Study setting

This study was conducted as part of the Global Health Fellows and Scholars Program, funded by the Fogarty International Center and other participating institutes and centers at the US National Institutes of Health. Specifically, we engaged trainees from the UJMT Consortium, comprising principal and collaborating faculty from the University of North Carolina at Chapel Hill, Johns Hopkins University, Morehouse School of Medicine, and Tulane University. The UJMT Consortium has been funded to support global health research training efforts since 2012 and has provided training opportunities to more than 200 trainees (described in greater detail elsewhere [16]). At the time this study was conducted, the UJMT Consortium was one of six such groups that supported US and international trainees in one-year mentored global health research attachments.

We engaged trainees from the prior five years of the program (2017–2022), including pre-doctoral and pre-professional trainees from HICs (i.e., US citizens or permanent residents) and postdoctoral and post-professional trainees from both the U.S. and LMICs. Training sites were located in Bangladesh, Democratic Republic of the Congo, Ghana, India, Malawi, Mexico, Peru, Sierra Leone, South Africa, Suriname, Uganda, Vietnam and Zambia. From 2017–2022, the top four training sites accounted for 76/111 (68%) of trainees (29 in Malawi; 19 in Peru; 16 in Uganda, and 12 in Zambia). All trainees in the UJMT Consortium have English language proficiency.

Sample

We created our sample from the complete list of trainees and mentors within the most recent five years of the UJMT program (2017–2022). We chose the most recent five years because attention to decolonizing global health is a relatively recent phenomenon and because surveying recent trainees was better suited to the broader projects’ goal of developing new ethics and decolonizing global health education materials (e.g., to reduce recall bias in surveying trainees from 10 years prior). All 111 trainees from this time period were sent emails asking them to participate. We also sent recruitment emails to 91 active mentors from the UJMT Consortium who have supported at least one trainee in the program, asking them to complete a survey.

Survey instrument

After reviewing the literature, we found no validated items related to the concept of decolonizing global health; therefore, we created two de novo survey instruments (one geared toward global health trainees; another, toward global health mentors) based on current literature and the research team’s experience. The goals of the survey were to understand individuals’ self-reported knowledge and attitudes toward decolonizing global health and to evaluate perceptions of the impact of historical colonialism on eight aspects of global health today: (1) funding available for global health; (2) research questions in global health; (3) research questions that get prioritized for funding; (4) developing collaborative relationships in global health; (5) data/statistical analyses used in global health; (6) authorship decisions on global health; (7) equal career advancement opportunities; and (8) global health research, overall.

The survey went through multiple rounds of feedback and revision to ensure face and content validity. First, it was reviewed by a Technical Advisory Panel comprised of nine individuals from diverse geographic locales and with diverse expertise in global health and global health education. After revision, we conducted five cognitive interviews with three global health trainees and two global health mentors, who represented the intended audience for our survey. Following iterative refinement, the final instrument included 36 core items and required 15–20 minutes to administer (see Appendix).

Survey administration

The survey was conducted online from February 2022 to April 2022 via Qualtrics (Provo, UT). It was optimized for both desktop and mobile format. Survey respondents were given the option of being entered into a random chance drawing to win a tablet computer device. We sent several email reminders by email to encourage participation.

Data processing and analysis

For analysis of quantitative data, responses were exported for analysis in Stata 17.0 (Stata Corporation, College Station, Texas, USA), which is a secure, password-protected software platform. Only fully completed surveys were included in the final analysis. One survey was only 40% complete and was removed from the sample. Another survey was missing only the self-identified gender item and was included.

Some variables were combined to aid in interpretation and to improve statistical power. For example, we assessed aspects of global health on a 7-point Likert scale from –3 (very negative) to +3 (very positive). Recognizing that this scale was not proven valid to differentiate, e.g., –3 versus –2, these data were analyzed by combining responses of –3, –2, and –1 into one category (“Negative”) and by combining +1, +2, and +3 into one category (“Positive”). Similarly, we assessed participants’ attitudes and emotions toward the concept of decolonizing global health on a 5-point Likert scale; these data were analyzed by combining “Strongly Disagree” and “Disagree” into a single category and by combining “Strongly Agree” and “Agree” into a single category. Unless otherwise noted, data were analyzed using Pearson Chi-squared tests and Fisher Exact Tests (for bivariate comparisons with small n’s).

For analysis of qualitative data, recognizing the limitations of open-ended survey response, we employed basic content analysis. Three members of the team (MD, MM, and LM) reviewed responses to generate initial codes of thematic content. Together they refined and elucidated definitions of these codes. Then, all the responses were re-coded independently with any disagreements in coding resolved by discussion to arrive at a final set of coded responses. For each response, we identified one primary theme (i.e., we did not allow for more than one coded theme per response). Given our interest in comparing LMIC to HIC trainees, coders were blinded to whether a response came from an LMIC or HIC trainee, in order to promote unbiased analysis.

Ethics approval

The study was reviewed and approved by Institutional Review Board at the University of North Carolina at Chapel Hill (IRB#21-2994). All participants gave their voluntary, informed consent to participate. The study was conducted in accord with the Declaration of Helsinki.

Characteristics of respondents

Characteristics of survey respondents are in Table 1. We noted that respondents from HICs were more likely to report female gender, a finding consistent with overall UJMT participation. Among all HIC trainees from 2017–2022, 55/69 (80%) reported female gender and 14/69 (20%) reported male gender. In comparison, over the same five years, 18/42 (43%) of LMIC trainees reported female gender and 24/42 (57%) LMIC trainees reported male gender. We observed that trainees from LMICs were more likely to report having no prior experiences in global health training (which we defined broadly to include both local and international programs, clinical and non-clinical). Trainees from LMICs were also more likely to report conducting implementation science research.

We were able to compare respondents to non-respondents within the UJMT program, and found no differences between respondents and non-respondents in terms of gender, LMIC versus HIC status, or program year, but we did find that respondents were more likely to be post-doctoral (i.e., 45/61 [74%] compared to 25/50 (50%) non-respondents). (See Supplementary Material Table S1).

Self-reported knowledge and attitudes toward decolonizing global health

Respondents’ self-reported knowledge and attitudes toward decolonizing global health are in Table 2. We found that 29/36 (81%) of respondents from HICs reported that they “know a little” about the concept of decolonizing global health, and all had heard of the concept. In contrast, participants from LMICs reported less overall knowledge of the concept; 5/25 (20%) had never heard of it and 16/25 (64%) reported that they “know a little.” This finding was statistically significant (p = 0.019).

In the set of questions related to respondents’ attitudes toward the concept, we found that, on the whole, most respondents from LMICs and HICs were supportive of the movement toward decolonization of global health (57/61, 93%), few were worried that it could do more harm than good (7/61, 12%), and most (54/61, 89%) were excited that decolonizing global health would give more opportunities for people in LMICs to succeed. There was relatively less support for the idea that decolonization would actually reduce global health inequalities, with 45/61 (74%) agreeing or strongly agreeing and 14/61 (23%) neutral on this topic.

The only statistically significant difference between HIC and LMIC respondents related to feelings of guilt. Respondents from HICs (15/36, 42%) were more likely to agree or strongly agree with the statement that they “feel guilty because I personally benefitted from colonialism” whereas only a single individual from a LMIC agreed with this statement and 17/25 (68%) disagreed or strongly disagreed (p = 0.001).

When examining the 26 total mentor surveys, this statement around feelings of guilt was an area where we noted a possible difference when compared to the trainee surveys. We observed a potential difference in the attitudes of HIC mentors. Compared to the 42% of HIC trainees who agreed or strongly agreed regarding guilt, only 2/16 (13%) of HIC mentors did so, and 9/16 (56%) disagreed or strongly disagreed with the statement that they felt guilty for having benefitted from colonialism.

Themes in how trainees define decolonizing global health

Our survey asked respondents to give a 1-sentence summary of what decolonizing global health means to them. Eight individuals responded, “I don’t know,” leaving 53 open-ended responses for analysis. Summary results of our qualitative content analysis are in Table 3.

We found that responses fitted into five general categories, with no clear difference in responses based on whether a trainee was from an LMIC or HIC. The most common responses related to active elimination of historical and colonial power imbalances. These responses included words and phrases such as the following: “intentional undoing,” “fights against the system of dominance,” “critique imperialism,” “dismantling colonial legacies,” or “removing the superiority of one population.”

The second most common category of responses focused on increasing the prioritization of local needs, without necessarily referencing the elimination of existing institutional structures. These responses included phrases such as, “re-centering global health priorities around local needs and interests,” or “reprioritizing global health studies to focus on the goals, values, and question in the population of study.”

The third most common theme related to an emphasis of local leadership and decision-making, again separate from the active elimination of power structures (and recognizing that prioritizing local needs may or may not imply local leadership). Here, responses emphasized “empowering developing countries to take lead on such collaborations” or “making global health a concept that the ‘global south’ actively participates in and influences.”

Perceptions of the impact of colonialism on global health

In Table 4, we present respondents’ perceptions of the impact of historical colonialism on global health today. In general, trainees from LMICs had mixed perceptions whereas trainees from HICs had generally negative perceptions of this impact. To illustrate, when asked about the impact of colonialism on which questions get asked in global health, 12/25 (48%) of LMIC respondents assessed this as positive compared to 5/36 (14%) of HIC respondents. However, 11/25 (44%) of LMIC respondents had a negative assessment. The differences in the overall distribution were statistically significant (p = 0.012). We observed a similar finding regarding the development of collaborative relationships in global health, where 16/25 (64%) LMIC respondents assessed this as positive and 25/36 or 69% of HIC respondents assessed this as negative (p < 0.001). Finally, we included an item intentionally chosen to be potentially seen as more neutral, that is, the effect of colonialism on the data/statistical analyses used in global health research. We found that HIC respondents saw this as predominantly neutral (22/36, 61%) whereas LMIC respondents saw this as positive (12/25, 48%; p < 0.001).

In Figure 1 we present graphically the comparisons in Table 4 between LMIC and HIC respondents’ perceptions of the impact of historical colonialism on global health today. (See Supplemental Materials Figure S1 for aggregate data from all respondents.) For better visualization, we plot these perceptions by removing the “Neutral” category from the eight items. For all but one prompt (“the total amount of funding available for global health today”), LMIC respondents more frequently had a positive assessment in response to our questions. This allows the overall trend of the greater perceived negative impact of colonialism among HIC respondents compared to LMIC respondents to be seen more clearly.

Within the items on perceptions of the impact of colonialism on global health, we observed a potential difference in the mentor responses. While limited by a small sample (n = 26 responses; 16 HIC, 10 LMIC), we did not see evidence that LMIC mentors were generally more positive than HIC mentors regarding the effect of colonialism. Across all items, at least 6/10 (60%) of LMIC mentors perceived the impact of colonialism to have been negative. On four of the eight items, 8/10 (80%) of LMIC mentors expressed negative perceptions of the impact of colonialism on global health. See Supplemental Materials Table S2 for tabular presentation of these data.

Our relatively small sample on a 7-point scale limits our ability to do additional analyses. However, we conducted several analyses to help ensure the validity of these findings. For example, we compared results when including the neutral category in the “Negative” impact category and when including the neutral category in the “Positive” impact category; neither affected our results. We also analyzed using the mean values on the full –3 to +3 scale; all three findings remained statistically significant. Using the Mann-Whitney test, only the question about collaborative relationships remained statistically significant (See Supplementary Materials Table S3).

We attempted to analyze for interactions between site of LMIC/HIC status and gender. We found no consistent findings in the direction or magnitude of the effect size (e.g., on two items where the LMIC/HIC difference was statistically significant, no gender effect was observed). For example, when comparing females to males within the HIC and LMIC categories, the differences seen were not statistically significant. However, our analysis was limited by our small sample size (more than half of the cells had an n < 5), which was not sufficient to produce reliable estimates on the basis of gender (See Supplemental Materials Table S3).