Trust in physicians assumes great significance in the context of illnesses like cancer that are characterized by increased patient vulnerability and uncertainty.1 Newly diagnosed cancer patients require support to understand complex medical information, cope with their diagnosis, and make difficult treatment decisions. The extent to which patients receive helpful informational, emotional, and decision-making support from physicians is thus likely to impact their trust perceptions.

Existing studies have shown physicians’ communication behavior to be significantly associated with patient trust.2–6 Studies focusing on cancer, however, are limited.3 Moreover, virtually all studies are cross-sectional, presenting only a snapshot of the relationship between physician behavior and patient trust. Using longitudinal data during the first year of diagnosis, we examined the association between receipt of helpful informational, emotional, and decision-making support from physicians and ratings of trust among women newly diagnosed with breast cancer.

METHODS

Study Design

We conducted secondary analyses of survey data from a randomized trial of the Comprehensive Health Enhancement Support System (CHESS), a computer-based system designed to assist individuals facing a health crisis. Data collection was completed by January 1998. Details about the trial have been published elsewhere.7 We analyzed data from all 246 women who completed the trial (121 received CHESS, 125 were controls). Participants provided data at baseline prior to the trial and at 2-month and 5-month follow-up assessments.

Measures

Trust in physicians

We assessed women’s trust in physicians using a single item asking them to rate the “confidence they had in their doctor(s)” during the past 7 days on a five-point scale ranging from not at all to very much. This item is part of the Functional Assessment of Cancer Therapy survey.8,9 Scores were linearly transformed to a 0–100 scale.

Physicians’ Supportive Behavior

We assessed patient perceptions of the helpfulness of informational, emotional, and decision-making support received from physicians at each time point. For each type of support, we asked two questions: (1) did the patient receive support in the past 2 months from her physicians? and (2) how helpful was the support that she received (not at all, a little bit, somewhat, quite a bit, or very much)?

We created a measure of perceived helpfulness of support by combining the responses from these questions into a single dichotomous indicator. As described elsewhere,10 a score of 1 was given to women who received support from physicians and rated its helpfulness to be quite a bit or very much (this was the “received very helpful support” group), and a score of 0 was given to the rest of the sample, i.e., women who either did not receive support from physicians or they received support but rated it not at all, only a little, or somewhat helpful (this was the “not received/received, but not very helpful support” group).

Analyses

Linear regression models estimated the association between helpful informational, emotional, and decision-making support from physicians and patient trust. The baseline model controlled for patients’ age, race, income, education, living status, insurance status, days since diagnosis, cancer stage, type of surgery before baseline, and receipt of adjuvant therapy before baseline.

The 2-month model evaluated the association of both concurrent (2-month) and prior (baseline) physician support with patients’ trust at 2-month follow-up. Besides the covariates in the baseline model, we controlled for baseline trust and receipt of any treatment between baseline and 2-month follow-up. We also controlled for potential impact of the CHESS intervention by including an indicator of group assignment (CHESS v/s control) as a covariate.

Similarly, the 5-month model evaluated the association of concurrent (5-month) and prior (2-month) physician support with patients’ trust at 5-month follow-up. Besides the covariates in the baseline model, this model also controlled for 2-month follow-up trust, receipt of any treatment between 2-month and 5-month assessments, and group assignment.

In all models, variables were entered hierarchically with covariates entered in the first block followed by physician support variables in the second block. To avoid over-fitting the models and to reduce type I error, we used the forward stepwise regression procedure in each block to identify the most parsimonious set of predictors for each model. Change over time in receipt of helpful physician support was evaluated by the non-parametric Cochran’s Q test. Change in patients’ trust was evaluated by a univariate repeated measures ANOVA with time as the main predictor.

RESULTS

Patient Characteristics

Table 1 describes the study sample. All women were ≤60 years old and had been diagnosed, on average, about 2 months prior to the baseline assessment.

Table 1 Sample Characteristics and Score Distribution (N = 246)
Full size table

Patient Trust and Physician Support Scores

As shown in Table 1, patients had relatively high levels of trust in their physicians; trust scores did not change over time (P > 0.05). At baseline, a majority of women reported receiving helpful informational (81%), emotional (62%), and decision-making (73%) support from physicians. However, these figures decreased over time for all types of support (P < 0.001).

Correlates of Patient Trust

As shown in Table 2, at baseline, receipt of helpful informational, emotional, and decision-making support from physicians were all significantly associated with patients’ trust. At the 2-month follow-up, helpful informational and emotional support from physicians in the 2 months between baseline and 2-month follow-up were significant along with helpful informational support at baseline. At the 5-month follow-up, only helpful emotional support from physicians in the 2 months prior to the 5-month follow-up assessment was significantly associated with patients’ trust. As per Cohen’s effect size definitions,11 effect sizes for the association between physician support and patient trust ranged from 0.26 (slightly more than a small effect) to 0.49 (almost a medium effect).

Table 2 Linear Regression Model Estimates Showing Association Between Perceived Helpfulness of Support Received from Physicians and Patients’ Trust at Baseline, 2-month, and 5-month Follow-up*
Full size table

DISCUSSION

In the period closer to diagnosis, informational, emotional, and decision-making support from physicians were all significantly associated with patient trust. However, 2 months later, decision-making support was not associated with trust. By the time of the 2-month assessment most women had completed surgery. Perhaps physician support for subsequent decisions such as receipt of adjuvant therapy, while important, may not have been as critical to patients’ trust perceptions as the initial “big” surgery decision.

While informational support continued to be associated with trust at 2-month follow-up, it was not significant at the 5-month assessment. Given that cancer patients/survivors report a number of information needs months and even years after diagnosis,12 this lack of association at 5-months needs further exploration.

Emotional support from physicians was most consistently associated with patient trust. Addressing patient emotions, however, remains an infrequently conducted task in oncology.13 Our findings underscore the need for skills training of health professionals and perhaps patients as well to ensure patients’ emotional needs are adequately addressed.

Interestingly, while perceived helpfulness of all types of physician support decreased over time, trust remained high and unchanged. We offer several potential explanations. First, it is possible that in the period closer to diagnosis when patients are the most vulnerable, receipt of any support from physicians is considered helpful and results in high trust. Over time, once past the initial crisis, patients might make more nuanced assessments of the helpfulness of support they receive from physicians. Perceptions of trust once formed, however, might remain invariant to such changes in perceived support. Second, it is possible that patients themselves may not expect as high levels of support from their physicians over time as they expected in the period closer to their diagnosis. Thus, decrease in support over time may not impact their trust ratings. Third, given that we lack information on responsiveness to change of our trust item, it is possible that the item could not detect actual changes in trust resulting from declining physician support. Future studies need to evaluate the reasons for these differential patterns of change in trust and support.

We note two major study limitations: (1) Our data were collected a decade ago; to what extent our findings may change in the current environment of increased consumerism needs examination. Our findings were, however, confirmed in a recent study that demonstrated an association between informational and emotional support from physicians and patient trust.14 (2) We were limited to assessing trust using a single item that was included in the original dataset of the CHESS trial. This item had good face validity as it assessed patients’ confidence in their physician, a core element of several definitions of trust;15–17 however, it has not been validated against existing multi-item trust scales. We do note that the score distribution of our item was similar to that reported for other trust scales,18,19 and the ceiling effect was less than that reported in another study that used a similar single item to assess patient trust.4 We acknowledge the need to replicate our findings with more comprehensive assessments of patient trust.

Despite potential limitations, we know of no other longitudinal study that has examined the association between physicians’ behavior and patient trust. Our findings suggest that the role of physician support in building and maintaining trust may change over time, with emotional support likely to be the most consistent long-term determinant of trust. Our study provides important insights for future longitudinal research needed to disentangle this relatively complex yet critical driver of quality of care.