Development of managed care in recent years has resulted in cost-containment practices such as gatekeeping, utilization review, and physician payment incentives, which are believed to undermine patient-physician trust [1–4]. Trust as a quality of healthcare measure [5–7] is important in medical treatment relationships and better health outcomes. Trust affects many important health attitudes, behaviors, and outcomes including medication adherence [8, 9], therapeutic effects [10], patient-physician communication, health promotion efforts [11], disputes, likelihood of malpractice claims [12, 13], and transaction costs [14].

The importance of trust and observation of declining trust [15, 16] in the healthcare setting has led to the examination of its predictors. Examined variables included patient and physician characteristics and behaviors [7, 16–19], quality and continuity of care [7, 18, 20–22], patient-physician communication [7, 12, 14, 23, 24], patient satisfaction [7, 14], physician payment methods [25], gatekeeping [26] and utilization review [27]. Many managed care plans, including health maintenance organizations (HMOs), have limited provider choice [28]. Studies [7, 9, 20, 22] have identified that the amount of physician choice is a predictor of, or was strongly associated with, provider trust. Hsu et al. found consumers reported greater trust in the provider on active physician choice, though the results became insignificant on adjustment of patient demographics [29]. However, the association of 'patient satisfaction with amount of physician choice' to 'trust in the provider' is not clear in the presence of other 'satisfaction with provider choice' variables.

Healthcare stakeholders have different agendas and conflicting goals in choice [30]. Consumers considered cost, physician choice, and coverage to be the most important criteria in choosing health plans [30–32]. Managed care plans are selective in provider contracting [33]. On the other hand, employers have applied fixed health contributions [34], flexible spending accounts, and health care reimbursement accounts [35] to contain costs. Consumers have to assume a greater responsibility in cost-sharing and more decisions in healthcare expense [36]. There is a predicament between cost, choice, and other consumer preferences. By identifying variables of 'satisfaction with amount of PCP choice', we can better understand 1) the association between physician choice and provider trust, and 2) if satisfaction with perceived amount of PCP choice was confounded by other PCP choice satisfaction variables. Findings would have implications for the enhancement of provider trust and would be of great interest to health plans, providers, or employers for the many benefits of provider trust.

The objectives of this study were to analyze domains that are predictive of provider choice satisfaction and to analyze the importance of variables that were categorized into these domains, including 'satisfaction with amount of PCP choice', on trust in the provider. A literature review driven conceptual framework that verified hypothetical domains predictive of provider choice satisfaction was adopted to identify variables for analyses [37]. The framework described five domains that may be related to PCP choice satisfaction: (1) consumer characteristics and health status, (2) information and decision, (3) trust in providers in general and trust in the insurer, (4) health plan financing and plan characteristics, and (5) provider characteristics and office management practices. These five domains each carry variables that may also be consumer preferences. Consumer preference was thus an implicit domain that overlapped with the five domains in the framework. For example, PCP being a specialist would be a provider characteristic and a consumer preference for a specialist. The five domains were abbreviated as 'consumer characteristics,' 'information and decision,' 'system and insurer trust,' 'plan characteristics,' and 'provider characteristics' respectively in this paper.

The specific aims of this study were: 1) to determine variables associated with the identified domains related to 'satisfaction with amount of PCP choice': consumer characteristics, information and decision, system and insurer trust, plan characteristics, and provider characteristics; 2) to determine the relative importance of each of these identified domains in predicting 'satisfaction with amount of PCP choice'; 3) to analyze the association between 'satisfaction with amount of PCP choice' and provider trust, after controlling for other provider choice satisfaction variables and potential confounders.

Two variables, 'income' and 'insurer trust,' involved 14% and 22% missing values respectively. Nevertheless, these two variables were rejected in the variable selection processes within domains and excluded from subsequent analyses. The mean, standard error of the mean, and standard deviation of 'insurer trust' before and after imputation were 36.5 [range possible: 11–55] (SE 0.37) SD = 7.80 and 36.6 (SE 0.29) SD = 6.89 respectively.

Table 1 shows the five models of specific aim 1 and their measurements. Variables with an asterisk indicated significant associations (p ≤ .05) within respective domains. The confounders (only related to specific aims 2 and 3) were listed in the same Table to show how they were measured. A total of 18 significant variables within four significant domains were found associated with 'satisfaction with amount of PCP choice.' The domain model of 'system and insurer trust', which had two identified variables 'trust in providers in general' and 'insurer trust', was only tested with the 'insurer trust' variable due to the dataset limitation of no overlapping observations. Coefficients of Hosmer-Lemeshow chi-square indicated good model fit of all models, except the single variable model 'system and insurer trust' and the domain was excluded from further analyses.

The 18 variables (Table 1, in bold) from the four significant domain models were: chronic health condition, visit to PCP > 10 times, speak to PCP outside medical office, main reason selected PCP (being assigned or little PCP choice), dispute with the PCP, would change PCP if could, years with the PCP, satisfaction with the PCP, satisfaction with the insurer, enough insurer choice, government insurance, insurer charged less if provider within network, insurer assigned a specific PCP, provider type, PCP always able to diagnose, PCP will listen with care and concern, PCP's care been extremely effective, and long waiting time to appointment. 'Assigned or little PCP choice' was one of the seven categories of the variable 'main reason selected PCP,' which varied in magnitude compared to the dichotomous variable, 'insurer assigned a specific PCP' in influencing the models. Variance Inflation Factor affirmed no multicollinearity concerns between these two variables and on all variables within the model.

In addition, consumers who reported satisfaction with their insurers and who reported insurer charged less if the provider was within network had 2.54 times (95% CI = 1.45, 4.46) and 2.44 time (95% CI = 1.35, 4.41) the odds respectively of reporting 'satisfaction with amount of PCP choice.' compared to those who did not. Consumers who reported they have enough insurer choice had 2.16 times (95% CI = 1.12, 4.17) the odds of reporting 'satisfaction with amount of PCP choice' compared to those who did not. For each additional year with the PCP, consumers had 1.06 times (95% CI = 1.01, 1.12) the odds of reporting 'satisfaction with the amount of PCP choice'. On the other hand, consumers who had dispute experience(s) with the PCP had a reduced odds of 74 percent (95% CI = 0.12, 0.59) in reporting 'enough PCP choice' compared to those who did not. Consumers who experienced 'long waiting time to appointment' also had a reduced odds of 65 percent (95% CI = 0.17, 0.68) in reporting 'satisfaction with amount of PCP choice' compared to those who did not.

The full and reduced models (n = 564) accounted for 47 percent and 37 percent of the variations in 'provider trust' (p < .001) respectively. A significant F (14, 510) = 19.68, p < .001 also indicated a linear relationship between the dependent and independent variables [45]. The Variance Inflation Factors of individual independent variables were ≤ 1.44, suggested no issue of multicollinearity among the variables. The full model showed eight significant variables associated with 'provider trust'. They were: 'enough PCP choice', 'dispute with PCP', 'years with the PCP', 'PCP's care has been extremely effective', 'long waiting time to appointment,' 'second opinion on PCP', consumer's gender being 'female', and consumer's 'education' level.

Consumers who felt that their 'PCP's care been extremely effective' was highly associated with provider trust (β = 7.1, robust SE 0.95, p < .001). The experience of having a second opinion on the PCP was negatively associated with provider trust (β = -6.44, robust SE = 0.94, p < .001). 'Enough PCP choice' has a positive association with 'provider trust' (β = 1.89, robust SE = 0.52, p < .001). 'Dispute with the PCP' and 'long waiting time to appointment' were both negatively associated with provider trust (β = -1.75, robust SE = 0.83, p < .035, and β = -1.19, robust SE = 0.56, p < .035 respectively). The consumer being female has some positive association with provider trust (β = 1.05, robust SE = 0.41, p < .01). Consumer's education level has some association with provider trust (β = 0.63, robust SE = 0.24, p < .01). 'Years with PCP' also has a slight association with provider trust (β = 0.07, robust SE = 0.03, p < .03).

In this study, a total of 18 variables have been identified as associated with 'satisfaction with amount of PCP choice' when analyzed in each of the five domains of the conceptual framework. The 'system and insurer trust' domain was not fully tested due to data limitation and poor model fit. However, the variable 'trust in providers in general' of this domain had strong potential influence on interpersonal provider trust [38]. It should be reconsidered for examination in future studies. Analyses validated four of the five domains to be associated with 'satisfaction with amount of PCP choice' when each domain was examined independently. When the significant variables of these four domains were examined together in a single model to identify their relationship with 'satisfaction with amount of PCP choice', the domain 'consumer characteristics' was no longer associated with 'satisfaction with amount of PCP choice'.

Having 'enough insurer choice' was associated with 'satisfaction with amount of PCP choice' but not 'provider trust'. This appeared to be logical as plan choice preceded and necessarily impacted provider choice. Eighty-three percent of participants (n = 150) who reported enough insurer choice also reported enough PCP choice. Among participants who reported not enough insurer choice, 33 percent (n = 414) reported not enough PCP choice. 'Enough insurer choice' appeared to have some relationship with 'enough PCP choice.' 'Enough PCP choice', on the other hand, was found to be associated with provider trust. This confirmed the existing knowledge that enough or some amount of PCP choice was associated with provider trust.

When a plan had the characteristic of charging less if the providers were within network, it was strongly associated with greater odds in the reporting of 'satisfaction with amount of PCP choice.' This result suggested that this plan characteristic, which has the connotation of cost saving, was relatively popular among consumer preferences. A plan characteristic and consumer preference match explained the greater odds reported. Early studies of consumer behavior in plan choice suggested consumers were cost sensitive [46] while others saw provider choice as more important than benefits coverage [47]. Since Preferred Provider Organization (PPO) plans were gaining popularity over HMO plans in past years [48], it was not unreasonable to postulate that consumers had to make trade-offs between plan cost and provider choice, and that consumer preference might oscillate between the two depending on other factors, such as the proportion of employer contribution in health benefits.

'Years with the PCP' has some association with 'satisfaction with amount of PCP choice.' The duration of the patient-physician relationship indicated that the patient volunteered to stay with the provider for that length of time, or that they were required to stay with the provider for some reasons. Patients would have switched providers if the relationships were not good and they have a choice or a better choice. Nevertheless, involuntary provider switch behavior appeared to be more an issue of "employer-imposed disruption". Voluntary disenrollment from PCP practice was, however, associated with trust [49]. In the provider trust model, the variable 'years with the PCP' was also associated with the outcome of provider trust. In this case, the association was likely due to the issue of continuity of care, which may enhance provider trust [50].

'Long waiting time to appointment' has some negative association with both 'satisfaction with amount of PCP choice' and 'provider trust.' The consumer might not know if a certain provider has a 'long waiting time to appointment' unless they had known the provider before or had made efforts to find out how long it normally takes to secure an appointment. The variable also reflected the consumer preference for access to prompt appointments. The association with 'provider trust' would likely be due to a patient's perception of the provider's level of control in providing efficient services. "Control" refers to the physician's autonomy in providing needed services to the patient in a timely manner, which is a building block of physician trust [37].

The association of 'second opinion on the PCP' (due to concern of care) with lower provider trust involved a temporal issue of whether provider trust was already low at the time a second opinion was sought or if the second opinion confirmed the consumer's concern and hence decreased trust. This is a limitation of the cross-sectional study design not being able to measure the dynamic dimension of trust. Regardless of whether trust was low, declined before, after, or continued to decline after a second opinion on the PCP, the experience of having sought a second opinion on the PCP demarcated a decrease in trust.

Eight variables associated with 'satisfaction with amount of PCP choice' were analyzed in the model of 'provider trust' with and without potential confounders. Consumers' perception of the PCP's care being extremely effective was strongly associated with both 'satisfaction with amount of PCP choice' and with 'provider trust'. This finding was consistent with the consumer preference for professionally qualified providers [19], for having information about the provider's performance including error rates and adverse outcomes [31], and with the definitions of trust that the provider has the competencies to enhance the technical aspects of care [37].

Contrary to the existing knowledge that minorities, especially Blacks and Hispanics, preferred a provider of the same ethnic background or who speaks the same language [51], this study found that 'racematch' was not significantly associated with 'satisfaction with amount of PCP choice'. It was also interesting to find that 'consumer education' was negatively associated with 'provider trust.' It suggested the higher the education level the more likely that the consumer would "challenge" the provider's competencies in patient care. It was not clear if the high proportion of female subjects (67%) in the sample might have contributed to the association of 'consumer being female' to 'satisfaction with amount of PCP choice' and to 'provider trust'.

Future studies

Results suggested the adapted conceptual framework can be modified. Figure 1, titled 'PCP choice satisfaction-related associations and outcomes with modifications based on empirical findings of this study', illustrates the relationships of the four relevant domains. 'Consumer characteristics' was no longer significant. The revised conceptual framework constituted four renamed domains: plan related, provider related, trust related, and information related. Other than plan or provider related variables, trust or information related variables require more exploration. Conflicting information and biased marketing materials often lead to information distrust [30]. 'Trust in provider performance reports' should also be explored under the 'trust related' domain in future studies to see its influence on consumer choice behaviors.

The provider's professionalism has never been explored as a predictor of provider trust before and it deserves attention in future studies. Future studies of healthcare public reporting, of provider specific information, and of trust in information available to assist choice will unveil the roles of information in consumer preferences, in consumer satisfaction, and in provider trust. If possible, examining all the known predictors of provider trust in one study may provide a holistic view of what is relatively important in predicting provider trust and in sustaining trust.