Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (46 results)

  • Thesis
  • Peer Reviewed
UC Berkeley Electronic Theses and Dissertations (2020)

Understanding of the health effects of specific interventions is a key objective of epidemiology. Policies, whether implemented by governments or other institutions, have immense potential to influence population health and thus represent important study exposures for epidemiologists. However, because policies are often implemented without specific consideration of or plans for evaluation, there are two key challenges to any attempt to evaluate their effects: data availability and the methodological challenges associated with estimating causal effects using observational data. Fortunately, both the number and diversity of data sources that are available to health and other social science researchers are vast and constantly growing. In addition, there is a rich history of methodological advancement across multiple disciplines (most notably economics, statistics, political science, epidemiology, and biostatistics) pertaining to the evaluation of intervention effects in observational settings. The availability of large, novel data sources and a diverse set of methodological tools, combined with recent calls for a more “consequentialist” epidemiology that focuses as much on the effects of interventions as it does on understanding etiology, makes it an exciting time to be an epidemiologist. As such, the studies that constitute this dissertation aim to evaluate the health effects of three distinct sets of policies using diverse administrative and observational data sources and methods developed across multiple disciplines.

The first study evaluates the impacts of two national policies that aimed to expand access to buprenorphine for the treatment of opioid use disorder. The United States continues to grapple with an unprecedented opioid overdose epidemic and expanding access to medications for the treatment of opioid use disorder (MOUD), such as buprenorphine, is a key priority in the national response to the epidemic. This study uses comprehensive administrative data sources from California to evaluate the effects of 2016 policies expanding the buprenorphine waiver program on buprenorphine prescribing and opioid-related health outcomes.

The second study evaluates the impact of two clinic-level policies that aimed to reduce opioid prescribing at a safety-net clinic in 2013 and 2014 in San Francisco, CA. In addition to expanding access to MOUD, a concurrent strategy for addressing the opioid crisis has been to reduce unnecessary prescribing of opioid medications. However, there is ongoing concern that inflexible one-size-fits-all reductions in opioid prescribing could ultimately harm patients, and even result in transitions from prescription opioid use and dependence to heroin use and dependence. In 2013 and 2014, the clinic under study established standard protocols for dispensing opioid refills and conducting urine toxicology testing and a new committee facilitating opioid treatment decisions for complex patient cases. Using data from a retrospective cohort study of patients on long-term opioid therapy, this study evaluates the effects of these policies on opioid prescribing and use of heroin and opioid analgesics not prescribed to the patient.

The third study evaluates the impact of a major California criminal justice policy on mortality among state prisoners. In October 2011, California Assembly Bill 109 (AB109) sought to decrease the prison population by prospectively shifting the custodial responsibility of non-violent, non-serious, and non-sexual offenders from state prisons to county jails and probation departments. Leveraging several complementary data sources, this study evaluates the effects of AB109 on California state prisoner mortality.

Collectively, these studies contribute new evidence regarding the effects of these three sets of policies. Furthermore, two of the studies demonstrate the utility of administrative data sources for assessing the impacts of real-world interventions. As these data sources expand and become more organized, rigorous analyses that leverage their strengths and acknowledge their limitations will grow increasingly valuable as a tool for knowledge creation among epidemiologists and other applied researchers.

    • Article
    • Peer Reviewed
    UC Berkeley Previously Published Works (2023)
    In 2011, a historic Supreme Court decision mandated that the state of California substantially reduce its prison population to alleviate overcrowding, which was deemed so severe as to preclude the provision of adequate healthcare. To comply, California passed the Public Safety Realignment Act (Assembly Bill [AB] 109), representing the largest ever court-ordered reduction of a prison population in U.S. history. AB109 was successful in reducing the state prison population; however, although the policy was precipitated by inadequate healthcare in state prisons, no studies have examined its effects on prisoner health. As other states grapple with overcrowded prisons and look to Californias experience with this landmark policy, understanding how it may have impacted prisoner health is critical. We sought to evaluate the effects of AB109 on prison mortality and assess the extent to which policy-induced changes in the age distribution of prisoners may have contributed to these effects. To do so, we used prison mortality data from the Bureau of Justice Statistics and the California Deaths in Custody reporting program and prison population data from the National Corrections Reporting Program to examine changes in overall prison mortality, the age distribution of prisoners, and age-adjusted prison mortality in California relative to other states before and after the implementation of AB109. Following AB109, California prisons experienced an increase in overall mortality relative to other states that attenuated within three years. Over the same period, California experienced a greater upward shift in the age distribution of its prisoners relative to other states, suggesting that the states increase in overall mortality may have been driven by this change in age distribution. Indeed, when accounting for this differential change in age distribution, mortality among California prisoners exhibited a greater reduction relative to other states in the third year after implementation. As other states seek to reduce their prison populations to address overcrowding, assessments of Californias experience with AB109 should consider this potential improvement in age-adjusted mortality.
      Cover page: Californias Public Safety Realignment Act and prisoner mortality.
      • Article
      • Peer Reviewed
      UC San Francisco Previously Published Works (2020)
      Despite elevated mortality in people with HIV (PWH) using drugs, drug-related deaths are poorly characterized. Among 6764 drug-related deaths, methamphetamine was more common in PWH than others. One in 4 deaths in PWH involved acute infection. Combatting mortality in PWH who use drugs should include stimulant-specific and infection prevention efforts.
        Cover page: Beyond Overdose: Drug-Related Deaths in People With and Without HIV in San Francisco, 2007–2018
        • Article
        • Peer Reviewed
        UC San Francisco Previously Published Works (2019)
        BACKGROUND: Single room occupancy (SRO) buildings, also known as residential hotels, are a form of affordable housing common to cities in North America, and residents of these buildings face elevated rates of substance use, physical and mental multimorbidity, and mortality. Identifying distinct populations at greater risk of overdose death is crucial to the planning of interventions aiming to reduce drug-related mortality, yet no studies have assessed the population burden of overdose mortality among SRO residents. The present study quantifies and characterizes drug overdose mortality among residents of SRO buildings in a large U.S. city. METHODS: We used mortality records and a database of SRO buildings to calculate rate ratios comparing overdose mortality due to opioids, cocaine, and methamphetamine among SRO residents and non-SRO residents in San Francisco, CA 2010-2017 and assessed bivariate differences in decedent and death location characteristics between SRO resident and other overdose decedents. RESULTS: There were 1,551 overdose deaths during the study period, with an overall rate of 21.3 per 100,000 residents (95%CI = 20.2-22.6). The rate among SRO residents (278.7, 95%CI = 252.9-306.5) was 19.3 (95%CI = 17.1-21.7) times that of non-SRO residents (21.3, 95%CI = 20.2-22.6). An additional 79 (5%) deaths among non-residents occurred in SRO buildings, and 86% of SRO resident decedents died at home compared to 64% of non-SRO residents (p < 0.05). CONCLUSIONS: Overdose mortality was substantially higher among SRO residents, who were also more likely to die from overdose at home, which highlights the need for resources and targeted interventions directed towards residents of SRO buildings.
          Cover page: Drug overdose mortality among residents of single room occupancy buildings in San Francisco, California, 2010-2017.
          • Article
          • Peer Reviewed
          UC Berkeley Previously Published Works (2022)
          AIMS: To evaluate associations between new types of buprenorphine waivers (nurse practitioner and physician assistant [NP/PA]; 275-patient limit [MD/DO-275]) and both buprenorphine prescribing and health outcomes. METHODS: Using comprehensive county-level data from California 2010-2018, we modeled quarterly associations between numbers of NP/PA and MD/DO-275 waivers and rates of buprenorphine prescribing, opioid-related deaths, emergency department (ED) visits, and hospitalizations among all counties and separately among metropolitan and nonmetropolitan counties using Poisson regression models with county and quarter fixed effects and adjusting for time-varying covariates. RESULTS: Each additional NP/PA and MD/DO-275 waiver was associated with a 2.6% (95%CI: 1.1-4.1%) and 5.8% (4.1-7.4%) increase in buprenorphine prescribing among nonmetropolitan counties, respectively. Each additional MD/DO-275 waiver was associated with a 2.8% (1.0%-4.6%) increase in buprenorphine among metropolitan counties. There were no statistically significant associations between NP/PA waivers and buprenorphine prescribing among metropolitan counties or among either waiver type and opioid-related health outcomes. CONCLUSIONS: NP/PA waivers were associated with increased buprenorphine prescribing among nonmetropolitan counties and MD/DO-275 waivers were associated with increased buprenorphine prescribing among both metropolitan and nonmetropolitan counties.
            Cover page: Evaluating buprenorphine prescribing and opioid-related health outcomes following the expansion the buprenorphine waiver program.
            • Article
            • Peer Reviewed
            UC San Francisco Previously Published Works (2018)

            Background

            Self-reported data are widely used in substance-use research, yet few studies have assessed the validity of self-reported methamphetamine use compared to biological assays.

            Objectives

            We sought to assess the validity and correlates of validity of self-reported methamphetamine use compared to urine toxicology (UTOX).

            Methods

            Using a sample of methamphetamine-dependent individuals enrolled in a randomized controlled pharmacotherapy trial in the United States (n = 327 visits among 90 participants), we calculated sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the kappa coefficient of self-reported methamphetamine use in the past 3 days compared to UTOX, as well as the NPV of self-reported methamphetamine use over an extended recall period of 1 month. We used multivariable logistic regression models to assess correlates of concordance between self-reported methamphetamine use and UTOX.

            Results

            The sensitivity of self-reported methamphetamine use in the past 3 days was 86.7% (95% confidence intervals (95%CI): 81.4%-91.4%), the specificity was 85.3% (77.7-91.3), the PPV was 91.5% (86.9-94.8), and the NPV was 78.0% (69.4-86.1), compared to UTOX (kappa = 0.71). The NPV over the extended recall period was 70.6% (48.0-85.7). In multivariable analyses, validity of self-reported methamphetamine use was higher for older participants but lower during follow-up compared to baseline and when polysubstance use or depressive symptoms were reported. Conclusions/Importance: Our sample of methamphetamine-dependent adults reported recent methamphetamine use with high validity compared to UTOX. Validity increased with age but decreased when participants reported depressive symptoms or polysubstance use as well as later in the study timeline and during longer recall periods.
              Cover page: Correlates of Validity of Self-Reported Methamphetamine Use among a Sample of Dependent Adults
              • Article
              • Peer Reviewed
              UC San Francisco Previously Published Works (2016)
              Men who have sex with men (MSM) are the demographic group most severely affected by HIV in the USA. Global association studies have shown that MSM who binge drink are more likely to engage in risky sexual behaviors and day- and event-level analyses have linked binge drinking to sexual risk behavior on specific days and during specific sexual encounters. Despite this strong foundation of research, no studies have examined the association between the frequency of situational binge drinking (i.e., binge drinking concurrent with sexual activity) and aggregated sexual risk over periods of longer duration. We used multivariable logistic regression to assess the relationship between situational binge drinking (i.e., binge drinking concurrent with anal intercourse) and condomless anal intercourse (CAI) and among a cross-sectional sample of 124 MSM in San Francisco, CA. There was a positive relationship between frequency of situational binge drinking and CAI (1-5 times vs. never: adjusted odds ratio = 2.78, 95% CI = 1.01-7.63; 6-10 times vs. never: 6.19, 1.27-30.22; more than 10 times vs. never: 11.88, 1.31-107.60). By filling a methodological gap and complementing existing global and event-level analyses, this positive situational relationship strengthens the evidence linking binge drinking and sexual risk, enhances the comparability of the existing literature, and further suggests that the integration of dual strategies that aim to prevent HIV and reduce binge drinking may be warranted.
                Cover page: Binge drinking concurrent with anal intercourse and condom use among men who have sex with men.
                • Article
                • Peer Reviewed
                UC Berkeley Previously Published Works (2022)

                Purpose

                Population-based surveys are possible sources from which to draw representative control data for case-control studies. However, these surveys involve complex sampling that could lead to biased estimates of measures of association if not properly accounted for in analyses. Approaches to incorporating complex-sampled controls in density-sampled case-control designs have not been examined.

                Methods

                We used a simulation study to evaluate the performance of different approaches to estimating incidence density ratios (IDR) from case-control studies with controls drawn from complex survey data using risk-set sampling. In simulated population data, we applied four survey sampling approaches, with varying survey sizes, and assessed the performance of four analysis methods for incorporating survey-based controls.

                Results

                Estimates of the IDR were unbiased for methods that conducted risk-set sampling with probability of selection proportional to survey weights. Estimates of the IDR were biased when sampling weights were not incorporated, or only included in regression modeling. The unbiased analysis methods performed comparably and produced estimates with variance comparable to biased methods. Variance increased and confidence interval coverage decreased as survey size decreased.

                Conclusions

                Unbiased estimates are obtainable in risk-set sampled case-control studies using controls drawn from complex survey data when weights are properly incorporated.
                  Cover page: Conducting density-sampled case-control studies using survey data with complex sampling designs: A simulation study
                  • Article
                  • Peer Reviewed
                  UC San Francisco Previously Published Works (2022)
                  INTRODUCTION: People with opioid use disorder (OUD) have high rates of discharge against medical advice from the hospital. Interventions for addressing these patient-directed discharges (PDDs) are lacking. We sought to explore the impact of methadone treatment for OUD on PDD. METHODS: Using electronic record and billing data from an urban safety-net hospital, we retrospectively examined the first hospitalization on a general medicine service for adults with OUD from January 2016 through June 2018. Associations with PDD compared to planned discharge were examined using multivariable logistic regression. Administration patterns of maintenance therapy versus new in-hospital initiation of methadone were examined using bivariate tests. RESULTS: During the study time period, 1,195 patients with OUD were hospitalized. 60.6% of patients received medication for OUD, of which 92.8% was methadone. Patients who received no treatment for OUD had a 19.1% PDD rate while patients initiated on methadone in-hospital had a 20.5% PDD rate and patients on maintenance methadone during the hospitalization had a 8.6% PDD rate. In multivariable logistic regression, methadone maintenance was associated with lower odds of PDD compared to no treatment (aOR 0.53, 95% CI 0.34-0.81), while methadone initiation was not (aOR 0.89, 95% CI 0.56-1.39). About 60% of patients initiated on methadone received 30 mg or less per day. CONCLUSIONS: In this study sample, maintenance methadone was associated with nearly a 50% reduction in the odds of PDD. More research is needed to assess the impact of higher hospital methadone initiation dosing on PDD and if there is an optimal protective dose.
                    Cover page: Methadone treatment and patient-directed hospital discharges among patients with opioid use disorder: Observations from general medicine services at an urban, safety-net hospital.
                    • Article
                    • Peer Reviewed
                    UC San Francisco Previously Published Works (2018)

                    Background

                    Opioids and stimulants (e.g., cocaine or methamphetamine/amphetamine [MAMP]) are major contributors to acute substance toxicity deaths. Causes of stimulant death have received little attention. We sought to characterize and compare causes of death and significant contributing conditions among persons who died from acute opioid, cocaine, or MAMP toxicity.

                    Methods

                    We identified all opioid, cocaine, or MAMP deaths in San Francisco from 2005 to 2015 through the California Electronic Death Reporting System. Multivariable logistic regression analyses were used to estimate associations between acute substance toxicity deaths (opioid versus stimulant; cocaine versus MAMP), additional reported causes of death, and significant contributing conditions most often linked to opioid and stimulant use.

                    Results

                    From 2005-2015, there were 1252 opioid deaths and 749 stimulant deaths. Cocaine accounted for most stimulant deaths. Decedents with cardiac or cerebral hemorrhage deaths had higher adjusted odds of death due to acute stimulant toxicity versus acute opioid toxicity (aOR = 4.79, 95%CI = 2.88-7.96, p < 0.01; aOR = 58.58, 95%CI = 21.06-162.91, p < 0.01, respectively); no statistically significant associations were found for cocaine compared to MAMP deaths. Significant contributing cardiac conditions were associated with higher adjusted odds of stimulant compared to opioid (aOR = 1.46, 95%CI = 1.19-1.79, p < 0.01) and cocaine compared to MAMP death (aOR = 1.66, 95%CI = 1.13-2.45, p = .01).

                    Conclusions

                    Stimulant compared to opioid deaths tended to involve cardiac or cerebrovascular causes of death, and cocaine deaths were more likely than MAMP deaths to involve significant contributing cardiac conditions. Mounting evidence suggests that stimulant use be considered a cardio/cerebrovascular risk factor and clinical care be adjusted to address this heightened risk.
                      Cover page: Cross-sectional cause of death comparisons for stimulant and opioid mortality in San Francisco, 2005-2015.
                      Top