Transcriptomic responses from improved murine sepsis models can better mimic human surgical sepsis.

Historically, murine models of inflammation in biomedical research have been shown to minimally correlate with genomic expression patterns from blood leukocytes in humans. In 2019, our laboratory reported an improved surgical sepsis model of cecal ligation and puncture (CLP) that provides additional daily chronic stress (DCS), as well as adhering to the Minimum Quality Threshold in Pre-Clinical Sepsis Studies (MQTiPSS) guidelines. This model phenotypically recapitulates the persistent inflammation, immunosuppression, and catabolism syndrome observed in adult human surgical sepsis survivors. Whether these phenotypic similarities between septic humans and mice are replicated at the circulating blood leukocyte transcriptome has not been demonstrated. Our analysis, in contrast with previous findings, demonstrated that genome-wide expression in our new murine model more closely approximated human surgical sepsis patients, particularly in the more chronic phases of sepsis. Importantly, our new model of murine surgical sepsis with chronic stress did not reflect well gene expression patterns from humans with community-acquired sepsis. Our work indicates that improved preclinical murine sepsis modeling can better replicate both the phenotypic and transcriptomic responses to surgical sepsis, but cannot be extrapolated to other sepsis etiologies. Importantly, these improved models can be a useful adjunct to human-focused and artificial intelligence-based forms of research in order to improve septic patients' morbidity and mortality.

Impact of an Emergency Department-to-Home Transitional Care Intervention on Health Service Use in Medicare Beneficiaries: A Mixed Methods Study.

BACKGROUND: Hospital-based acute care [emergency department (ED) visits and hospitalizations] that is preventable with high-quality outpatient care contributes to health care system waste and patient harm. OBJECTIVE: To test the hypothesis that an ED-to-home transitional care intervention reduces hospital-based acute care in chronically ill, older ED visitors. RESEARCH DESIGN: Convergent, parallel, mixed-methods design including a randomized controlled trial. SETTING: Two diverse Florida EDs. SUBJECTS: Medicare fee-for-service beneficiaries with chronic illness presenting to the ED. INTERVENTION: The Coleman Care Transition Intervention adapted for ED visitors. MEASURES: The main outcome was hospital-based acute care within 60 days of index ED visit. We also assessed office-based outpatient visits during the same period. RESULTS: The Intervention did not significantly reduce return ED visits or hospitalizations or increase outpatient visits. In those with return ED visits, the Intervention Group was less likely to be hospitalized than the Usual Care Group. Interview themes describe a cycle of hospital-based acute care largely outside patients' control that may be difficult to interrupt with a coaching intervention. CONCLUSIONS AND RELEVANCE: Structural features of the health care system, including lack of access to timely outpatient care, funnel patients into the ED and hospital admission. Reducing hospital-based acute care requires increased focus on the health care system rather than patients' care-seeking decisions.

Use of negative controls in a prescription sequence symmetry analysis to reduce time-varying bias.

PURPOSE: There is an increased use in the (prescription) sequence symmetry analysis (PSSA); however, limited studies have incorporated a negative control, and no study has formally quantified and controlled for within-patient time-varying bias using a negative control. Our aim was to develop a process to incorporate the effect of negative controls into the main analysis of a PSSA. METHODS: Using a previously assessed dihydropyridine calcium channel blocker (DH-CCB) and loop diuretic PSSA, we directly compared the adjusted sequence ratios (aSRs) of DH-CCBs to each of the two negative control index drugs (levothyroxine and angiotensin converting enzyme [ACE] inhibitor/angiotensin-2 receptor blocker [ARB]) using the ratio of the aSRs to estimate a relative aSR with a Z test. Further, we utilized the relative aSR in stratum-specific analyses and varying exposure windows. RESULTS: The relative aSR of DH-CCBs decreased from 1.87 to 1.72 (95% CI 1.66-1.78) using levothyroxine as a negative control index drug. ACE inhibitor/ARB negative control index drug resulted in an aSR of 1.27 thus reducing the relative aSR for DH-CBB from 1.84 to 1.45 (95% CI 1.41-1.49). When restricting the exposure window to 180 and 90 days, the relative aSR of DH-CCBs increased to 1.68 (95% CI 1.62-1.74) and 1.86 (95% CI 1.78-1.94), respectively, relative to the ACE inhibitor/ARB negative control index drug. CONCLUSION: We illustrated how to incorporate negative control index drugs into a PSSA and generate relative aSRs. Stratum-specific assessments and varying the exposure windows while using negative control index drugs can yield more informative results.

Fidelity of a Traffic Safety Education Intervention for Combat Veterans.

Equipoise, feasibility, and fidelity were studied for the control condition of an occupational therapy driving intervention in a randomized controlled trial. We ranked equipoise and feasibility of six traffic safety education methods and created an implementation fidelity competency checklist. Education method selection was informed using the proportion of concordant ranks analysis while literature and a peer review informed competency checklist development. A proctored-online course delivery had the highest rater agreement (equipoise = .96 [.87-1.00]; feasibility = .99 [.97-1.00]). Implementation fidelity was supported by a 19-component training and evaluation checklist. This study supports promoting the scientific rigor of the RCT via - equipoise, feasibility, and implementation fidelity.

Phenotypic heterogeneity by site of infection in surgical sepsis: a prospective longitudinal study.

BACKGROUND: The role of site of infection in sepsis has been poorly characterized. Additionally, sepsis epidemiology has evolved. Early mortality has decreased, but many survivors now progress into chronic critical illness (CCI). This study sought to determine if there were significant differences in the host response and current epidemiology of surgical sepsis categorized by site of infection. STUDY DESIGN: This is a longitudinal study of surgical sepsis patients characterized by baseline predisposition, insult characteristics, serial biomarkers, hospital outcomes, and long-term outcomes. Patients were categorized into five anatomic sites of infection. RESULTS: The 316 study patients were predominantly Caucasian; half were male, with a mean age of 62 years, high comorbidity burden, and low 30-day mortality (10%). The primary sites were abdominal (44%), pulmonary (19%), skin/soft tissue (S/ST, 17%), genitourinary (GU, 12%), and vascular (7%). Most abdominal infections were present on admission and required source control. Comparatively, they had more prolonged proinflammation, immunosuppression, and persistent organ dysfunction. Their long-term outcome was poor with 37% CCI (defined as > 14 in ICU with organ dysfunction), 49% poor discharge dispositions, and 30% 1-year mortality. Most pulmonary infections were hospital-acquired pneumonia. They had similar protracted proinflammation and organ dysfunction, but immunosuppression normalized. Long-term outcomes are similarly poor (54% CCI, 47% poor disposition, 32% 1-year mortality). S/ST and GU infections occurred in younger patients with fewer comorbidities, less perturbed immune responses, and faster resolution of organ dysfunction. Comparatively, S/ST had better long-term outcomes (23% CCI, 39% poor disposition, 13% 1-year mortality) and GU had the best (10% CCI, 20% poor disposition, 10% 1-year mortality). Vascular sepsis patients were older males, with more comorbidities. Proinflammation was blunted with baseline immunosuppression and organ dysfunction that persisted. They had the worst long-term outcomes (38% CCI, 67% poor disposition, 57% 1-year mortality). CONCLUSION: There are notable differences in baseline predisposition, host responses, and clinical outcomes by site of infection in surgical sepsis. While previous studies have focused on differences in hospital mortality, this study provides unique insights into the host response and long-term outcomes associated with different sites of infection.

Validation of algorithms to estimate gestational age at birth in the Medicaid Analytic eXtract-Quantifying the misclassification of maternal drug exposure during pregnancy.

PURPOSE: Accurate ascertainment of gestational age (GA) has been a challenge in perinatal epidemiologic research. To date, no study has validated GA algorithms in Medicaid Analytic eXtract (MAX). METHODS: We linked livebirths of mothers enrolled in Medicaid ≥30 days after delivery in 1999-2010 MAX to state birth certificates. We used clinical/obstetric estimate of gestation on the birth certificates as gold standard to validate claims-based GA algorithms. We calculated the proportions of deliveries with algorithm-estimated GA within 1-/2-weeks of the gold standard, the sensitivity, specificity, and positive/negative predictive value (PPV/NPV) of exposure to select medications during specific gestation windows, and quantified the impact of exposure misclassification on hypothetical relative risk (RR) estimates. RESULTS: We linked 1 336 495 eligible deliveries. Within 1-week agreement was 77%-80% overall and 47%-56% for preterm deliveries. The trimester-specific drug exposure status had high sensitivities and PPVs (88.5%-98.5%), and specificities and NPVs (>99.0%). Assuming a hypothetical RR of 2.0, bias associated with exposure misclassification during first trimester ranged from 10% to 40% under non-differential/differential misclassification assumptions. CONCLUSIONS: Claims-based GA algorithms had good agreement with the gold standard overall, but lower agreement among preterm deliveries, potentially resulting in biased risk estimated for pregnancy exposure evaluations.

Increased Risk of Acute Pancreatitis with Codeine Use in Patients with a History of Cholecystectomy.

BACKGROUND: Codeine has a spasmodic effect on sphincter of Oddi and is suspected to cause acute pancreatitis in patients with a history of cholecystectomy. AIMS: To assess the association between codeine use and acute pancreatitis in patients with a previous cholecystectomy. METHODS: We conducted a retrospective nested case-control study using the 2005-2015 MarketScan® Commercial Claims and Encounters Database. The cohort included patients aged 18-64; cohort entry began 365 days after cholecystectomy. Odds ratios (ORs) and 95% CIs for acute pancreatitis hospitalization were estimated comparing use of codeine with non-use of codeine. In a secondary analysis, use of codeine was compared with an active comparator: use of non-steroidal anti-inflammatory drugs (NSAIDs). RESULTS: Of the 664,083 patients included in the cohort, 1707 patients were hospitalized for acute pancreatitis (incidence 1.1 per 1000 person-years) and were matched to 17,063 controls. Compared with non-use of codeine, use of codeine was associated with an increased risk of acute pancreatitis (OR 2.67; 95% CI 1.63, 4.36), particularly elevated in the first 15 days of codeine use (OR 5.37; 95% CI 2.70, 10.68). Compared with use of NSAIDs, use of codeine was also associated with an increased risk of acute pancreatitis (OR 2.64; 95% CI 1.54, 4.52). CONCLUSION: Codeine is associated with an increased risk of acute pancreatitis in patients who have previously undergone cholecystectomy; greater clinician awareness of this association is needed.

The Development of Chronic Critical Illness Determines Physical Function, Quality of Life, and Long-Term Survival Among Early Survivors of Sepsis in Surgical ICUs.

OBJECTIVES: This study sought to examine mortality, health-related quality of life, and physical function among sepsis survivors who developed chronic critical illness. DESIGN: Single-institution, prospective, longitudinal, observational cohort study assessing 12-month outcomes. SETTING: Two surgical/trauma ICUs at an academic tertiary medical and level 1 trauma center. PATIENTS: Adult critically ill patients that survived 14 days or longer after sepsis onset. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Baseline patient characteristics and function, sepsis severity, and clinical outcomes of the index hospitalization were collected. Follow-up physical function (short physical performance battery; Zubrod; hand grip strength) and health-related quality of life (EuroQol-5D-3L, Short Form-36) were measured at 3, 6, and 12 months. Hospital-free days and mortality were determined at 12 months. We compared differences in long-term outcomes between subjects who developed chronic critical illness (≥ 14 ICU days with persistent organ dysfunction) versus those with rapid recovery. The cohort consisted of 173 sepsis patients; 63 (36%) developed chronic critical illness and 110 (64%) exhibited rapid recovery. Baseline physical function and health-related quality of life did not differ between groups. Those who developed chronic critical illness had significantly fewer hospital-free days (196 ± 148 vs 321 ± 65; p < 0.0001) and reduced survival at 12-months compared with rapid recovery subjects (54% vs 92%; p < 0.0001). At 3- and 6-month follow-up, chronic critical illness patients had significantly lower physical function (3 mo: short physical performance battery, Zubrod, and hand grip; 6 mo: short physical performance battery, Zubrod) and health-related quality of life (3- and 6-mo: EuroQol-5D-3L) compared with patients who rapidly recovered. By 12-month follow-up, chronic critical illness patients had significantly lower physical function and health-related quality of life on all measures. CONCLUSIONS: Surgical patients who develop chronic critical illness after sepsis exhibit high healthcare resource utilization and ultimately suffer dismal long-term clinical, functional, and health-related quality of life outcomes. Further understanding of the mechanisms driving the development and persistence of chronic critical illness will be necessary to improve long-term outcomes after sepsis.

Model-based standardization using an outcome model with random effects.

Model-based standardization uses a statistical outcome model or exposure model to estimate a population-average association that is unconfounded by selected covariates. With it, one can compare groups using a distribution of confounders identical in each group to that of a standard population. We develop an approach based on an outcome model, in which the mean of the outcome is modeled conditional on the exposure and the confounders. In our approach, there is a confounder that clusters the observations into a very large number of categories. We treat the parameters for the clusters as random effects. We use a between-within model to account for the association of the random effects not only with the exposure but also with the cluster population sizes. We review alternative approaches presented in the literature, and we compare the outcome-modeling approach to recently proposed exposure-modeling approaches incorporating random effects. To illustrate, we use 2014 to compare proportions of acute respiratory tract infection diagnoses with an antibiotic prescription for emergency department versus outpatient visits, adjusting for confounding by unmeasured patient level variables and measured diagnosis-level variables. We also present results of a simulation study.

Prognostic value of NT-proBNP levels in the acute phase of sepsis on lower long-term physical function and muscle strength in sepsis survivors.

BACKGROUND: Sepsis survivors often develop chronic critical illness (CCI) and demonstrate the persistent inflammation, immunosuppression, and catabolism syndrome predisposing them to long-term functional limitations and higher mortality. There is a need to identify biomarkers that can predict long-term worsening of physical function to be able to act early and prevent mobility loss. N-terminal pro-brain natriuretic peptide (NT-proBNP) is a well-accepted biomarker of cardiac overload, but it has also been shown to be associated with long-term physical function decline. We explored whether NT-proBNP blood levels in the acute phase of sepsis are associated with physical function and muscle strength impairment at 6 and 12 months after sepsis onset. METHODS: This is a retrospective analysis conducted in 196 sepsis patients (aged 18-86 years old) as part of the University of Florida (UF) Sepsis and Critical Illness Research Center (SCIRC) who consented to participate in the 12-month follow-up study. NT-proBNP was measured at 24 h after sepsis onset. Patients were followed to determine physical function by short physical performance battery (SPPB) test score (scale 0 to12-higher score corresponds with better physical function) and upper limb muscle strength by hand grip strength test (kilograms) at 6 and 12 months. We used a multivariate linear regression model to test an association between NT-proBNP levels, SPPB, and hand grip strength scores. Missing follow-up data or absence due to death was accounted for by using inverse probability weighting based on concurrent health performance status scores. Statistical significance was set at p ≤ 0.05. RESULTS: After adjusting for covariates (age, gender, race, Charlson comorbidity index, APACHE II score, and presence of CCI condition), higher levels of NT-proBNP at 24 h after sepsis onset were associated with lower SPPB scores at 12 months (p < 0.05) and lower hand grip strength at 6-month (p < 0.001) and 12-month follow-up (p < 0.05). CONCLUSIONS: NT-proBNP levels during the acute phase of sepsis may be a useful indicator of higher risk of long-term impairments in physical function and muscle strength in sepsis survivors.

Featured Article: Behavior Interventions Addressing Obesity in Rural Settings: The E-FLIP for Kids Trial.

OBJECTIVE: To assess the effectiveness of behavioral parent-only (PO) and family-based (FB) interventions on child weight, dietary intake, glycated hemoglobin, and quality of life in rural settings. METHODS: This study was a three-armed, randomized controlled trial. Participants were children (age 8-12 years) with overweight or obesity and their parents. A FB (n = 88), a PO (n = 78) and a health education condition (HEC) (n = 83) each included 20 group contacts over 1 year. Assessment and treatment contacts occurred at Cooperative Extension Service offices. The main outcome was change in child body mass index z-score (BMIz) from baseline to year 2. RESULTS: Parents in all conditions reported high treatment satisfaction (mean of 3.5 or higher on a 4-point scale). A linear mixed model analysis of change in child BMIz from baseline to year 1 and year 2 found that there were no significant group by time differences in child BMIz (year 2 change in BMIz for FB = -0.03 [-0.1, 0.04], PO = -0.01 [-0.08, 0.06], and HEC = -0.09 [-0.15, -0.02]). While mean attendance across conditions was satisfactory during months 1-4 (69%), it dropped during the maintenance phase (42%). High attendance for the PO intervention was related to greater changes in child BMIz (p < .02). Numerous barriers to participation were reported. CONCLUSION: Many barriers exist that inhibit regular attendance at in-person contacts for many families. Innovative delivery strategies are needed that balance treatment intensity with feasibility and acceptability to families and providers to facilitate broad dissemination in underserved rural settings.ClinicalTrials.gov Identifier: NCT01820338.

Human Myeloid-derived Suppressor Cells are Associated With Chronic Immune Suppression After Severe Sepsis/Septic Shock.

OBJECTIVE: We hypothesized that after sepsis in humans, MDSCs will be persistently increased, functionally immunosuppressive, and associated with adverse clinical outcomes. BACKGROUND: Cancer and sepsis have surprisingly similar immunologic responses and equally dismal long term consequences. In cancer, increased myeloid-derived suppressor cells (MDSCs) induce detrimental immunosuppression, but little is known about the role of MDSCs after sepsis. METHODS: Blood was obtained from 74 patients within 12 hours of severe sepsis/septic shock (SS/SS), and at set intervals out to 28 days, and also in 18 healthy controls. MDSCs were phenotyped for cell surface receptor expression and enriched by cell sorting. Functional and genome-wide expression analyses were performed. Multiple logistic regression analysis was conducted to determine if increased MDSC appearance was associated with in-hospital and long-term outcomes. RESULTS: After SS/SS, CD33CD11bHLA-DR MDSCs were dramatically increased out to 28 days (P < 0.05). When co-cultured with MDSCs from SS/SS patients, antigen-driven T-cell proliferation and TH1/TH2 cytokine production were suppressed (P < 0.05). Additionally, septic MDSCs had suppressed HLA gene expression and up-regulated ARG1 expression (P < 0.05). Finally, SS/SS patients with persistent increased percentages of blood MDSCs had increased nosocomial infections, prolonged intensive care unit stays, and poor functional status at discharge (P < 0.05). CONCLUSIONS: After SS/SS in humans, circulating MDSCs are persistently increased, functionally immunosuppressive, and associated with adverse outcomes. This novel observation warrants further studies. As observed in cancer immunotherapy, MDSCs could be a novel component in multimodality immunotherapy targeting detrimental inflammation and immunosuppression after SS/SS to improve currently observed dismal long-term outcomes.

Estimating the Effect of School Water, Sanitation, and Hygiene Improvements on Pupil Health Outcomes.

BACKGROUND: We conducted a cluster-randomized water, sanitation, and hygiene trial in 185 schools in Nyanza province, Kenya. The trial, however, had imperfect school-level adherence at many schools. The primary goal of this study was to estimate the causal effects of school-level adherence to interventions on pupil diarrhea and soil-transmitted helminth infection. METHODS: Schools were divided into water availability groups, which were then randomized separately into either water, sanitation, and hygiene intervention arms or a control arm. School-level adherence to the intervention was defined by the number of intervention components-water, latrines, soap-that had been adequately implemented. The outcomes of interest were pupil diarrhea and soil-transmitted helminth infection. We used a weighted generalized structural nested model to calculate prevalence ratio. RESULTS: In the water-scarce group, there was evidence of a reduced prevalence of diarrhea among pupils attending schools that adhered to two or to three intervention components (prevalence ratio = 0.28, 95% confidence interval: 0.10, 0.75), compared with what the prevalence would have been had the same schools instead adhered to zero components or one. In the water-available group, there was no evidence of reduced diarrhea with better adherence. For the soil-transmitted helminth infection and intensity outcomes, we often observed point estimates in the preventive direction with increasing intervention adherence, but primarily among girls, and the confidence intervals were often very wide. CONCLUSIONS: Our instrumental variable point estimates sometimes suggested protective effects with increased water, sanitation, and hygiene intervention adherence, although many of the estimates were imprecise.

Adjusting for unmeasured confounding due to either of two crossed factors with a logistic regression model.

Motivated by an investigation of the effect of surface water temperature on the presence of Vibrio cholerae in water samples collected from different fixed surface water monitoring sites in Haiti in different months, we investigated methods to adjust for unmeasured confounding due to either of the two crossed factors site and month. In the process, we extended previous methods that adjust for unmeasured confounding due to one nesting factor (such as site, which nests the water samples from different months) to the case of two crossed factors. First, we developed a conditional pseudolikelihood estimator that eliminates fixed effects for the levels of each of the crossed factors from the estimating equation. Using the theory of U-Statistics for independent but non-identically distributed vectors, we show that our estimator is consistent and asymptotically normal, but that its variance depends on the nuisance parameters and thus cannot be easily estimated. Consequently, we apply our estimator in conjunction with a permutation test, and we investigate use of the pigeonhole bootstrap and the jackknife for constructing confidence intervals. We also incorporate our estimator into a diagnostic test for a logistic mixed model with crossed random effects and no unmeasured confounding. For comparison, we investigate between-within models extended to two crossed factors. These generalized linear mixed models include covariate means for each level of each factor in order to adjust for the unmeasured confounding. We conduct simulation studies, and we apply the methods to the Haitian data. Copyright © 2016 John Wiley & Sons, Ltd.

Model-based standardization to adjust for unmeasured cluster-level confounders with complex survey data.

Model-based standardization uses a statistical model to estimate a standardized, or unconfounded, population-averaged effect. With it, one can compare groups had the distribution of confounders been identical in both groups to that of the standard population. We develop two methods for model-based standardization with complex survey data that accommodate a categorical confounder that clusters the individual observations into a very large number of subgroups. The first method combines a random-intercept generalized linear mixed model with a conditional pseudo-likelihood estimator of the fixed effects. The second method combines a between-within generalized linear mixed model with census data on the cluster-level means of the individual-level covariates. We conduct simulation studies to compare the two approaches. We apply the two methods to the 2008 Florida Behavioral Risk Factor Surveillance System survey data to estimate standardized proportions of people who drink alcohol, within age groups, adjusting for measured individual-level and unmeasured cluster-level confounders.

Monitoring water sources for environmental reservoirs of toxigenic Vibrio cholerae O1, Haiti.

An epidemic of cholera infections was documented in Haiti for the first time in more than 100 years during October 2010. Cases have continued to occur, raising the question of whether the microorganism has established environmental reservoirs in Haiti. We monitored 14 environmental sites near the towns of Gressier and Leogane during April 2012-March 2013. Toxigenic Vibrio cholerae O1 El Tor biotype strains were isolated from 3 (1.7%) of 179 water samples; nontoxigenic O1 V. cholerae was isolated from an additional 3 samples. All samples containing V. cholerae O1 also contained non-O1 V. cholerae. V. cholerae O1 was isolated only when water temperatures were ≥31°C. Our data substantiate the presence of toxigenic V. cholerae O1 in the aquatic environment in Haiti. These isolations may reflect establishment of long-term environmental reservoirs in Haiti, which may complicate eradication of cholera from this coastal country.

Methods of estimating or accounting for neighborhood associations with health using complex survey data.

Reasons for health disparities may include neighborhood-level factors, such as availability of health services, social norms, and environmental determinants, as well as individual-level factors. Investigating health inequalities using nationally or locally representative data often requires an approach that can accommodate a complex sampling design, in which individuals have unequal probabilities of selection into the study. The goal of the present article is to review and compare methods of estimating or accounting for neighborhood influences with complex survey data. We considered 3 types of methods, each generalized for use with complex survey data: ordinary regression, conditional likelihood regression, and generalized linear mixed-model regression. The relative strengths and weaknesses of each method differ from one study to another; we provide an overview of the advantages and disadvantages of each method theoretically, in terms of the nature of the estimable associations and the plausibility of the assumptions required for validity, and also practically, via a simulation study and 2 epidemiologic data analyses. The first analysis addresses determinants of repeat mammography screening use using data from the 2005 National Health Interview Survey. The second analysis addresses disparities in preventive oral health care using data from the 2008 Florida Behavioral Risk Factor Surveillance System Survey.

Using structural-nested models to estimate the effect of cluster-level adherence on individual-level outcomes with a three-armed cluster-randomized trial.

Much attention has been paid to estimating the causal effect of adherence to a randomized protocol using instrumental variables to adjust for unmeasured confounding. Researchers tend to use the instrumental variable within one of the three main frameworks: regression with an endogenous variable, principal stratification, or structural-nested modeling. We found in our literature review that even in simple settings, causal interpretations of analyses with endogenous regressors can be ambiguous or rely on a strong assumption that can be difficult to interpret. Principal stratification and structural-nested modeling are alternative frameworks that render unambiguous causal interpretations based on assumptions that are, arguably, easier to interpret. Our interest stems from a wish to estimate the effect of cluster-level adherence on individual-level binary outcomes with a three-armed cluster-randomized trial and polytomous adherence. Principal stratification approaches to this problem are quite challenging because of the sheer number of principal strata involved. Therefore, we developed a structural-nested modeling approach and, in the process, extended the methodology to accommodate cluster-randomized trials with unequal probability of selecting individuals. Furthermore, we developed a method to implement the approach with relatively simple programming. The approach works quite well, but when the structural-nested model does not fit the data, there is no solution to the estimating equation. We investigate the performance of the approach using simulated data, and we also use the approach to estimate the effect on pupil absence of school-level adherence to a randomized water, sanitation, and hygiene intervention in western Kenya.

Adjusting for confounding by neighborhood using generalized linear mixed models and complex survey data.

When investigating health disparities, it can be of interest to explore whether adjustment for socioeconomic factors at the neighborhood level can account for, or even reverse, an unadjusted difference. Recently, we proposed new methods to adjust the effect of an individual-level covariate for confounding by unmeasured neighborhood-level covariates using complex survey data and a generalization of conditional likelihood methods. Generalized linear mixed models (GLMMs) are a popular alternative to conditional likelihood methods in many circumstances. Therefore, in the present article, we propose and investigate a new adaptation of GLMMs for complex survey data that achieves the same goal of adjusting for confounding by unmeasured neighborhood-level covariates. With the new GLMM approach, one must correctly model the expectation of the unmeasured neighborhood-level effect as a function of the individual-level covariates. We demonstrate using simulations that even if that model is correct, census data on the individual-level covariates are sometimes required for consistent estimation of the effect of the individual-level covariate. We apply the new methods to investigate disparities in recency of dental cleaning, treated as an ordinal outcome, using data from the 2008 Florida Behavioral Risk Factor Surveillance System (BRFSS) survey. We operationalize neighborhood as zip code and merge the BRFSS data with census data on ZIP Code Tabulated Areas to incorporate census data on the individual-level covariates. We compare the new results to our previous analysis, which used conditional likelihood methods. We find that the results are qualitatively similar.