Characterizing multimorbidity in ALIVE: Comparing single and ensemble clustering methods.

Multimorbidity, defined as having 2 or more chronic conditions, is a growing public health concern, but research in this area is complicated by the fact that multimorbidity is a highly heterogenous outcome. Individuals in a sample may have a differing number and varied combinations of conditions. Clustering methods, such as unsupervised machine learning algorithms, may allow us to tease out the unique multimorbidity phenotypes. However, many clustering methods exist and choosing which to use is challenging because we do not know the true underlying clusters. Here, we demonstrate the use of 3 individual algorithms (partition around medoids, hierarchical clustering, and probabilistic clustering) and a clustering ensemble approach (which pools different clustering approaches) to identify multimorbidity clusters in the AIDS Linked to the Intravenous Experience cohort study. We show how the clusters can be compared based on cluster quality, interpretability, and predictive ability. In practice, it is critical to compare the clustering results from multiple algorithms and to choose the approach that performs best in the domain(s) that aligns with plans to use the clusters in future analyses.

A Simulation Study Comparing the Performance of Time-Varying Inverse Probability Weighting and G-Computation in Survival Analysis.

Inverse probability weighting (IPW) and g-computation are commonly used in time-varying analyses. To inform decisions on which to use, we compared these methods using a plasmode simulation based on data from the Effects of Aspirin in Gestation and Reproduction (EAGeR) Trial (June 15, 2007-July 15, 2011). In our main analysis, we simulated a cohort study of 1,226 individuals followed for up to 10 weeks. The exposure was weekly exercise, and the outcome was time to pregnancy. We controlled for 6 confounding factors: 4 baseline confounders (race, ever smoking, age, and body mass index) and 2 time-varying confounders (compliance with assigned treatment and nausea). We sought to estimate the average causal risk difference by 10 weeks, using IPW and g-computation implemented using a Monte Carlo estimator and iterated conditional expectations (ICE). Across 500 simulations, we compared the bias, empirical standard error (ESE), average standard error, standard error ratio, and 95% confidence interval coverage of each approach. IPW (bias = 0.02; ESE = 0.04; coverage = 92.6%) and Monte Carlo g-computation (bias = -0.01; ESE = 0.03; coverage = 94.2%) performed similarly. ICE g-computation was the least biased but least precise estimator (bias = 0.01; ESE = 0.06; coverage = 93.4%). When choosing an estimator, one should consider factors like the research question, the prevalences of the exposure and outcome, and the number of time points being analyzed.

Geographic variation in HCV treatment penetration among people who inject drugs in Baltimore, MD.

We evaluated geographic heterogeneity in hepatitis C virus (HCV) treatment penetration among people who inject drug (PWID) across Baltimore, MD since the advent of direct-acting antivirals (DAAs) using space-time clusters of HCV viraemia. Using data from a community-based cohort of PWID, the AIDS Linked to the IntraVenous Experience (ALIVE) study, we identified space-time clusters with higher-than-expected rates of HCV viraemia between 2015 and 2019 using scan statistics. We used Poisson regression to identify covariates associated with HCV viraemia and used the regression-fitted values to detect adjusted space-time clusters of HCV viraemia in Baltimore city. Overall, in the cohort, HCV viraemia fell from 77% in 2015 to 64%, 49%, 39% and 36% from 2016 to 2019. In Baltimore city, the percentage of census tracts where prevalence of HCV viraemia was ≥85% dropped from 57% to 34%, 25%, 22% and 10% from 2015 to 2019. We identified two clusters of higher-than-expected HCV viraemia in the unadjusted analysis that lasted from 2015 to 2017 in East and West Baltimore and one adjusted cluster of HCV viraemia in West Baltimore from 2015 to 2016. Neither differences in age, sex, race, HIV status, nor neighbourhood deprivation were able to explain the significant space-time clusters. However, residing in a cluster with higher-than-expected viraemia was associated with age, sex, educational attainment and higher levels of neighbourhood deprivation. Nearly 4 years after DAAs became available, HCV treatment has penetrated all PWID communities across Baltimore city. While nearly all census tracts experienced improvements, change was more gradual in areas with higher levels of poverty.

Estimation of the Time-Varying Incremental Effect of Low-dose Aspirin on Incidence of Pregnancy.

BACKGROUND: In many research settings, the intervention implied by the average causal effect of a time-varying exposure is impractical or unrealistic, and we might instead prefer a more realistic target estimand. Instead of requiring all individuals to be always exposed versus unexposed, incremental effects quantify the impact of merely shifting each individual's probability of being exposed. METHODS: We demonstrate the estimation of incremental effects in the time-varying setting, using data from the Effects of Aspirin in Gestation and Reproduction trial, which assessed the effect of preconception low-dose aspirin on pregnancy outcomes. Compliance to aspirin or placebo was summarized weekly and was affected by time-varying confounders such as bleeding or nausea. We sought to estimate what the incidence of pregnancy by 26 weeks postrandomization would have been if we shifted each participant's probability of taking aspirin or placebo each week by odds ratios (OR) between 0.30 and 3.00. RESULTS: Under no intervention (OR = 1), the incidence of pregnancy was 77% (95% CI: 74%, 80%). Decreasing women's probability of complying with aspirin had little estimated effect on pregnancy incidence. When we increased women's probability of taking aspirin, estimated incidence of pregnancy increased, from 83% (95% confidence interval [CI] = 79%, 87%) for OR = 2 to 89% (95% CI = 84%, 93%) for OR=3. We observed similar results when we shifted women's probability of complying with a placebo. CONCLUSIONS: These results estimated that realistic interventions to increase women's probability of taking aspirin would have yielded little to no impact on the incidence of pregnancy, relative to similar interventions on placebo.

Impact of Hepatitis C Treatment Uptake on Cirrhosis and Mortality in Persons Who Inject Drugs : A Longitudinal, Community-Based Cohort Study.

BACKGROUND: Hepatitis C virus (HCV) infection can be cured, and the United States has joined the World Health Organization in calling for HCV elimination by 2030. However, historically low uptake of HCV treatment among people who inject drugs (PWID) threatens HCV elimination and exacerbates social and racial health disparities. OBJECTIVE: To assess whether all-oral HCV treatments were accessed by PWID and reduced liver disease burden and mortality. DESIGN: Community-based, longitudinal cohort study of persons with a history of injection drug use. SETTING: Baltimore, Maryland. PARTICIPANTS: 1323 participants enrolled in the ALIVE (AIDS Linked to the IntraVenous Experience) study from 2006 to 2019 and chronically infected with HCV. MEASUREMENTS: Liver stiffness measures (LSMs) by transient elastography, HCV RNA, and mortality from the National Death Index. RESULTS: Among 1323 persons with evidence of chronic HCV infection at baseline, the median age was 49 years. Most were Black (82%), male (71%), and HIV-negative (66%). The proportion in whom HCV RNA was detected decreased from 100% (by definition) in 2006 to 48% in 2019. Across 10 350 valid LSMs, cirrhosis was detected in 15% of participants in 2006, 19% in 2015, and 8% in 2019. Undetectable HCV RNA was significantly associated with reduced odds of cirrhosis (adjusted odds ratio, 0.28 [95% CI, 0.17 to 0.45]) and reduced all-cause mortality risk (adjusted hazard ratio, 0.54 [CI, 0.38 to 0.77]). LIMITATION: Noninvasive markers of liver fibrosis have not been validated in persons with sustained virologic response. CONCLUSION: Many community-based PWID in Baltimore are receiving HCV treatment, which is associated with sharp decreases in liver disease and mortality. Additional efforts will be needed to reduce residual barriers to treatment and to eliminate HCV as a public health threat for PWID. PRIMARY FUNDING SOURCE: National Institutes of Health.

Illustrating How to Simulate Data From Directed Acyclic Graphs to Understand Epidemiologic Concepts.

Simulation methods are a powerful set of tools that can allow researchers to better characterize phenomena from the real world. As such, the ability to simulate data represents a critical set of skills that epidemiologists should use to better understand epidemiologic concepts and ensure that they have the tools to continue to self-teach even when their formal instruction ends. Simulation methods are not always taught in epidemiology methods courses, whereas causal directed acyclic graphs (DAGs) often are. Therefore, this paper details an approach to building simulations from DAGs and provides examples and code for learning to perform simulations. We recommend using very simple DAGs to learn the procedures and code necessary to set up a simulation that builds on key concepts frequently of interest to epidemiologists (e.g., mediation, confounding bias, M bias). We believe that following this approach will allow epidemiologists to gain confidence with a critical skill set that may in turn have a positive impact on how they conduct future epidemiologic studies.

Estimation of the Average Causal Effect in Longitudinal Data With Time-Varying Exposures: The Challenge of Nonpositivity and the Impact of Model Flexibility.

There are important challenges to the estimation and identification of average causal effects in longitudinal data with time-varying exposures. Here, we discuss the difficulty in meeting the positivity condition. Our motivating example is the per-protocol analysis of the Effects of Aspirin in Gestation and Reproduction (EAGeR) Trial. We estimated the average causal effect comparing the incidence of pregnancy by 26 weeks that would have occurred if all women had been assigned to aspirin and complied versus the incidence if all women had been assigned to placebo and complied. Using flexible targeted minimum loss-based estimation, we estimated a risk difference of 1.27% (95% CI: -9.83, 12.38). Using a less flexible inverse probability weighting approach, the risk difference was 5.77% (95% CI: -1.13, 13.05). However, the cumulative probability of compliance conditional on covariates approached 0 as follow-up accrued, indicating a practical violation of the positivity assumption, which limited our ability to make causal interpretations. The effects of nonpositivity were more apparent when using a more flexible estimator, as indicated by the greater imprecision. When faced with nonpositivity, one can use a flexible approach and be transparent about the uncertainty, use a parametric approach and smooth over gaps in the data, or target a different estimand that will be less vulnerable to positivity violations.

The Role of the Natural Course in Causal Analysis.

The average causal effect compares counterfactual outcomes if everyone had been exposed versus if everyone had been unexposed, which can be an unrealistic contrast. Alternatively, we can target effects that compare counterfactual outcomes against the factual outcomes observed in the sample (i.e., we can compare against the natural course). Here, we demonstrate how the natural course can be estimated and used in causal analyses for model validation and effect estimation. Our example is an analysis assessing the impact of taking aspirin on pregnancy, 26 weeks after randomization, in the Effects of Aspirin in Gestation and Reproduction trial (United States, 2006-2012). To validate our models, we estimated the natural course using g-computation and then compared that against the observed incidence of pregnancy. We observed good agreement between the observed and model-based natural courses. We then estimated an effect that compared the natural course against the scenario in which participants assigned to aspirin always complied. If participants had always complied, there would have been 5.0 (95% confidence interval: 2.2, 7.8) more pregnancies per 100 women than was observed. It is good practice to estimate the natural course for model validation when using parametric models, but whether one should estimate a natural course contrast depends on the underlying research questions.

Estimating Associations Between Annual Concentrations of Particulate Matter and Mortality in the United States, Using Data Linkage and Bayesian Maximum Entropy.

BACKGROUND: Exposure to fine particulate matter (PM2.5) is an established risk factor for human mortality. However, previous US studies have been limited to select cities or regions or to population subsets (e.g., older adults). METHODS: Here, we demonstrate how to use the novel geostatistical method Bayesian maximum entropy to obtain estimates of PM2.5 concentrations in all contiguous US counties, 2000-2016. We then demonstrate how one could use these estimates in a traditional epidemiologic analysis examining the association between PM2.5 and rates of all-cause, cardiovascular, respiratory, and (as a negative control outcome) accidental mortality. RESULTS: We estimated that, for a 1 log(µg/m3) increase in PM2.5 concentration, the conditional all-cause mortality incidence rate ratio (IRR) was 1.029 (95% confidence interval [CI]: 1.006, 1.053). This implies that the rate of all-cause mortality at 10 µg/m3 would be 1.020 times the rate at 5 µg/m3. IRRs were larger for cardiovascular mortality than for all-cause mortality in all gender and race-ethnicity groups. We observed larger IRRs for all-cause, nonaccidental, and respiratory mortality in Black non-Hispanic Americans than White non-Hispanic Americans. However, our negative control analysis indicated the possibility for unmeasured confounding. CONCLUSION: We used a novel method that allowed us to estimate PM2.5 concentrations in all contiguous US counties and obtained estimates of the association between PM2.5 and mortality comparable to previous studies. Our analysis provides one example of how Bayesian maximum entropy could be used in epidemiologic analyses; future work could explore other ways to use this approach to inform important public health questions.

Mortality Among Persons Entering HIV Care Compared With the General U.S. Population : An Observational Study.

BACKGROUND: Understanding advances in the care and treatment of adults with HIV as well as remaining gaps requires comparing differences in mortality between persons entering care for HIV and the general population. OBJECTIVE: To assess the extent to which mortality among persons entering HIV care in the United States is elevated over mortality among matched persons in the general U.S. population and trends in this difference over time. DESIGN: Observational cohort study. SETTING: Thirteen sites from the U.S. North American AIDS Cohort Collaboration on Research and Design. PARTICIPANTS: 82 766 adults entering HIV clinical care between 1999 and 2017 and a subset of the U.S. population matched on calendar time, age, sex, race/ethnicity, and county using U.S. mortality and population data compiled by the National Center for Health Statistics. MEASUREMENTS: Five-year all-cause mortality, estimated using the Kaplan-Meier estimator of the survival function. RESULTS: Overall 5-year mortality among persons entering HIV care was 10.6%, and mortality among the matched U.S. population was 2.9%, for a difference of 7.7 (95% CI, 7.4 to 7.9) percentage points. This difference decreased over time, from 11.1 percentage points among those entering care between 1999 and 2004 to 2.7 percentage points among those entering care between 2011 and 2017. LIMITATION: Matching on available covariates may have failed to account for differences in mortality that were due to sociodemographic factors rather than consequences of HIV infection and other modifiable factors. CONCLUSION: Mortality among persons entering HIV care decreased dramatically between 1999 and 2017, although those entering care remained at modestly higher risk for death in the years after starting care than comparable persons in the general U.S. population. PRIMARY FUNDING SOURCE: National Institutes of Health.

Simulation as a Tool for Teaching and Learning Epidemiologic Methods.

In aspiring to be discerning epidemiologists, we must learn to think critically about the fundamental concepts in our field and be able to understand and apply many of the novel methods being developed today. We must also find effective ways to teach both basic and advanced topics in epidemiology to graduate students, in a manner that goes beyond simple provision of knowledge. Here, we argue that simulation is one critical tool that can be used to help meet these goals, by providing examples of how simulation can be used to address 2 common misconceptions in epidemiology. First, we show how simulation can be used to explore nondifferential exposure misclassification. Second, we show how an instructor could use simulation to provide greater clarity on the correct definition of the P value. Through these 2 examples, we highlight how simulation can be used to both clearly and concretely demonstrate theoretical concepts, as well as to test and experiment with ideas, theories, and methods in a controlled environment. Simulation is therefore useful not only in the classroom but also as a skill for independent self-learning.

Incremental Propensity Score Effects for Time-fixed Exposures.

When causal inference is of primary interest, a range of target parameters can be chosen to define the causal effect, such as average treatment effects (ATEs). However, ATEs may not always align with the research question at hand. Furthermore, the assumptions needed to interpret estimates as ATEs, such as exchangeability, consistency, and positivity, are often not met. Here, we present the incremental propensity score (PS) approach to quantify the effect of shifting each person's exposure propensity by some predetermined amount. Compared with the ATE, incremental PS may better reflect the impact of certain policy interventions and do not require that positivity hold. Using the Nulliparous Pregnancy Outcomes Study: monitoring mothers-to-be (nuMoM2b), we quantified the relationship between total vegetable intake and the risk of preeclampsia and compared it to average treatment effect estimates. The ATE estimates suggested a reduction of between two and three preeclampsia cases per 100 pregnancies for consuming at least half a cup of vegetables per 1,000 kcal. However, positivity violations obfuscate the interpretation of these results. In contrast, shifting each woman's exposure propensity by odds ratios ranging from 0.20 to 5.0 yielded no difference in the risk of preeclampsia. Our analyses show the utility of the incremental PS effects in addressing public health questions with fewer assumptions.

Defining and Identifying Per-protocol Effects in Randomized Trials.

In trials with noncompliance to assigned treatment, researchers might be interested in estimating a per-protocol effect-a comparison of two counterfactual outcomes defined by treatment assignment and (often time-varying) compliance with a well-defined treatment protocol. Here, we provide a general counterfactual definition of a per-protocol effect and discuss examples of per-protocol effects that are of either substantive or methodologic interest. In doing so, we seek to make more concrete what per-protocol effects are and highlight that one can estimate per-protocol effects that are more than just a comparison of always taking treatment in two distinct treatment arms. We then discuss one set of identifiability conditions that allow for identification of a causal per-protocol effect, highlighting some potential violations of those conditions that might arise when estimating per-protocol effects.

Incarceration and Number of Sexual Partners After Incarceration Among Vulnerable US Women, 2007-2017.

Objectives. To examine whether women's incarceration increases numbers of total and new sexual partners.Methods. US women with or at risk for HIV in a multicenter cohort study answered incarceration and sexual partner questions semiannually between 2007 and 2017. We used marginal structural models to compare total and new partners at visits not following incarceration with all visits following incarceration and visits immediately following incarceration. Covariates included demographics, HIV status, sex exchange, drug or alcohol use, and housing instability.Results. Of the 3180 participants, 155 were incarcerated. Women reported 2 partners, 3 or more partners, and new partners at 5.2%, 5.2%, and 9.3% of visits, respectively. Relative to visits not occurring after incarceration, odds ratios were 2.41 (95% confidence interval [CI] = 1.20, 4.85) for 2 partners, 2.03 (95% CI = 0.97, 4.26) for 3 or more partners, and 3.24 (95% CI = 1.69, 6.22) for new partners at visits immediately after incarceration. Odds ratios were similar for all visits following incarceration.Conclusions. Women had more total partners and new partners immediately and at all visits following incarceration after confounders and loss to follow-up had been taken into account.

Estimating Human Immunodeficiency Virus (HIV) Prevention Effects in Low-incidence Settings.

BACKGROUND: Randomized controlled trials (RCTs) for determining efficacy of preexposure prophylaxis (PrEP) in preventing human immunodeficiency virus (HIV) infection have not been conducted among US women because their lower HIV incidence requires impractically large studies. Results from higher-incidence settings, like Sub-Saharan Africa, may not apply to US women owing to differences in age, sexual behavior, coinfections, and adherence. METHODS: We propose a novel strategy for evaluating PrEP efficacy in the United States using data from both settings to obtain four parameters: (1) intention-to-treat (ITT) and (2) per-protocol effects in the higher-incidence setting, (3) per-protocol effect generalized to the lower-incidence setting, and (4) back-calculated ITT effect using adherence data from the lower-incidence setting. To illustrate, we simulated two RCTs comparing PrEP against placebo: one in 4000 African women and another in 500 US women. We estimated all parameters using g-computation and report risk ratios averaged over 2000 simulations, alongside the 2.5th and 97.5th percentiles of the simulation results. RESULTS: Twelve months after randomization, the African ITT and per-protocol risk ratios were 0.65 (0.47, 0.88) and 0.20 (0.08, 0.34), respectively. The US ITT and per-protocol risk ratios were 0.42 (0.20, 0.62) and 0.17 (0.03, 0.38), respectively. These results matched well the simulated true effects. CONCLUSIONS: Our simple demonstration informs the design of future studies seeking to estimate the effectiveness of a treatment (like PrEP) in lower-incidence settings where a traditional RCT would not be feasible. See video abstract at, http://links.lww.com/EDE/B506.

Using Animations of Risk Functions to Visualize Trends in US All-Cause and Cause-Specific Mortality, 1968-2016.

OBJECTIVES: To use dynamic visualizations of mortality risk functions over both calendar year and age as a way to estimate and visualize patterns in US life spans. METHODS: We built 49 synthetic cohorts, 1 per year 1968 to 2016, using National Center for Health Statistics (NCHS) mortality and population data. Within each cohort, we estimated age-specific probabilities of dying from any cause (all-cause analysis) or from a particular cause (cause-specific analysis). We then used Kaplan-Meier (all-cause) or Aalen-Johansen (cause-specific) estimators to obtain risk functions. We illustrated risk functions using time-lapse animations. RESULTS: Median age at death increased from 75 years in 1970 to 83 years in 2015. Risk by age 100 years of cardiovascular mortality decreased (from a risk of 55% in 1970 to 32% in 2015), whereas risk attributable to other (i.e., nonrespiratory and noncardiovascular) causes increased in compensation. CONCLUSIONS: Our findings were consistent with the trends published in the NCHS 2015 mortality report, and our dynamic animations added an efficient, interpretable tool for visualizing US mortality trends over age and calendar time.

Correction to: Parametric assumptions equate to hidden observations: comparing the efficiency of nonparametric and parametric models for estimating time to AIDS or death in a cohort of HIV-positive women.

.

Parametric assumptions equate to hidden observations: comparing the efficiency of nonparametric and parametric models for estimating time to AIDS or death in a cohort of HIV-positive women.

BACKGROUND: When conducting a survival analysis, researchers might consider two broad classes of models: nonparametric models and parametric models. While nonparametric models are more flexible because they make few assumptions regarding the shape of the data distribution, parametric models are more efficient. Here we sought to make concrete the difference in efficiency between these two model types using effective sample size. METHODS: We compared cumulative risk of AIDS or death estimated using four survival models - nonparametric, generalized gamma, Weibull, and exponential - and data from 1164 HIV patients who were alive and AIDS-free in 1995. We added pseudo-observations to the sample until the spread of the 95% confidence limits for the nonparametric model became less than that for the parametric models. RESULTS: We found the 3-parameter generalized gamma to be a good fit to the nonparametric risk curve, but the 1-parameter exponential both underestimated and overestimated the risk at different times. Using two year-risk as an example, we had to add 354, 593, and 3960 observations for the nonparametric model to be as efficient as the generalized gamma, Weibull, and exponential models, respectively. CONCLUSIONS: These added observations represent the hidden observations underlying the efficiency gained through parametric model form assumptions. If the model is correctly specified, the efficiency gain may be justified, as appeared to be the case for the generalized gamma model. Otherwise, precision will be improved, but at the cost of specification bias, as was the case for the exponential model.

At-Risk Alcohol Use Among HIV-Positive Patients and the Completion of Patient-Reported Outcomes.

Heavy drinking is prevalent among people living with HIV. Studies use tools like patient-reported outcomes (PROs) to quantify alcohol use in a detailed, timely manner. However, if alcohol misuse influences PRO completion, selection bias may result. Our study included 14,145 adult HIV patients (133,036 visits) from CNICS who were eligible to complete PROs at an HIV primary care visit. We compared PRO completion proportions between patients with and without a clinical diagnosis of at-risk alcohol use in the prior year. We accounted for confounding by baseline and visit-specific covariates. PROs were completed at 20.8% of assessed visits. The adjusted difference in PRO completion proportions was -3.2% (95% CI -5.6 to -0.8%). The small association between receipt of an at-risk alcohol use diagnosis and decreased PRO completion suggests there could be modest selection bias in studies using the PRO alcohol measure.