OBJECTIVE: This paper presents a protocol of a bibliometric analysis aimed at offering the public insights into the current state and emerging trends in research related to the use of chatbot technology for promoting health.
METHODS: In this bibliometric analysis, we will select published papers from the databases of CINAHL, IEEE Xplore, PubMed, Scopus, and Web of Science that pertain to chatbot technology and its applications in health care. Our search strategy includes keywords such as "chatbot," "virtual agent," "virtual assistant," "conversational agent," "conversational AI," "interactive agent," "health," and "healthcare." Five researchers who are AI engineers and clinicians will independently review the titles and abstracts of selected papers to determine their eligibility for a full-text review. The corresponding author (ZN) will serve as a mediator to address any discrepancies and disputes among the 5 reviewers. Our analysis will encompass various publication patterns of chatbot research, including the number of annual publications, their geographic or institutional distribution, and the number of annual grants supporting chatbot research, and further summarize the methodologies used in the development of health-related chatbots, along with their features and applications in health care settings. Software tool VOSViewer (version 1.6.19; Leiden University) will be used to construct and visualize bibliometric networks.
RESULTS: The preparation for the bibliometric analysis began on December 3, 2021, when the research team started the process of familiarizing themselves with the software tools that may be used in this analysis, VOSViewer and CiteSpace, during which they consulted 3 librarians at the Yale University regarding search terms and tentative results. Tentative searches on the aforementioned databases yielded a total of 2340 papers. The official search phase started on July 27, 2023. Our goal is to complete the screening of papers and the analysis by February 15, 2024.
CONCLUSIONS: Artificial intelligence chatbots, such as ChatGPT (OpenAI Inc), have sparked numerous discussions within the health care industry regarding their impact on human health. Chatbot technology holds substantial promise for advancing health care systems worldwide. However, developing a sophisticated chatbot capable of precise interaction with health care consumers, delivering personalized care, and providing accurate health-related information and knowledge remain considerable challenges. This bibliometric analysis seeks to fill the knowledge gap in the existing literature on health-related chatbots, entailing their applications, the software used in their development, and their preferred functionalities among users.
INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): PRR1-10.2196/54349.
METHODS: This study prospectively evaluated mortality after prison release and whether methadone initiated before release increased survival after release in a sample of men with HIV and OUD (n = 291). We linked national death records to data from a controlled trial of prerelease methadone initiation conducted from 2010 to 2014 with men with HIV and OUD imprisoned in Malaysia. Vital statistics were collected through 2015. Allocation to prerelease methadone was by randomization (n = 64) and participant choice (n = 246). Cox proportional hazards models were used to estimate treatment effects of prerelease methadone on postrelease survival.
RESULTS: Overall, 62 deaths occurred over 872.5 person-years (PY) of postrelease follow-up, a crude mortality rate of 71.1 deaths per 1000 PY (95% confidence interval [CI] 54.5-89.4). Most deaths were of infectious etiology, mostly related to HIV. In a modified intention-to-treat analysis, the impact of prerelease methadone on postrelease mortality was consistent with a null effect in unadjusted (hazard ratio [HR] 1.3, 95% CI 0.6-3.1) and covariate-adjusted (HR 1.2, 95% CI 0.5-2.8) models. Predictors of mortality were educational level (HR 1.4, 95% CI 1.0-1.8), pre-incarceration alcohol use (HR 2.0, 95% CI 1.1-3.9), and lower CD4+ T-lymphocyte count (HR 0.8 per 100-cell/mL increase, 95% CI 0.7-1.0).
CONCLUSIONS: Postrelease mortality in this sample of men with HIV and OUD was extraordinarily high, and most deaths were likely of infectious etiology. No effect of prerelease methadone on postrelease mortality was observed, which may be due to study limitations or an epidemiological context in which inadequately treated HIV, and not inadequately treated OUD, is the main cause of death after prison release.
TRIAL REGISTRATION: NCT02396979. Retrospectively registered 24/03/2015.
METHODS: From 2010 to 2014, men with HIV (N = 212) and opioid dependence before incarceration were enrolled in MMT within 6 months of release from Malaysia's largest prison and followed for 12-months post-release. As a prospective trial, allocation to MMT was at random and later by preference design (predictive nonetheless). MMT dosing was individually targeted to minimally achieve 80 mg/day. Time-to-event analyses were conducted to model linkage to MMT after release.
FINDINGS: Of the 212 participants allocated to MMT, 98 (46 %) were prescribed higher dosages (≥80 mg/day) before release. Linkage to MMT after release occurred in 77 (36 %) participants and significantly higher for those prescribed higher dosages (46% vs 28 %; p = 0.011). Factors associated with higher MMT dosages were being married, on antiretroviral therapy, longer incarceration periods, having higher levels of depression, and methadone preference compared to randomization. After controlling for other variables, being prescribed higher methadone dosage (aHR: 2.53, 95 %CI: 1.42-4.49) was the only independent predictor of linkage to methadone after release.
INTERPRETATION: Higher doses of methadone prescribed before release increased the likelihood of linkage to MMT after release. Methadone dosing should be introduced into international guidelines for treatment of opioid use disorder in prisons and further post-release benefits should be explored.
FUNDING: National Institute of Drug Abuse (NIDA).
METHODS: Five graph models were fit using data from 1574 people who inject drugs in Hartford, CT, USA. We used a degree-corrected stochastic block model, based on goodness-of-fit, to model networks of injection drug users. We simulated transmission of HCV and HIV through this network with varying levels of HCV treatment coverage (0%, 3%, 6%, 12%, or 24%) and varying baseline HCV prevalence in people who inject drugs (30%, 60%, 75%, or 85%). We compared the effectiveness of seven treatment-as-prevention strategies on reducing HCV prevalence over 10 years and 20 years versus no treatment. The strategies consisted of treatment assigned to either a randomly chosen individual who injects drugs or to an individual with the highest number of injection partners. Additional strategies explored the effects of treating either none, half, or all of the injection partners of the selected individual, as well as a strategy based on respondent-driven recruitment into treatment.
FINDINGS: Our model estimates show that at the highest baseline HCV prevalence in people who inject drugs (85%), expansion of treatment coverage does not substantially reduce HCV prevalence for any treatment-as-prevention strategy. However, when baseline HCV prevalence is 60% or lower, treating more than 120 (12%) individuals per 1000 people who inject drugs per year would probably eliminate HCV within 10 years. On average, assigning treatment randomly to individuals who inject drugs is better than targeting individuals with the most injection partners. Treatment-as-prevention strategies that treat additional network members are among the best performing strategies and can enhance less effective strategies that target the degree (ie, the highest number of injection partners) within the network.
INTERPRETATION: Successful HCV treatment as prevention should incorporate the baseline HCV prevalence and will achieve the greatest benefit when coverage is sufficiently expanded.
FUNDING: National Institute on Drug Abuse.
METHODS: Opioid dependent PWID were interviewed and tested for HIV and HCV in five Ukrainian cities from January 2014 to March 2015. Logistic regression was used to examine the independent correlates of two cascade steps: a) anti-HCV positive status awareness; b) chronic HCV confirmation; and of c) annual HCV testing for PWID.
RESULTS: Among 1613 PWID, 1002 (62.1%) had anti-HCV positive test result, of which 568 (56.7%) were aware of it before the study and 346 (34.5%) reported previous confirmatory testing for chronic HCV. Independent correlates of being aware they had anti-HCV positivity included: current [AOR: 3.08; 95%CI: 2.16-4.40] or prior [AOR: 1.85; 95%CI: 1.27-2.68] opioid agonistic treatment (OAT) experience, relative to no prior OAT, living in Lviv [AOR: 0.50; 95%CI: 0.31-0.81] or Odesa [AOR: 2.73; 95%CI: 1.51-4.93] relative to Kyiv and being aware of having HIV [AOR: 4.10; 95%CI: 2.99-5.62]. Independent correlates of confirming HCV infection among those who were aware of their anti-HCV positive status included: current OAT [AOR: 2.00; 95%CI: 1.24-3.23], relative to prior OAT, the middle income category [AOR: 1.74, 95%CI: 1.15-2.63], relative to the lowest, and receiving ART [AOR: 4.54; 95%CI: 2.85-7.23]. Among 1613 PWID, 918 (56.9%) were either HCV negative or not aware of their HCV positive status, of which 198 (21.6%) reported recent anti-HCV test (during last 12 month). Recent anti-HCV test in this group was associated with current [AOR: 7.17; 95%CI: 4.63-11.13] or prior [AOR: 2.24; 95%CI: 1.32-3.81] OAT experience, relative to no prior OAT.
CONCLUSION: Encouraging PWID to participate in OAT may be an effective strategy to diagnose and link PWID who are HCV positive to care. Among HIV negative participants, regular HCV testing may be ensured by participation in OAT. More studies are needed to assess HCV treatment utilization among PWID in Ukraine and OAT as a possible way to retain them in treatment.
OBJECTIVE: The aim of this study was to identify the barriers to and facilitators of Malaysian MSM's acceptance of an AI chatbot designed to assist in HIV testing and prevention in relation to its perceived benefits, limitations, and preferred features among potential users.
METHODS: We conducted 5 structured web-based focus group interviews with 31 MSM in Malaysia between July 2021 and September 2021. The interviews were first recorded, transcribed, coded, and thematically analyzed using NVivo (version 9; QSR International). Subsequently, the unified theory of acceptance and use of technology was used to guide data analysis to map emerging themes related to the barriers to and facilitators of chatbot acceptance onto its 4 domains: performance expectancy, effort expectancy, facilitating conditions, and social influence.
RESULTS: Multiple barriers and facilitators influencing MSM's acceptance of an AI chatbot were identified for each domain. Performance expectancy (ie, the perceived usefulness of the AI chatbot) was influenced by MSM's concerns about the AI chatbot's ability to deliver accurate information, its effectiveness in information dissemination and problem-solving, and its ability to provide emotional support and raise health awareness. Convenience, cost, and technical errors influenced the AI chatbot's effort expectancy (ie, the perceived ease of use). Efficient linkage to health care professionals and HIV self-testing was reported as a facilitating condition of MSM's receptiveness to using an AI chatbot to access HIV testing. Participants stated that social influence (ie, sociopolitical climate) factors influencing the acceptance of mobile technology that addressed HIV in Malaysia included privacy concerns, pervasive stigma against homosexuality, and the criminalization of same-sex sexual behaviors. Key design strategies that could enhance MSM's acceptance of an HIV prevention AI chatbot included an anonymous user setting; embedding the chatbot in MSM-friendly web-based platforms; and providing user-guiding questions and options related to HIV testing, prevention, and treatment.
CONCLUSIONS: This study provides important insights into key features and potential implementation strategies central to designing an AI chatbot as a culturally sensitive digital health tool to prevent stigmatized health conditions in vulnerable and systematically marginalized populations. Such features not only are crucial to designing effective user-centered and culturally situated mobile health interventions for MSM in Malaysia but also illuminate the importance of incorporating social stigma considerations into health technology implementation strategies.
METHODS: A convenience sample of 96 HIV-positive and 104 HIV-negative incarcerated men who met pre-incarceration criteria for opioid dependence was interviewed using a structured questionnaire to examine participant characteristics and attitudes toward MMT. Factors associated with interest in prison-based MMT initiation were identified using logistic regression analysis.
RESULTS: Among all participants, 85 (42.5%) were interested in receiving MMT within prison. Independent correlates of interest in prison-based MMT were being previously married (AOR=4.15, 95% CI: 1.15, 15.02), previously incarcerated (AOR=5.68, 95% CI: 1.54, 21.02), depression (AOR=3.66, 95% CI: 1.68, 7.98), daily heroin use in the 30days prior to incarceration (AOR=5.53, 95% CI: 1.65, 18.58), and more favorable attitudes toward MMT (AOR=19.82, 95% CI: 6.07, 64.74).
CONCLUSIONS: Overall, interest in receiving prison-based MMT was low, and was associated with adverse social, mental health, and drug use consequences. Incarceration provides a unique opportunity to initiate MMT for those who need it, however, optimal scale-up efforts must be systemic and address modifiable factors like improving attitudes toward and motivation for MMT. Informed or shared decision-making tools may be useful in improving expectations and acceptability of MMT.
METHODS: A total of 436 physicians at two major university medical centers in Kuala Lumpur, Malaysia, completed an online survey. Sociodemographic characteristics, stigma-related constructs, and intentions to discriminate against transgender people were measured. Bivariate and multivariate linear regression were used to evaluate independent covariates of discrimination intent.
RESULTS: Medical doctors who felt more fearful of transgender people and more personal shame associated with transgender people expressed greater intention to discriminate against transgender people, whereas doctors who endorsed the belief that transgender people deserve good care reported lower discrimination intent. Stigma-related constructs accounted for 42% of the variance and 8% was accounted for by sociodemographic characteristics.
CONCLUSIONS: Constructs associated with transgender stigma play an important role in medical doctors' intentions to discriminate against transgender patients. Development of interventions to improve medical doctors' knowledge about and attitudes toward transgender people are necessary to reduce discriminatory intent in healthcare settings.
METHODS: Between June 2015 and August 2016, 50 HIV-positive TGW were recruited in Lima, Peru. Multivariable logistic regression was used to identify factors associated with viral suppression (<200 copies/mL) among the TGW.
RESULTS: Among TGW, 85% achieved viral suppression. Approximately half (54%) reported anal sex with more than five partners in the past 6 months, 38% reported sex work, 68% had not disclosed their HIV status to one or more of their partners, and 38% reported condomless sex with their last partner. The prevalence of alcohol use disorders was high (54%), and 38% reported use of drugs in the past year. Moderate-to-severe drug use significantly reduced odds of achieving viral suppression (adjusted odds ratio 0.69; 95% confidence interval: 0.48-0.98).
CONCLUSION: Our findings highlight the need for integrated treatment for substance disorders in HIV care to increase the viral suppression rate among TGW in Lima, Peru.
METHODS: A randomized, double-blind, placebo-controlled trial was conducted among 100 HIV+ prisoners with AUDs. Participants were randomized 2:1 to receive 6 monthly injections of XR-NTX or placebo starting one week prior to release. Using multiple imputation strategies for data missing completely at random, data were analyzed for the 6-month post-incarceration period. Main outcomes included: time to first heavy drinking day; number of standardized drinks/drinking day; percent of heavy drinking days; pre- to post-incarceration change in average drinks/day; total number of drinking days; and a composite alcohol improvement score comprised of all 5 parameters.
RESULTS: There was no statistically significant difference overall between treatment arms for time-to-heavy-drinking day. However, participants aged 20-29 years who received XR-NTX had a longer time to first heavy drinking day compared to the placebo group (24.1 vs. 9.5days; p<0.001). There were no statistically significant differences between groups for other individual drinking outcomes. A sub-analysis, however, found participants who received ≥4 XR-NTX were more likely (p<0.005) to have improved composite alcohol scores than the placebo group. Post-hoc power analysis revealed that despite the study being powered for HIV outcomes, sufficient power (0.94) was available to distinguish the observed differences.
CONCLUSIONS: Among CJS-involved PLH with AUDs transitioning to the community, XR-NTX lengthens the time to heavy drinking day for younger persons; reduces alcohol consumption when using a composite alcohol consumption score; and is not associated with any serious adverse events.
DESIGN: A randomized, double-blind, placebo-controlled trial was conducted among incarcerated individuals with HIV and AUDs transitioning to the community from 2010 through 2016.
METHODS: Eligible participants (N = 100) were randomized 2:1 to receive 6 monthly injections of XR-NTX (n = 67) or placebo (n = 33) starting at release and continued for 6 months. The primary and secondary outcomes were the proportion that maintained or improved VS at <200 and <50 copies per milliliter from baseline to 6 months, respectively, using an intention-to-treat analysis.
RESULTS: Participants allocated to XR-NTX improved VS from baseline to 6 months for <200 copies per milliliter (48.0%-64.2%, P = 0.024) and for <50 copies per milliliter (31.0%-56.7%, P = 0.001), whereas the placebo group did not (<200 copies/mL: 64%-42.4%, P = 0.070; <50 copies/mL: 42.0%-30.3%, P = 0.292). XR-NTX participants were more likely to achieve VS than the placebo group at 6 months (<200 copies/mL: 64.2% vs. 42.4%; P = 0.041; <50 copies/mL: 56.7% vs. 30.3%; P = 0.015). XR-NTX independently predicted VS [<200 copies/mL: adjusted odds ratio (aOR) = 2.68, 95% confidence interval (CI) = 1.01 to 7.09, P = 0.047; <50 copies/mL: aOR = 4.54; 95% CI = 1.43 to 14.43, P = 0.009] as did receipt of ≥3 injections (<200 copies/mL: aOR = 3.26; 95% CI = 1.26 to 8.47, P = 0.010; <50 copies/mL: aOR = 6.34; 95% CI = 2.08 to 19.29, P = 0.001). Reductions in alcohol consumption (aOR = 1.43, 95% CI = 1.03 to 1.98, P = 0.033) and white race (aOR = 5.37, 95% CI = 1.08 to 27.72, P = 0.040) also predicted VS at <50 copies per milliliter.
CONCLUSIONS: XR-NTX improves or maintains VS after release to the community for incarcerated people living with HIV and AUDs.