Is Self-Referral Associated with Higher Quality Care?

Abstract

Objective: To assess the extent to which patients self-refer to cancer specialists and whether self-referral is associated with better experiences and quality of care.

Data sources: Data from surveys and medical record abstraction collected through the Cancer Care Outcomes Research and Surveillance Consortium.

Study design: Observational study of patients with lung and colorectal cancer diagnosed from 2003 through 2005 in five geographically defined regions and five integrated health care delivery systems.

Methods: Multivariable logistic regression models used to assess factors associated with self-referral and propensity score-weighted doubly robust models to test the association between self-referral and experiences/quality of care.

Principal findings: Among 5,882 patients, 9.7 percent of lung cancer patients and 14.9 percent of colorectal cancer patients self-referred to at least one cancer specialist. Black patients were less likely to self-refer than white patients (odds ratio: 0.48, 95 percent confidence interval: 0.35, 0.64); patients with high incomes (vs. low) and with a college degree (vs. non-high school graduates) were significantly more likely to self-refer. Self-referral was associated with lower ratings of overall physician communication for patients with lung cancer but, conversely, higher odds of curative surgery among patients with stage I/II lung cancer.

Conclusions: A small but significant proportion of patients self-referred to their cancer specialists; rates varied by patient race and socioeconomic status. To the extent that self-referral is associated with quality, it may reinforce disparities in care.

Keywords: Self-referral; colorectal cancer; lung cancer; referral.

Figure 1

Cohort Definition

Figure 2

Referral Pathway for Patients to Their Surgeon, Radiation Oncologist, and Medical Oncologist, Stratified by Lung and Colorectal Cancer

Figure 3

Adjusted Association between Self-Referral and (a) Better Experiences with the Interpersonal Aspects of Care and (b) Patient/Surrogate-Reported Experiences of Care and Quality of Care for Patients with Lung and Colorectal Cancer Using Doubly Robust Propensity Score-Weighted Models

1. Note: Doubly robust models adjust for propensity weights and sociodemographic, clinical, insurance, and health care delivery-related factors. Due to collinearity, the following variables were removed: from the overall quality of care, we omitted whether the patient had a PCP; for adjuvant chemotherapy for stage III colon cancer, we omitted age, insurance, education, income, marital status, PCP status, and number of different providers referred to; for adjuvant chemotherapy and radiation for rectal cancer, we omitted age, insurance, education, income, marital status, comorbidity, PCP status, number of different providers referred to, and CanCORs site; for curative surgery for early-stage lung cancer, we omitted age, comorbidity, and PCP status.