IMPACT OF THE INTERNATIONAL COLLABORATIVE ADDICTION MEDICINE RESEARCH FELLOWSHIP ON PHYSICIANS’ FUTURE ENGAGEMENT IN ADDICTION RESEARCH

Abstract

Background:

To evaluate how an international one-year intensive research training program for addiction medicine physicians contributed to subsequent research involvement and productivity.

Methods:

We prospectively compared addiction medicine physician fellows admitted to a one-year training program with non-admitted controls, using baseline questionnaire and peer-reviewed publication data. Participants’ publication activity was assessed from fellowship application date onwards using biomedical databases (e.g., PubMed, Embase).

Results:

Between July 2014 and June 2020, which is six years of cohorts, 56 (39 women) physicians, both fellows (n = 25) and non-admitted applicants (n = 31), were observed and included in the study, contributing 261 person-years of observation. At baseline, in the fellows’ cohort: 76% of participants (19/25) reported past research involvement, 24% (6/25) had one or more advanced graduate degrees (e.g., MPH), and the median number of peer-reviewed, first author publications was one (Interquartile Range [IQR] = 0–2). At baseline, in the controls’ cohort: 84% of participants (26/31) reported past research involvement, 39% (12/31) had one or more advanced graduate degrees, and the median number of peer-reviewed, first author publications was zero. The physicians’ training included internal medicine (n = 8), family medicine (n = 33), psychiatry (n = 5) and others (n = 4). At follow up, there was a significant difference between fellows (n = 25) and controls (n = 31) in total number of publications (Rate Ratio [RR] = 13.09, 95% Confidence Interval [CI], 5.01 − 34.21, p < 0.001), as well as first author publications (RR = 5.59, 95% CI, 2.23 − 14.06, p < 0.001).

Conclusion:

In the six-year observation period, fellows’ productivity indicates undertaking this fellowship was associated with significant research outputs in comparison to controls, signaling successful training of addiction physicians to help recruit addiction medicine physicians to participate in addiction research.

INTRODUCTION

Recent discoveries in addiction science can enhance the treatment, prevention and management of substance use disorders. Their ultimate beneficial impact on individuals, however, often stagnates for a range of reasons, including most medical professionals neither being “sufficiently trained to diagnose or treat them,”^1–4 nor having sufficient research training to understand the importance of employing novel interventions to narrow the treatment gap for substance use disorders. Indeed, research involvement has been linked to an increased likelihood of adopting alcohol pharmacotherapies by programs affiliated with the National Institute on Drug Abuse’s Clinical Trials Network (CTN) over a two-year period even though the CTN, until then, has not conducted research on alcohol pharmacotherapies.⁵ A key gap is therefore a dearth of skilled clinician-scientists in addiction medicine.^1,6 Despite this group’s potential to bridge the gap between science and clinical practice, cultivating these professionals remains a challenge. A lack of practical experience in the field during core training, exposure to supportive research environments, mentors, and career advice are key factors that hamper research training, involvement, and productivity among early career professionals working in the field of addiction medicine.^7–13

To address these barriers, the National Institute on Drug Abuse (NIDA) sponsored several programs in North America for clinicians to gain experience in research.⁹ One such program created by the British Columbia Center on Substance Use (BCCS) in Vancouver, Canada is the International Collaborative Addiction Medicine Research Fellowship (ICAMRF) embedded in an extensive addiction research and training environment.^14,15 This one-year intensive fellowship for physicians with preexisting clinical experience in addiction medicine emphasizes individualized mentorship teams and hands-on research experience, taking advantage of an extensive NIDA-funded platform of qualitative, observational and experimental research.

Preliminary evaluations from this and other sponsored programs found increased productivity after training completion.^9,15 However, impact on future engagement in addiction research has not been fully characterized. Therefore, using a controlled, nonrandomized comparison design, we sought to prospectively evaluate how an international one-year intensive research training program for addiction medicine physicians contributed to subsequent research involvement and productivity.

METHODS

Setting

Building on a strong track record of training graduate students and other non-physician trainees at the center, (ICAMRF) developed a standardized part-time program that includes a total of six months of training which occurs over a one-year timeframe. The program is based on published research findings and feedback from faculty members that recommended: (a) dedicated time, (b) basic instruction in addiction research methods, (c) practical research experience in the field, (d) peer-reviewed publication development, and (e) interdisciplinary mentorship.^16–19

The program addresses the inflexibility of graduate school programs that may preclude clinicians from continuing clinical care activities, a commonly reported barrier to pursuing research training among practising physicians.^20,21 The curriculum is therefore flexible, with free afternoons for independent activities and clinical duties that are not part of the fellowship. Meetings and workshops are in-person, where possible, or virtual, subject to public health regulations on social distancing. A stipend allows clinicians to devote time for research. The overall structure of the half-day curriculum involves: (1) addiction research immersion training, (2) longitudinal research methods, (3) scholarly activity, and (4) career development. (1) The immersive addiction research training includes opportunities in a variety of National Institute on Drug Abuse (NIDA)-funded addiction research programs, such as: the Vancouver Injection Drug Users Study (VIDUS); the AIDS Care Cohort to Evaluate Exposure to Survival Services (ACCESS); the At-Risk Youth Study (ARYS); Ethnographic, epidemiological, and qualitative research; randomized controlled trials; and addiction research training experiences with US partners (e.g., the Research Scholars in Addiction Medicine – RAMS hosted five fellows to date). (2) Training in longitudinal research includes: Didactic lectures on research methods; Ethics teaching activities to prepare Fellows for navigating ethical challenges in addiction research; Monthly journal club meetings to critically review addiction medicine literature; Peer-review teaching; and Academic presentation skills development, which culminates in an opportunity to present at the annual BC Substance Use Conference and Research Day at St. Paul’s Hospital. (3) Scholarly activities during the fellowship include: Developing first-author articles, including original research, systematic reviews, commentaries, and case reports;^14,15 and Teaching academic seminars. (4) Career development starts when each trainee is paired up with two mentors: an addiction medicine clinician-scientist mentor and a PhD scientist mentor who supervise the scholarly activity and provide guidance in career development.

We are committed to fostering an environment of diversity, equity, and inclusion within our addiction research fellowship program, and applications are open to all individuals regardless of race, gender, ethnicity, disability, culture, or socioeconomic background. Upon fellowship completion, alumni are recruited to continue collaborating with the host institution and the fellowship program remotely, from wherever they secure their next position, or as junior faculty mentors for incoming research fellows if they work locally.

Recruitment

Email and poster advertisements were circulated through networks including the following: clinical training programs; addiction medicine societies and meetings; core faculty and collaborating partners; email mailing lists; fellowships and host institution’s websites; lay media; scientific journals; newsletters; social media; and word-of-mouth. The fellowship’s Advisory Committee guided ongoing, year-round recruitment activities. The fellowship received approximately 15–20 applications from Canadian and international candidates each year for four training positions. Advertising for the first year (2014/2015) was limited in that there was only a two-month lead-up time between fellowship promotion and program start. Fellows are physicians who were admitted into the fellowship program, and controls were rejected applicants. The fellowship admission committee considered the following admission criteria: Motivation for helping people with substance use disorders; Demonstrated interest in Addiction Medicine: clinical, research, and/or advocacy work; Potential for future impact in the field; Academic performance in Canadian College of Family Physicians (CCFP), or Royal College program, or in practice/Honors; Other academic work; Two Letters of Reference (performance with patients; skill on a team or work with colleagues; strength of recommendation; overall impression and fit for the program).

Data sources

This study utilized data from the first six cohorts of the program (2014 − 2020, 98% [56/57] response rate). Their past and new research productivity was assessed by a single baseline questionnaire at the time of application to the fellowship program and by annual searches at the fellowship conclusion of publicly available biomedical databases, respectively. The questionnaire also collected self-reported training and background in research. The primary outcome measure for statistical analysis was the number of first-author peer-reviewed publications between the sequential waves of admitted fellows and non-admitted applicants. This has provided a total of 261 person-years of follow-up over six years, calculated by summing of the time between study entry and end of the study period (i.e., 2020) of each individual. Following Campbell and colleagues’ recommendation for evaluating addiction research training programs,⁹ this analysis focused on one domain of the adapted Kirkpatrick’s conceptual model – performance change. In line with Curran and Fleet’s adaptation of Kirkpatrick’s model for levels of summative evaluation,²² the primary outcome of interest was the number of first-author publications, which represents a widely used performance measure in academia. The secondary outcomes included the numbers of last-authored, coauthored and all publications in the period after their application to the fellowship program. We verified all past and new publications online using PubMed ID (PMID) or digital object identifiers (DOI), where PMID was unavailable. At baseline, the publications provided by applicants in their application packages were identified online using electronic search engines such as PubMed or Google Scholar. At follow-up, participant names were used to search these databases by trained research assistants who retrieved any new publications in the last 12 months. We included all peer-reviewed publications, regardless of their topics, article types, or study designs. Additionally, semi-structured interviews were recorded with fellows annually and the results reported periodically.²³

Data analysis

We compared groups at baseline using Fisher’s exact test for categorical variables and Mann-Whitney U test for continuous variables. Poisson regression compared the rate of peer-reviewed publications between fellows and controls. We examined bivariable and multivariable associations between the publication rate and explanatory variables. We a priori adjusted the Poisson regression for potential confounders: (i) number of all past publications at baseline, (ii) advanced graduate degree, e.g., MPH, PhD (yes vs. no), (iii) past research experience, i.e., research assistantship/coordination (yes vs. no), and (iv) past research funding that the person has personally received for research-related activities in the past five years (yes vs. no). Age and publication rates were treated as time-varying variables. All hypothesized confounders were included in the fixed multivariable model. All p-values were two-sided. All statistical analyses were performed using the SAS software version 9.3 (SAS, Cary, NC, USA). The Providence Health Care/University of British Columbia Behavioral Research Ethics Board has approved this study (Certificate no. H14-00244). The primary research question and analysis plan have not been pre-registered, and therefore results should be considered exploratory.

Results

Between 1 July 2014 and 30 June 2020, 56 (39 women) physicians, both fellows (n = 25) and non-admitted applicants (n = 31), entered the study (Table 1), and were followed annually at the conclusion of fellowship. At baseline, in the fellows’ cohort: 76% of participants (19/25) reported past research involvement, 24% (6/25) had one or more advanced graduate degrees (e.g., MPH), and median number of peer-reviewed, first author publications was one (Interquartile Range [IQR] = 0–2). At baseline, in the controls’ cohort: 84% of participants (26/31) reported past research involvement, 39% (12/31) had one or more advanced graduate degrees, and median number of peer-reviewed, first author publications was zero. They were internal medicine physicians (n = 8), family physicians (n = 33), psychiatrists (n = 5) and others (n = 4). Significantly more controls than fellows graduated from a Non-Canadian medical school (88 vs. 58%), and reported Psychiatry residency program (0 vs. 16%). Among fellows, higher proportion of them completed Family (68%) and Internal (24%) medicine residency. The median length of time in the study was four years (IQR: 3–6).

Table 1.

Baseline Characteristics of the International Collaborative Addiction Medicine Research Fellowship Study Participants Stratified by the Fellow and Control Cohorts – 2014–2020 (n = 56).

Characteristic	Value	TotalN (%)56 (100)	Fellowsn (%)25 (45)	Controlsn (%)31 (55)	p-Value
Age, years	Median (IQR)	34 (30–39)	32 (30–36)	34 (31–40)	0.281
Gender	Male	17 (30.4)	8 (32.0)	9 (29.0)	0.810
	Non-male	39 (69.6)	17 (68.0)	22 (71.0)
Year of medical school graduation	Median (IQR)	2010 (2006–2013)	2011 (2007–2013)	2010 (2005–2013)	0.382
Medical residency program
	Internal	8 (14.3)	6 (24.0)	2 (6.5)	0.040b
	Family	33 (58.9)	17 (68.0)	16 (51.6)
	Psychiatry	5 (8.9)	0 (0)	5 (16.1)
	Otherc	4 (7.1)	1 (4.0)	3 (9.7)
Canadian Medical School	Yes	40 (71.4)	22 (88.0)	18 (58.1)	0.032b
	“No”	15 (26.8)	3 (12.0)	12 (38.7)
No. first-author papers at the time of application *	Median (IQR)	1 (0–2)	1 (0–2)	0 (0–2)	0.468
No. coauthor papers at the time of application *	Median (IQR)	1 (0–3)	1 (0–3)	0 (0–3)	0.612
Advanced graduate degree *	Yes	18 (32.1)	6 (24.0)	12 (38.7)	0.241
	“No”	38 (67.9)	19 (76.0)	19 (61.3)
Past research experience *	Yes	45 (80.4)	19 (76.0)	26 (83.9)	0.514b
	“No”	11 (19.6)	6 (24.0)	5 (16.1)
Received research funding in USD a	Yes	6 (10.7)	2 (8.0)	4 (12.9)	0.682b
	“No”	50 (89.3)	23 (92.0)	27 (87.1)
Received research funding in CAD a	Yes	16 (28.6)	9 (36.0)	7 (22.6)	0.269
	“No”	40 (71.4)	16 (64.0)	24 (77.4)
No. scholarly presentationss a	Median (IQR)	3 (1–8)	3 (1–6)	4 (1–8)	0.766

^*Per participant, as self-reported on the date of application/informed consent, i.e., 3 May 2014–13 December 2019. Five (20%) participants also undertook the RAMS (Research in Addiction Medicine Scholars) Scholar program at Boston University, MA.

^aDenotes activities in the past 5 years before the date of application.

^bFisher’s exact test due to small cell counts. For categorical variables, the test is chi-square test. If the cell count is small, Fisher’s exact test is used, as mentioned in the methods. For continuous variables, Wilcoxon rank sum test is used.

^cThere are 6 missing data points for medical residency program and 1 missing data point for Canadian medical school.

The participants authored a total of 260 peer-reviewed articles (of those, 191 were unique papers) in the period after their application to the fellowship program, 239 (67 first-authored) in the fellows’ group and 21 (13 first-authored) in the control group, published in 48 peer-reviewed journals (Table 2). At follow up, the fellows had a median of two peer-reviewed, first author publications (IQR = 1 − 4), and the controls had a median of zero first author publications. There was a significant difference between fellows (n = 25) and controls (n = 31) in total number of all publications (Rate Ratio [RR] = 13.09, 95% Confidence Interval [CI], 5.01 − 34.21, p < 0.001), as well as first author publications (RR = 5.59, 95% CI, 2.23 − 14.06, p < 0.001). RR for last-author publications and coauthor publications were not calculated due to the small number of publications in the control group (i.e., 2 publications).

Table 2.

Publication Characteristics Stratified by the Fellow and Control Cohorts – 2014–2020 (n = 56).

Publication characteristics	Total papers N (%) 56 (100)	Fellows’ papers n (%) 25 (45)	Controls’ papers n (%) 31 (55)
Total number	260 (100)	239 (91.9)	21 (8.1)
First authored	80 (30.8)	67 (28.0)	13 (61.9)
Last authored	43 (16.5)	41 (17.2)	2 (9.5)
Coauthored*	137 (52.7)	131 (54.8)	6 (28.6)
Total person-years	261	127	134
Rate (95% confidence interval)Number of publications per person-year
Total publications	1.00	1.88	0.16
	“(0.56–1.78)”	(1.07–3.30)	(0.07–0.33)
First authored	0.31	0.53	0.10
	“(0.22–0.43)”	(0.39–0.71)	(0.04–0.23)
Last authored	0.16	0.32	0.01
	“(0.06–0.43)”	(0.12–0.84)	(0.01–0.06)
Coauthored	0.52	1.03	0.04
	“(0.27–1.04)”	(0.54–1.96)	(0.02–0.10)

Notes: Four participants switched groups. They were counted as controls at the beginning and then were counted as fellows starting from the year when they switched to the fellow group. While they were counted twice, this was in different time points. All accepted fellows completed the one-year fellowship to date.

^*For each individual, we counted the number of first author papers, coauthor papers, and last author papers and then summed them up as the total number of publications for each individual. Each individual could only have one role in a paper.

In a multivariable Poisson model adjusted for fellowship admission criteria, such as training and experience in research and socio-demographic factors, the number of first author publications at follow up was positively associated with greater likelihood of being in the fellow cohort (Adjusted RR [ARR] = 4.93, 95% CI: 2.11–11.53, p < 0.001), having baseline first-author paper experience (ARR = 2.39, 95% CI: 1.73–4.15, p = 0.002), and having a baseline research experience (ARR = 3.01, 95% CI: 1.86 − 4.86, p < 0.001) (Table 3).

Table 3.

Unadjusted and Adjusted Poisson Regression Analysis of Factors Associated With the Number of First Author Publications at Follow Up – 2014–2020 Cohorts.

		Unadjusted		Adjusted
Characteristic	Rate ratio(95% CI)	p – Value	Rate ratio(95% CI)	p – Value
Group
(fellows vs. controls)	5.59 (2.23–14.06)	<0.001	4.93 (2.11–11.53)	<0.001
Age
(per year increase)	0.98 (0.93–1.04)	0.485	1.00 (0.97–1.03)	0.813
Gender
(male vs. non-male)	0.89 (0.44–1.84)	0.761	0.92 (0.51–1.67)	0.789
Baseline first-author paper experience *
(yes vs. no)	2.66 (1.34–5.25)	0.005	2.39 (1.37–4.15)	0.002
Baseline coauthor paper experience *
(Yes vs. no)	1.80 (0.85–3.79)	0.123	0.69 (0.36–1.32)	0.268
Advanced graduate degree †
(Yes vs. no)	0.75 (0.33–1.71)	0.491	0.68 (0.36–1.26)	0.220
Past research experience †
(yes vs. no)	2.75 (1.14–6.59)	0.024	3.01 (1.86–4.86)	<0.001
Received funding (USD/CAD) †
(yes vs. no)	1.91 (0.96–3.82)	0.066	1.11 (0.63–1.96)	0.723

^*The baseline self-reported number of First-author publications and coauthor publications were dichotomized to binary variables (yes vs. no).

^†Self-reported on the date of application/consent, i.e., between 3 May 2014 and 13 December 2019.

DISCUSSION

The first six years of fellows’ productivity indicate undertaking this fellowship was independently associated with greater research outputs in comparison to controls, signaling progress in the training of addiction physicians to advance participation of physicians in addiction research. As expected based on preliminary data,¹⁵ the number of first author publications at follow up was positively associated with greater likelihood of being in the fellow cohort, having baseline first-author paper experience, and having a baseline research experience. Here, qualitative interviews with mentors suggested that fellows who had past research experience were perhaps better prepared to publish for the purpose of improving their track record.²³ This high impact on productivity among fellows can also be attributed to the efficacy of key features of this training program which are aligned with the scientific evidence: instruction in basic addiction research methods, hands-on experience in research, clear guidance regarding development of research products, interdisciplinary mentorship and protected time for research.^16–19,23 As described in the exploratory qualitative study, which was published elsewhere,23 the presence of a supportive learning environment, flexibility in timelines, and clear guidance regarding the development of research products facilitated clinician-scientist training.

Future research should investigate the impact of research training programs on the sustainability of the findings and long-term outcomes, such as success with securing faculty positions and obtaining grant funding, using tools such as the NIH reporter for more distal fellow cohorts. It would also be interesting in the future to do more specific analyses on how much each of the various components of the training contributes to future research productivity (research instruction vs. fellowship protected time vs. collaboration opportunities, etc.). While this analysis focused on a shorter term outcome of academic productivity, this metric is widely accepted as academic progress in literature and is a pre-requisite for securing faculty positions and grants. The growth of the alumni cohort will allow evaluation of longer-term outcomes in the future and impacts of the continued relationship of alumni with the host site on productivity rates. Furthermore, robust designs to adequately appraise impacts of research training among future cohorts of clinician-scientists are needed. For instance, other research training programs for early-career scientists include the Research Scholars in Addiction Medicine (RAMS),²⁴ the Chief Resident Immersion Training (CRIT) Program in addiction medicine,²⁵ the Substance Abuse Research Education and Training (SARET),²⁶ and the Writers’ Task Force.²⁷ The RAMS reported that over five cycles (2012–2018), 24 scholars had a total of 201 publications (82 were first- and 30 senior-authored);24 the CRIT reported at 11-month follow-up 86% (56/64) chief residents describing the impact of their teaching project (which included lecture/teaching activity, new screening protocols, etc.) on their residency program curriculum;²⁵ the SARET reported more positive attitudes about research, and, in some cases, change in career plans among 30 students of nursing, medicine, and dentistry who completed research mentorships (25 summers, 5 yearlong);26 and, the Writers’ Task Force reported that in 14 years there were 113 postdoctoral fellow enrollments in a six-month writing seminar. Of the 113 participant enrollments, 97 (86%) submitted a manuscript for publication, and 87 participants (77%) published their manuscript.²⁷ The differences in productivity among these programs might be attributable, in part, to the characteristics of the program, mentors, trainees, and the training environment. These programs would also benefit from evaluation including controlled design and with consistent outcome measures.⁹

We acknowledge limitations of this study design and results. In particular, that individuals were not randomized for entry into the fellowship, i.e., the fellows were admitted based on higher-ranking at the time of application to the program. To address potential for confounding based on prior training and research experience, a range of factors were controlled for in the multivariable analyses. The factors included certain admission criteria, training, and experience in research as well as socio-demographic factors. Nevertheless, we were unable to control for the individual and socio-structural factors that may impact research productivity and career success in this field (e.g., motivation, personality type, or self-selection bias, which is common for this type of nonrandomized studies).²⁷ Nevertheless, since the fellowship provides structured mentorship in research productivity, as well as protected time for learning about research, access to research data, research infrastructure, and research collaborators, we argue that the findings are at least in part attributable to the impacts of the research training fellowship.

Based on presented six-year data, we found that, after adjusting for background propensity to contribute to research, undertaking this research fellowship program was associated with significant research outputs in comparison to controls. Ultimately, it is hoped that the fellowship will contribute to the next generation of addiction clinician-scientists who can help close the implementation gap between addiction science and clinical practice.