Beyond the Numbers: Finding Hidden HIV Risk in Sub-Saharan Africa

Dr. Katherine Rucinski and Yuanqi Mi explore their newly published article,“Defining strata of women at risk for HIV in Sub-Saharan Africa: a pooled latent class analysis,” published in the October 2026 (Vol. 122) issue of Annals of Epidemiology. This study investigates the different profiles and levels of HIV risk among female sex workers in sub-Saharan Africa using combined survey data.

Read the full article here:
https://www.sciencedirect.com/science/article/abs/pii/S1047279726001912

Episode Credits:

• Executive Producer: Sabrina Debas (Episodes 1-18) and Sofina Tran (19-)
• Technical Producer: Paula Burrows
• Annals of Epidemiology is published by Elsevier
  
  
  

Transcript

Copy Transcript Copied

Patrick Sullivan 0:11

Hello, you're listening to EpiTalk Behind the Paper, a podcast from the Annals of Epidemiology. I'm Patrick Sullivan, editor-in-chief of the journal, and in this series, we'll take you behind the scenes of some of the latest epidemiologic research featured in our journal. Today, we're here with Dr. Kate Rucinski and Yuan qi Mi to highlight their article, Defining Strata of Women at Risk for HIV in Sub-Saharan Africa: A pooled latent class analysis. You can read the full article online in volume 122, the October 2026 issue of the journal at www.annalsof epidemiology.org. So I'll introduce our guests, Katherine or Kate Rucinski is an associate scientist in the Department of International Health and Core Faculty with the program for implementation and equity research at the Johns Hopkins Bloomberg School of Public Health. An epidemiologist by training, her work addresses the social and behavioral dimensions of HIV and sexual and reproductive health globally, with a particular focus on women and girls. Her research aims to improve the delivery of HIV prevention and treatment, contraception, and other sexual and reproductive health services in restores-constrained settings through collaborative, co-designed work in which communities shape, refine, and advocate for their research priorities. Her work supports implementation-oriented outputs designed to optimize existing programs for populations she serves, conducting this work alongside implementing partners throughout Southern Africa and in the United States. Dr. Rucinski holds a PhD in epidemiology from the Gillings School of Public Health at the University of North Carolina Chapel Hill, and an MPH in international health from New York University. Yuanqi Mi is a research data analyst in the Department of Epidemiology at the Johns Hopkins Bloomberg School of Public Health. Her research focuses on HIV prevention and care and health disparities among marginalized populations with particular interests in structural determinants of health, stigma, and social and policy environments that shape HIV prevention and care outcomes. She received her bachelor's degree in nursing from Peking Union Medical College in China and her Master of Science degree in epidemiology from the Johns Hopkins University. Thank you both so much for joining us today on EPITalk: Behind the Paper.

Dr. Kate Rucinski 2:43

Thank you so much for the invitation. We're really excited to talk about our paper and are delighted to be here today.

Patrick Sullivan 2:48

So can you start just by giving some background about the problem described in your paper? Why was this an important topic to take on?

Dr. Kate Rucinski 2:58

Yeah, I'm happy to start us off. You know, I think we know that female sex workers globally really are disproportionately affected by HIV. And in sub-Saharan Africa, we really see a high incidence and prevalence of HIV across sex worker communities. And I think importantly, this sort of large incidence and burden of HIV is reflective of sort of larger historical issues related to policy, legal frameworks, and stigma. And I think these sort of trickle down and impact service delivery needs as well. You know, I think importantly, communities have really been effective at reaching sex workers with programming specific to HIV prevention and treatment. But I think we're always dealing with resource constraints. And if the last year has taught us anything, we really need to do more with less. And so I think trying to understand communities of sex workers that may benefit from more targeted intense service delivery approaches was really at the core of this problem and analysis that we wanted to do.

Patrick Sullivan 4:02

So say a little bit more about the analysis. You sort of introduced the topic, but in this particular analysis, what was the key question and how did you choose the methodology that you chose to answer this particular question?

Yuanqi Mi 4:16

I'm happy to take that question. So we use latent class analysis or LCA, which is essentially a person-centered approach that allows us to identify underlying groups of individuals based on patterns across multiple characteristics. In our study, rather than defining someone as being at high risk based on a single behavior such as number of clients or inconsistent condom use, we look at multiple dimensions together, including structural determinants such as stigma and violence they experience. So LCA allows us to identify groups of women who share similar patterns across these characteristics.

Patrick Sullivan 4:59

And so what is the thought? I'm jumping ahead a little bit here to link to how your research will impact health, but once you get these latent classes, you know, how could that be used to improve health?

Dr. Kate Rucinski 5:11

Yeah, I'm happy to take that. And I think it's a conversation that we have been having both in the context of this paper and then also more broadly as we think about opportunities to leverage existing data sources with sort of novel and innovative methods. You know, I think at the sort of beginning of this work, we thought about could we somehow translate our findings from these latent class analyses to some sort of operationalized tool in a clinical setting? So for folks that were already accessing programs, would there be an opportunity to pull out some of these sort of more salient factors that we perhaps were not measuring in the context of routine services, but we thought would be sort of core indicators of how women would engage with services moving forward. And I think that still holds true. But I also think, you know, some of the findings from this paper also point to sort of subgroups of sex worker communities that are really not regularly engaging with programs, even community-based programs. And so thinking about ultimately using these findings to inform who is not being served, and potentially as we think about restructuring programs and reallocating resources, who we should be targeting who are not otherwise engaged with programs already.

Patrick Sullivan 6:23

Great. So that does sort of bring us back to this question about latent class analysis, either maybe not even so much in the technical sense, but like why latent class analysis is the right kind of approach to ask the question that you're asking, which is about really, it's a really pragmatic question about identifying women who might benefit from services and how you do that with data, right? So why does latent class analysis make sense with that kind of question?

Dr. Kate Rucinski 6:52

I mean, I think it's a great question. And I would maybe push back a little bit and say that I don't know if there's ever a perfect analytical or methodological approach with data. But I think one of the reasons that we were really drawn to using more of a latent variable approach here is that we weren't necessarily interested in sort of like single indicators around risk or single indicators that were measuring these sort of larger higher order determinants. I think we wanted to understand what these underlying constructs were that really existed across multiple indicators. And so I think that was really the motivation here for using latent class analysis. I'll also say that we initially embarked on this multi-country latent class analysis where we were really specifically including heterogeneity across these different countries in the model, and we ran into some convergence issues. And so we are able to interrogate that heterogeneity a little bit across the different countries, but the actual analysis differs a little bit from what we had hypothesized to begin with.

Patrick Sullivan 7:51

Which is okay, right? All right. So with that methodologic background in mind, what were some of the key findings that you describe in the paper?

Dr. Kate Rucinski 8:01

So I think there's a few different levels of the findings. I think the most obvious finding is that we came up with four discrete classes, which were defined by different conditional probabilities across these indicators. So the classes were defined by indicators that reported the sort of highest probability to really sort of paint a picture in terms of how these classes differed. So the first was women that experienced a high probability of alcohol and violence. The second was women that experienced a high frequency of client turnover. The third was defined by women who reported, on average, a lower client volume relative to their peers. And then this fourth class that reported limited use of condoms and lubricants. And then I think within these classes, we were really interested to understand how, again, some of these sort of higher order determinants potentially varied across class membership to give a little bit of insight as to what was going on under the hood. And so I won't go through all of them, but I think a couple of really salient findings are that women in this class that experienced a high probability of alcohol use and violence also experienced stigma related to accessing legal services, stigma related to speaking with their family members about their role as sex workers. And so I think potentially suggesting that those women were particularly vulnerable above and beyond the alcohol and violence use that we captured as part of those indicators. We saw similar patterns for women that were experiencing a high degree of client turnover. And then in this class where we had limited condoms and lubricant, we actually found that this group was the group where women were on average the youngest. And I think while we can't confirm this from the data that we collected, our thoughts are these are women that are potentially newer to sex work, not as networked as their older counterparts, and also less likely to be engaging in services with community-based partners.

Patrick Sullivan 9:56

So when you think about those findings, how do you make the connection then to like the public health action or like a health equity framework that we apply in many cases in this setting of sub-Saharan Africa? How would you package up that finding and take it and propose a change that would maybe mitigate some of these vulnerabilities?

Yuanqi Mi 10:22

I think one contribution is that the study challenges the idea that a general prevention approach will work equally well for all the female sex worker in Sub-Saharan Africa. And we found meaningful differences in both HIV-related vulnerabilities and engagement with prevention services. So, from a health equity perspective, I think it's important because equity doesn't necessarily mean providing exactly the same services to everyone. It means recognizing that different groups may face different barriers and may need different types of support. And one of our key findings suggests that more differentiated HIV services could help identify and reach women who are currently underserved. For example, women with limited condom and lubricant use who appear to be less engaged with HIV testing. And I think that could be generalized across different contexts.

Dr. Kate Rucinski 11:17

Maybe just also to add there, thinking about how to pragmatically operationalize some of this work. You know, I think there is this theme across our findings that we find that younger sex workers are particularly vulnerable. And as we think about what makes sense from a prevention standpoint, I think we know that HIV prevalence only increases with age. And so folks that are younger, I think are inherently ripe for interventions. And so thinking about ways, particularly in this new situation where we find ourselves with constricted funding, to really sort of focus on those younger community members who may benefit potentially from more intense services, thinking about adapting services that historically have targeted key populations to be more youth-friendly, to sort of merge what we know about, you know, providing HIV prevention for youth with sort of how we deliver HIV prevention for sex workers is, I think, something that we critically could explore moving forward with our interventions.

Patrick Sullivan 12:16

And it is very pragmatic, you know, this kind of latent class analysis approach, because you're not making assumptions up front about what those associations might be, but the kinds of things that you're you're talking about as both of you interpret this really are actionable in terms of being focused or having at least a place to start and prioritize provision of services that's an empiric, you know, basis for that. So it's a really nice sort of follow-through from the methods to the application to improve public health, which in the end is what it's about. So if you don't make that step, then the good in the world doesn't manifest. So I just want to ask one more question about the analysis, and then we'll talk a little bit about the how, which is what kind of limitations or weaknesses did you think about and mention in the article, or just for this kind of research and this analysis in particular, what are limitations and weaknesses would you think are important to recognize?

Yuanqi Mi 13:13

Um, I can take that question. So, one question that came up during the analysis was whether it was appropriate to pull data across countries. And we ultimately decided to pull the survey data collected across nine countries and several years because it allows us to identify the broader patterns of heterogeneity across different settings. But since the social and programmatic context differ across these countries, I think we should be careful not to interpret these four classes as fixed categories that will necessarily look exactly the same in every setting. Another limitation was that some indicators were operationalized somewhat differently across settings. And as with any latent class analysis, the classes were we identified depended on the indicators that were available and measured consistently enough to include in the poll analysis. So there might be other important dimensions of HIV vulnerability that were we were unable to capture.

Patrick Sullivan 14:17

So you talked about the opportunity to sort of pool data and you get, in some ways, a broader inference and you get more data. Is there a role for to take this kind of analysis to a smaller geographic level? And do you think there might be differences in the patterns within a specific country? And might that be relevant for what prevention or programmatic recommendations are?

Dr. Kate Rucinski 14:40

So I think it's a great question. And I think, you know, importantly, despite all of the limitations that Yuanqi really nicely identified, we did do quite a few sensitivity analyses with our data. And so one of those analyses is we sort of did this leave one-out method where we repeated the analysis, excluding countries iteratively. And I think importantly, we generally saw that the class structure was maintained across these sensitivity analyses. But to your point, Dr. Sullivan, you know, I do think context really matters. And even if the classes hold across different settings, it's possible that their overall prevalence might look very different. And so we might be thinking from sort of a resource perspective, maybe focusing on a specific class, such as the alcohol and violence class, relative to some of the other classes that we identified across this analysis and other settings.

Patrick Sullivan 15:25

Great. Okay.

Behind the Paper

Patrick Sullivan 15:27

So it's great to have both of you on the podcast and to talk about your roles, but I wonder if you could talk a little bit about how the collaboration on this project came to be and how you came to take on the roles that you did working together as someone maybe who's earlier in their career and someone who's mid-career or later in their career. I'm always interested in how those kinds of collaborations come together and turn out to produce great work like you've done here.

Yuanqi Mi 15:54

How should I say this? I got my master at Johns Hopkins and I stay in the same team after graduation as a full-time research data analyst. And this project is actually not the project that I start with, but I kind of join in the middle of the analysis. And my main responsibility is writing the paper and make sure it aligns with the initial purpose of the analysis. And I think this kind of work style is one of the most interesting working style that I have experienced because I get to know like how a faculty member writes papers. And it's really different from what I originally write papers, because I see uh Kate would normally outline the important points that she wants to stress in the introduction, and I would kind of elaborate on that. But it's a really interesting experience and I learned a lot from it.

Dr. Kate Rucinski 16:50

Maybe if I could also just chime in here, because I think sometimes we tend to be humble, but you know, Yuanqi has really joined our team with such gusto. And I think with this type of work, the pie is really big. And so there are just sort of more opportunities than we have person power for. And I know Yuanqi had been leading on other late-in-class analyses, both during her time as a student and as she sort of started as an analyst with our team. And this work was already underway. And so she came to me and she said, I would be really interested in working on this. And I said, I don't actually have the capacity to write the paper myself. I've drafted a lot of it, I've done a lot of the analysis. But like if you want to come in and support a sort of co-first author, we would only welcome that opportunity. And so it's been a really nice collaboration between the two of us. And without Yuanchi's support, I don't think this paper would have ever made it to publication. So we're so grateful that she volunteered and stepped on board.

Patrick Sullivan 17:50

Even when we record these podcasts, we do it on Zoom. And so I can see the smiles and nods that the listeners won't see, but there's a really genuine sense of shared purpose and appreciation here that just comes through in every way.

Yuanqi Mi 18:02

So I'll also just say that Yuanqi is really, like many of the other team members we've worked with and mentored historically, interested in pursuing PhD work. And so I think making sure that she's engaging in these analytic opportunities early on and using this as a moment to actually define her own ideas and research agenda has been a really nice byproduct of this collaboration and this specific analysis in particular.

Patrick Sullivan 18:27

So we've talked about the research that you published and the sort of implications of that for health, but I wonder what each of you thinks about next steps or next questions. Great research always maybe answers one question and spawns three more. So in this area, what else are you interested in thinking about in the future, either in terms of applying these methods to related issues or what other kinds of EBI questions need to be asked about your substantive findings that would help take this knowledge and translate it to improvements in health?

Dr. Kate Rucinski 19:01

I think this is a really important question in terms of where our field is going. And I think one of the real value adds of this analysis is that we were working with data that were collected really over the past 10 years. And I think it subscribes to this ethos of what can we do with existing data and how we can pool together data that have been collected across different time periods and settings, fully appreciating that the value that we can gain from some of these larger pool data really outweighs some of the limitations. You know, I think as a field, we are really drilling down on the importance of implementation science. And I think that's exactly in line with our philosophy around how we approach this analysis. And so, you know, we know that lots of effective interventions exist for HIV prevention and treatment. We know that populations that most need these interventions are not able to access them. And so, how can we use existing data, whether it's through surveys or through program partners, to really understand why that is and how we can actually potentially change how we're delivering services in the field to make sure that they are more fully accessible? And so again, I think this movement towards doing more with less and leveraging existing data and working collaboratively with partners that are in the field, collecting data day to day is quite important. Yuanchi, do you have some thoughts here?

Yuanqi Mi 20:25

Yeah, I think another point that I want to mention is regarding the potential next step. I've been to a lot of conferences and I noticed that there are more and more sections around machine learning. And we have a lot of unutilized data in our group that are like biobehavioral data sets. And I think there are potentials there to use that data like to generate machine learning behavioral profiles in different regions.

Patrick Sullivan 20:51

Great. So thinking about other kinds of analytic approaches that could give you some more insights. And it's probably true that like different kinds of analytic approaches are likely to different or surface different aspects of these relationships. So cool idea. So Yuanqi, you are an early career professional and working. What are you most looking forward to as your career evolves? What are you excited about in terms of new methods to learn or new topics to study? And what do you see in your professional future?

Yuanqi Mi 21:23

I think what I most look forward to is developing the ability to become a more independent researcher and eventually being able to lead my own projects. And I'm really excited about the opportunity to further develop my own research interests and strengthen my methodological skills. In particular, I'm interested in learning about theories of structural stigma and other structural determinants of health, how we can conceptualize and measure these broader social and policy environments, and how they can translate into individual level experiences and health outcomes. I've learned a lot from the researchers I've worked with so far, and I'm looking forward to conducting research that can help promote health equity among marginalized populations.

Patrick Sullivan 22:12

And Kate, I'm going to pitch you a sideball here, which is uh what have you learned from working with Yuan Chi on this analysis?

Dr. Kate Rucinski 22:20

I've learned a lot from working with Yuanqi on this analysis. I think sometimes we are siloed a little bit in terms of how we're working on papers, and we're very focused on how we approach things both analytically and then also from a writing standpoint. And as we talked about earlier, this has really been such a nice and truly collaborative opportunity. And so I think Yuanchi, being a little bit newer in her career and closer to methodological instruction, had a really keen eye for approaching how we handled this analysis and all of the assorted limitations, and also was able to propose solutions for how to interrogate some of those assumptions and potential limitations in a way that I probably would not have done had it not been for her keen insight. So I think that's been really fantastic. I think she's made me a better writer and forced me to be a little bit more clear in my prose and sort of translating findings from the analyses to real world implications. And I'm very grateful for that.

Patrick Sullivan 23:18

I love your frankness and thoughtfulness about this because I find the same thing, which is in some cases, it's that earlier career people are closer to their classroom instruction and have heard things recently that maybe we haven't heard in a while. But it's almost always true in these kinds of earlier career and later career collaborations that, and I asked these questions in podcasts before, that the learning is really goes in both directions. And I think sometimes we think structurally that it's more one way, but it almost never is. And so this idea that earlier career colleagues have been in the classwork more recently and thinking about maybe the analysis approaches in ways that are developed after we took classes, later career folks took classes, but also just seeing the systematic approach, like when you're collaborating with someone, I think in any kind of collaboration, it forces you to get more systematic and organized about what you're going to do as opposed to being we can feel a little ad hoc sometimes when we're on our own. So for all those reasons, I think that it's so nice to be able to spotlight the work that came out of this collaboration. And in this podcast, it's very generous of both of you to open up to talk a little bit about the professional collegial side of this and how you work together, which I think is such an important part of how impactful work gets done. So I think that's a good place to wrap up. Is there anything else that either of you would like to share with our listeners as we conclude this episode?

Dr. Kate Rucinski 24:54

I've been asked a lot very recently from people that are newer to epidemiology and newer to global health if there's room at the table for them, given sort of the big disruptions we've seen over the past two years. And I think it's a really hard moment to be in our field. But I really believe that this is an opportunity. And I think unfortunately, we have created sort of seismic harms through the communities that we serve. And I think that ultimately will present substantial opportunities to ask really meaningful public health research questions moving forward. And so I think sort of a closing message that I'd love to offer is please come and join our team. We really need this next generation of leaders and public health professionals to come work on these really hard, complex problems.

Yuanqi Mi 25:44

No, but I think Kate covers that really well.

Patrick Sullivan 25:47

Yeah. Inspirational. I love it. I'm your host, Patrick Sullivan. Thanks for tuning in to this episode and see you next time on Epitalk. Brought to you by Annals of Epidemiology, the official journal of the American College of Epidemiology. For a transcript of this podcast or to read the article featured on this episode and more from the journal, you can visit us online at www.annals of epidemiology.org.