High-Value, Cost-Conscious Care Attitudes in the Graduate Medical Education Learning Environment: Various Stakeholder Attitudes That Residents Misjudge

We conducted a cross-sectional survey (distributed between June 2017 and December 2018) of residents, faculty, administrators, and patients across the Netherlands. The Ethical Review Board of the Netherlands Association for Medical Education approved this study (no. NERB814, amendment no. NERB817) before launch.

The Netherlands and the USA are among the ten highest healthcare expenditure per capita countries in the world.⁴⁸ In the Netherlands, 12.9% of the gross domestic product was spent on healthcare in 2018. In the USA, this percentage was 17.8. The Dutch healthcare system largely relies on public resources, whereas the American healthcare system depends mainly on the private sector.⁴⁹ In the Netherlands, residents are trained in the workplace under supervision by faculty and have internships within several non(academic) hospitals, which is similar to residency education in the USA. Administrators in both contexts attend to the financial concerns of healthcare organizations. For the patient population in the Netherlands, there is a uniform approach where every patient has insurance and relatively low out-of-pocket costs. In the USA, nearly 1 in 12 persons are uninsured,⁵⁰ health insurance is much more variable, and out-of-pocket costs can be substantial.⁵¹

Survey items included the Dutch version of the Maastricht HVCCC Attitude Questionnaire (MHAQ), which measures key dimensions of HVCCC⁴ on three subscales: (1) high-value care, the degree to which the respondent thinks physicians should be responsible for the provision of high-value care; (2) cost incorporation, the degree to which the respondent thinks physicians should integrate costs in daily clinical practices; and (3) perceived drawbacks, which reflects the respondent’s beliefs about potential negative consequences of HVCCC (see supplemental Appendix A for a description and example item for each subscale⁵²). For the first two subscales, a higher score reflects a more positive attitude; for the third subscale, a lower score reflects a more positive attitude. Respondents indicated their extent of agreement using a four-point Likert scale (1 = strongly disagree, 4 = strongly agree).

Demographic items for residents and faculty addressed age, gender, hospital name, training region, clinical experience (in years), training location, and specialty (surgical, non-surgical, and supportive, see Supplemental Appendix B). Items for administrators included age, gender, hospital name, training region, and type of administrator. Items for patients included age, gender, hospital name, medical specialty visited, number of inpatient admissions and/or outpatient visits (average per year), estimated number of physicians treating the patient, and self-perceived health (rated using a seven-point scale, from 1 = very bad to 7 = very good).

The resident version of the survey also included two items asking respondents to estimate the HVCCC attitudes of other stakeholders. Specifically, we asked residents to indicate the extent to which faculty, administrators, and patients would agree with a central item from subscale 2 and subscale 3 using the same four-point Likert scale. We decided not to ask an item from subscale 1 because general guidelines in medicine indicate physicians should provide high-value care to patients.⁵³

Finally, residents answered the job demands-resources questionnaire,⁵⁴ a 40-item instrument that measures perceptions of five job demands, including work pressure, cognitive demands, emotional demands, role conflict, and hassles, and five job resources, including autonomy, social support, feedback, opportunities for development, and coaching (see supplemental Appendix C for a description and example item for each subscale). Residents responded to these items using one of two five-point Likert scales (either 1 = never, 5 = very often, or 1 = strongly disagree, 5 = strongly agree).

We approached hospital education committees from all academic training regions (N = 8), after which these committees recruited residents and faculty in several hospital meetings and formal educational events. We also recruited residents and faculty via a digital newsletter of the Dutch “Bewustzijnsproject,” a project advocating HVCCC in postgraduate education.² Authors F.S. and L.S. recruited administrators from several hospitals. We approached patients via patient platforms (e.g., patient panel of a hospital), via a health insurance company in the Netherlands, and before and after consults in hospitals (after gaining approval from the hospital).

All residents, faculty, and administrators answered the questionnaire online; patients also had the option to answer on hardcopy. All respondents received an information letter and signed an informed consent.

Descriptive summary statistics were reported as means with standard deviations or frequencies with percentages, as appropriate. We performed multiple imputations to deal with missing values.⁵⁵ As participants were nested in hospitals, we used multilevel analyses to answer the research questions. We used multilevel ANOVA’s to compare the means of different stakeholders’ attitude scores on the subscales of the MHAQ. We assessed residents’ estimations of other stakeholders’ HVCCC attitudes and compared these with the actual HVCCC attitudes of these stakeholders, using multilevel independent t tests for the aforementioned two items. We used multilevel regression analyses to examine the relations between HVCCC attitudes scores and the measured independent variables. We used a backward elimination procedure, as we had not set a sequence of adding variables a priori.⁵⁶ For all analyses, we estimated the population mean (μ) and accompanying standard deviation (SD). We calculated the regression slope using β, determined statistical significance at p < 0.05, and used confidence intervals (CI) when appropriate. Additionally, we calculated the effect size Hedges’ g (g) and considered an effect size above 0.8 as large.⁵⁷ We conducted all analyses using IBM SPSS Statistics 25.0 for Windows (Armonk, NY: IBM Corp.).

Figure 1 visualizes the mean scores per subscale and stakeholder group. Overall, faculty (abbreviated f) and administrators (abbreviated a) demonstrated more positive attitudes toward high-value care than patients (abbreviated p) (μ_f-p = 0.08, p = 0.002, g = 4.19; μ_f-p = 0.11, p = 0.03, g = 5.97) (Table 1). They were also more likely to agree physicians should integrate costs in daily clinical practices compared with residents (abbreviated r) (μ_f-r = 0.10, p = 0.005, g = 3.33; μ_a-r = 0.21, p = 0.001, g = 6.13) and patients (μ_f-p = 0.15, p = < 0.001, g = 6.54; μ_a-p = 0.27, p = < 0.001, g = 7.36). Residents endorsed more drawbacks to HVCCC than faculty (μ_r-f = 0.13, p = < 0.001, g = 4.91) and administrators (μ_r-a = 0.28, p = < 0.001, g = 8.20), but fewer drawbacks than patients (μ_r-p = − 0.10, p = 0.006, g = 5.33). Faculty, in turn, believed HVCCC had more potential drawbacks than administrators (μ_f-a = 0.15, p = 0.03, g = 4.29), while both faculty and administrators believed HVCCC has fewer drawbacks than patients (μ_f-p = − 0.23, p = < 0.001, g = 12.07; μ_a-p = − 0.38, p = < 0.001, g = 18.16).

Residents underestimated the attitudes of faculty regarding the physician’s duty to incorporate costs in daily practice (p = < 0.001, g = 8.03) and overestimated faculty’s attitude toward the drawbacks of HVCCC (p = < 0.001, g = 8.30) (Table 2). We found no significant differences between residents’ estimations on administrators’ attitudes and administrators’ embraced attitudes. Residents underestimated patients’ beliefs that HVCCC has drawbacks (p = < 0.001, g = 6.55); patients identified more drawbacks than residents were aware of.

Among residents, male gender and a non-surgical specialty were associated with more favorable attitudes toward high-value care (β = 0.09, 95% CI = 0.03, 0.16, p = 0.006; p = 0.006). However, male residents also endorsed more potential drawbacks (β = 0.13, 95% CI = 0.04, 0.21, p = 0.004), as did younger residents (β = 0.02, 95% CI = 0.03, 0.00, p = 0.01) and residents working in a surgical specialty compared with a supportive specialty (p = 0.03).

Among faculty, practicing in a supportive specialty was associated with more favorable attitudes toward high-value care compared with a surgical specialty (p = 0.002) and toward incorporation costs in daily practice compared with a surgical and non-surgical specialty (p = 0.02; p = 0.02). Older faculty members were more likely to believe physicians should incorporate costs (β = 0.01, 95% CI = 0.00, 0.008, p = 0.01) and less likely to endorse drawbacks of HVCCC (β = − 0.01, 95% CI = − 0.01, − 0.00, p = 0.001). Hospital region was significantly associated with faculty’s attitudes toward high-value care (p = 0.002) and cost incorporation (p = 0.004), but not with their beliefs about potential drawbacks. Among administrators, none of the variable measures predicted their attitudes toward HVCCC.

Among patients, more frequent hospital visits were associated with less favorable attitudes toward high-value care (β = − 0.00, 95% CI = − 0.01, − 0.00, p = 0.03) and cost incorporation (β = − 0.01, 95% CI = − 0.01, − 0.00, p = 0.04). A higher number of physicians treating the patients was associated with more favorable attitudes toward cost incorporation (β = − 0.03, 95% CI = − 0.05, − 0.00, p = 0.03). More positive perceptions regarding their own health condition were associated with more favorable attitudes toward high-value care (β = 0.03, 95% CI = 0.01, 0.06, p = < 0.001) and cost incorporation (β = 0.06, 95% CI = 0.03, 0.08, p = < 0.001), and less endorsement of drawbacks of HVCCC (β = − 0.07, 95% CI = − 0.10, − 0.05, p = < 0.001).

The more autonomy residents reported, the more positive their attitudes toward high-value care (β = 0.05, 95% CI = 0.01, 0.10, p = 0.03) and cost incorporation (β = 0.07, 95% CI = 0.02, 0.11, p = 0.008), and the fewer drawbacks they perceived (β = − 0.11, 95% CI = − 0.16, − 0.05, p = < 0.001). Higher perceived work pressure also related to more positive attitudes toward high-value care (β = 0.05, 95% CI = 0.02, 0.09, p = 0.004) and cost incorporation (β = 0.09, 95% CI = 0.04, 0.15, p = < 0.001). Opportunities for development positively related to high-value care (β = 0.07, 95% CI = 0.02, 0.12, p = 0.01), while supervisory coaching negatively related to high-value care (β = − 0.06, 95% CI = − 0.12, − 0.01, p = 0.02). Additionally, greater cognitive demands related to more favorable attitudes toward cost incorporation (β = − 0.09, 95% CI = − 0.16, − 0.02, p = 0.01), and the more emotional demands residents perceived, the more drawbacks they perceived (β = 0.08, 95% CI = 0.02, 0.14, p = 0.01). See Supplemental Appendix E for a complete overview of measured independent variables for each stakeholder group.

Our findings have several implications. First, it may be beneficial to forthrightly address the variation of HVCCC attitudes in the learning environment to help residents navigate the varying interests and perspectives they encounter. Senior faculty and specialty-specific teaching⁷² could play a role in this process. Second, the discrepancies between faculty’s attitudes and residents’ estimation of their attitudes suggest a need for faculty to more explicitly share their attitudes toward HVCCC and demonstrate these through their behavior, thereby setting good examples. Such role modeling is critical, as it represents a key mechanism by which social and occupational norms are transmitted.^{18, 35, 37} Patients also set norms for residents in postgraduate education,⁷³ but are misjudged by residents. Training residents effective and empathic strategies for eliciting patients’ concerns and communicating the benefits of HVCCC may lead to more accurate perceptions and help boost patients’ confidence in such care. Finally, more autonomy at the workplace might contribute to more favorable HVCCC attitudes and behaviors, although further studies are needed to better understand the direction of associations with autonomy and work pressure, and explore potential trade-offs (e.g., related to patient safety or resident well-being).

A strength of this study is that we surveyed multiple stakeholders from (non)academic hospitals geographically distributed over the Netherlands, supporting the study’s generalizability. We pursued several options recruiting respondents to reduce selection bias, and the large number of respondents per stakeholder group supports the representativeness of each sample. However, our study also has limitations. Despite the large sample, we cannot assure that participants represent the entire population, because of bias in who decided to fill out the questionnaire. This bias is hard to prevent, but we hope our large sample warrants good quality of data. We saw an acceptable but considerable rate of missing responses with patients, due to patients indicating not knowing the answer. As missing values were at random across all items, it was unlikely that this affected our results. Additionally, we are unable to report a response rate given the nature of our sampling strategy. We also did not include all potential stakeholders in the clinical learning environment, e.g., nurses might also affect residents’ thinking about HVCCC and could be involved in future research.