Students enrolled in MD, PA, and NP programs from two US academic medical centers were recruited to participate in this cross-sectional study. One institution is private and located in the Northeast and the other is public and in the South. Both institutions have traditional campus-based MD programs and a combination of online or hybrid and/or campus-based PA and NP programs. MD-PhD and PA-MPH students were included. Additionally, while nursing students pursuing either a masters (Master of Science in Nursing) or doctoral degree (Doctor of Nursing Practice) were included, PhD in Nursing Science students at both institutions were excluded as their programs are not clinical in nature. Similarly, students in other clinical degree-seeking programs were excluded (e.g., dental, pharmacy, midwifery) given the narrower scope of patient care. Students across all years in the identified programs were included. A target response of 318 was calculated using a 95% confidence interval with a 5% margin of error based on a population size of 1820 [20]. Survey recruitment occurred between August 6 and October 9, 2023, through a combination of emails, flyers, and live in-person and online sessions.

Vanderweele’s Secure Flourish Index (SFI) has operationalized his concept of human flourishing, offering a method to begin to quantify this complex concept [10]. The traditional SFI scoring approach provides zero to ten points for each of 12 questions, divided into six domains of two questions each: Happiness and Life Satisfaction, Mental and Physical Health, Meaning and Purpose, Character and Virtue, Close Social Relationships, Financial and Material Stability. This produces an overall flourishing score with a possible range of 0–120 and six domain-specific flourishing scores with possible score ranges of 0–20 each. While providing a strong foundation for the assessment of flourishing, individuals may value the relative importance of each of these six domains differently, which is not accounted for in the current scoring approach that equally weights the relative value of each domain [13, 21].

For example, two individuals may both score a question on being content with their current social relationships a 7/10. However, one individual may not believe that social relationships strongly influence their ability to flourish while the other may believe that social relationships are foundational in flourishing. With the traditional scoring method, both individuals receive a score of 7/10, but this does not reflect the individual, perceived importance of social relationships to their ability to flourish. This study proposes an alternative scoring method where the participant provides relative domain weights totaling 100% across the six domains to create a new, self-weighted SFI score taking into account the individuals’ perceived importance of each domain (Table 1).

Survey data were collected and managed using REDCap electronic data capture tools hosted at the Medical University of South Carolina [22]. Students who agreed to the statement of research were asked to complete a series of validated instruments and demographic questions. Participants completed the SFI and then were asked to apply a relative percentage weight (0–100%) to each of the established six flourishing domains based on their perceived relative importance to their individual flourishing [10]. Domain weights (percentages) were then applied to SFI responses to produce a novel, self-weighted SFI score. The established scoring approach will be referred to as the traditional approach (tSFI) and the novel approach will be referred to as the self-weighted approach (swSFI). An example of the comparative scoring approaches, where the participant assigns different weights for each domain, is outlined in Table 1.

Using this example, the self-weighted SFI scoring approach results in a flourishing score that is four points higher than the traditional scoring approach. It also reveals that this individual placed a much higher relative value on Meaning and Purpose (45.0%) than on Close Social Relationships (5.0%).

Cases missing more than one tSFI response(s) were filtered out of the data set to avoid making multiple assumptions regarding the missing values. For cases missing an entry for one question in the tSFI, the score was imputed by duplicating the response for the second question in the domain for the missing score. For example, if question one was missing a response, the score from question two (also within the domain of Happiness and Life Satisfaction) was used as the score for the missing question one data. Following imputation, participant data was included only if the participant also completed the self-weighed SFI sections completely and correctly (swSFI total must equate to 100%), as complete responses were needed to accurately calculate scores (Table 1). Additional cases with missing data for demographic questions were not included in the respective analysis involving these questions; however, they were included in analyses of other questions and instruments that had complete data. The complete survey can be found in Appendix 1.

Descriptive statistics were used to summarize demographic characteristics, including means, standard deviations, medians, and interquartile ranges when applicable. The Bland–Altman (BA) plot was used to determine the agreement between the tSFI and the swSFI scoring approaches [23]. Bland–Altman plot graphs the difference (bias) between the methods (traditional SFI score − self-weighted SFI score) against the mean flourishing score (traditional SFI score + self-weighted SFI score)/2 [23]. Acceptable limits of agreement (LoA; precision) were not set a priori given the exploratory nature of the project and post hoc judgement was used to determine sufficiency of the LoA with 95% confidence limits (mean difference ± 1.96). The assumption of normality was met for each the tSFI, swSFI, and calculated differences in scores through visualization of approximate normal distribution using histograms. The Preiss-Fisher [24] procedure was conducted with 100 random pairings of the tSFI and swSFI scores to verify the data range was sufficiently wide to conduct the BA test, supporting that the results are unique to the pairing of the original data set and BA plots are appropriate for analyses. Data sets with small ranges may produce practically acceptable precision results secondary to any pairing of the data points, removing the value of the BA test. Surveys were only obtained once from each participant; therefore, analysis of repeatability is not applicable. A scatterplot was used to plot the mean differences (bias) on the vertical axis against the mean of the flourishing scores on the horizontal axis. Linear regression was carried out to identify trends across measurement sizes, addressing whether the agreement between the two scoring approaches varies when mean flourishing scores are low versus high [25]. The analysis was repeated using the natural log of the scores to further confirm the results.

The cases lying outside of the LoA were filtered from the data set and descriptive statistics were obtained for the outliers (n = 18), with the adjusted primary data set without the outliers (new n = 263), to investigate whether participants in these groups differed in terms of demographic characteristics. Means were compared using the Kruskal–Wallis test due to the unequal and small sample sizes. Data were analyzed using Statistical Package Social Sciences (SPSS) version 28 (IBM; Armonk, NY) and Microsoft Excel Version 16.48; significance was set at alpha of 0.05.

The study was approved by the Institutional Review Boards at the Medical University of South Carolina (protocol #00129125) and Yale University (protocol #2000035757).