From goal to outcome: Analyzing the progression of biomedical sciences PhD careers in a longitudinal study using an expanded taxonomy

Abstract Biomedical sciences PhDs pursue a wide range of careers inside and outside academia. However, there is little data regarding how career interests of PhD students relate to the decision to pursue postdoctoral training or to their eventual career outcomes. Here, we present the career goals and career outcomes of 1452 biomedical sciences PhDs who graduated from Vanderbilt University between 1997 and 2021. We categorized careers using an expanded three‐tiered taxonomy and flags that delineate key career milestones. We also analyzed career goal changes between matriculation and doctoral defense, and the reasons why students became more‐ or less‐interested in research‐intensive faculty careers. We linked students' career goal at doctoral defense to whether they did a postdoc, the duration of time between doctoral defense and the first non‐training position, the career area of the first non‐training position, and the career area of the job at 10 years after graduation. Finally, we followed individual careers for 10 years after graduation to characterize movement between different career areas over time. We found that most students changed their career goal during graduate school, declining numbers of alumni pursued postdoctoral training, many alumni entered first non‐training positions in a different career area than their goal at doctoral defense, and the career area of the first non‐training position was a good indicator of the job that alumni held 10 years after graduation. Our findings emphasize that students need a wide range of career development opportunities and career mentoring during graduate school to prepare them for futures in research and research‐related professions.


| INTRODUCTION
The biomedical research community has been debating the appropriate size and composition of the PhD-trained workforce since Congress passed the National Research Service Awards (NRSA) Act of 1974.The legislation required periodic reviews of the program to ensure it was funding an appropriate number of pre-and postdoctoral trainees to sustain the biomedical research workforce, which was primarily centered in academic institutions at the time. 1 Indeed, in the early 1980s when the NSFsponsored Survey of Doctorate Recipients first began tracking tenure-line faculty appointments among U.S. citizens, nearly two-thirds of biomedical PhDs worked in academia and 44% were employed as tenured or tenuretrack professors. 2 Over the next 50 years, the number of life sciences PhDs working in the U.S. quadrupled without a corresponding increase in employment in faculty jobs. 1,3Much of this growth was absorbed by the private sector, with PhDs finding work in both research-and research-related jobs in the flourishing biotechnology industry.][6][7] Historically, the main source of data about the biomedical research workforce comes from two NSF-sponsored surveys: the Survey of Earned Doctorates given to doctoral students at graduation since 1957, and the longitudinal Survey of Doctoral Recipients, administered since 1973. 8hile these surveys provide useful information about the number of PhDs granted and the general types of jobs and sectors in which biomedical PhDs worked, they have numerous limitations 4,8 and do not provide granular data about individual career goals or trajectories.Few institutions have employed systematic approaches to collecting and analyzing data about the careers of their program graduates, and few have adapted their training programs to provide comprehensive career information and preparation to their trainees or meet the needs of the evolving biomedical research workforce.
It is against this backdrop that we began to conduct exit surveys with Vanderbilt University School of Medicine PhD students at the time of their doctoral defense, collecting information about their immediate next steps as well as their long-term career goal.We also began longitudinal tracking of all Vanderbilt University School of Medicine doctoral alumni who matriculated since 1992 and created a classification system to categorize biomedical PhD career outcomes.Our career outcomes taxonomy resembles recently published classification systems, 9,10 with a few notable differences, including the use of novel flags to distinguish key milestones in the career trajectories of alumni (Appendix S1).We aimed to develop an accurate understanding of how our PhD students' career plans and careers evolve over time to inform the alignment of our biomedical PhD training programs with workforce needs and to inform our career development initiatives for trainees.
With nearly 25 years of alumni outcomes data and 15 years of PhD student exit survey data in hand, we undertook the present study of the early career trajectories of 1452 biomedical PhD alumni who matriculated since 1992.We analyzed how career goals changed during graduate school, and the reasons students changed career goals away from, or toward, being an academic PI.We also examined how career goal at the time of doctoral defense related to whether alumni did a postdoc or not, and for how long.Consistent with national trends, 11 we observed a small but steady decline in the percentage of biomedical alumni pursuing postdocs each year.Finally, we followed alumni career trajectories after PhD completion and evaluated how career goal at doctoral defense related to their first non-training position and their career area at 1, 3, 5, and 10 years after PhD completion.

| Study population
The population examined in this paper included graduates of Vanderbilt University's biomedical sciences PhD programs who (a) were admitted in 1992 or after through one of two umbrella PhD admissions programs, the Interdisciplinary Graduate Program (IGP) or the Quantitative and Chemical Biology Program (QCB), or (b) were directly admitted or transferred to a program affiliated with career mentoring during graduate school to prepare them for futures in research and research-related professions.

K E Y W O R D S
biomedical research, career choice, PhD training, postdoctoral, science, workforce IGP/QCB programs and (c) graduated with a PhD between 1997 and August 31, 2021.Data from six deceased individuals was excluded from all datasets examined.Students who were enrolled in Vanderbilt's dual MD/PhD program were not part of the study population as their career outcomes 10 years after earning a PhD can be very different from those of the population reported in this paper.Sixtynine alumni were excluded from all analyses because they did not have an exit survey, a first known position after PhD, and a known position for at least one of the timepoints examined in this analysis (1, 3, 5, or 10 years after PhD).
Of the 1452 PhD alumni who graduated during this time period, 99% earned a PhD from one of the following current or legacy biomedical PhD programs: Biochemistry, Biological Sciences, Cancer Biology, Cell Biology, Cell and Developmental Biology, Chemical and Physical Biology, Human Genetics, Microbe-Host Interactions, Microbiology and Immunology, Molecular Biology, Molecular Pathology and Immunology, Molecular Physiology and Biophysics, Neuroscience, or Pharmacology; the remaining 1% of students earned a PhD in Biomedical Engineering, Biomedical Informatics, Chemical Engineering, Chemistry, Interdisciplinary, Physics, or Psychology.The n for analyses varied based on the timepoint or milestone restrictions for the analysis and the maximum cohort size for any figure was n = 1413.Table 1 shows the inclusion/ exclusion criteria for each figure.All cohorts analyzed had the following demographic breakdown: 52%-60% female, 8%-15% underrepresented in the biomedical sciences, and 14%-19% foreign nationals with a temporary US visa, and no demographic group was over-or underrepresented relative to our population of biomedical sciences PhD students at Vanderbilt.

| Exit survey
Starting on January 1, 2007, all biomedical PhD students at Vanderbilt University completed a voluntary online exit survey (IRB #171216) within 2 to 3 weeks after their dissertation defense.The exit survey was implemented to collect student feedback on their PhD training, career goals, and immediate employment plans.A total of 944 students completed the exit survey, including 20 alumni who defended in 2005 and 2006 and served as a focus group to help validate the survey.The overall survey response rate from January 1, 2007 to September 8, 2021 was 90%, calculated as the [(number of students who completed the survey) ÷ (# of students invited to complete the survey)].
In the exit survey, students were asked the following questions about their career goals at matriculation and doctoral defense: (1) "Which best describes your primary long-term career goal when you STARTED graduate school?" and (2) "Which best describes your primary longterm career goal NOW, as you finish graduate school?".Answer options for both questions were: Faculty position in academia (primarily research), Faculty position in academia (primarily teaching), Science Outreach/K-12 Education, Research position in industry, Research position in government/nonprofit lab, other (please specify); "I didn't know what I wanted to do after graduate school" was also an answer option for the question about career goals at matriculation.In 2009, the following answer options were added: Science or Medical Writing/Editing/ Publishing, Patent Law, Science Policy, and Undecided.The exit survey also contains an optional, free-text question to gather information about why students changed their career goals during graduate school: "If your career goals changed since you started your PhD program, what factor has most affected these goals?"While asking students to report their career goal from several years earlier may have the potential for recall bias, our results on changing career goals during PhD training are consistent with previous studies that examined student career goals at the specific timepoints in question 12,13 as well as studies that asked students to recall their career goals from earlier timepoints [14][15][16] .

| Coding of free-text responses
We analyzed free-text responses about reasons for changing career goals from two groups: those whose career goals between matriculation and doctoral defense changed (1) away from or (2) toward becoming a research-intensive faculty member in academia.Out of 944 exit survey respondents, there were 171 individuals in the first category, 153 of whom provided comments.There were 100 individuals in the second category, 68 of whom provided reasons for switching career goals.
The coding process began with a member of the research team using inductive thematic analysis to identify semantic themes in the comments of both groups. 17In a first reading of comments, two main types of themes were identified that we called either: (1) external factors, relating to the respondents' perceptions about the job of being a PI or working in academia and (2) internal factors, relating to the respondents' perceptions about themselves or fit with a career option, or their personal experience in graduate school.In a second reading of comments, eight subthemes (four external and four internal) were noted for individuals whose career goals changed away from being a research-intensive faculty member in academia, and four subthemes (all internal) were identified for individuals whose career goals changed toward being      3 Not applicable Table 4 Table 5A Table 5B Table 6 Not applicable Table 7 Table 8 Table 9 Table 10 Table 11 Table 12 Graduation dates (MM/DD/YYYY) a research-intensive faculty member in academia.To be defined as a subtheme, a sentiment's prevalence across the data set was considered, and a sentiment had to be expressed by at least 10 people within either group of survey respondents.Themes and subthemes were verified by a second member of the analysis team, then both team members independently coded each comment.A single comment could be assigned multiple codes if the survey respondent mentioned more than one theme or subtheme in their survey response.The two coders resolved discrepancies in coding by clarifying subtheme definitions and consensus building until there was complete agreement in coding.Nonspecific responses for both groups were coded, "Not descriptive enough to categorize."

| Career outcomes
The career outcomes of biomedical PhD alumni were collected using Google searches of publicly available information, including PubMed (RRID: SCR_004846), ResearchGate (RRID:SCR_006505), ORCID (RRID:SCR_008700), LinkedIn (www.linkedin.com), and institutional websites.Alumni who could not be found by a single member of our team were included in follow-up searches by two additional team members.Specific data captured for each position held by an alumnus included job title, employer, location of employment, and inferred start date.Position start dates allowed us to determine the length of time alumni remained in each position by calculating the time from the start date of the first position to the start date of the second position.We were able to find career outcomes for 95% of our alumni who graduated between January 1, 1997, and August 31, 2021.Employment information was stored in a database we developed and optimized for tracking employment outcomes over time.

| Relational database and data processing
A PostgreSQL 18 relational database was created to store static academic records and dynamic outcomes data about our graduates.The database is deployed onto a secure institutional server, provides multi-user access, and maintains project data integrity.To generate reports and visualize the data, we used commercially available software tools, Aqua Data Studio (https://www.aquafold.com) and Tableau Desktop (RRID:SCR_013994); database revision history and process documentation are stored in a wiki, Atlassian Confluence (https://www.atlassian.com/softw are/confl uence).

| Career categorization
To categorize alumni careers, we used a three-tiered taxonomy developed in-house in 2012 to classify career outcomes based on sector of employment, type of career, and job function (Appendix S1).Our strategy to classify each position by sector and career activity was based on the categorization of STEM PhD careers in the NSF Survey of Doctoral Recipients.Our initial list of job functions was developed organically by grouping similar careers together, then subsequently modified to achieve greater consistency with the classifications used in crossinstitutional surveys by the NIH Broadening Experiences in Scientific Training (BEST) consortium. 19Between 2012 and 2015, we refined our categorization system to include an additional taxonomy for classifying faculty jobs, capturing secondary jobs (e.g., adjunct professorships or starting a side-business; only primary positions were included in this study), flagging career milestones of the first position after PhD completion and the first non-training position, assigning Carnegie IDs to classify employment in US institutions of higher education, and flagging positions that are science-related, entrepreneurial, or executive-level.
The classification system we developed is similar to the three-tiered Unified Career Outcomes Taxonomy 9 and further optimized by Stayart et al. (2020), 10 with differences described in Appendix S1.At least two experienced career outcomes professionals independently categorized every job, and every position that was not identically classified was reviewed to reach a consensus category.
For the analyses in this paper, we also used our taxonomy to categorize student career goals at the time of matriculation and doctoral defense.Respondents who answered, "Other (please specify)" were asked to describe their career goals in free text, and each of these responses was coded to the best of our ability.Some "Other" responses were very specific and easy to code in all three tiers of our taxonomy, for example, "lab manager in academia" was coded as Career Sector: "Academia", Career Type: "Primarily Research", Job Function: "Research Staff or Technical Director".In contrast, some "Other" responses were nonspecific and could not be coded in all three tiers of the taxonomy.In these cases, we coded the available tier(s) and coded the remaining tier(s) as, "Unspecified."For example, an "Other" response described as, "Industry" was coded as Career Sector: "For-profit", Career Type: "Unspecified" and Job Function "Unspecified".Free-text responses that indicated that the student was unsure (e.g., "I don't know") or considering multiple options (e.g., "science writing or clinical research") were coded as, "Undecided."

| Timepoints and milestones
This study examined alumni career outcomes at four timepoints after graduation, and at two key milestones that are important for understanding the career trajectories of biomedical scientists: the first position after completion of the PhD and the first non-training position.The career outcome at 1, 3, 5 or 10 years after graduation was defined as the job held by an alumnus on the exact date 1, 3, 5, or 10 years after the date of degree conferral recorded by the Vanderbilt Graduate School; Vanderbilt University has three standard graduation dates annually, in May, August, and December.The two milestones were manually designated (or "flagged") by career outcomes coders.The first non-training position was defined as the first job that is not postdoctoral training, a degree-granting program, or some other term-limited training position (e.g., medical residency or fellowship.) The first position after completion of the PhD was defined as the first job or postdoctoral position that is not in the same lab in which the alumnus completed their PhD research.An exception to this rule was made if a postdoctoral position in the dissertation lab lasted longer than 6 months.We developed this procedure because about 30% of our alumni did a short postdoc of 6 months or less in their dissertation lab prior to taking a job or full-term postdoc in another lab.Alumni reported in the exit survey that they did these short postdocs for a variety of reasons, often to time a move to coincide with the start date of their next position or to coincide with the needs of their family or partner.Although short postdocs in a PhD lab were not flagged as the first position after PhD completion unless they were longer than 6 months, the duration of these short postdocs was included in the calculation of time to first non-training position and duration of postdoctoral training.Under our classification system, positions were categorized as "postdoctoral" based on the position title.Most postdoctoral positions carried unambiguous titles like "postdoc" or "postdoctoral fellow" and were easy to classify.Ambiguous titles like, "fellow" or "research associate," were investigated further on institutional websites and in postdoctoral handbooks to ascertain the appropriate category.

| Statistics
Logistic regression was used to examine the significance of trends in alumni entering postdoctoral training as their first position after PhD completion.For regression analysis, graduation year was treated as the independent variable, and choice to enter postdoctoral training as a binary (yes/no) dependent variable.Regression analysis and visualization, likelihood ratio testing, and calculation of odds ratios were conducted in R v. 4.2.2 20 using lmtest 21 and ggplot2 22 packages.Percent decrease in probability was calculated as the difference in probabilities of entering a postdoctoral position at the initial full-year timepoint (1997) and end full-year timepoint (2020), all divided by the probability at the initial timepoint.
For alumni who entered at least one postdoctoral training position, the length of time between a graduate's doctoral defense and entry into their first nontraining position was calculated.These calculated time-to-event values were used to generate cumulative incidence curves grouped by individuals' career goals at doctoral defense, which represent an individual's probability of entering a first non-training position at any point in time following doctoral defense.Cumulative incidence curves were generated using (1 − survival probability) determined by the Kaplan-Meier survival calculation method. 23This method is robust in that it allows for measurements from individuals who have not entered a first non-training position by the end of the observation period ("censored" individuals) to contribute to calculations for the probability of entering a first non-training position.Observations used in Kaplan-Meier calculations were coded into two separate groups: 1) non-censored (observations from alumni who entered their first non-training position), and 2) censored (observations from alumni who defended their doctoral thesis but had not entered their first non-training position by 5/1/2020, the cutoff date for recorded observations).For non-censored observations, time-to-event was measured as the difference between doctoral thesis defense and entry into first non-training position.For censored observations, time-to-event was measured as the difference between doctoral thesis defense and the cutoff date for recorded observations.One individual pursued a first non-training position prior to defending their thesis, resulting in a negative time in training calculation; this individual was not included in the analysis.Kaplan-Meier calculations make use of both censored and non-censored observations, providing probability inputs for the final cumulative incidence curve.Probability calculations and visualizations for cumulative incidence curves were conducted using GraphPad Prism v. 9.5.1 for Windows (GraphPad Software.RRID:SCR_002798) For alumni with known career outcomes 10 years after graduation, Kruskal-Wallis with Dunn's multiple comparisons testing was conducted on data reflecting graduates' total time spent in active postdoctoral training.Statistical significance was determined using an alpha threshold of 0.05.Both analyses were conducted using GraphPad Prism.

| Data visualization
To visualize how career goals and career outcomes evolved over time, the Sector and Career Type tiers for research-focused careers were grouped as either Academic Research, For-profit Research, or Government or Nonprofit Research (Government/Nonprofit Research).Non-research careers in all sectors were grouped together and presented as either Teaching careers or Administrative, Managerial, or Operational (Administrative/ Managerial/Operational). Examples of job titles held by individuals in these other career areas are shown in Table S2.Students with Undecided or Unspecified career goals at doctoral defense were grouped together for visualization as "Undecided/Unspecified." 2.10 | Not employed, unknown, and not science-related Some alumni are hard to locate, for a variety of reasons.In instances where one team member could not locate an alumnus, two other people attempted a search and verified that the individual could not be found.During the time period covered in this study (1997-2021), 10 alumni were not employed, and 88 alumni had unknown positions at one or more of the defined milestones or timepoints of interest (first position after PhD, first non-training position, or 1, 3, 5, or 10 years after graduation with a PhD).These individuals were included in the figures in which they had known positions and excluded from figures and data analyses in which they had a missing milestone or timepoint.Details of the inclusion/exclusion criteria for cohorts represented in each figure and table are described in Table 1.The 26 alumni (1.7% of our population) who held one or more jobs that were flagged as "not science-related" were included in the visualizations.

| Career goal at matriculation and doctoral defense
Since 2007, we have administered an exit survey to graduates of our institution's biomedical sciences PhD programs in which they were asked to indicate their primary career goal at two timepoints: the time they matriculated into the PhD program and the time they completed the PhD program (Figure 1).62% of students recalled being interested in a Research career in the Academic, For-profit, or Government or Nonprofit sectors when they started graduate school (n = 589).14% (n = 128) of students recalled that they were interested in Teaching careers, and 4% (n = 35) were interested in careers categorized as Administrative, Managerial, or Operational jobs (Administrative/Managerial/Operational).The remaining students (20%, n = 192) reported that they were Undecided or Unspecified in their career goal (Undecided/Unspecified).
Between matriculation and doctoral defense, 56% of all students, including those Undecided/Unspecified at matriculation, had changed their career goal (Figure 1 and Table 2).47% of students who initially selected Academic Research changed their primary career goal to another career type, and 153 of these 172 individuals provided free-text comments about why their career goal changed away from being a principal investigator (PI) in academia (Table 3).Two-thirds of the reasons cited by students for becoming less interested in F I G U R E 1 Comparison of biomedical sciences PhD students' career goal at matriculation and doctoral defense.Sankey diagram showing the career goal of biomedical sciences PhD students at matriculation (left nodes) and doctoral defense (right nodes).The number of students with a goal in each career area, at matriculation and at doctoral defense, is indicated in the corresponding nodes.The links between the nodes show how many students maintained or changed their original career goal, as reported retrospectively on the exit survey.The thickness of each link is proportional to the number of students represented in the link and the colors of the links correspond to the students' career goal at matriculation.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.Students who graduated between January 1, 2005, and August 31, 2021, and completed the exit survey were included in this analysis (n = 944, Cohort A).PI careers were negative perceptions about academia or being a PI (Table 3: External Factors).Among the most frequently noted reasons were concerns about the competitive funding environment for biomedical research, the lifestyle of academic PIs, and the competition for PI jobs.The remaining one-third of comments were more self-directed factors, such as a perceived mismatch between being a PI and the respondent's strengths or values or interests, or conversely, a perceived better fit with another type of career (Table 3: Internal Factors).Other internal subthemes included exposure to other career options during graduate school, and occasionally, a negative graduate school experience.
There were also 103 students whose primary career goal shifted toward becoming an academic PI, 68 of whom provided comments about their reasons.About 25% of reasons involved positive perceptions of the PI role or working in academia such as research independence and lifestyle flexibility (Table 3: External Factors).The remaining 75% of comments related to internal factors; most frequently mentioned were enjoying research, increased confidence about succeeding in an academic career, and encouraging role model/mentor(s).
Similar to Academic Research, 47% of students who recalled being interested in For-profit Research at the start of graduate school became interested in another type of career by doctoral defense (Table 2).Also, nearly 60% of the students who selected Government/Nonprofit Research or Teaching at matriculation changed their primary career goal by doctoral defense.Persistence of career goal was highest in the Administrative/Managerial/Operational group, with only 26% of the 35 students changing their primary goal by doctoral defense.7% of students who entered graduate school interested in a specific career path became Undecided/ Unspecified by doctoral defense (n = 53); most of these students (n = 30) switched away from being interested in Academic Research.
As might be predicted, a large number of students who recalled being Undecided/Unspecified about their career goal at matriculation were able to identify a career goal by the time their PhD training ended.Specifically, of the 192 students who were Undecided/Unspecified at matriculation, 84% reported having a defined career goal at doctoral defense (Table 2).These students developed an interest in pursuing Academic Research (22%), Forprofit Research (18%), Government/Nonprofit Research (6%), Teaching (14%), or Administrative/Managerial/ Operational (23%).
In summary, a majority of students changed their primary career goal over the course of graduate training.We observed this shift in career goal regardless of the type of career students recalled being interested in at T A B L E 2 PhD students' career goal at matriculation (rows) and doctoral defense (columns).

| First position after PhD
Traditionally, it has been the norm for biomedical PhD graduates to enter a postdoctoral research position immediately after completion of the PhD.Out of 1413 students who graduated between 1997 and August 31, 2021, 75% entered postdoctoral training as their first position (Figure 2A).84% of postdoctoral positions were in Academic Research, 14% were in Government/Nonprofit, 2% were in For-profit Research, and three individuals entered postdoctoral positions focused on Teaching or Administrative/Managerial/Operational (0.3%).An additional 12 individuals eventually did one or more postdocs, primarily in Academic Research, but it was not the first position they held after completing their PhD.For this analysis, we classified positions as "postdoctoral" based on the job title, and defined the first position after completion of the PhD as the first job or postdoctoral position that was not in the same lab in which the alumnus completed their PhD research.We adopted this practice because we observed that about 30% of our alumni did a short postdoc in their dissertation lab, usually to fill a gap in time until the start date of a job or full-term postdoc in a new lab, or to time a move with a partner.An exception to this rule was made if a postdoctoral position in the dissertation lab lasted longer than 6 months.The next steps of the 347 alumni who did not immediately do a postdoc were varied.17% of these individuals pursued further educational degrees in law, medicine, nursing, veterinary medicine, or business, and the remainder entered employment directly after graduating.Of the 288 who took jobs, 11% entered Academic Research, 19% entered For-profit Research, 3% entered Government/Nonprofit Research, 11% entered Teaching, and 40% entered Administrative/Managerial/ Operational positions.Between 1997 and 2021, there was year-to-year variability in the percentage of graduates who pursued postdoctoral training as a first position after PhD completion (Figure 2B).Logistic regression was conducted on fullyear data (1997-2020) to create a model estimating the probability that a graduate began postdoctoral training.The model presented here, which incorporated graduation year as a predictor of postdoctoral training, fit these data better than an "intercept-only" model that did not use graduation year as a predictor (χ 2 (1) = 8.22, p = 0.004, likelihood ratio test).The odds ratio corresponding to graduation year was 0.97 with 95% confidence interval (CI) [0.95, 0.99].The upper limit of this 95% CI is less than 1, indicating the odds of a graduate entering postdoctoral training decreased as graduation year increased.From our model, the probability of entering postdoctoral training started at 83% with 95% CI [78%, 87%] in 1997 and dropped to 71% with 95% CI [66%, 75%] by 2020.Based on this modeling, the probability of students entering postdoctoral training decreased by approximately 15% over time from 1997 to 2020.

| Career goal and first non-training position
To characterize the early career paths of our alumni, we related their career goal at doctoral defense to their first position after PhD completion and their first non-training  position (Figure 3 and Tables 4, 5A,B).We restricted this analysis to alumni who graduated before 2015 to avoid biasing the cohort toward individuals who entered Administrative/Managerial/Operational or Teaching for their first non-training position, as these individuals were more likely to do short postdocs or skip postdoctoral training altogether.There were 418 alumni who graduated between 2005 and 2014 for whom we had all three of these data points.Overall, 78% of this cohort of 418 alumni did a postdoc and 22% did not.
We found that participation in postdoctoral training correlated with career goal at doctoral defense.Almost all individuals with some type of Research career goal at doctoral defense pursued postdoctoral training (88%; 204 of 231; Figure 3 and Table 4).Further, as shown in Table 5A, 71% of this cohort who did a postdoc went on to some type of Research job for their first non-training position (144 of 204), while 6% (13 of 204) entered a Teaching position, and 23% entered an Administrative/Managerial/Operational job upon completing their postdoctoral training (47 of 204).
73% of individuals interested in Teaching at doctoral defense did a postdoc as their first position after PhD completion (49 of 67; Figure 3 and Table 4).As shown in Table 5A, 41% of these individuals (20 of 49) entered a Teaching job for their first non-training position, but 37% (18 of 49) entered a Research career of some type and 22% (11 of 49) entered an Administrative/Managerial/Operational job as their first non-training position.Another group with high participation in postdoctoral training comprised individuals who were Undecided/Unspecified at the time of doctoral defense; 75% of this group (30 of 40) also did a postdoc (Table 4), and 67% of these went on to a research career (20 of 30; Table 5A).
Those with Administrative/Managerial/Operational career goals at the time of doctoral defense were the least likely to pursue postdoctoral training, but 53% of these individuals (42 of 80) still opted to do a postdoc (Table 4).Of these 42 individuals, 62% (26) eventually entered an Administrative/Managerial/Operational career, 33% (14)  entered some type of Research job, and 5% (2) started a Teaching job after completing postdoctoral training (Table 5A).
For alumni who entered postdoctoral training, the length of time between doctoral defense and the first non-training position varied by students' career goal at  doctoral defense (Figure 4A).This analysis was intentionally limited to those alumni who entered a postdoc immediately after PhD completion, allowing this measure to serve as an approximation of total time spent in postdoctoral training.4B).Ninety-three alumni in this cohort of 418 graduates (cohort C) chose not to do a postdoc (Figure 3 and Tables 4 and 5B).
T A B L E 4 Proportion of PhD alumni who did a postdoc for their first position after PhD completion, based on career goal at doctoral defense.# of alumni with goal x at doctoral defense who did a postdoc.( # of alumni with goal x at doctoral defense who did a postdoc) + ( # of alumni with goal x at doctoral defense who did not do a postdoc) where x is the career area of the goal at doctoral defense.While there is considerable similarity between the relative distribution of students' career goal at doctoral defense and the relative distribution of first non-training positions (Figure 3), a direct comparison of individuals' career goal at doctoral defense to their first non-training position revealed that only 51% of graduates entered a first non-training position that was the same as their career goal at doctoral defense (Table 6).The alignment of career goal at doctoral defense and first non-training position was highest among those interested in Administrative/Managerial/Operational careers, with 75% of individuals who were interested in an Administrative/Managerial/Operational career at doctoral defense entering an Administrative/Managerial/Operational career as a first non-training job.About half of those interested in either Academic Research or Teaching careers at doctoral defense entered those career paths for their first non-training position (52% and 49%, respectively).Only 35% of those interested in For-profit Research careers, and only 21% of those interested in Government/Nonprofit Research careers, entered those types of jobs for their first non-training role.Overall, the match between career goal at doctoral defense and first non-training position was lower for those who did a postdoc compared to those who did not (45% vs. 71%, respectively; data not shown).Taken together, the results suggest that goals at doctoral defense are not a reliable indicator of the career area of the first non-training position, especially when there is an intervening period of postdoctoral training.

| Careers during first 10 years after PhD
By following individual careers longitudinally, we were able to see how the careers of 654 alumni progressed throughout the 10 years following graduation with a PhD (Figure 5, cohort E).At 1 year after graduation (Y1), 64% of alumni were in Academic Research.During the following 9 years, 59% of those who initially worked in Academic     7).Despite this attrition, 53% of alumni were still employed in some type of Research career at Y10, with the majority of these in Academic Research, and 24% held tenure-track faculty appointments by Y10.About two-thirds of tenure-track positions were research-focused positions in academia, government, or nonprofit organizations and one-third were teaching-focused positions (Table 8).
When alumni were grouped by career area at Y10, individuals who were employed in Academic Research, For-profit Research, and Government/Nonprofit Research were significantly more likely to have spent a longer time in postdoc positions than those who were employed in Administrative/Managerial/Operational at Y10 (Table 9 lists p values for significant pairwise comparisons; H (4) = 69.59,p < 0.0001, Kruskal-Wallis test).Likewise, those employed in Academic Research at Y10 were significantly more likely to have spent a longer time in postdoc positions than those employed in Teaching at Y10 (Table 9).Overall, those employed in Academic Research positions at Y10 had the longest median postdoctoral training (4.9 years) and those employed in Administrative/ Managerial/Operational positions at Y10 had the shortest (2.8 years).
During the 10 years following graduation, movement out of, or between, different types of Research careers was more common than movement out of Teaching or Administrative/Managerial/Operational careers, where alumni tended to remain once they entered jobs in those areas.There was a net growth in employment in For-profit Research careers from 4% to 14% (Figure 5), but overall alumni tended to migrate from Research careers into Administrative/Managerial/Operational careers at each timepoint examined.Since transitions out of Teaching or Administrative/Managerial/ Operational careers were infrequent compared to transitions into these career areas, employment in Teaching careers grew from 4% to 10%, employment in Administrative/ Managerial/Operational careers grew from 9% to 36% as alumni gained experience and started to move away from the bench into managerial roles, and the number of alumni pursuing further education decreased over time.
To gain insight into how alumni careers evolve, we examined the proportion of alumni at each timepoint whose career area was identical to the career they held at Y10  Note: Alumni who graduated 1/1/997 to 12/31/2014 were included in this analysis (n = 654; Cohort E).Individuals were counted once, at the first timepoint at which they held the faculty position.† In addition to the 141 individuals shown in the table, there were an additional 14 alumni who held a tenure-track position for some time between 5-and 10-years post-PhD who are not represented in the table (6 in research-focused positions and 8 in teaching-focused positions).AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.
after graduation.Across all career areas, a progressively greater proportion of were employed in the same career area at Y1, Y3, and Y5, respectively, as the job they would eventually hold at Y10 (Table 10 and Table S2).At Y1, 42% of alumni were employed in the same career area in which they would be employed in Y10; this grew to 54% and 69% at Y3 and Y5, respectively.Notably, 80% of alumni who were employed in a Teaching career at Y1, and nearly 100% of alumni who were employed in an Administrative/Managerial/Operational career at Y1, were still employed in those career areas at Y10.The proportion of individuals who would remain employed in some type of Research career area at Y10 grew from 64% at Y1 to 80% by Y5, though there was still considerable attrition from, and switching between Academic Research, For-profit Research, and Government/Nonprofit Research during the first 5 years after graduation (Figure 5).Taken together, these comparisons show that alumni who pursue Teaching and Administrative/Managerial/Operational careers immediately after PhD completion are more likely to persist in the identical career at Y10 than alumni who enter research career areas.

| Evolution of early career relative to career goal and first non-training position
For the subset of alumni for whom we had an exit survey and a Y10 career outcome (Cohort F: 271 alumni who graduated between 2005 and 2011), we analyzed whether their career goal at doctoral defense matched their Y10 career area.Except for Administrative/Managerial/Operational careers, the distribution of career outcomes at Y10 appeared comparable to the distribution T A B L E 1 0 Percent of alumni at 1, 3, or 5-years post-PhD whose career area is identical to their career area at Y10. Note: Alumni who graduated 1/1/1997 to 12/31/2011 were included in this analysis (n = 654; Cohort E).The first row shows the percentage of alumni whose career area at Y10 matched their career area at an earlier timepoint across all career areas, regardless of the specific career area of employment.The second row shows the percentage of alumni with Some Type of Research career at Y10 as well as the earlier timepoint, but not necessarily the same Research career area.The raw number of alumni underlying each percentage calculation is shown in Table S2.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.

Years
of career at doctoral defense (Figure 6).However, across all career areas, only 51% students' career goals at doctoral defense matched their career outcome at Y10 (Table 11).Alignment of career goal at doctoral defense with career area at Y10 was highest for those with Administrative/Managerial/Operational or Academic Research career goals (80% and 52%, respectively) and lowest for those interested in Government/Nonprofit careers (13%).At Y10, more than half the 27 individuals who were Undecided/Unspecified at doctoral defense were in some kind of Research career (Table 11).
We also examined the match between the career area of alumni first non-training position and their Y10 job (Figure 7 and Table 12).There were 669 alumni who graduated from 1997 and 2011 for whom we had both their first non-training position milestone and their career outcome at the Y10 timepoint.Seven of these alumni were in Further Education at Y10 so they were excluded from the analysis.
In the remaining cohort, the relative distribution of first non-training positions was very similar to the  Indicates totals that exclude individuals with an Undecided/Unspecified career goal at doctoral defense and individuals who were pursuing further education at Y10. AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.The "Proportion of alumni whose goal at doctoral defense matched their career area at Y10" was calculated as: # whose goal at doctoral defense matched their Y10 career area (shaded box) Total,goal at doctoral defense in the corresponding career area.

| DISCUSSION
Here, we present the results of a longitudinal study that the career goals of biomedical PhD students at matriculation and doctoral defense to their career outcomes at key career milestones and timepoints up to ten years after completing PhD training.Although our study was restricted to 1452 alumni from a single institution, the demographics of our student population are similar to that of other institutions 14,24 and our students' time to degree and total publication number are similar to those at other BEST institutions, 25 suggesting that our findings may be generalizable.Several groups have reported on either PhD student career goals 12,14,15,[26][27][28] or career trajectories after PhD completion, 13,24,29,30 but to our knowledge, this is the first study to examine the two in combination.This type of analysis is rare because it requires longitudinal data collection for an extended period as well as an in-depth understanding of sciencerelated careers to curate alumni career outcomes accurately.However, we hope that other institutions will be inspired to track and report similar career goal and milestone data, thus enabling cross-institutional analyses to detect trends in the early careers of biomedical research scientists.
Our career outcomes taxonomy (Appendix S1) expands on previous classification systems 9,10 with the addition of flags to document faculty positions as well as the first position after PhD completion and the first non-training position of PhD graduates.The use of these novel flags enabled us to report on career milestones and better understand the early career trajectories of our alumni.Major findings include that nearly equal numbers of alumni were in Academic Research and Administrative/Managerial/Operational careers at Y10, and the best indicator of alumni career area 10 years after graduation with a PhD was their first non-training position, whether that occurred immediately after PhD completion or after a period of postdoctoral training.

| Changing career goals
More than half the PhD students in our study population changed their primary long-term career goal between matriculation and doctoral defense.These data were gathered from an exit survey administered at the time of doctoral defense, in which students were asked to report both their current career goal as well as recall their career goal from the time they started graduate school.It has not escaped our attention that asking students to report their career goal from several years earlier may have the potential for recall bias.][16]26,28,31 Whether conducted longitudinally or retrospectively, these studies all showed that a large fraction of graduate students changed career goals during graduate school, strongly suggesting that this is common behavior rather than the exception to the rule.
As reported by Wood et al. ( 2020), 26 we found that students who were interested in research-intensive faculty careers at doctoral defense were a mix of those who were interested in PI careers since matriculation and those who became interested in being a PI during graduate training.14][15][16]28 The reasons cited by our students for changing career goals during graduate school fit well within the framework of Social Cognitive Career Theory (SCCT). 32SCCT posits that an individual's career interests are formed through a dynamic interplay between self-efficacy in a particular domain and outcomes expectations about careers in that area.In our analysis, the internal factors cited by students as reasons for changing career goals contribute to formation of self-efficacy, and the external factors cited by students as reasons for changing career goals align with their outcomes expectations about attaining a career as a PI.Several studies designed to examine career interest formation through the lens of SCCT have also shown that social identity influences career interests of biomedical scientists, 15,16,28,33,34 though we did not attempt to analyze the role of social identity in our population, as it is beyond the scope of this study.
Changes in career interest also occur during postdoctoral training, as we found that only 51% of our alumni had a first non-training position in a career area that matched their career goal at doctoral defense.We did not survey our alumni directly to ask why they changed directions during postdoctoral training, but a recent survey study by Lambert and colleagues (2020) 33 suggested that success in research during postdoctoral training played an important role in shaping a postdoc's self-efficacy and intent pursue a research-intensive faculty career.This large survey of postdocs from over 80 institutions also found that social identity, personal values, and perceptions about faculty careers were significant factors in determining postdocs' career interests.
The fluidity of PhD student career interests during graduate school, and the subsequent shift in career trajectory that occurs yet again during training, points to the critical need for research mentors and training programs to provide both student and postdoc trainees with career mentoring and encourage experienlearning about research-and research-related careers, including that of academic scientists.Historically, many institutions and faculty did not promote career development activities for trainees because of perceived conflict of effort with the NIH funding sources from which many biomedical trainee stipends are paid. 35However, shortly after the NIH created the NIH Director's Broadening Experiences in Scientific Training (BEST) program, which funded 17 institutions to test and evaluate the impact of providing grad students and postdocs with experiential training for a range of research-and research-related careers, the NIH clarified that it is acceptable for PhD students and postdocs supported by NIH grants to participate in career development activities. 36ther faculty were historically reluctant to encourage trainees to participate in career development activities because they believed that spending time on career development activities would diminish trainee productivity and increase time in training.These concerns were debunked by a robust cross-institutional analysis from 10 NIH BEST institutions, who pooled their data on student participation in career development activities to demonstrate conclusively that engaging in such activities did not extend time-to-degree in graduate school or decrease the number of first-author or total publications of graduate students. 25urthermore, the NIH BEST experiment demonstrated that there are a variety of ways for PhD students and postdocs to engage in meaningful experiential career learning opportunities in non-academic laboratory work environments, many of which do not require extended time away from the laboratory. 37-43

| Most biomedical PhD alumni did a postdoc, regardless of career goal
Our data show that between 1997 and 2021, the majority of our biomedical sciences PhD graduates entered a postdoctoral position, primarily in Academic Research.During postdoctoral training, biomedical scientists build out their CV, expand their scientific expertise, and develop autonomy that is valuable preparation for research-intensive careers, and especially for careers as PIs who oversee an independent research program. 44,45ostdoctoral training is also considered an asset for many teaching-intensive colleges that expect faculty to engage undergraduates in laboratory research.Given the benefits of postdoctoral training, it would be easy to assume that alumni who do postdocs intend to pursue a research-or teaching-intensive faculty position that involves significant research.Indeed, nearly 90% of our students with a Research career goal at doctoral defense, and nearly 75% with a Teaching career goal, pursued postdoctoral training.Furthermore, alumni with a Research career goal who did a postdoc were more likely to enter a Research career for their first non-training position than those who did not do a postdoc (84% compared to 52%).
However, the 76% of our alumni who did a postdoc far exceeded the 31% who had an Academic Research career goal at doctoral defense, a finding consistent with previous studies and reports. 44,46This imbalance was partly because 53% of individuals with an Administrative/Managerial/ Operational career goal and 74% of those who were Undecided/Unspecified at doctoral defense chose to do postdocs.Our alumni with an Administrative/Managerial/ Operational career goal were more likely to attain a first non-training position quickly compared to alumni with an Academic Research career goal, but economic analyses have shown that time spent in postdoctoral training has negative financial ramifications, especially for individuals who do not pursue Academic Research careers. 29,45,47,48lthough postdoctoral training may be beneficial experience for some Administrative/Managerial/Operational positions (e.g., editorial careers at scientific journals), it is not a requirement for many Administrative/Managerial/ Operational jobs.We did observe that alumni in our study who were in Administrative/Managerial/Operational or Teaching careers at Y10 spent less time in postdoctoral training than individuals who pursued Research careers, but the opportunity cost of doing a postdoc on lifetime earnings should prompt institutions to help PhD students who do not intend to pursue Research careers to move quickly into their first non-training position, skipping the postdoctoral training period if it is not necessary or of benefit for their intended career.

| The pursuit of postdoctoral training declined over time
We observed that the percentage of alumni who pursued postdoctoral training has declined steadily over time, dropping to below 60% for the first time in 2019.0][51] Concern about declining numbers of postdocs is the basis for the creation of a new Advisory Committee to the NIH Director that is charged with providing recommendations to address the causes and consequences of declining numbers of postdocs. 11ost of our PhD students who did not do a postdoc immediately after PhD completion entered the career area matched their goal at doctoral defense, suggesting that postdoctoral training was not necessary for them to attain their preferred career.The recent trend toward fewer of our PhD alumni pursuing postdoctoral training may be an encouraging sign that PhD students are defining their career goals earlier in their scientific training rather than deferring career decision-making and pursuing postdoctoral training by default. 44If so, interventions like the NIH BEST program, which were designed to broaden PhD and postdoc training and perspectives to better align with the realities of the biomedical research workforce, appear to be working.The resulting decline in postdocs is an understandable cause for concern among academic institutions that have relied upon postdoctoral scientists and early career researchers to make biomedical research discoveries.However, as noted in a recent editorial by Dzirasa (2023), 52 the solution to this problem requires robust policy responses and economic investment aimed at making academic research careers more sustainable and attractive.

| Evolution of early career pathways
Following individual careers longitudinally for the first 10 years post-PhD revealed that our alumni gradually moved out of Academic Research and into other career areas.This shift over time is consistent with data from a study of Wayne State University STEM alumni. 24An important distinction is that we followed career outcomes for the same individuals over time, whereas the Mathur study reported career outcomes for three different groups of alumni, at three distinct timepoints.Following the same individuals over time allowed us to determine that the shift of our alumni out of the Academic Research career area in the first 10 years after graduation was largely due to alumni finishing postdoctoral training.Within 10 years of earning their doctoral degree, our alumni were employed in Academic Research or Teaching and held tenure-track faculty jobs at rates comparable to biomedical PhDs nationally. 4 We were particularly interested in comparing patterns of alumni employment 10 years after graduation to earlier timepoints and milestones to determine the point at which alumni become relatively set in their career path.
Not surprisingly, only 51% of alumni were employed in a career area at Y10 that matched their career goal at doctoral defense.While an increasing percentage of alumni at Y1, Y3, and Y5 were employed in the same career area as they were employed at Y10, the continued movement of alumni out of postdoc positions meant that none of these earlier timepoints were particularly good indicators of the career area of employment of alumni at Y10.It is important to note that except for the transition out of postdoctoral training, our analyses focused on the movement of alumni between different career areas, not between jobs per se.For example, an alumnus who changed from being a medical writer at Y3 to being a medical science liaison at Y5 would be represented in the Administrative/Managerial/Operational career area at both timepoints.Thus, our visualizations do not depict the number of job changes alumni had in the first 10 years after graduation.
Interestingly, the career area that alumni entered for their first non-training position was identical to their career area at Y10 for 81% of alumni, suggesting that this milestone is a good indicator of the early career pathway of alumni.This observation has several implications for policymakers, trainees, and research training programs.Currently, the NIH requires institutions with T32 training grants to track the career outcomes and publications of PhD alumni for 15 years after degree conferral.The amount of data collection and reporting that is required for T32 training grant submissions is time-consuming and represents a substantial administrative burden.Our findings suggest that alumni could be tracked to their first non-training position, whenever that occurs, as a reasonable proxy for the broad career area in which they will likely be employed in the future.Training programs could continue to collect the publication record of alumni who pursue Research careers for the purpose of understanding the impact of the T32 training program on scientific research advances, but the administrative burden of tracking alumni careers beyond their first non-training position could be eliminated.Many institutions will undoubtedly continue tracking alumni for other purposes with a frequency that makes sense for their internal needs.
For trainees, it is interesting to note that the career area in which one works during their first job outside of training-whether it occurs after graduate school or after a postdoc-is likely to be the career area in which one will work at Y10; this is especially true if one's first step out of training is an Administrative/Managerial/Operational or Teaching career.While there is a decreasing trend toward pursuing postdoctoral training, PhD students who are interested in a Research career should strongly consider postdoctoral training, as we observed that alumni whose primary long-term career goal at doctoral defense was a Research career, but who did not do a postdoc, were less likely to a Research career for their first non-training position.Conversely, trainees who have ruled out careers in Research should be actively supported by their institutions and training programs to skip postdoctoral training and pursue other career options that utilize their transferable skills, biomedical sciences expertise, and research training in ways that better align with their interests.Doctoral training provides scientists with many transferable skills are valued by PhD alumni and employers in a range of professions, 53,54 and PhD scientists contribute to the biomedically trained workforce in myriad productive ways, both in research and research-related careers.Training programs should aim to help connect trainees to these careers quickly to maximize their satisfaction and contributions to the scientific enterprise and keep graduate school an attractive path for high potential individuals.

| Study limitation
This study is based on biomedical PhD alumni from a single institution in the United States, Vanderbilt University School of Medicine.This makes it difficult to detect employment trends that may be observable in larger datasets from multiple institutions, such as declining numbers of alumni employed in tenure-track faculty jobs or increasing numbers of alumni employed in the For-Profit sector.Additionally, although Vanderbilt University's biomedical PhD programs are representative of many U.S. research universities, these findings may not fully apply to PhD programs at other institutions, to graduate students or alumni in other disciplines, or to graduate programs and alumni abroad.

T A B L E 1
Figure 1

F I G U R E 2
First position after PhD completion.(A) Stacked bar chart showing the career area of alumni who did (top) or did not do (bottom) a postdoc as their first position after PhD completion.Students who graduated between 1/1/1997 and 8/31/2021 and completed the exit survey were included in this analysis (n = 1413, cohort B).AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.(B) Line graph (black) indicating percent of biomedical sciences PhD alumni who conducted a postdoc immediately after PhD completion, grouped by year of graduation.Full-year data (1997-2020) are represented with filled dots, and half-year data (2021) are represented with an open dot.The probability of alumni conducting a postdoc as their first position after PhD completion was modeled using logistic regression (gray line with shaded 95% confidence interval).Students who graduated 1/1/1997-8/31/2021 were included in the line graph (n = 1413, cohort B), and students who graduated 1/1/1997-12/31/2020 were used for logistic regression (n = 1369, full-year subset of cohort B).

F I G U R E 3
Comparison of career goal at doctoral defense with first position after PhD completion and first non-training position.Sankey diagram showing the career goal of biomedical sciences PhD students at doctoral defense (left nodes), first position after PhD completion (middle nodes, grayscale), and first non-training position (right nodes).The links between the left and the middle nodes represent the movement of alumni from each career goal area to their first position after PhD completion, categorized as either a postdoc or not a postdoc position.The links from the middle to the right nodes show the movement of alumni from their first position after PhD completion to their first non-training position.The thickness of each link is proportional to the number of students represented in the link and the colors of the links between two nodes correspond to the color of the respective career category of the earlier node.For the 93 individuals who did not do a postdoc (middle node, light gray), their first position after PhD was their first non-training position.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.Students who graduated from 1/1/2005 to 12/31/2014 and completed the exit survey were included in this analysis (n = 418, cohort C).

5 AT A L E 5 B
First non-training position of alumni who did a postdoc as their first position after their PhD, based on their career goal at the time of doctoral defense.Career progression of PhD alumni in Cohort C (graduated 1/1/2005 to 12/31/2014 and completed the exit survey; n = 418) who did a postdoc as their first position after PhD completion (204 of 418).The blue shaded boxes are the number of students with Some Type of Research goal at doctoral defense who did a postdoc, consisting of students with goals in Academic Research + For-profit Research + Govt./NonprofitResearch.The green shaded boxes are the number of alumni in Some Type of Research role for their first non-training position.AMO is Administrative/Managerial/ Operational.Govt./NonprofitResearch is Government/Nonprofit Research.The "proportion of alumni in career area of goal at doctoral defense for first non-training position" was calculated as: # in career area for 1st non − training position # with goal at doctoral defense.First non-training position of alumni who did not do a postdoc as their first position after PhD completion, based on their career goal at the time of doctoral defense.Goal at doctoral defense (Number with goal at graduation who did not do a postdoc) Number of alumni in each career area for first nontraining position Proportion of alumni in career area of goal at doctoral defense for first non-training position Some Type of Research goal at doctoral defense, and did not do a postdoc (27)

F I G U R E 4
Cumulative incidence curves showing the estimated probability of entering a first non-training position at timepoints following doctoral defense.Curve color corresponds to career area designated in legend with graphs showing (A) all career areas, and (B) Academic Research vs. Administrative/ Managerial/Operational (AMO) with 95% confidence bands.Alumni who graduated 1/1/2005-8/31/2019 were included in the probability calculations if they entered a postdoctoral training position immediately after their doctoral defense and had either i.) completed training and entered their first non-training position, or ii.) were still in training and had not entered a first non-training position at the cutoff date for recorded observations (a subset of Cohort D, n = 773).The 229 individuals in Cohort D who did not do a postdoc immediately after doctoral defense but instead entered a first non-training position were not included in the analysis.Individuals per career goal at doctoral defense and median time after doctoral defense to first non-training position (time at which curve crosses 50% probability threshold): Academic Research (n = 230; 5.5 years), For-profit Research (n = 132; 4.1 years), Government/Nonprofit Research (n = 40; 3.8 years), Teaching (n = 84; 3.4 years), AMO (n = 71; 3.0 years), Undecided/Unspecified (n = 57; 4.3 years).
complete by 5 years after graduation (Y5; Table

T A B L E 7 T A B L E 8
Count of postdocs, by career area of postdoc, at 1, 3, 5, or 10 years post-PhD.Timepoint at which alumni first held a tenure-track faculty position.

F I G U R E 6
Comparison of career goal at doctoral defense to career area at 10 years after graduation.Sankey diagram showing career goals of biomedical sciences PhD students at doctoral defense (left nodes) compared to career area at 10 years after graduation (right nodes).The color of the links corresponds to the students' career goal at doctoral defense and the thickness of each link is proportional to the number of alumni represented in the link.AMO is Administrative/Managerial/Operational.Govt./Nonprofit Research is Government/Nonprofit Research.Students who graduated between 1/1/2005 and 12/31/2011 and completed the exit survey were included in this analysis (n = 271, cohort F).
Goal at doctoral defense compared to career area at Y10. Alumni who graduated 1/1/2005 to 12/31/2011 who completed the exit survey were included in this analysis (n = 271; Cohort F).The shaded number at the intersection of each matching row/column is the number of individuals whose career goal at doctoral defense matched their career area at 10-years post-PhD.Individuals with an Undecided/Unspecified career goal at doctoral defense and individuals who were pursuing further education at Y10 are shown in parentheses because they were excluded from the totals prior to the calculation of percentages.†

Table 2 Table
The analyses shown in each figure or table were based on subsets of Vanderbilt University biomedical sciences PhD alumni, termed Cohorts, who were admitted in 1992 or after and graduated with a PhD between January 1, 1997 and August 31, 2021.Cohorts were defined based on graduation dates and were comprised of the largest group of alumni for whom we had an exit survey (if required) and career outcomes data at Note: specific milestones (first position after PhD or first non-training position) and/or timepoints (1, 3, 5, or 10-years post-PhD, designated Y1, etc.).The range of graduation dates and number of alumni in each Cohort is shown in yellow.The exit survey, milestone, and timepoint requirements for each Cohort are shown in green.The demographic breakdown of each Cohort is shown in gray.
The shaded number at the intersection of each matching row/column is the number of individuals whose career goal at matriculation matched their career goal at doctoral defense.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.
Note:PhD students who graduated 1/1/2005 to 8/31/2021 and completed the exit survey were included in this analysis (n = 944, Cohort A).The "Percent who changed goal between matriculation and doctoral defense" is the percent of students with a particular goal at matriculation, x, who changed to some other goal by the time of doctoral defense, calculated as: [Total,goal x

AWAY from being a research-focused faculty member (156 comments from 153 individuals) TOWARD being a research-focused faculty member (88 comments from 68 individuals)
Frequency with which exit survey respondents cited external or internal themes as reasons for changing their career goal away from or toward becoming a research-focused faculty member in academia.
T A B L E 3Note: PhD students who graduated between 1/1/2005 to 8/31/2021 and completed the exit survey were included in this analysis (n = 944, Cohort A).Counts were generated by categorizing free-text responses from individuals who switched their career goal away from, or toward, becoming a research-focused faculty member in academia, and chose to answer the optional exit survey question prompt, "If your career goal changed since you started your PhD program, what factor has most affected these goals?"Individual responses were coded across multiple themes if necessary.
= 418) who did not do a postdoc as their first position after PhD completion (93 of 418).The blue shaded boxes are the number of students with Some Type of Research goal at doctoral defense who did not do a postdoc, consisting of students with goals in Academic Research The green shaded boxes are the number of alumni in Some Type of Research role for their first non-training position.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.The "proportion of alumni in career area of goal at doctoral defense for first non-training position" was calculated as: Goal at doctoral defense compared to first non-training position.Note: PhD students who graduated 1/1/2005 to 12/31/2014 and completed the exit survey were included in this analysis (n = 418; Cohort C).The shaded number at the intersection of each matching row/column is the number of alumni whose career goal at doctoral defense matched the career area of their first non-training position (192 individuals across all career areas).Alumni with an Undecided/Unspecified career goal at doctoral defense are shown in parentheses because they were excluded from the totals prior to the calculation of percentages.† Indicates totals that excluded individuals with Undecided/Unspecified career goals at doctoral defense.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.The "proportion of alumni whose first non-training position matched their goal at doctoral defense" was calculated as: Movement of PhD alumni between different career areas in the first 10 years after graduation.Sankey diagram showing the career area of biomedical sciences PhD alumni at four timepoints from one to 10 years after graduation.Nodes show the career area at each timepoint, and links show the movement of alumni between career areas.The color of the link between two timepoints corresponds to the Research (246 out of 418) and 88% of those who initially worked in Government/Nonprofit Research (75 out of 85) transitioned to other career types.The gradual attrition from Ac or Government/Nonprofit Research careers to other career types parallels the movement of alumni out of postdoctoral training into jobs, a transition that is largely T A B L E 6 # with career goal x at doctoral defense who were in same career area x for their first non − training position (shaded boxes).Total,goal x at doctoral defense F I G U R E 5 career area at the earlier timepoint and the thickness of each link is proportional to the number of alumni represented in the link.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.Students who graduated 1/1/1997 to 12/31/2011 were included in this analysis (n = 654, cohort E).
Alumni who graduated 1/1/1997 to 12/31/2014 were included in this analysis (n = 654; Cohort E).Of this cohort of 654 alumni, only the 482 who did a postdoc are shown in this table.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research. Note: Duration of postdoctoral training of individuals who conducted postdocs, based on career outcome at Y10. Alumni who graduated 1/1/1997 to 12/31/2011 were included in this analysis (n = 654; Cohort E).From this cohort, only those individuals who entered postdoctoral training as the first position after PhD completion were included in the calculation of median length of postdoc based on Y10 career outcome (n = 509 of 654).Pairwise comparisons were made between all groups using Dunn's multiple comparisons testing, and the four statistically significant pairwise comparisons observed are noted in the table.AMO is Administrative/Managerial/Operational.Govt./NonprofitResearch is Government/Nonprofit Research.
T A B L E 9Note: Cohort G).The shaded number at the intersection of each matching row/column is the number of alumni with matching career areas in the first non-training position and at Y10.Individuals who were pursuing further education or training at Y10 are shown in parentheses because they were excluded from the totals prior to the calculation of percentages.† Indicates totals that excluded individuals pursuing further education or training at Y10. AMO is Administrative/ Managerial/Operational. Govt./NonprofitResearch is Government/Nonprofit Research.The "Proportion of alumni with matching career area in the first non-training position and Y10" was calculated as: