The research team, led by Kirsten Paulus and Sarah Bauerle Bass of Temple University College of Public Health, applied the Health Belief Model, a long-standing behavioral framework, to understand pandemic decision-making among people who use drugs. The model proposes that a person’s belief in the threat of a disease, together with their belief in the effectiveness of a proposed protective behavior, predicts the likelihood that they will actually perform that behavior. It rests on six constructs: perceived susceptibility, perceived severity, perceived benefits, perceived barriers, cues to action, and self-efficacy. Public health professionals have used the model for decades to design interventions that increase uptake of behaviors like disease screening, and during the COVID-19 pandemic it was successfully applied to predict vaccine intention and adherence to prevention practices across many population groups. What it had never been used for, until now, was understanding pandemic behavior among people who use drugs.

To fill that gap, the researchers conducted a cross-sectional survey between September and December 2021 with 75 clients of a large harm reduction organization in Philadelphia, an agency that serves more than 25,000 unique clients with services including medication for opioid use disorder, behavioral health care, infectious disease prevention, syringe exchange, medical treatment, and housing support. Research staff approached clients while they were receiving services, obtained informed consent in a private area, and administered the survey verbally on an iPad, entering responses into REDCap. Participants received a paper survey and a laminated scale sheet to improve visibility and comprehensibility, and each session took roughly 15 minutes under strict COVID-19 protocols. The instrument combined validated measures with study-specific items derived from earlier qualitative interviews, covering 18 sociodemographic questions and 22 thematic sections on COVID-19, vaccination, harm reduction, sex work, drug use, and basic needs. Five sections mapped directly onto Health Belief Model constructs, with respondents rating statements from zero for highly disagree to ten for highly agree.

The analytic centerpiece was a k-means cluster analysis, a non-hierarchical statistical method that discerns latent subgroups within a sample by assigning individual cases to clusters according to their proximity to the nearest centroid of constituent items, iterating until the desired cluster solution is achieved. Classification was based on four items capturing each participant’s relationship with COVID-19: whether they had ever tested positive, whether they knew someone who had the disease, whether they believed they were susceptible to getting it in the future, and a sum score of their perceived resiliency against difficult things. After accounting for missing data on 15 cases, the final analytic sample was 60 participants. Convergence was achieved after eight iterations, and a two-cluster solution emerged that was independent of every demographic variable measured.

The first cluster, 28 participants or 47 percent of the sample, was more likely to believe they had not been significantly affected by COVID-19 and that it posed no potential future threat, scoring a mean of 2 on the perceived impact item, and showed high overall resiliency with a mean score of 21. The researchers labeled this group Less COVID impact/High resilience, or NoCOV/HR. The second cluster, 32 participants or 53 percent, believed they had been significantly affected by COVID-19 or would be in the future, with a mean of 5, and had lower resiliency, with a mean of 13. This group was labeled High COVID impact/Low resilience, or COV/LR. Strikingly, the two clusters did not differ on age, gender, race, education, employment, income, housing status, vaccination status, or COVID-19 knowledge. Instead, the segmentation was defined by what the authors call psychographic characteristics, meaning perceptions and attitudes rather than demographic descriptors, a novel approach to understanding populations in COVID-19 research.

When the researchers compared the clusters on Health Belief Model constructs using Student’s t-tests and chi-square tests at a 0.05 significance level, meaningful differences appeared across four of the five construct sections. The COV/LR cluster more strongly agreed that people they know had gotten COVID-19, scoring a mean of 8.09 versus 5.61 for the NoCOV/HR group, indicating higher perceived susceptibility. The low-resilience cluster was also far more likely to agree that it is difficult to follow prevention instructions such as mask wearing and social distancing, with a mean of 3.94 versus 0.63, and more likely to report lacking things to keep them safe, such as masks or hand sanitizer, at 2.16 versus 0.67. On the severity side, 58.1 percent of the COV/LR cluster agreed they had to continue working in close contact with people who might be infected, compared with 26.9 percent of the high-resilience cluster, and 71.9 percent reported an increase in mental health problems or symptoms such as mood, anxiety, or stress, versus 46.4 percent.

The high-resilience cluster showed its advantages on the self-efficacy and cues-to-action constructs. Those participants agreed more strongly that they know what is safe and not safe to protect themselves from COVID-19, scoring a mean of 9.29 versus 8.38, and that messages about how to protect themselves and others are easy to understand, at 9.71 versus 8.81. In other words, the psychologically resilient group not only felt less personally impacted by the pandemic but also felt confident in its ability to act on protective information, while the low-resilience group, despite clearly fearing COVID-19’s impact, was less confident in its ability to adhere to preventative behaviors. The authors note that respondents in both clusters often fell on the same end of the agree-disagree scale, diverging mainly in magnitude, which suggests that messaging or interventions aimed at improving resiliency for everyone should yield improvement across both groups.

The findings also illuminate the connection between resilience and mental wellbeing during the pandemic. Those with lower resilience reported an increase in mental health problems, an association consistent with prior work showing that the experience of a high perceived threat can itself erode resilience. COVID-19 restrictions may have exacerbated these negative outcomes, the authors write, because government-enforced curfews and decreased capacity of community agencies reduced access to life-protecting harm reduction services, leading to increased drug use, feelings of isolation, and diminished resilience. Notably, the clusters did not differ on any non-HBM topic blocks, including knowledge of COVID-19, trust of information sources, coping, vaccine beliefs, and neighborhood and community needs, which the authors say highlights how important the Health Belief Model constructs and resiliency are to consider, and illustrates the stability of the two-cluster solution.

Practically, the results suggest concrete tools for harm reduction practitioners. The authors propose that easy screeners could be incorporated to assess overall resilience or beliefs in disease severity and susceptibility, enabling staff to quickly gauge where a client is in their thinking and tailor communication or intervention accordingly. Increasing perceived threat of COVID-19 so that people who use drugs can act upon it is one piece of the puzzle, they write, but increasing resiliency and perceived self-efficacy to use preventative behaviors is key. This may include targeting the susceptibility and severity constructs by ensuring messages are catered to the way this population does or does not believe it is susceptible to an infectious disease, then boosting self-efficacy through cues to action or increased access to protective supplies, alongside community-building techniques to create a resilient community at large.

The study carries important limitations. Because the data are cross-sectional, temporality cannot be inferred, and the reliance on self-reported data means social desirability bias should be considered. Results may not generalize to the broader population of people who use drugs, since the study was conducted only in Philadelphia among clients of a large social services agency that adapted its operations during the pandemic, likely leaving the sample feeling resourced and knowledgeable about staying safe, which may not hold for people without such access. The relatively small sample size limited the use of alternate statistical models and covariate controls, and the clustering variables overlapped conceptually with the perceived susceptibility construct. Still, the core message stands: resilience appears central to protecting this vulnerable population from future infectious disease outbreaks, and interventions aimed at increasing it may improve preventative behavior uptake and decrease disease burden when the next pandemic arrives.

Subject of Research: Health Belief Model constructs and resilience in COVID-19 prevention behavior among people who use drugs

Article Title: Using health belief model constructs to understand the role of perceived disease threat and resilience in responding to COVID-19 among people who use drugs: a cluster analysis

Article References: Paulus, K., Bass, S. B., Kelly, P. J. A., Pilla, J., Otor, A., Scialanca, M., Arroyo, A., & Faison, N. (2024). Using health belief model constructs to understand the role of perceived disease threat and resilience in responding to COVID-19 among people who use drugs: a cluster analysis. Advances in Drug and Alcohol Research, 4, Article 12197. https://doi.org/10.3389/adar.2024.12197

Image Credits: AI Generated

DOI: 10.3389/adar.2024.12197

Keywords: people who use drugs, Health Belief Model, COVID-19, resilience, cluster analysis, harm reduction, perceived susceptibility, self-efficacy, pandemic preparedness, mental health, vulnerable populations, infectious disease