Faculty Bibliography

UNLABELLED: HIGHLIGHTS/UNASSIGNED:Diminishing marginal lifespan utility (DMLU) means that the value of extending lifespan may differ based on the duration of life already lived.DMLU is not typically considered in cost-effectiveness analyses.Not considering DMLU may bias cost-effectiveness analyses against equity.Not considering DMLU may reduce the accuracy of distributive cost-effectiveness analyses and other approaches to consider equity along with efficiency.

Important barriers to the use of QALYs in the United States include concerns about disability and age discrimination.Modifications to the utility function underlying QALYs have been proposed to mitigate these concerns, but some find them challenging to consider and/or to apply.Unrelated to these concerns, QALYs have been adapted within the framework of distributional cost-effectiveness analysis to allow consideration of inequality as well as efficiency.I outline how this framework can also remediate concerns about disability and age discrimination.

BACKGROUND:Alcohol use disorder (AUD) and major depressive disorder (MDD) drive HIV transmission in many sub-Saharan African settings. The impact of screening and treating AUD and MDD on HIV outcomes is unknown. We aimed to identify the cost-effectiveness of AUD and MDD interventions in Zimbabwe, and their potential contribution to reaching Zimbabwe's Ending the HIV Epidemic 2030 goal. METHODS:Using a validated HIV compartmental transmission model in Zimbabwe, we compared four policy scenarios: prevention as usual (baseline); implement AUD screening (using AUDIT) and treatment (motivational interviewing and cognitive-behavioral therapy); implement MDD screening (using PHQ-9) and treatment (cognitive-behavioral therapy); and implement screening and treatment for both. Outcomes were HIV incidence projections, infections averted through 2030, quality-adjusted life-years gained, cost per infection averted, and cost per QALY gained. Analyses considered "spillover," when treatment for AUD also results in an improvement in MDD and the converse. Sensitivity analyses identified cost reductions necessary for AUD and MDD interventions to be as cost-effective as other HIV interventions, particularly the scale-up of long-acting PrEP. RESULTS:AUD and MDD combined will be responsible for 21.1% of new HIV infections in Zimbabwe by 2030. Without considering spillover, compared to the baseline, MDD intervention can reduce new infections by 5.4% at $2039/infection averted and $3186/QALY. AUD intervention can reduce new infections by 5.8%, but at $2,968/infection averted and $4753/QALY, compared to baseline. Both MDD and AUD interventions can reduce new infections by 11.1% at $2810/infection averted and $4229/QALY, compared to baseline. Considering spillover, compared to the baseline, MDD intervention can reduce new infections by 6.4% at $1714/infection averted and $2630/QALY. AUD intervention can reduce new infections by 7.4%, but at $2299/infection averted and $3560/QALY compared to baseline. Both MDD and AUD interventions can reduce new infections by 11.9% at $2247/infection averted and $3382/QALY compared to baseline. For MDD intervention to match the cost-effectiveness of scaling long-acting PrEP, the cost of MDD intervention would need to be reduced from $16.64 to $12.88 per person. CONCLUSIONS:Implementing AUD and MDD interventions can play an important role in HIV reduction in Zimbabwe, particularly if intervention cost can be decreased while preserving effectiveness.

In mid-2022, New York City (NYC) became the epicenter of the US mpox outbreak. We provided real-time mpox case forecasts to the NYC Department of Health and Mental Hygiene to aid in outbreak response. Forecasting methodologies evolved as the epidemic progressed. Initially, lacking knowledge of at-risk population size, we used exponential growth models to forecast cases. Once exponential growth slowed, we used a Susceptible-Exposed-Infectious-Recovered (SEIR) model. Retrospectively, we explored if forecasts could have been improved using an SEIR model in place of our early exponential growth model, with or without knowing the case detection rate. Early forecasts from exponential growth models performed poorly, as 2-week mean absolute error (MAE) grew from 53 cases/week (July 1-14) to 457 cases/week (July 15-28). However, when exponential growth slowed, providing insight into susceptible population size, an SEIR model was able to accurately predict the remainder of the outbreak (7-week MAE: 13.4 cases/week). Retrospectively, we found there was not enough known about the epidemiological characteristics of the outbreak to parameterize an SEIR model early on. However, if the at-risk population and case detection rate were known, an SEIR model could have improved accuracy over exponential growth models early in the outbreak.

We used results from an optimization randomized controlled trial which tested five behavioral intervention components to support HIV antiretroviral adherence/HIV viral suppression, grounded in the multiphase optimization strategy and using a fractional factorial design to identify intervention components with cost-effectiveness sufficiently favorable for scalability. Results were incorporated into a validated HIV computer simulation to simulate longer-term effects of combinations of components on health and costs. We simulated the 32 corresponding long-term trajectories for viral load suppression, health related quality of life (HRQoL), and costs. The components were designed to be culturally and structurally salient. They were: motivational interviewing counseling sessions (MI), pre-adherence skill building (SB), peer mentorship (PM), focused support groups (SG), and patient navigation (short version [NS], long version [NL]. All participants also received health education on HIV treatment. We examined four scenarios: one-time intervention with and without discounting and continuous interventions with and without discounting. In all four scenarios, interventions that comprise or include SB and NL (and including health education) were cost effective (< $100,000/quality-adjusted life year). Further, with consideration of HRQoL impact, maximal intervention became cost-effective enough to be scalable. Thus, a fractional factorial experiment coupled with cost-effectiveness analysis is a promising approach to optimize multi-component interventions for scalability. The present study can guide service planning efforts for HIV care settings and health departments.

OBJECTIVE:We aimed to better understand patients' treatment preferences and quantify the level of cancer risk at which treatment preferences change (risk threshold) to inform better counseling of patients with intraductal papillary mucinous neoplasms (IPMNs). SUMMARY BACKGROUND DATA/BACKGROUND:The complexity of IPMN management provides an opportunity to align treatment with individual preference. METHODS:We surveyed a sample of healthy volunteers simulating a common scenario: undergoing an imaging study that incidentally identifies an IPMN. In the scenario, the estimated risk of cancer in the IPMN was 5%. Patients were asked their treatment preference (surgery or surveillance), to quantify the level of cancer risk in the IPMN at which their treatment preference would change (i.e. risk threshold), and their level of cancer anxiety as measured on a 5-point Likert scale. We examined associations between participant characteristics, treatment preferences, and risk threshold using multivariable linear regression. RESULTS:The median risk threshold among the 520 participants was 25% (IQR 2.3-50%). The risk threshold had a bimodal distribution: 40% of participants had a risk threshold between 0-10% and 47% had a risk threshold above 30%. When informed that the risk of cancer was 5%, 62% of participants (n=323) preferred surveillance, and the remaining 38% (n=197) preferred surgery. After adjusting for potential confounders, participants who expressed "worry" or "extreme worry" about the malignancy risk of IPMN had significantly lower risk thresholds than participants who were "not at all worried" (Coefficient -12, 95%CI -21 to -2, P=0.015 and Coefficient -18, 95%CI -29 to -8, P<0.001, respectively). CONCLUSIONS:Participants varied in treatment preference and risk threshold of incidentally identified IPMNs. Given the uncertainty in estimating the true malignant potential of IPMNs, a better understanding of a patient's risk threshold, as influenced by patient concern about malignancy, will help inform the shared decision-making process.

BACKGROUND:People in Connecticut are now more likely to die of a drug-related overdose than a traffic accident. While Connecticut has had some success in slowing the rise in overdose death rates, substantial additional progress is necessary. METHODS:We developed, verified, and calibrated a mechanistic simulation of alternative overdose prevention policy options, including scaling up naloxone (NLX) distribution in the community and medications for opioid use disorder (OUD) among people who are incarcerated (MOUD-INC) and in the community (MOUD-COM) in a simulated cohort of people with OUD in Connecticut. We estimated how maximally scaling up each option individually and in combinations would impact 5-year overdose deaths, life-years, and quality-adjusted life-years. All costs were assessed in 2021 USD, employing a health sector perspective in base-case analyses and a societal perspective in sensitivity analyses, using a 3% discount rate and 5-year and lifetime time horizons. RESULTS:Maximally scaling NLX alone reduces overdose deaths 20% in the next 5 years at a favorable incremental cost-effectiveness ratio (ICER); if injectable rather than intranasal NLX was distributed, 240 additional overdose deaths could be prevented. Maximally scaling MOUD-COM and MOUD-INC alone reduce overdose deaths by 14% and 6% respectively at favorable ICERS. Considering all permutations of scaling up policies, scaling NLX and MOUD-COM together is the cost-effective choice, reducing overdose deaths 32% at ICER $19,000/QALY. In sensitivity analyses using a societal perspective, all policy options were cost saving and overdose deaths reduced 33% over 5 years while saving society $338,000 per capita over the simulated cohort lifetime. CONCLUSIONS:Maximally scaling access to naloxone and MOUD in the community can reduce 5-year overdose deaths by 32% among people with OUD in Connecticut under realistic budget scenarios. If societal cost savings due to increased productivity and reduced crime costs are considered, one-third of overdose deaths can be reduced by maximally scaling all three policy options, while saving money.

HIV infections disproportionately impact Latinx populations in the United States, yet oral pre-exposure prophylaxis (PrEP) uptake is low. This study was a secondary gendered analysis of interviews with Latina cisgender women (n = 20) recruited from an urban safety net hospital inNew York City between August 2019 and October 2022. All women were indicated for PrEP by the provider. In-depth interviews were conducted with participants in English and Spanish and asked about social determinants of health, sexual partnerships and behaviors, and PrEP-specific enablers and barriers. Secondary thematic content analysis was conducted to identify gender-related factors influencing PrEP uptake. The following themes emerged from the data:structural factors (e.g., employment), partner-related factors, low sexual health knowledge, and resilience and empowerment. Partner-related factors were the most salient; partner infidelity served as reasons for initiating PrEP. Despite being constrained by low power in relationships, women made empowered choices to initiate PrEP and protect themselves. Findings indicated that the impact of gender inequity was an important factor in Latina women's PrEP decision making, pointing to a need to address partner-driven HIV risk, imbalance of power in relationships, and gender norms.

Background Given the disproportionate rates of incarceration and lower life expectancy (LE) among Black sexual minority men (BSMM) and Black transgender women (BTW) with HIV, we modeled the impact of decarceration and screening for psychiatric conditions and substance use on LE of US BSMM/BTW with HIV. Methods We augmented a microsimulation model previously validated to predict LE and leading causes of death in the US with estimates from the HPTN 061 cohort and the Veteran's Aging Cohort Studies. We estimated independent associations among psychiatric and substance use disorders, to simulate the influence of treatment of one condition on improvement on others. We used this augmented simulation to estimate LE for BSMM/BTW with HIV with a history of incarceration under alternative policies of decarceration (i.e., reducing the fraction exposed to incarceration), screening for psychiatric conditions and substance use, or both. Results Baseline LE was 61.3 years. Reducing incarceration by 25%, 33%, 50%, and 100% increased LE by 0.29 years, 0.31 years, 0.53 years, and 1.08 years, respectively, versus no reductions in incarceration. When reducing incarceration by 33% and implementing screening for alcohol, tobacco, substance use, and depression, in which a positive screen triggers diagnostic assessment for all psychiatric and substance use conditions and linkage to treatment, LE increased by 1.52 years compared to no screening or decarceration. Discussion LE among BSMM/BTW with HIV is short compared with other people with HIV. Reducing incarceration and improving screening and treatment of psychiatric conditions and substance use could substantially increase LE in this population.