/AIDS [11]. In some of these clinical settings, physician demographic characteristics, including sex and race, have been associated with the tendency to consider these non-clinical factors [6].

The purpose of this study was to investigate the extent to which medical oncologists consider non-clinical patient factors (ability to follow instructions, level of social support, and level of support and flexibility at work) in their decisions to recommend adjuvant chemotherapy for breast cancer. These three factors were chosen because of the potential of each to affect the successful delivery of a complete course of adjuvant chemotherapy. As described in greater detail below in Methods, we hypothesized that physician and practice characteristics would be associated with the degree to which medical oncologists consider each non-clinical factor.

In summary, we found that medical oncologists frequently report consideration of non-clinical patient factors in their decisions to recommend breast cancer adjuvant chemotherapy, with two-thirds of oncologists reporting strong consideration of one or more non-clinical factor. Oncologists reported strong consideration of a patient’s ability to follow instructions at a higher frequency than level of support. This may be due to a perception that ability to follow instructions has a greater direct impact on the successful delivery of a course of adjuvant chemotherapy than the patient’s support environment. Alternatively, oncologists may place more emphasis on ability to follow instructions because they have greater confidence in being able to accurately assess this as compared to a patient’s social environment.

The associations between years in practice and consideration of a patient’s ability to follow instructions, and extent of specialization in breast cancer and consideration of level of social support, may be the result of greater physician experience and confidence in his/her ability to support the patient through chemotherapy treatment regardless of these non-clinical patient factors. Our finding that physicians with more experience are less likely to incorporate these non-clinical factors may partly explain previous research on the relationship between years in practice and patient outcomes [17,18].

The association between tumor board availability and strong consideration of social support and work support and flexibility may be related to an enhanced awareness of social vulnerability of patients. In clinical vignette studies, social vulnerability has been shown to play a role in treatment recommendation decision-making. For example, in one study, black men and those who were former injection drug users were rated by physicians as less likely to adhere to highly active antiretroviral therapy (HAART) for HIV. This perception was associated with a lower likelihood that the responding physician would prescribe HAART [19]. In another study using written vignettes describing patients with prostate cancer, urologists were less likely to recommend a radical prostatectomy to patients who were described as low-income and widowed compared to those who were middle-income and married [20]. Multidisciplinary tumor boards are a forum in which multiple providers describe patient cases, often including descriptions of behavioral and social issues. Discussions about patients’ financial and social challenges before a medical oncologist meets a patient may heighten awareness of issues that are perceived as impacting whether or not a patient will actively seek supportive care during chemotherapy.

Although we observed an association between non-white oncologist race and greater likelihood of strongly considering level of social support and level of work support and flexibility, the non-white race group was heterogeneous. Further work investigating the relationship between physician race and the role that non-clinical patient factors play in decisions to recommend adjuvant chemotherapy for breast cancer treatment may be warranted.

Our findings could be interpreted as evidence that medical oncologists are providing patient-centered care. According to the Institute of Medicine, provision of patient-centered care requires that physicians must know the patient as a whole person, including the social circumstances and ability to follow instructions [21]. However, in order for patient-centered care to result in desirable outcomes, the physician assessment of these non-clinical factors and the impact these non-clinical factors will have on treatment delivery must be accurate. The issue is further complicated because decisions to recommend adjuvant chemotherapy are, by definition, made under conditions of uncertainty since the degree of benefit achieved in an individual patient is uncertain [22]. Therefore, consideration of non-clinical patient factors could incorrectly result in adjuvant chemotherapy not being recommended in patients with a higher risk of recurrence for whom treatment offers greater benefit.

This study has several limitations. First, a limited number of medical oncologists in only two areas of the country were surveyed. Respondents may not be representative of all the medical oncologists in the Detroit and Los Angeles SEER catchment areas. Second, it is possible that the degree of consideration of non-clinical factors was overestimated because medical oncologists felt that to consider non-clinical patient factors is the socially desirable response. Third, we do not know whether oncologists’ survey responses correspond to their actual practice patterns. Finally, the analyses were not specifically powered to detect significant associations for all of the medical oncologist and practice characteristics being investigated. Thus, caution should be exercised when interpreting the non-significant findings.

The rates of missing data differed across gender, race and participation in a practice network to examine treatment variation. To reduce potential missing data bias, we performed multiple imputation on the missing values [23]. This allowed us to include all 204 oncologists in our analyses to gain efficiency and be able to generalize the conclusions to the original sample population. The multiple imputation results are similar to the analyses limited to oncologists with complete data. Although the association between access to a tumor board and non-white race with level of work support and flexibility change slightly after imputation, the directions of all associations were consistent with and without multiple imputation, and both factors remain significantly associated with strong consideration of support and flexibility at work.

Quality of care is affected by errors of omission that result from failure to provide indicated treatments [24]. Consideration of non-clinical patient factors may increase errors of omission due to inaccurate assessment of a patient’s true social context and incorporation of these non-clinical patient factors into decisions whether or not to recommend adjuvant chemotherapy treatment in cases where they should not be considered. Decisions made by medical oncologists to recommend adjuvant chemotherapy are not made over time. Rather, they are made after one or, at most, two patient-provider encounters. Decisions that incorporate non-clinical factors after few meetings with the patient are prone to error. In a study of simulated first-encounter primary care office visits, 88% of the physicians made what are referred to as “contextual errors” in which they did not adequately assess a patient's environment or behavior (context) when faced with complicated scenarios [25]. Furthermore, tumor boards provide the opportunity for medical oncologists to be primed with potentially inaccurate information about a patient’s social situation even prior to meeting the patient. Therefore, in the setting of adjuvant chemotherapy decision-making, errors in assessment of a patient’s context may significantly increase the risk of omission of important therapy. The use of assessment tools to identify patients who have problems following instructions, and those with minimal support, may increase the accuracy of oncologists’ knowledge of a patient’s social context. Incorporating tools such as these into a healthcare system in which individual providers and practices are prepared to meet the needs of patients whom providers perceive to be at risk during chemotherapy will improve the quality of cancer care.