[Show abstract][Hide abstract]ABSTRACT: High concentrations of household air pollution (HAP) due to biomass fuel usage with unvented, insufficient combustion devices are thought to be an important health risk factor in South Asia population. To better characterize the indoor concentrations of particulate matter (PM2.5) and carbon monoxide (CO), and to understand their impact on health in rural southern Nepal, this study analyzed daily monitoring data collected with DataRAM pDR-1000 and LASCAR CO data logger in 2980 households using traditional biomass cookstove indoor through the Nepal Cookstove Intervention Trial-Phase I between March 2010 and October 2011. Daily average PM2.5 and CO concentrations collected in area near stove were 1,376 (95% CI, 1,331-1,423) μg/m3 and 10.9 (10.5-11.3) parts per million (ppm) among households with traditional cookstoves. The 95th percentile, hours above 100μg/m3 for PM2.5 or 6ppm for CO, and hours above 1000μg/m3 for PM2.5 or 9ppm for CO were also reported. An algorithm was developed to differentiate stove-influenced (SI) periods from non-stove-influenced (non-SI) periods in monitoring data. Average stove-influenced concentrations were 3,469 (3,350-3,588) μg/m3 for PM2.5 and 21.8 (21.1-22.6) ppm for CO. Dry season significantly increased PM2.5 concentration in all metrics; wood was the cleanest fuel for PM2.5 and CO, while adding dung into the fuel increased concentrations of both pollutants. For studies in rural southern Nepal, CO concentration is not a viable surrogate for PM2.5 concentrations based on the low correlation between these measures. In sum, this study filled a gap in knowledge on HAP in rural Nepal using traditional cookstoves and revealed very high concentrations in these households.
Full-text available · Article · Jul 2016 · PLoS ONE
[Show abstract][Hide abstract]ABSTRACT: Epidemiology is concerned with determining the distribution and causes of disease. Throughout its history, epidemiology has
drawn upon statistical ideas and methods to achieve its aims. Because of the exponential growth in our capacity to measure
and analyze data on the underlying processes that define each person's state of health, there is an emerging opportunity for
population-based epidemiologic studies to influence health decisions made by individuals in ways that take into account the
individuals’ characteristics, circumstances, and preferences. We refer to this endeavor as “individualized health.” The present
article comprises 2 sections. In the first, we describe how graphical, longitudinal, and hierarchical models can inform the
project of individualized health. We propose a simple graphical model for informing individual health decisions using population-based
data. In the second, we review selected topics in causal inference that we believe to be particularly useful for individualized
health. Epidemiology and biostatistics were 2 of the 4 founding departments in the world's first graduate school of public
health at Johns Hopkins University, the centennial of which we honor. This survey of a small part of the literature is intended
to demonstrate that the 2 fields remain just as inextricably linked today as they were 100 years ago.
Article · Feb 2016 · American Journal of Epidemiology
[Show abstract][Hide abstract]ABSTRACT: The Pneumonia Etiology Research for Child Health (PERCH) study seeks to use
modern measurement technology to infer the causes of pneumonia for which
gold-standard evidence is unavailable. The paper describes a latent variable
model designed to infer from case-control data the etiology distribution for
the population of cases, and for an individual case given his or her
measurements. We assume each observation is drawn from a mixture model for
which each component represents one cause or disease class. The model addresses
a major limitation of the traditional latent class approach by taking account
of residual dependence among multivariate binary outcome given disease class,
hence reduces estimation bias, retains efficiency and offers more valid
inference. Such "local dependence" on a single subject is induced in the model
by nesting latent subclasses within each disease class. Measurement precision
and covariation can be estimated using the control sample for whom the class is
known. In a Bayesian framework, we use stick-breaking priors on the subclass
indicators for model-averaged inference across different numbers of subclasses.
Assessment of model fit and individual diagnosis are done using posterior
samples drawn by Gibbs sampling. We demonstrate the utility of the method on
simulated and on the motivating PERCH data.
[Show abstract][Hide abstract]ABSTRACT: Study objective:
Randomized controlled trials report inconsistent findings when comparing the initial success rate of peripheral intravenous cannulation using landmark versus ultrasonography for patients with difficult venous access. We sought to determine which method is superior for patients with varying levels of intravenous access difficulty.
We conducted a 2-group, parallel, randomized, controlled trial and randomly allocated 1,189 adult emergency department (ED) patients to landmark or ultrasonography, stratified by difficulty of access and operator. ED technicians performed the peripheral intravenous cannulations. Before randomization, technicians classified subjects as difficult, moderately difficult, or easy access according to visible or palpable veins and perception of difficulty with a landmark approach. If the first attempt failed, we randomized subjects a second time. We compared the initial and second-attempt success rates by procedural approach and difficulty of intravenous access, using a generalized linear mixed regression model, adjusted for operator.
The 33 participating technicians enrolled a median of 26 subjects (interquartile range 9 to 55). The initial success rate was 81% but varied significantly by technique and difficulty of access. The initial success rate by ultrasonography was higher than landmark for patients with difficult access (48.0 more successes per 100 tries; 95% confidence interval [CI] 35.6 to 60.3) or moderately difficult access (10. 2 more successes per 100 tries; 95% CI 1.7 to 18.7). Among patients with easy access, landmark yielded a higher success rate (10.6 more successes per 100 tries; 95% CI 5.8 to 15.4). The pattern of second-attempt success rates was similar.
Ultrasonographic peripheral intravenous cannulation is advantageous among patients with difficult or moderately difficult intravenous access but is disadvantageous among patients anticipated to have easy access.
[Show abstract][Hide abstract]ABSTRACT: summaryIn population studies on the aetiology of disease, one goal is the estimation of the fraction of cases that are attributable to each of several causes. For example, pneumonia is a clinical diagnosis of lung infection that may be caused by viral, bacterial, fungal or other pathogens. The study of pneumonia aetiology is challenging because directly sampling from the lung to identify the aetiologic pathogen is not standard clinical practice in most settings. Instead, measurements from multiple peripheral specimens are made. The paper introduces the statistical methodology designed for estimating the population aetiology distribution and the individual aetiology probabilities in the Pneumonia Etiology Research for Child Health study of 9500 children for seven sites around the world. We formulate the scientific problem in statistical terms as estimating the mixing weights and latent class indicators under a partially latent class model (PLCM) that combines heterogeneous measurements with different error rates obtained from a case–control study. We introduce the PLCM as an extension of the latent class model. We also introduce graphical displays of the population data and inferred latent class frequencies. The methods are tested with simulated data, and then applied to Pneumonia Etiology Research for Child Health data. The paper closes with a brief description of extensions of the PLCM to the regression setting and to the case where conditional independence between the measures is relaxed.
Article · Mar 2015 · Journal of the Royal Statistical Society Series C Applied Statistics
[Show abstract][Hide abstract]ABSTRACT: Acute lower respiratory infections (ALRI) are a leading cause of death among young children in low and middle income countries. Low birthweight is highly prevalent in South Asia and is associated with increased risks of mortality, morbidity, and poor motor and cognitive development. High levels of indoor household air pollution caused by open burning of biomass fuels such as wood, animal dung, and crop waste are common in these settings and are associated with high rates of ALRI and low birthweight. Alternative stove designs that burn biomass fuel more efficiently have been proposed as one method for reducing these high exposures and lowering the rates of these disorders. We designed two randomized trials to test this hypothesis.Methods/design: We conducted a pair of community-based, randomized trials of alternative cookstove installation a rural district in southern Nepal. Phase one was a cluster randomized, modified step-wedge design using an alternative biomass stove with a chimney to vent smoke to the exterior. A pre-installation period of morbidity assessment and household environmental assessment was conducted for six months in all households. This was followed by a one year step-wedge phase with 12 monthly steps for clusters of households to receive the alternative stove. The timing of alternative stove introduction was randomized. This step-wedge phase was followed in all households by another six month follow-up phase. Eligibility criteria for phase one included household informed consent, the presence of a married woman of reproductive age (15-30 yrs) or a child < 36 months. Children were followed until 36 months of age or the end of the trial and then discharged. Pregnancies were identified and followed until completion or end of the trial.Phase two was an individually randomized trial of the same alternative biomass stove versus liquid propane gas stove installation in a subset of households that participated in phase one. Follow-up for phase two was 12 months following stove installation. Eligibility criteria included the same components as phase one except children were only enrolled for morbidity follow-up if they were less than 24 months are the start.The primary outcomes included: the incidence of ALRI in children and birthweight among newborn infants.
We have presented the design and methods of two randomized trials of alternative cookstoves on rates of acute lower respiratory infection and birthweight in a rural population in southern Nepal.Trial registration: Clinicaltrials.gov (NCT00786877, Nov. 5, 2008).
Full-text available · Article · Dec 2014 · BMC Public Health
[Show abstract][Hide abstract]ABSTRACT: In 2006, Massachusetts expanded insurance coverage to many low-income individuals.
This study aimed to estimate the change in emergency department (ED) utilization per individual among a cohort who qualified for subsidized health insurance following the Massachusetts health care reform.
We obtained Massachusetts public health insurance enrollment data for the fiscal years 2004-2008 and identified 353,515 adults who enrolled in Commonwealth Care, a program that subsidizes insurance for low-income adults. We merged the enrollment data with statewide ED visit claims and created a longitudinal file that indicated each enrollee's ED visits and insurance status each month during the preenrollment and postenrollment periods.
We estimated the ratio in an individual's odds of an ED visit during the postperiod versus preperiod by conditional logistic regression.
Among the 112,146 CommCare enrollees who made at least 1 ED visit during the study period, an individual's odds of an ED visit decreased 4% [odds ratio (OR)=0.96; 95% confidence interval (CI), 0.94, 0.98] postenrollment. However, it varied significantly depending on preenrollment insurance status. A person's odds of an ED visit was 12% higher in the postperiod among enrollees not publicly insured prior (OR=1.12; 95% CI, 1.10, 1.25), but was 18% lower among enrollees who transitioned from the Health Safety Net, a program that pays for limited services for low-income individuals (OR=0.82; 95% CI, 0.78, 0.85).
Expanding subsidized health insurance did not uniformly change ED utilization for all newly insured low-income adults in Massachusetts.
[Show abstract][Hide abstract]ABSTRACT: In population studies on the etiology of disease, one goal is the estimation
of the fraction of cases attributable to each of several causes. For example,
pneumonia is a clinical diagnosis of lung infection that may be caused by
viral, bacterial, fungal, or other pathogens. The study of pneumonia etiology
is challenging because directly sampling from the lung to identify the
etiologic pathogen is not standard clinical practice in most settings. Instead,
measurements from multiple peripheral specimens are made. This paper introduces
the statistical methodology designed for estimating the population etiology
distribution and the individual etiology probabilities in the Pneumonia
Etiology Research for Child Health (PERCH) study of 9; 500 children for 7 sites
around the world. We formulate the scientific problem in statistical terms as
estimating the mixing weights and latent class indicators under a
partially-latent class model (pLCM) that combines heterogeneous measurements
with different error rates obtained from a case-control study. We introduce the
pLCM as an extension of the latent class model. We also introduce graphical
displays of the population data and inferred latent-class frequencies. The
methods are tested with simulated data, and then applied to PERCH data. The
paper closes with a brief description of extensions of the pLCM to the
regression setting and to the case where conditional independence among the
measures is relaxed.
[Show abstract][Hide abstract]ABSTRACT: Great uncertainty exists around indoor biomass burning exposure-disease relationships due to lack of detailed exposure data in large health outcome studies. Passive nephelometers can be used to estimate high particulate matter (PM) concentrations during cooking in low resource environments. Since passive nephelometers do not have a collection filter they are not subject to sampler overload. Nephelometric concentration readings can be biased due to particle growth in high humid environments and differences in compositional and size dependent aerosol characteristics. This paper explores relative humidity (RH) and gravimetric equivalency adjustment approaches to be used for the pDR-1000 used to assess indoor PM concentrations for a cookstove intervention trial in Nepal. Three approaches to humidity adjustment performed equivalently (similar root mean squared error). For gravimetric conversion, the new linear regression equation with log-transformed variables performed better than the traditional linear equation. In addition, gravimetric conversion equations utilizing a spline or quadratic term were examined. We propose a humidity adjustment equation encompassing the entire RH range instead of adjusting for RH above an arbitrary 60% threshold. Furthermore, we propose new integrated RH and gravimetric conversion methods because they have one response variable (gravimetric PM2.5 concentration), do not contain an RH threshold, and is straightforward.
Full-text available · Article · Jun 2014 · International Journal of Environmental Research and Public Health
[Show abstract][Hide abstract]ABSTRACT: Clustered data analysis is characterized by the need to describe both systematic variation in a mean model and cluster-dependent random variation in an association model. Marginalized multilevel models embrace the robustness and interpretations of a marginal mean model, while retaining the likelihood inference capabilities and flexible dependence structures of a conditional association model. Although there has been increasing recognition of the attractiveness of marginalized multilevel models, there has been a gap in their practical application arising from a lack of readily available estimation procedures. We extend the marginalized multilevel model to allow for nonlinear functions in both the mean and association aspects. We then formulate marginal models through conditional specifications to facilitate estimation with mixed model computational solutions already in place. We illustrate the MMM and approximate MMM approaches on a cerebrovascular deficiency crossover trial using SAS and an epidemiological study on race and visual impairment using R. Datasets, SAS and R code are included as supplemental materials.
[Show abstract][Hide abstract]ABSTRACT: Increasing evidence, including publication of the Transfusion Requirements in Critical Care trial in 1999, supports a lower hemoglobin threshold for RBC transfusion in ICU patients. However, little is known regarding the influence of this evidence on clinical practice over time in a large population-based cohort.
Retrospective population-based cohort study.
Thirty-five Maryland hospitals.
Seventy-three thousand three hundred eighty-five nonsurgical adults with an ICU stay greater than 1 day between 1994 and 2007.
The unadjusted odds of patients receiving an RBC transfusion increased from 7.9% during the pre-Transfusion Requirements in Critical Care baseline period (1994-1998) to 14.7% during the post-Transfusion Requirements in Critical Care period (1999-2007). A logistic regression model, including 40 relevant patient and hospital characteristics, compared the annual trend in the adjusted odds of RBC transfusion during the pre- versus post-Transfusion Requirements in Critical Care periods. During the pre-Transfusion Requirements in Critical Care period, the trend in the adjusted odds of RBC transfusion did not differ between hospitals averaging > 200 annual ICU discharges and hospitals averaging ≤ 200 annual ICU discharges (odds ratio, 1.07 [95% CI, 1.01-1.13] annually and 1.03 [95% CI, 0.99-1.07] annually, respectively; p = 0.401). However, during the post-Transfusion Requirements in Critical Care period, the adjusted odds of RBC transfusion decreased over time in higher ICU volume hospitals (odds ratio, 0.96 [95% CI, 0.93-0.98] annually) but continued to increase in lower ICU volume hospitals (odds ratio, 1.10 [95% CI, 1.08-1.13] annually), p < 0.001.
In this population-based cohort of ICU patients, the unadjusted odds of RBC transfusion increased in both higher and lower ICU volume hospitals both before and after Transfusion Requirements in Critical Care publication. After adjusting for relevant characteristics, the odds continued to increase in lower ICU volume hospitals in the post-Transfusion Requirements in Critical Care period, but it decreased in higher ICU volume hospitals. This suggests that evidence supporting restrictive RBC transfusion thresholds may not be uniformly translated into practice in different hospital settings.
[Show abstract][Hide abstract]ABSTRACT: Study objective:
We determine whether prescription information or services improve the medication adherence of emergency department (ED) patients.
Adult patients treated at one of 3 EDs between November 2010 and September 2011 and prescribed an antibiotic, central nervous system, gastrointestinal, cardiac, or respiratory drug at discharge were eligible. Subjects were randomly assigned to usual care or one of 3 prescription information or services intervention groups: (1) practical services to reduce barriers to prescription filling (practical prescription information or services); (2) consumer drug information from MedlinePlus (MedlinePlus prescription information or services); or (3) both services and information (combination prescription information or services). Self-reported medication adherence, measured by primary adherence (prescription filling) and persistence (receiving medicine as prescribed) rates, was determined during a telephone interview 1 week postdischarge.
Of the 3,940 subjects enrolled and randomly allocated to treatment, 86% (N=3,386) completed the follow-up interview. Overall, primary adherence was 88% and persistence was 48%. Across the sites, primary adherence and persistence did not differ significantly between usual care and the prescription information or services groups. However, at site C, subjects who received the practical prescription information or services (odds ratio [OR]=2.4; 95% confidence interval [CI] 1.4 to 4.3) or combination prescription information or services (OR=1.8; 95% CI 1.1 to 3.1) were more likely to fill their prescription compared with usual care. Among subjects prescribed a drug that treats an underlying condition, subjects who received the practical prescription information or services were more likely to fill their prescription (OR=1.8; 95% CI 1.0 to 3.1) compared with subjects who received usual care.
Prescription filling and receiving medications as prescribed was not meaningfully improved by offering patients patient-centered prescription information and services.
[Show abstract][Hide abstract]ABSTRACT: Study objective:
We determine the validity of self-reported prescription filling among emergency department (ED) patients.
We analyzed a subgroup of 1,026 patients enrolled in a randomized controlled trial who were prescribed at least 1 medication at ED discharge, were covered by Medicaid insurance, and completed a telephone interview 1 week after the index ED visit. We extracted all pharmacy and health care use claims information from a state Medicaid database for all subjects within 30 days of their index ED visit. We used the pharmacy claims as the criterion standard and evaluated the accuracy of self-reported prescription filling obtained during the follow-up interview by estimating its sensitivity, specificity, positive likelihood ratio and negative likelihood ratio tests. We also examined whether the accuracy of self-reported prescription filling varied significantly by patient and clinical characteristics.
Of the 1,635 medications prescribed, 74% were filled according to the pharmacy claims. Subjects reported filling 90% of prescriptions for a difference of 16% (95% confidence interval [CI] 14% to 18%). The self-reported data had high sensitivity (0.96; 95% CI 0.95 to 0.97) but low specificity (0.30; 95% CI 0.26 to 0.34). The positive likelihood ratio (1.37; 95% CI 1.29 to 2.46) and negative likelihood ratio (0.13; 95% CI 0.09 to 0.17) tests indicate that self-reported data are not a good indicator of prescription filling but are a moderately good indicator of nonfulfillment. Several factors were significantly associated with lower sensitivity (drug class and over-the-counter medications) and specificity (drug class, as needed, site and previous ED use).
Self-reported prescription filling is overestimated and associated with few factors.