Osarogiagbon, Raymond U. MBBS*; Phelps, Gregory MD†; McFarlane, Joshua MD‡; Bankole, Olufunsho MBBS§
Optimal care of patients with lung cancer and other thoracic tumors requires the active involvement of many specialists, including radiologists, pulmonologists, thoracic surgeons, pathologists, medical, and radiation oncologists. Each specialist brings a particular perspective and skill set to the diagnosis, staging, and treatment of thoracic tumors. Distinguishing between benign and malignant lesions, and choosing the most effective means of management require interaction between the key specialists. Because of fears that a system of serial referrals and autonomous decision making might result in delayed, fragmented, and suboptimal care,1 there has been great interest in formal multidisciplinary fora where thoracic cases can be discussed and optimal recommendations for treatment made.2
However, the evidence for benefit from multidisciplinary care is mostly empirical.3 Yet, setting up multidisciplinary programs with the active involvement of the key specialists is challenging because of the different practice culture and schedules of the specialists; the need to ensure the accuracy of clinical information and the appropriateness of management recommendations; and the need for timely and accurate communication of recommendations to the responsible clinicians. We hypothesized that if multidisciplinary recommendations represent the ideal management plan, deviation from these recommendations should lead to worse patient outcomes. If this proves true, it would be necessary to understand the extent and causes of such deviation and to devote effort to minimize it.
We evaluated the clinical benefit of our Multidisciplinary Thoracic Oncology Conference (MTOC) at the University of Tennessee Cancer Institute by comparing MTOC recommendations with actual clinical management. We sought to characterize the extent and etiology of deviations from recommended care and analyzed the impact on patient outcomes.
PATIENTS AND METHODS
After approval of our study by the Institutional Review Board of the University of Tennessee, we reviewed records of all patients presented for discussion at our MTOC from its onset in February 2006 to the date of data censorship in October 2009.
Structure of the MTOC
From February 2006 to February 2007, the MTOC was held twice a month. After March 2007, the conference was held weekly, immediately after a multidisciplinary clinic (MDC) to facilitate multidisciplinary patient contact before MTOC patient discussions. Regular participants at the conference included two thoracic surgeons, a radiologist, two pulmonologists, two medical oncologists, a radiation oncologist, a palliative care specialist, a nurse coordinator, a nurse attached to each involved specialty, and a clinical research coordinator. Pathologists participated on an ad hoc basis, when a pathology problem was to be discussed. The MDC consisted of a thoracic surgeon, medical oncologist, pulmonologist, and the MTOC nurse coordinator.
We created an open referral system, in which providers could submit patients for discussion in the MTOC. After creation of the MDC, providers were encouraged (but not mandated) to have their patients evaluated in the clinic before discussion. Patients who were not evaluated in the MDC had to be presented by the referring provider who was, therefore, an active participant in the clinical care debate. Patients seen in the MDC were presented by the referring physician or one of the clinic participants.
MTOC Decision Making and Communication with Referring Physicians
After conference discussion, consensus recommendations were enumerated for each patient, and an MTOC note was placed in each patient's electronic medical record. In addition, the nurse coordinator directly contacted each referring physician with the consensus recommendations within 24 hours, irrespective of the physician's presence in the conference. Nevertheless, recommendations were not binding on the referring physician.
Data on patient demographics, clinical history (including performance status, smoking history, histology, and clinical stage), reason for presentation, name of referring and presenting physician, prior evaluation in the MDC, and management recommendations of the MTOC were prospectively recorded on a standardized template. We retrospectively collected information on actual clinical management and outcomes from the University of Tennessee Cancer Institute electronic medical records system, supplemented by a search of the Social Security Death Index. We abstracted the cause of discordant care from the medical records.
We compared the recommendations of the MTOC with the actual clinical care provided. We identified care that matched recommendations as concordant and mismatched care as discordant. For statistical reasons, patients who were presented multiple times were assigned on the basis of compliance with recommendations made at the initial discussion. We compared the characteristics of patients who received concordant and discordant care. We also compared outcomes, including the time to definitive care, overall survival, and progression-free survival after MTOC discussion.
Categorical variables were compared by the χ2 test and Fisher's exact test. Survival was estimated by the Kaplan-Meier method, and survival curves were compared by the log-rank test. The Cox-proportional hazards model was used for multiple regression analysis of determinants of survival. We excluded patients who died or were lost to follow-up before any clinical intervention from the survival analysis to avoid bias against the discordant care cohort.
From February 2006 to October 2009, 376 patients were presented for discussion at our MTOC. Two hundred thirty-five patients (63%) received concordant care and 141 (37%) received discordant care. The demographic and clinical characteristics of patients were similar between the groups, except for race (p < 0.05) and health insurance status (p < 0.001) (Table 1). Whites were more likely to receive concordant care than black patients. Those with commercial insurance were the most likely to receive concordant care and those with no insurance, least likely. A slight majority of the uninsured received discordant care (Figure 1). Ninety-four percent of patients had an Eastern Cooperative Oncology Group performance status of 0 to 1, and 99% had Eastern Cooperative Oncology Group 0 to 2.
One hundred twenty-eight patients (34% of the whole cohort) were seen in the MDC before MTOC discussion. Sixty-five percent of these patients received concordant care, compared with 61% of those who were not evaluated in the MDC before their initial presentation.
MTOC Discussions and Recommendations
Four hundred fifty-four sets of recommendations were made for the 376 patients. Sixty-three patients (17%) were presented on multiple occasions, with a median of 2 (range: 2–3) discussions. The reasons for discussion in the whole cohort were advice on options for diagnosis of an intrathoracic lesion, 147 (32%); advice on first-line treatment for newly diagnosed cancer, 132 (29%); advice on salvage therapy in patients with progressing or relapsed cancer 74 (16%); advice on staging newly diagnosed cancer, 48 (11%); advice on postoperative management options after resection of lung cancer, 33 (7%); and follow-up discussions of the outcomes of previously recommended interventions, 19 (4%).
Recommendations were classified into four groups (Table 2). Definitive therapy was the most frequent recommendation (49%), followed by biopsy (27%), radiologic studies (19%), and “other” therapy (6%). The 454 discussions with recommendations led to 396 identifiable clinical actions. Fifty-seven discussions (12.6%) did not lead to any identifiable clinical activity. We also evaluated the performance of recommended invasive procedures (Table 3). Fifty invasive procedure recommendations (22.3%) were not followed by any identifiable invasive procedures. There were neither glaring differences in the relative proportion of clinical management choices (radiologic study, biopsy, treatment, or “other”) nor in the utilization of invasive tests in the whole cohort (Tables 2 and 3). Nevertheless, when we matched the actual clinical management to the 454 sets of MTOC recommendations, we observed concordance in only 282 (62%) and discordance in 171 (38%) cases.
The reason for conference discussion did not seem to have any impact on the concordance rate of the 376 patients. However, the concordance rate varied significantly with specific recommendations overall and also with recommendations for invasive procedures (Table 4). The highest rates of concordance occurred with recommendations for radiotherapy (80%) and radiologic procedures (79%), whereas recommendations for biopsy and combined modality therapy had the lowest concordance (50%). In the subset of patients with recommendations involving invasive procedures, recommendations for surgery had the highest concordance rate (68%), and recommendations for mediastinoscopy (48%) and “other” procedures (20%) had the lowest.
Outcome of Care
Patients with concordant clinical management had a significantly shorter time to clinical intervention than those who received discordant care (median: 14 days versus 25 days, p < 0.002). They also had a longer median overall (2.1 years versus 1.3 years, log rank p < 0.01) and progression-free survival (1.3 years versus 0.8 years, log rank p < 0.05) than those who received discordant care (Table 5, Figures 2A, B). In the Cox proportional hazards model, the hazard ratios for overall and progression-free survival in the concordant cohort were 1.8 and 1.5, respectively. These differences in survival remained significant after adjustment for stage of disease at the time of presentation (Table 5).
We identified clinician decision as the cause in 104 of 171 (61%) instances of discordant care. Other causes were patient loss to follow-up care (19%) or refusal of recommended care (15%). In nine instances (5%), the patient died before the recommended care could be provided. The reasons for clinician-induced discordance were listed as “clinical contraindication” in 61, “comorbidity” in 14, “insurance problem” in 11, “stage discrepancy” in 11, and “poor performance status” in seven instances. In most cases, we were unable to determine the reasons why patients refused recommended care.
Many advocate multidisciplinary care for optimal management of patients with thoracic tumors because of the complicated sets of clinical decisions involved. Support for the benefit of this model of care has mostly been based on intuition and “commonsense,” with little objective evidence of superior outcomes in patients who participate.3 Superior outcomes, if achieved, should arise from effective implementation of multidisciplinary recommendations. We hypothesized that if multidisciplinary recommendations represent optimal care, deviation from these best-case recommendations should be associated with demonstrably inferior outcomes.
One third of patients discussed at our MTOC received care discordant with recommendations. Health insurance status and race were significantly associated with higher rates of discordant care. Patients who received discordant care had significant delays in the onset of definitive treatment. They also had poorer survival. These findings seem to support our hypothesis that MTOC recommendations reflect optimal management, and deviation from these recommendations may lead to inferior treatment. Our study further suggests that racial minorities and the uninsured still experience health care disparities, even after gaining access to a comprehensive multidisciplinary care environment.
We questioned the high proportion of patients receiving discordant care in our study. Although recommendations made at MTOC represent ideal proposals for management, they are not binding on clinicians at the point of care. MTOC recommendations emphasize histologic staging with a view to stage-appropriate treatment, careful evaluation of eligibility for surgical resection, and the use of multimodality therapy for patients with more locally advanced disease.4 These approaches, which optimize risk stratification and selection for potentially curative therapy, may not be as readily pursued by clinicians who are not intimately involved in the MTOC process.5 The cost, risk, and added time required for histologic confirmation of stage, which often entails the use of highly invasive and specialized techniques, may serve as a disincentive to clinicians and patients to follow through on recommendations for these procedures. For example, only 50% of recommendations for mediastinoscopy and 25% of recommendations for endobronchial ultrasound-guided mediastinal lymph node biopsy were carried out. Similarly, concordance was low (50%) with recommendations for combined modality therapy, which improves outcomes in patients with locally advanced disease but requires coordination between clinicians from different specialties. This level of discordance occurred even though these procedures were available through the specialists involved in the MTOC.
The recommended care may have been inappropriate for some patients in the discordant care group, for example, because of inaccurate or incomplete information at the time of discussion. The referring physicians may have correctly recognized contraindications to the care proposed by the MTOC and wisely opted out. Under this theory, the poorer outcomes of the discordant care group may have been due to poor patient condition—comorbid illness and poor performance status—rather than deviation from optimal care. We designed the MTOC to foster active participation of the referring physicians, open discussion, and a consensus decision-making approach to guard against this possibility.
Our findings are compatible with those of Leo et al.6 who found a trend toward a survival benefit from implementing recommendations of MTOC in their single-year prospective study of the MTOC process in a region of France. However, they found a discordance rate of only 5%, compared with 37% in our study. This difference may reflect the availability of universal health care coverage and the legal mandate for multidisciplinary care in France, neither of which exist in the United States. Others have failed to show any benefit from multidisciplinary care. Riedel et al.,7 in their study of the Veterans Affairs population, found no difference in outcomes between an era of MDC care and a subsequent era without the clinic. However, their clinic did not involve a thoracic surgeon, and the nonclinic era had an active MTOC throughout, which probably confounded their study.
In the absence of demonstrable survival advantage, some investigators have pointed to surrogate benefits, such as improved timeliness of care and increased rates of referral for potentially curative surgery,8 as evidence of benefit from multidisciplinary programs. Although we found shorter time from discussion to definitive treatment in those who received concordant care, it has been pointed out that timeliness of care for lung cancer does not necessarily translate to improved survival.1,9 It is likely that delays in the discordant care group were partly due to known and unknown group characteristics such as poorer insurance status and greater clinical complexity and may not have been a direct benefit of the MTOC process itself.
Our study contributes evidence of a quantifiable and clinically meaningful survival benefit from effective participation in multidisciplinary care. We demonstrate the need for quality control measures to ensure that recommendations are implemented with a high level of fidelity and point to the need to better understand the reasons for discordant care. The success of any multidisciplinary program requires provision of accurate data for discussion, recommendations reflective of clinically available options, timely communication with point-of-care clinicians, and effective discussion with patients regarding the rationale for management plans. Disruption at any of these levels may cause a deviation from the recommendations. For example, uninsured patients and their providers may have no choice but to deviate from recommendations, if there is no realistic avenue to provide access to the care proposed. Such recommendations, however, “optimal,” are arguably not reflective of clinically available options for those particular patients.
The severity of adverse consequences of deviation from MTOC recommendations may depend on the cause and the extent of deviation. For example, inability to follow recommendations because of significant unforeseen comorbidities might not have as much negative impact as inability to follow recommendations because of lack of insurance coverage or errors in communicating recommendations or clinician misunderstanding. We have previously reported our experience with poor follow-through on recommendations for invasive procedures during a pilot phase of our MTOC.10 During this phase, provider failure to carry out recommended procedures was the main identified cause of discordance. We were able to reduce this discrepancy in a pilot study of a group clinic model, in which era the few discrepancies found were due to patient refusal of recommended interventions.11 We used the lessons learned to further develop our multidisciplinary program in its current form.
The major limitation of our study is the retrospective design and the absence of a case-mix adjustment between the concordant and discordant care groups. For example, our attempt to analyze the reasons for discordance is hampered by the absence of direct, real-time information. The reasons we have identified may not be entirely accurate. As an illustration, although the uninsured made up 17% of the whole cohort, they contributed 27% of the patients who received discordant care, but we identified “insurance problem” as the reason for discordance in only 11 of 171 (6%) instances. We tried to minimize the adverse effect of retrospective review by including all patients discussed, by prospectively determining the data items to be collected, and by standardizing our data collection tools. Another potential criticism of our study is the inclusion of all cases discussed, rather than separation by type of cancer, for example, focusing on the patients with lung cancer. Nevertheless, this is the reality of the MTOC experience. All patients with potentially neoplastic intrathoracic pathology are referred for multispecialty debate and decision making without restriction by specific diagnosis.
The optimal test of benefit from multidisciplinary care would entail a prospective comparison between patients receiving serial specialty care and those receiving care through a multidisciplinary program. It is possible that our discordant care cohort was inherently destined for poorer outcomes due to unidentified health and socioeconomic factors that in combination is reflected in the social state of “lack of health insurance,”12 and the outcome differences we report are not a result of benefit from multidisciplinary care. A retrospective study such as this cannot definitively correct for this possibility. Nevertheless, we can point out the proportional similarity of all key disease factors such as performance status, histology, and stage between the two cohorts of patients. In addition, the key demographic characteristics such as age, sex, and employment status were similar between the two groups. Only race and insurance status were significantly different.
Our study suggests the potential for significant clinical benefits from proper implementation of multidisciplinary thoracic oncology programs, in which care is taken to maximize concordance with best practice recommendations. Comparative prospective studies of the multidisciplinary thoracic oncology process are sorely needed. Such studies will more accurately quantify the benefits of multidisciplinary care and delineate the causes of discordant care. Incentives should be provided to encourage development of effective multidisciplinary care programs designed to improve the quality of care provided to patients with tumors of the thorax.
The authors thank Matthew Smeltzer for help with statistical analysis and Salman Sarwar for help with data collection.