Missing covariate data in clinical research: when and when not to use the missing-indicator method for analysis

Informative missingness in electronic health record systems: the curse of knowing

Electronic health records provide a potentially valuable data source of information for developing clinical prediction models. However, missing data are common in routinely collected health data and often missingness is informative. Informative missingness can be incorporated in a clinical prediction model, for example by including a separate category of a predictor variable that has missing values. The predictive performance of such a model depends on the transportability of the missing data mechanism, which may be compromised once the model is deployed in practice and the predictive value of certain variables becomes known. Using synthetic data, this phenomenon is explained and illustrated.

Relation between in utero arsenic exposure and growth during the first year of life in a New Hampshire pregnancy cohort

BACKGROUND

We have previously reported that in utero arsenic exposure is associated with increased length and other anthropometric outcomes at birth in a U.S. cohort. However, it is unknown whether these anthropometric differences persist through early life.

OBJECTIVES

We assessed in utero arsenic exposure in relation to attained anthropometry and growth trajectories through the first year of life.

METHODS

Among 760 mother-infant pairs from the New Hampshire Birth Cohort Study, we assessed in utero arsenic exposure using maternal second trimester urinary arsenic and assessed infant growth from medical records.

RESULTS

Median maternal second trimester total urinary arsenic (tAs; inorganic arsenic + monomethylarsonic acid + dimethylarsinic acid) was 3.96 μg/L (IQR: 2.02, 6.72). In adjusted linear mixed effects models, each doubling of maternal urinary tAs was associated with a 0.05 increase in length WHO Z score (95% CI: 0, 0.09) over the first year of life which corresponds to an approximately 0.12 cm increase in males and 0.13 cm increase in females at 12 months. No associations were observed between urinary tAs and attained weight, weight-for-length, or head circumference. In adjusted piecewise linear mixed effects models, each doubling of urinary tAs was associated with a 0.07 (95% CI: 0.02, 0.12) cm per month decreased length growth rate through 3.5 months with no evidence of an association thereafter. No associations were observed between urinary tAs and infant weight gain or change in weight-for-length and head circumference through one year.

CONCLUSIONS

On average, infants exposed to higher in utero arsenic attained modestly longer length during the first year, despite having slower linear growth in the first 3.5 months of life. This suggests that the previously demonstrated arsenic-associated longer length among study infants at birth persists through the first year of life. No other anthropometric associations with in utero arsenic exposure were observed across the full study population.

A Latent Class Analysis of the Social Determinants of Health Impacting Heavy Alcohol Consumption Among Women Living with HIV in Canada: The Canadian HIV Women's Sexual and Reproductive Health Cohort Study

We used longitudinal data from the 2013-2017 Canadian HIV Women's Sexual and Reproductive Health Cohort Study (N = 1422) to assess the clustered impact of social determinants of health (SDoH) on hazardous drinking. Two measures of alcohol use were defined: (i) weekly alcohol use, with > 7 drinks/week as heavy drinking, and (ii) monthly binge drinking (≥ 6 drinks at one sitting), with ≥ 1/month as frequent binging. Twelve SDoH indicators were classified using latent class analysis: no/least adversities, discrimination/stigma, economic hardship, and most SDoH adversities. Inverse-probability weighted multinomial logistic regression was used to report relative-risk ratio (RRR). Women living with HIV (WLWH) in no/least adversity class had a substantially lower likelihood of both heavy weekly alcohol use and frequent binging than those in discrimination/stigma, economic hardship, and most SDoH adversities classes, with RRR estimates ranging from 0.02 to 0.18. Findings indicate the need to address SDoH to reduce hazardous drinking among WLWH.

Interest in Research Participation Among Caregivers of Children with Neurodevelopmental Disorders

The goal of this study was to examine caregiver agreement to hear about local research opportunities by joining a clinical research registry. Data from this cross-sectional study were gathered, between 2014 and 2017, across two outpatient clinics: (1) a multidisciplinary Autism Spectrum Disorder (ASD) clinic (N = 5228) and (2) a general psychology clinic serving youth with, or at risk for, a neurodevelopmental disorder (NDD; N = 5040). Overall, more than 8 in 10 caregivers agreed to join the registry. Several child clinical characteristics, as well as racial and sociodemographic factors, were predictive of parental agreement. Findings suggest caregivers of youth with ASD and NDD are amenable to joining the local research enterprise, however further work is needed to understand why some caregivers decline.

Area-level and individual correlates of active transportation among adults in Germany: A population-based multilevel study

This study aimed at estimating the prevalence in adults of complying with the aerobic physical activity (PA) recommendation through transportation-related walking and cycling. Furthermore, potential determinants of transportation-related PA recommendation compliance were investigated. 10,872 men and 13,144 women aged 18 years or older participated in the cross-sectional 'German Health Update 2014/15 - EHIS' in Germany. Transportation-related walking and cycling were assessed using the European Health Interview Survey-Physical Activity Questionnaire. Three outcome indicators were constructed: walking, cycling, and total active transportation (≥600 metabolic equivalent, MET-min/week). Associations were analyzed using multilevel regression analysis. Forty-two percent of men and 39% of women achieved ≥600 MET-min/week with total active transportation. The corresponding percentages for walking were 27% and 28% and for cycling 17% and 13%, respectively. Higher population density, older age, lower income, higher work-related and leisure-time PA, not being obese, and better self-perceived health were positively associated with transportation-related walking and cycling and total active transportation among both men and women. The promotion of walking and cycling among inactive people has great potential to increase PA in the general adult population and to comply with PA recommendations. Several correlates of active transportation were identified which should be considered when planning public health policies and interventions.

Risk of Myocardial Infarction Among New Users of Calcium Supplements Alone or Combined With Vitamin D: A Population-Based Case-Control Study

A population‐based case‐control study was conducted to evaluate the risk of acute myocardial infarction among new users of calcium supplements either in monotherapy (CaM) or in combination with vitamin D (CaD). A total of 23,025 cases and 114,851 controls randomly sampled from the underlying cohort and matched with cases by age, sex, and index date were included. New users of CaM and CaD were categorized as current users, recent users, past users, and nonusers. We computed adjusted odds ratios (AORs) and their 95% confidence intervals (CIs) among current users as compared with nonusers through a conditional logistic regression. No increased risk was associated with CaM overall (59 cases (0.26%) and 273 controls (0.24%); AOR = 0.80; 95% CI 0.59–1.09), nor was it found in any of the conditions examined. Instead, the use of CaD was associated with a decreased risk (275 cases (1.19%) and 1,160 controls (1.45%); AOR = 0.78; 95% CI 0.67–0.90), dose and duration‐dependent, and particularly evident in patients with a high cardiovascular risk (AOR = 0.59; 95% CI 0.43–0.81).

Development and validation of a novel risk stratification algorithm for relapsed multiple myeloma

Multiple myeloma (MM) is a malignancy with varying survival outcomes and drivers of disease progression. Existing MM staging tools were developed using data from newly diagnosed patients. As patient characteristics and disease-related factors change between diagnosis and the initiation of second-line (2L) treatment, an unmet need exists for a tool that can evaluate risk of death at first relapse. We have developed a risk stratification algorithm (RSA) using data from patients with MM who were at 2L. Hazard ratios for independent predictors of overall survival (OS) were derived from a Cox models, and individual patient scores were calculated for total risk. K-adaptive partitioning for survival was used to stratify patients into groups based on their scores. Relative risk doubled with ascending risk group; median OSs for patients in group 1 (lowest risk)-4 (highest risk) were 61·6, 29·6, 14·2 and 5·9 months, respectively. Differences in OS between risk groups were significant. Similar stratification was observed when the RSA was applied to an external validation data set. In conclusion, we have developed a validated RSA that can quantify total risk, frailty risk and disease aggressiveness risk, and stratify patients with MM at 2L into groups with profoundly different survival expectations.

A Cross-Sectional Analysis of Differences in Physical Activity Levels between Stroke Belt and Non-Stroke Belt US Adults

BACKGROUND

The Stroke Belt is a region of the United States with elevated stroke incidence and prevalence of stroke risk factors. Physical inactivity is an important stroke risk factor, but little is known about whether current physical activity levels differ between Stroke Belt and non-Stroke Belt states. In this nationally representative study, we determined whether unadjusted and adjusted physical activity levels differ between the Stroke Belt region and the rest of the United States.

METHODS

Using 2017 Behavioral Risk Factor Surveillance System data, we conducted bivariate analyses to obtain unadjusted physical activity levels in Stroke Belt and non-Stroke Belt states. Logistic regressions that controlled for sociodemographic and stroke risk factors were created to estimate adjusted associations between Stroke Belt residence and physical activity.

RESULTS

A higher percentage of Stroke Belt residents were inactive (Stroke Belt: 35.3%, non-Stroke Belt: 29.4%) and failed to meet physical activity guidelines (Stroke Belt: 53.7%, non-Stroke Belt: 47.8%) compared to non-Stroke Belt residents. Stroke Belt residence was significantly associated with lower odds of meeting physical activity guidelines in a model that adjusted for sociodemographic factors only (odds ratio [OR]: 0.85, 95% confidence interval [CI]: 0.78-0.91) and one that adjusted for both sociodemographic and stroke risk factors (OR: 0.87, 95% CI: 0.81-0.93).

CONCLUSIONS

The considerably lower physical activity levels and likelihood of meeting physical activity guidelines in Stroke Belt residents compared to their non-Stroke Belt counterparts demonstrates a need for clinician attention and public health interventions to increase regular physical activity as part of a stroke reduction strategy in this region.

Consequences of Model Misspecification for Maximum Likelihood Estimation with Missing Data

Researchers are often faced with the challenge of developing statistical models with incomplete data. Exacerbating this situation is the possibility that either the researcher’s complete-data model or the model of the missing-data mechanism is misspecified. In this article, we create a formal theoretical framework for developing statistical models and detecting model misspecification in the presence of incomplete data where maximum likelihood estimates are obtained by maximizing the observable-data likelihood function when the missing-data mechanism is assumed ignorable. First, we provide sufficient regularity conditions on the researcher’s complete-data model to characterize the asymptotic behavior of maximum likelihood estimates in the simultaneous presence of both missing data and model misspecification. These results are then used to derive robust hypothesis testing methods for possibly misspecified models in the presence of Missing at Random (MAR) or Missing Not at Random (MNAR) missing data. Second, we introduce a method for the detection of model misspecification in missing data problems using recently developed Generalized Information Matrix Tests (GIMT). Third, we identify regularity conditions for the Missing Information Principle (MIP) to hold in the presence of model misspecification so as to provide useful computational covariance matrix estimation formulas. Fourth, we provide regularity conditions that ensure the observable-data expected negative log-likelihood function is convex in the presence of partially observable data when the amount of missingness is sufficiently small and the complete-data likelihood is convex. Fifth, we show that when the researcher has correctly specified a complete-data model with a convex negative likelihood function and an ignorable missing-data mechanism, then its strict local minimizer is the true parameter value for the complete-data model when the amount of missingness is sufficiently small. Our results thus provide new robust estimation, inference, and specification analysis methods for developing statistical models with incomplete data.

Clopidogrel and Low-Dose Aspirin, Alone or Together, Reduce Risk of Colorectal Cancer

BACKGROUND & AIMS

The antiplatelet effect of low-dose aspirin, via inhibition of cyclooxygenase-1, might contribute to its ability to reduce the risk of colorectal cancer (CRC). Antiplatelet agents with a different mechanism, such as clopidogrel, might have the same effects. We aimed to quantify the effects of low-dose aspirin and clopidogrel on the risk of CRC in a Mediterranean population.

METHODS

We performed a nested case-control study using a primary care database (Base de datos para la Investigación Farmacoepidemiológica en Atención Primaria) in Spain. We collected data, from 2001 through 2014, on 15,491 incident cases of CRC and 60,000 randomly selected individuals (controls), frequency-matched to cases by age, sex, and year. To estimate the association between exposure to different antiplatelet agents and the risk of colorectal cancer, we built multiple logistic regression models and computed the adjusted-odds ratios (AORs) and their respective 95% CIs.

RESULTS

Use of low-dose aspirin was associated with a reduced risk of CRC overall (AOR, 0.83; 95% CI, 0.78-0.89) and in patients receiving treatment for more than 1 year (AOR, 0.79; 95% CI, 0.73-0.85). Use of clopidogrel was associated with a decreased risk of CRC overall (AOR, 0.8; 95% CI, 0.69-0.93) and in patients receiving treatment for more than 1 year (AOR, 0.65; 95% CI, 0.55-0.78). Dual antiplatelet therapy had the same effect as either drug taken as monotherapy. No modification by sex or age was observed.

CONCLUSIONS

In a nested case-control study of a primary care database in Spain, we found clopidogrel use, alone or in combination with low-dose aspirin, to reduce the risk of CRC by 20% to 30%, a magnitude similar to that of low-dose aspirin alone. These data support the concept that inhibiting platelets is an effective strategy for prevention of CRC.

Population-based case-control study: chemoprotection of colorectal cancer with non-aspirin nonsteroidal anti-inflammatory drugs and other drugs for pain control

BACKGROUND

Inflammation and overexpression of cyclooxygenase-2 (COX-2) have been described to play a key role in the progression from nonpathologic intestinal mucosa to colorectal cancer (CRC).

AIMS

To assess the chemoprotective effect of non-aspirin nonsteroidal anti-inflammatory drugs (NA-NSAIDs) under different patterns of use in a Mediterranean population and to explore the potential effect of symptomatic slow-acting drugs for osteoarthritis (SYSADOAs; chondroitin sulfate and glucosamine) and metamizole (or dipyrone), also reported to influence COX-2 activity.

METHODS

We performed a case-control study nested in a cohort extracted from the primary care database, BIFAP. From 2001 to 2014, we included 15 491 incident cases and 60 000 random controls. To estimate the association between the drugs of interest and CRC, we built logistic regression models to compute the adjusted-odds ratios (AOR) and 95% confidence intervals (CI).

RESULTS

NA-NSAIDs use was associated with a reduced risk of CRC (AOR = 0.67; 95% CI: 0.63-0.71) and increased linearly with duration of treatment (p for trend <0.001). The effect diminished upon discontinuation but persisted statistically significant up to 1 year. All individual NA-NSAIDs examined showed a decreased risk. The concomitant use of proton-pump inhibitors (PPI) had no impact on the protective effect of NA-NSAIDs; AOR_{PPI + NSAID}  = 0.64; 0.58-0.71. SYSADOA use was associated with a reduced risk (0.79; 0.69-0.90) but disappeared after the exclusion of NSAID users during the previous 1 or 3 years (0.85; 0.70-1.04 and 1.00; 0.76-1.31 respectively). Metamizole did not show a chemoprotective effect.

CONCLUSIONS

NA-NSAID use is associated with a duration-dependent risk reduction of CRC not shared by SYSADOAs or metamizole.

Patterns of social determinants of health associated with drug use among women living with HIV in Canada: a latent class analysis

BACKGROUND AND AIMS

Identifying typologies of social determinants of health (SDoH) vulnerability influencing drug use practices among women living with HIV (WLWH) can help to address associated harms. This research aimed to explore the association of SDoH clusters with drug use among WLWH.

DESIGN

Latent class analysis (LCA) was used to identify the distinct clusters of SDoH. Inverse probability weighting (IPW) was employed to account for confounding and potential selection bias. Associations were analyzed using generalized linear model with log link and Poisson distribution, and then weighted risk ratio (RR) and 95% confidence intervals (CI) were reported.

SETTING AND PARTICIPANTS

Data from 1422 WLWH recruited at time-point 1 of the Canadian HIV Women's Sexual and Reproductive Health Cohort Study (CHIWOS, 2013-15), with 1252 participants at 18 months follow-up (time-point 2).

MEASUREMENTS

Drug use was defined as use of illicit/non-prescribed opioids/stimulants in the past 6 months. SDoH indicators included: race discrimination, gender discrimination, HIV stigma, social support, access to care, food security, income level, employment status, education, housing status and histories of recent sex work and incarceration.

FINDINGS

LCA identified four SDoH classes: no/least SDoH adversities (6.6%), discrimination/stigma (17.7%), economic hardship (30.8%) and most SDoH adversities (45.0%). Drug use was reported by 17.5% and 17.2% at time-points 1 and 2, respectively. WLWH with no/least SDoH adversities were less likely to report drug use than those in economic hardship class (weighted RR = 0.13; 95% CIs = 0.03, 0.63), discrimination/stigma class (weighted RR = 0.15; 95% CIs = 0.03, 0.78), and most SDoH adversities class (weighted RR = 0.13; 95% CIs = 0.03, 0.58).

CONCLUSIONS

Social determinants of health vulnerabilities are associated with greater likelihood of drug use, underscoring the significance of addressing interlinked social determinants and drug use through the course of HIV care and treatment.

Risk Factors for Delayed Presentation Among Patients with Musculoskeletal Injuries in Malawi

BACKGROUND

The burden of injuries is high in low-income and middle-income countries such as Malawi, where access to musculoskeletal trauma care is limited. Delayed treatment can worsen trauma-related disability. Understanding risk factors for delayed hospital presentation will assist in guiding trauma system development.

METHODS

We examined the records of 1,380 pediatric and adult patients with fractures who presented to the orthopaedic clinics of 2 urban referral hospitals and 2 rural district hospitals in Malawi. We used multivariate Poisson regression to evaluate the association between presentation to a hospital ≥2 days after the injury (delayed presentation) and 11 covariates: age, sex, education level, occupation, season of injury, day of injury, injury mechanism, injury type or extremity of injury, referral status, hospital of presentation, and estimated travel time.

RESULTS

Twenty-eight percent of pediatric patients and 34% of adult patients presented late. In the pediatric cohort, fall (relative risk [RR], 1.40 [95% confidence interval (CI), 1.02 to 1.93]), sports injuries (RR, 1.65 [95% CI, 1.09 to 2.49]), tibial or fibular injuries (RR, 1.36 [95% CI, 1.05 to 1.77]), injury over the weekend (RR, 2.30 [95% CI, 1.88 to 2.80]), estimated travel time of ≥20 minutes (RR, 1.45 [95% CI, 1.16 to 1.81]), referral from another facility (RR, 1.46 [95% CI, 1.05 to 2.02]), and presentation to Kamuzu Central Hospital, Mangochi District Hospital, or Nkhata Bay District Hospital (RR, 1.34 [95% CI, 1.07 to 1.69]) independently increased the risk of delayed presentation. In the adult cohort, fall (RR, 1.85 [95% CI, 1.38 to 2.46]), injury over the weekend (RR, 1.80 [95% CI, 1.38 to 2.36]), estimated travel time ≥20 minutes (RR, 1.36 [95% CI, 1.03 to 1.80]), and presentation to Kamuzu Central Hospital (RR, 1.74 [95% CI, 1.30 to 2.33]) independently increased the risk of delayed presentation.

CONCLUSIONS

Delayed presentation to the hospital after a musculoskeletal injury is common in Malawi. Interventions are needed to improve access to musculoskeletal trauma care, especially for pediatric patients with tibial or fibular injuries, all patients after falls, patients injured over the weekend, and patients living far from health facilities.

Impact of maternal body mass index and gestational weight gain on pregnancy complications: an individual participant data meta-analysis of European, North American and Australian cohorts

OBJECTIVE

To assess the separate and combined associations of maternal pre-pregnancy body mass index (BMI) and gestational weight gain with the risks of pregnancy complications and their population impact.

DESIGN

Individual participant data meta-analysis of 39 cohorts.

SETTING

Europe, North America, and Oceania.

POPULATION

265 270 births.

METHODS

Information on maternal pre-pregnancy BMI, gestational weight gain, and pregnancy complications was obtained. Multilevel binary logistic regression models were used.

MAIN OUTCOME MEASURES

Gestational hypertension, pre-eclampsia, gestational diabetes, preterm birth, small and large for gestational age at birth.

RESULTS

Higher maternal pre-pregnancy BMI and gestational weight gain were, across their full ranges, associated with higher risks of gestational hypertensive disorders, gestational diabetes, and large for gestational age at birth. Preterm birth risk was higher at lower and higher BMI and weight gain. Compared with normal weight mothers with medium gestational weight gain, obese mothers with high gestational weight gain had the highest risk of any pregnancy complication (odds ratio 2.51, 95% CI 2.31- 2.74). We estimated that 23.9% of any pregnancy complication was attributable to maternal overweight/obesity and 31.6% of large for gestational age infants was attributable to excessive gestational weight gain.

CONCLUSIONS

Maternal pre-pregnancy BMI and gestational weight gain are, across their full ranges, associated with risks of pregnancy complications. Obese mothers with high gestational weight gain are at the highest risk of pregnancy complications. Promoting a healthy pre-pregnancy BMI and gestational weight gain may reduce the burden of pregnancy complications and ultimately the risk of maternal and neonatal morbidity.

TWEETABLE ABSTRACT

Promoting a healthy body mass index and gestational weight gain might reduce the population burden of pregnancy complications.

Associations between parental broader autism phenotype and child autism spectrum disorder phenotype in the Study to Explore Early Development

The autism spectrum disorder phenotype varies by social and communication ability and co-occurring developmental, behavioral, and medical conditions. Etiology is also diverse, with myriad potential genetic origins and environmental risk factors. Examining the influence of parental broader autism phenotype-a set of sub-clinical characteristics of autism spectrum disorder-on child autism spectrum disorder phenotypes may help reduce heterogeneity in potential genetic predisposition for autism spectrum disorder. We assessed the associations between parental broader autism phenotype and child phenotype among children of age 30-68 months enrolled in the Study to Explore Early Development (N = 707). Child autism spectrum disorder phenotype was defined by a replication of latent classes derived from multiple developmental and behavioral measures: Mild Language Delay with Cognitive Rigidity, Mild Language and Motor Delay with Dysregulation (e.g. anxiety/depression), General Developmental Delay, and Significant Developmental Delay with Repetitive Motor Behaviors. Scores on the Social Responsiveness Scale-Adult measured parent broader autism phenotype. Broader autism phenotype in at least one parent was associated with a child having increased odds of being classified as mild language and motor delay with dysregulation compared to significant developmental delay with repetitive motor behaviors (odds ratio: 2.44; 95% confidence interval: 1.16, 5.09). Children of parents with broader autism phenotype were more likely to have a phenotype qualitatively similar to broader autism phenotype presentation; this may have implications for etiologic research.

Dietary patterns and the risk of pregnancy-associated hypertension in the Danish National Birth Cohort: a prospective longitudinal study

OBJECTIVE

To examine the association between mid-pregnancy dietary patterns and pregnancy-associated hypertension (PAH).

DESIGN

A prospective longitudinal cohort study.

SETTING

Denmark.

POPULATION

About 55 139 Danish women with single enrolments and recorded food frequency questionnaire dates with complete information on dietary intake.

METHODS

Women were eligible if they could speak Danish and were planning to carry to term. Diet was assessed using a validated semi-quantitative 360-item food frequency questionnaire and dietary patterns were derived using factor analysis.

MAIN OUTCOME MEASURES

Gestational hypertension (GH) and pre-eclampsia (PE).

RESULTS

Disease prevalence was 14% for GH (5491/39 362); 2% for PE (1168/54 778), and 0.4% for severe PE (234/55 086). Seven dietary patterns were characterised in the population, of which two were associated with PAH. The Seafood diet characterised by high consumption of fish and vegetables was inversely associated with the odds of developing GH [odds ratio (OR) 0.86; 95% CI 0.77-0.95)] and PE (OR 0.79; 95% CI 0.65-0.97). The Western diet characterised by high consumption of potatoes (including French fries), mixed meat, margarine and white bread increased the odds of developing GH (OR 1.18; 95% CI 1.05-1.33) and PE (OR 1.40; 95% CI 1.11-1.76). No association was seen with severe PE.

CONCLUSIONS

We found protective associations of Seafood diet and harmful associations of Western diet with PAH. Dietary interventions encouraging the reduction of Western diet may contribute to a decrease of PAH.

TWEETABLE ABSTRACT

Western diet increases (Seafood diet decreases) the likelihood of developing pre-eclampsia among Danish pregnant women.

A New Insight Into Missing Data in Intensive Care Unit Patient Profiles: Observational Study

Background

The data missing from patient profiles in intensive care units (ICUs) are substantial and unavoidable. However, this incompleteness is not always random or because of imperfections in the data collection process.

Objective

This study aimed to investigate the potential hidden information in data missing from electronic health records (EHRs) in an ICU and examine whether the presence or missingness of a variable itself can convey information about the patient health status.

Methods

Daily retrieval of laboratory test (LT) measurements from the Medical Information Mart for Intensive Care III database was set as our reference for defining complete patient profiles. Missingness indicators were introduced as a way of representing presence or absence of the LTs in a patient profile. Thereafter, various feature selection methods (filter and embedded feature selection methods) were used to examine the predictive power of missingness indicators. Finally, a set of well-known prediction models (logistic regression [LR], decision tree, and random forest) were used to evaluate whether the absence status itself of a variable recording can provide predictive power. We also examined the utility of missingness indicators in improving predictive performance when used with observed laboratory measurements as model input. The outcome of interest was in-hospital mortality and mortality at 30 days after ICU discharge.

Results

Regardless of mortality type or ICU day, more than 40% of the predictors selected by feature selection methods were missingness indicators. Notably, employing missingness indicators as the only predictors achieved reasonable mortality prediction on all days and for all mortality types (for instance, in 30-day mortality prediction with LR, we achieved area under the curve of the receiver operating characteristic [AUROC] of 0.6836±0.012). Including indicators with observed measurements in the prediction models also improved the AUROC; the maximum improvement was 0.0426. Indicators also improved the AUROC for Simplified Acute Physiology Score II model—a well-known ICU severity of illness score—confirming the additive information of the indicators (AUROC of 0.8045±0.0109 for 30-day mortality prediction for LR).

Conclusions

Our study demonstrated that the presence or absence of LT measurements is informative and can be considered a potential predictor of in-hospital and 30-day mortality. The comparative analysis of prediction models also showed statistically significant prediction improvement when indicators were included. Moreover, missing data might reflect the opinions of examining clinicians. Therefore, the absence of measurements can be informative in ICUs and has predictive power beyond the measured data themselves. This initial case study shows promise for more in-depth analysis of missing data and its informativeness in ICUs. Future studies are needed to generalize these results.

PROBAST: A Tool to Assess Risk of Bias and Applicability of Prediction Model Studies: Explanation and Elaboration

Prediction models in health care use predictors to estimate for an individual the probability that a condition or disease is already present (diagnostic model) or will occur in the future (prognostic model). Publications on prediction models have become more common in recent years, and competing prediction models frequently exist for the same outcome or target population. Health care providers, guideline developers, and policymakers are often unsure which model to use or recommend, and in which persons or settings. Hence, systematic reviews of these studies are increasingly demanded, required, and performed. A key part of a systematic review of prediction models is examination of risk of bias and applicability to the intended population and setting. To help reviewers with this process, the authors developed PROBAST (Prediction model Risk Of Bias ASsessment Tool) for studies developing, validating, or updating (for example, extending) prediction models, both diagnostic and prognostic. PROBAST was developed through a consensus process involving a group of experts in the field. It includes 20 signaling questions across 4 domains (participants, predictors, outcome, and analysis). This explanation and elaboration document describes the rationale for including each domain and signaling question and guides researchers, reviewers, readers, and guideline developers in how to use them to assess risk of bias and applicability concerns. All concepts are illustrated with published examples across different topics. The latest version of the PROBAST checklist, accompanying documents, and filled-in examples can be downloaded from www.probast.org.

Mid-arm muscle area and anthropometry predict low birth weight and poor pregnancy outcomes in Tanzanian women with HIV

BACKGROUND

An observational study was conducted to examine the role of maternal anthropometry, including mid-arm muscle area (MAMA) and others, as risk factors for low birth weight (LBW), small for gestational age (SGA) and preterm births in human immunodeficiency virus (HIV) infected pregnant women. HIV-positive women (N = 2369), between 12 and 32 weeks gestation were followed through delivery in Tanzania, from 2003 to 2008. Participants were women enrolled in a randomized, double-blind, placebo-controlled, clinical trial who delivered live births.

METHODS

Binomial regression analysis was used to evaluate the association of maternal nutritional indicators of MAMA, mid-upper arm circumference (MUAC), body mass index (BMI) and maternal weight with LBW, SGA and preterm in multivariate analysis.

RESULTS

Higher MAMA was associated with a 32% lower risk of LBW compared to lower measurements (RR = 0.68, 95% CI = 0.50-0.94). Similar protective associations were noted for higher BMI (RR = 0.58, 95% CI = 0.42-0.79); maternal weight (RR = 0.50, 95% CI = 0.36-0.69) and MUAC (RR = 0.62, 95% CI = 0.45-0.86). Higher MAMA was also associated with lower risk of SGA (RR = 0.78, 95% CI = 0.68-0.90) and marginally associated with preterm (RR = 0.85, 95% CI = 0.69-1.04). Beneficial associations of MUAC, BMI and maternal weight with SGA and preterm were also observed.

CONCLUSION

MAMA performs comparably to MUAC, maternal weight and BMI, as a predictor of LBW and SGA in HIV-infected women. The possible role of MAMA and other indicators in screening HIV positive women at risk of adverse pregnancy outcomes should be investigated.

Epidemiological characteristics and determinants of dengue transmission during epidemic and non-epidemic years in Fortaleza, Brazil: 2011-2015

BACKGROUND

After being eliminated during the 1950s, dengue reemerged in Brazil in the 1980s. Since then, incidence of the disease has increased, as serotypes move within and between cities. The co-circulation of multiple serotypes contributes to cycles of epidemic and interepidemic years, and a seasonal pattern of transmission is observed annually. Little is known regarding possible differences in the epidemiology of dengue under epidemic and interepidemic scenarios. This study addresses this gap and aims to assess the epidemiological characteristics and determinants of epidemic and interepidemic dengue transmission, utilizing data from the 5th largest city in Brazil (Fortaleza), at fine spatial and temporal scales.

METHODS/PRINCIPAL FINDINGS

Longitudinal models of monthly rates of confirmed dengue cases were used to estimate the differential contribution of contextual factors to dengue transmission in Fortaleza between 2011 and 2015. Models were stratified by annual climatological schedules and periods of interepidemic and epidemic transmission, controlling for social, economic, structural, entomological, and environmental factors. Results revealed distinct seasonal patterns between interepidemic and epidemic years, with persistent transmission after June in interepidemic years. Dengue was strongly associated with violence across strata, and with poverty and irregular garbage collection during periods of low transmission, but not with other indicators of public service provision or structural deprivation. Scrapyards and sites associated with tire storage were linked to incidence differentially between seasons, with the strongest associations during transitional precipitation periods. Hierarchical clustering analysis suggests that the dengue burden concentrates in the southern periphery of the city, particularly during periods of minimal transmission.

CONCLUSIONS/SIGNIFICANCE

Our findings have direct programmatic implications. Vector control operations must be sustained after June even in non-epidemic years. More specifically, scrapyards and sites associated with tires (strongly associated with incidence during periods of minimal transmission), require sustained entomological surveillance, particularly during interepidemic intervals and in the urban periphery. Intersectoral collaborations that address urban violence are critical for facilitating the regular activities of vector control agents.

Hospital Volume and Mortality following Diagnostic Bronchoscopy in Lung Cancer Patients: Data from a National Inpatient Database in Japan

BACKGROUND

Recent advances in bronchoscopy utilizing endobronchial ultrasound (EBUS) as well as lung cancer therapy may have driven physicians to perform diagnostic bronchoscopy (DB) for high-risk patients.

OBJECTIVES

The aim of this study was to clarify the relationship between hospital volume (HV) and outcomes of DB.

METHODS

We collected data on inpatients with lung cancer who underwent DB from July 2010 to March 31, 2014. The annual HV of DB was classified as "very low" (≤50 cases/year), "low" (51-100 cases/year), "high" (101-300 cases/year), or "very high" (> 300 cases/year). The primary outcome was all-cause 7-day mortality after DB. Multivariable logistic regression fitted with a generalized estimation equation was performed to evaluate the association between HV and all-cause 7-day mortality after DB, adjusted for patient background factors.

RESULTS

We identified a total of 77,755 eligible patients in 954 hospitals. All-cause 7-day mortality was 0.5%. Compared with the low-volume group, 7-day mortality was significantly lower in the high-volume group (odds ratio [OR] = 0.69, 95% confidence interval [CI]: 0.52-0.92, p = 0.010), and a similar trend was shown in the very-high-volume group (OR = 0.67; 95% CI: 0.43-1.05, p = 0.080). Radial EBUS with the guide sheath method and EBUS-guided transbronchial needle aspiration showed a significantly lower 7-day mortality.

CONCLUSIONS

All-cause 7-day mortality was inversely associated with HV. The risk of DB in patients with lung cancer should be recognized, and the exploitation of EBUS may help reduce mortality after DB.

A comparison of different methods to handle missing data in the context of propensity score analysis

Propensity score analysis is a popular method to control for confounding in observational studies. A challenge in propensity methods is missing values in confounders. Several strategies for handling missing values exist, but guidance in choosing the best method is needed. In this simulation study, we compared four strategies of handling missing covariate values in propensity matching and propensity weighting. These methods include: complete case analysis, missing indicator method, multiple imputation and combining multiple imputation and missing indicator method. Concurrently, we aimed to provide guidance in choosing the optimal strategy. Simulated scenarios varied regarding missing mechanism, presence of effect modification or unmeasured confounding. Additionally, we demonstrated how missingness graphs help clarifying the missing structure. When no effect modification existed, complete case analysis yielded valid causal treatment effects even when data were missing not at random. In some situations, complete case analysis was also able to partially correct for unmeasured confounding. Multiple imputation worked well if the data were missing (completely) at random, and if the imputation model was correctly specified. In the presence of effect modification, more complex imputation models than default options of commonly used statistical software were required. Multiple imputation may fail when data are missing not at random. Here, combining multiple imputation and the missing indicator method reduced the bias as the missing indicator variable can be a proxy for unobserved confounding. The optimal way to handle missing values in covariates of propensity score models depends on the missing data structure and the presence of effect modification. When effect modification is present, default settings of imputation methods may yield biased results even if data are missing at random.

Situational expectations mediate the effect of global beliefs on depressive symptoms: a 1-year prospective study using a student sample

PURPOSE OF REVIEW

Recent research suggests that expectations might be a particularly important subgroup of cognitions in major depression. The present study aimed to further specify the role of situation-specific dysfunctional expectations (SDEs) in the development of depressive symptoms. It was hypothesized that the effects of intermediate beliefs and dispositional optimism on depressive symptoms would be mediated via SDEs.

RECENT FINDINGS

We examined 125 students (age M = 22.05, SD = 4.00, 74.6% women) using a longitudinal design with a 1-year follow-up. Questionnaires used in the study included the Depressive Expectations Scale, Dysfunctional Attitudes Scale, Life Orientation Test and Beck's Depression Inventory. When considering SDEs at baseline as mediator variable, neither intermediate beliefs nor dispositional optimism had significant direct effects on depressive symptoms. Instead, their effects were fully mediated via SDEs. When considering SDEs at the follow-up as the mediator variable, SDEs partly mediated the effects of intermediate beliefs and dispositional optimism on depressive symptoms.

SUMMARY

The study provides further evidence for a cognitive mediational model, suggesting that SDEs mediate the effects of rather global beliefs (such as intermediate beliefs and dispositional optimism) on depressive symptoms. Due to their high level of situational specificity, SDEs might be an effective target for cognitive-behavioural interventions.

Biomarkers of Systemic Inflammation and Growth in Early Infancy are Associated with Stunting in Young Tanzanian Children

Stunting can afflict up to one-third of children in resource-constrained countries. We hypothesized that low-grade systemic inflammation (defined as elevations in serum C-reactive protein or alpha-1-acid glycoprotein) in infancy suppresses the growth hormone–insulin-like growth factor (IGF) axis and is associated with subsequent stunting. Blood samples of 590 children from periurban Dar es Salaam, Tanzania, were obtained at 6 weeks and 6 months of age as part of a randomized controlled trial. Primary outcomes were stunting, underweight, and wasting (defined as length-for-age, weight-for-age and weight-for-length z-scores < −2) between randomization and endline (18 months after randomization). Cox proportional hazards models were constructed to estimate hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) of time to first stunting, underweight, and wasting as outcomes, with measures of systemic inflammation, insulin-like growth factor-1 (IGF-1) and insulin-like growth factor binding protein-3 (IGFBP-3) as exposures, adjusting for numerous demographic and clinical variables. The incidences of subsequent stunting, underweight, and wasting were 26%, 20%, and 18%, respectively. In multivariate analyses, systemic inflammation at 6 weeks of age was significantly associated with stunting (HR: 2.14, 95% CI: 1.23, 3.72; p = 0.002). Children with higher levels of IGF-1 at 6 weeks were less likely to become stunted (HR: 0.58, 95% CI: 0.37, 0.93; p for trend = 0.019); a similar trend was noted in children with higher levels of IGF-1 at 6 months of age (HR: 0.50, 95% CI: 0.22, 1.12; p for trend = 0.07). Systemic inflammation occurs as early as 6 weeks of age and is associated with the risk of future stunting among Tanzanian children.

Association of Broad-Based Genomic Sequencing With Survival Among Patients With Advanced Non-Small Cell Lung Cancer in the Community Oncology Setting

IMPORTANCE

Broad-based genomic sequencing is being used more frequently for patients with advanced non-small cell lung cancer (NSCLC). However, little is known about the association between broad-based genomic sequencing and treatment selection or survival among patients with advanced NSCLC in a community oncology setting.

OBJECTIVE

To compare clinical outcomes between patients with advanced NSCLC who received broad-based genomic sequencing vs a control group of patients who received routine testing for EGFR mutations and/or ALK rearrangements alone.

DESIGN, SETTING, AND PARTICIPANTS

Retrospective cohort study of patients with chart-confirmed advanced NSCLC between January 1, 2011, and July 31, 2016, and who received care at 1 of 191 oncology practices across the United States using the Flatiron Health Database. Patients were diagnosed with stage IIIB/IV or unresectable nonsquamous NSCLC who received at least 1 line of antineoplastic treatment.

EXPOSURES

Receipt of either broad-based genomic sequencing or routine testing (EGFR and/or ALK only). Broad-based genomic sequencing included any multigene panel sequencing assay examining more than 30 genes prior to third-line treatment.

MAIN OUTCOMES AND MEASURES

Primary outcomes were 12-month mortality and overall survival from the start of first-line treatment. Secondary outcomes included frequency of genetic alterations and treatments received.

RESULTS

Among 5688 individuals with advanced NSCLC (median age, 67 years [interquartile range, 41-85], 63.6% white, 80% with a history of smoking); 875 (15.4%) received broad-based genomic sequencing and 4813 (84.6%) received routine testing. Among patients who received broad-based genomic sequencing, 4.5% received targeted treatment based on testing results, 9.8% received routine EGFR/ALK targeted treatment, and 85.1% received no targeted treatment. Unadjusted mortality rates at 12 months were 49.2% for patients undergoing broad-based genomic sequencing and 35.9% for patients undergoing routine testing. Using an instrumental variable analysis, there was no significant association between broad-based genomic sequencing and 12-month mortality (predicted probability of death at 12 months, 41.1% for broad-based genomic sequencing vs 44.4% for routine testing; difference -3.6% [95% CI, -18.4% to 11.1%]; P = .63). The results were consistent in the propensity score-matched survival analysis (42.0% vs 45.1%; hazard ratio, 0.92 [95% CI, 0.73 to 1.11]; P = .40) vs unmatched cohort (hazard ratio, 0.69 [95% CI, 0.62 to 0.77]; log-rank P < .001).

CONCLUSIONS AND RELEVANCE

Among patients with advanced non-small cell lung cancer receiving care in the community oncology setting, broad-based genomic sequencing directly informed treatment in a minority of patients and was not independently associated with better survival.

Risk Factors for Institutionalization After Traumatic Brain Injury Inpatient Rehabilitation

OBJECTIVE

To create a profile of individuals with traumatic brain injury (TBI) who received inpatient rehabilitation and were discharged to an institutional setting using characteristics measured at rehabilitation discharge.

METHODS

The Traumatic Brain Injury Model Systems National Database is a prospective, multicenter, longitudinal database for people with moderate to severe TBI. We analyzed data for participants enrolled from January 2002 to June 2012 who had lived in a private residence before TBI. This cross-sectional study used logistic regression analyses to identify sociodemographic factors, lengths of stay, and cognitive and physical functioning levels that differentiated patients discharged to institutional versus private settings.

RESULTS

Older age, living alone before TBI, and lower levels of function at rehabilitation discharge (independence in locomotion, bladder management, comprehension, and social interaction) were significantly associated with higher institutionalization rates and provided the best models identifying factors associated with institutionalization. Institutionalization was also associated with decreased independence in bed-chair-wheelchair transfers and increased duration of posttraumatic amnesia.

CONCLUSIONS

Individuals institutionalized after inpatient rehabilitation for TBI were older, lived alone before injury, had longer posttraumatic amnesia durations, and were less independent in specific functional characteristics. Research evaluating the effect of increasing postdischarge support and improving treatment effectiveness in these functional areas is recommended.

A mean score method for sensitivity analysis to departures from the missing at random assumption in randomised trials

Most analyses of randomised trials with incomplete outcomes make untestable assumptions and should therefore be subjected to sensitivity analyses. However, methods for sensitivity analyses are not widely used. We propose a mean score approach for exploring global sensitivity to departures from missing at random or other assumptions about incomplete outcome data in a randomised trial. We assume a single outcome analysed under a generalised linear model. One or more sensitivity parameters, specified by the user, measure the degree of departure from missing at random in a pattern mixture model. Advantages of our method are that its sensitivity parameters are relatively easy to interpret and so can be elicited from subject matter experts; it is fast and non-stochastic; and its point estimate, standard error and confidence interval agree perfectly with standard methods when particular values of the sensitivity parameters make those standard methods appropriate. We illustrate the method using data from a mental health trial.

Effect of MRI on preterm infants and their families: a randomised trial with nested diagnostic and economic evaluation

BACKGROUND

We tested the hypothesis that routine MRI would improve the care and well-being of preterm infants and their families.

DESIGN

Parallel-group randomised trial (1.1 allocation; intention-to-treat) with nested diagnostic and cost evaluations (EudraCT 2009-011602-42).

SETTING

Participants from 14 London hospitals, imaged at a single centre.

PATIENTS

511 infants born before 33 weeks gestation underwent both MRI and ultrasound around term. 255 were randomly allocated (siblings together) to receive only MRI results and 255 only ultrasound from a paediatrician unaware of unallocated results; one withdrew before allocation.

MAIN OUTCOME MEASURES

Maternal anxiety, measured by the State-Trait Anxiety inventory (STAI) assessed in 206/214 mothers receiving MRI and 217/220 receiving ultrasound. Secondary outcomes included: prediction of neurodevelopment, health-related costs and quality of life.

RESULTS

After MRI, STAI fell from 36.81 (95% CI 35.18 to 38.44) to 32.77 (95% CI 31.54 to 34.01), 31.87 (95% CI 30.63 to 33.12) and 31.82 (95% CI 30.65 to 33.00) at 14 days, 12 and 20 months, respectively. STAI fell less after ultrasound: from 37.59 (95% CI 36.00 to 39.18) to 33.97 (95% CI 32.78 to 35.17), 33.43 (95% CI 32.22 to 34.63) and 33.63 (95% CI 32.49 to 34.77), p=0.02. There were no differences in health-related quality of life. MRI predicted moderate or severe functional motor impairment at 20 months slightly better than ultrasound (area under the receiver operator characteristic curve (CI) 0.74; 0.66 to 0.83 vs 0.64; 0.56 to 0.72, p=0.01) but cost £315 (CI £295-£336) more per infant.

CONCLUSIONS

MRI increased costs and provided only modest benefits.

TRIAL REGISTRATION

ClinicalTrials.gov NCT01049594 https://clinicaltrials.gov/ct2/show/NCT01049594. EudraCT: EudraCT: 2009-011602-42 (https://www.clinicaltrialsregister.eu/).

Risk of prostate cancer across different racial/ethnic groups in men with diabetes: a retrospective cohort study

AIM

To examine the associations between prostate cancer, diabetes and race/ethnicity.

METHODS

Using administrative data from British Columbia, Canada for the period 1994 to 2012, we identified men aged ≥50 years with and without diabetes. Validated surname algorithms identified men as Chinese, Indian or of other race/ethnicity. Multivariable Cox regression was used to estimate adjusted risks of prostate cancer according to diabetes status and race/ethnicity.

RESULTS

Our cohort of 160 566 men had a mean (sd) age of 64.7 (9.4) years and a median of 9 years' follow-up. The incidence rates of prostate cancer among those with and without diabetes were 177.4 (171.7-183.4) and 216.0 (209.7-222.5) per 1000 person-years, respectively. The incidence among Chinese men was 120.9 (109.2-133.1), among Indian men it was 144.1 (122.8-169.0) and in men of other ethnicity it was 204.8 (200.2-209.5). Diabetes was independently associated with a lower risk of prostate cancer (adjusted hazard ratio 0.82, 95% CI 0.78-0.86), as was Chinese (adjusted hazard ratio 0.54, 95% CI 0.46,0.63) and Indian (adjusted hazard ratio 0.66, 95% CI 0.49,0.89) race/ethnicity; however, there was no statistically significant interaction between diabetes status and race/ethnicity (all P>0.1).

CONCLUSION

Diabetes and Chinese and Indian race/ethnicity were each independently associated with a lower risk of prostate cancer.

Parental Socioeconomic Status and Weight Faltering in Infants in Japan

Background: Previous studies in the UK and Denmark found no significant association between low socioeconomic status (SES) and weight faltering. However, to our knowledge, there are no studies from other developed countries. We examined the association between parental SES and weight faltering in infants up to 1.5 years of age, and investigated whether the inequalities changed between 2001 and 2010 in Japan. Methods: We used data from two Japanese population-based birth cohorts started in 2001 (n = 34,594) and 2010 (n = 21,189). Parental SES was assessed as household income and parental education when the infant was 6 months old. Weight faltering was defined as the slowest weight gaining in 5% of all children in each cohort. Logistic regression analyses were conducted with adjustment for covariates. The relative index of inequality was used to assess relative impact of parental SES on weight faltering. Results: Infants in the lowest quartile of household income were 1.29 (95% confidence interval [CI]: 1.10, 1.52) and 1.27 (95% CI: 1.03, 1.56) times more likely to experience weight faltering than those in the highest income quartile both in the 2001 and 2010 cohorts, respectively. The relative index of inequality for household income was 1.66 (95% CI: 1.36, 1.96) in 2001 and 1.86 (95% CI: 1.42, 2.31) in 2010. Conclusions: Infants from lower income families have a greater risk of weight faltering in Japan. Additionally, the income-related inequalities in weight faltering did not change between the two cohorts. Social policies to address maldistribution of weight faltering due to household income are needed.

Associations of anxiety sensitivity and emotional symptoms with the subjective effects of alcohol, cigarettes, and cannabis in adolescents

Maladaptive emotional traits (anxiety sensitivity [AS], fear of anxiety-related sensations and consequences) and symptoms (major depressive disorder [MDD] and generalized anxiety disorder [GAD] symptoms) could play a role in altering sensitivity to the subjective effects of drugs of abuse in adolescents. Data were drawn from a longitudinal study of high school students in Los Angeles, CA, USA who completed surveys and reported past six-month use of alcohol (n=1054), cigarettes (n=297), or cannabis (n=706). At each of the four semi-annual waves during mid-adolescence (14-16years old), students reported positive and negative subjective drug effects experienced in the prior six-months. Controlling for covariates and the simultaneous covariance across the three domains of emotional dysfunction, AS was associated with more positive and negative cannabis effects (βs=0.09-0.16, ps<0.05), and MDD symptoms were associated with fewer negative cigarette effects (β=-0.13, p=0.04) and more negative cannabis effects (β=0.10, p=0.004). The acceleration of positive alcohol and cannabis effects over time was slower among adolescents with higher baseline MDD (MDD×time: β=-0.04, p=0.044) and GAD (GAD×time: β=-0.05, p=0.03) symptoms, respectively. These findings suggest that emotional dysfunction factors show differential and overlapping effects on subjective drug effects, which may vary across time. Future research should investigate emotional dysfunctions and subjective drug effects in relation to substance use across adolescence and emerging adulthood.

Association of Radiotherapy Boost for Ductal Carcinoma In Situ With Local Control After Whole-Breast Radiotherapy

Importance

The use of a radiotherapy (RT) boost to the tumor bed after whole-breast RT (WBRT) for ductal carcinoma in situ (DCIS) is largely extrapolated from invasive cancer data, but robust evidence specific to DCIS is lacking.

Objective

To compare ipsilateral breast tumor recurrence (IBTR) in women with DCIS treated with vs without the RT boost after breast-conserving surgery and WBRT.

Design, Setting, and Participants

This retrospective analysis pooled deidentified patient-level data from 10 academic institutions in the United States, Canada, and France from January 1, 1980, through December 31, 2010. All patients had newly diagnosed pure DCIS (no microinvasion), underwent breast-conserving surgery, and received WBRT with or without the boost with a minimum of 5 years of follow-up required for inclusion in the analysis. Given the limited events after WBRT, an a priori power analysis was conducted to estimate the DCIS sample size needed to detect the anticipated benefit of the boost. Data were uniformly recoded at the host institution and underwent primary and secondary reviews before analysis. Sample size calculations (ratio of patients who received the boost dose to those who did not, 2:1; α = .05; power = 80%) estimated that 2982 cases were needed to detect a difference of at least 3%. The final analysis included 4131 patients (2661 in the boost group and 1470 in the no-boost group) with a median follow-up of 9 years and media boost dose of 14 Gy. Data were collected from July 2011 through February 2014 and analyzed from March 2014 through August 2015.

Interventions

Radiotherapy boost vs no boost.

Main Outcomes and Measures

Ipsilateral breast tumor recurrence.

Results

The analysis included 4131 patients (median [SD] age, 56.1 [10.9] years; range, 24-88 years). Patients with positive margins, unknown estrogen receptor status, and comedo necrosis were more likely to have received an RT boost. For the entire cohort, the boost was significantly associated with lower IBTR (hazard ratio [HR], 0.73; 95% CI, 0.57-0.94; P = .01) and with IBTR-free survival (boost vs no-boost groups) of 97.1% (95% CI, 0.96-0.98) vs 96.3% (95% CI, 0.95-0.97) at 5 years, 94.1% (95% CI, 0.93-0.95) vs 92.5% (95% CI, 0.91-0.94) at 10 years, and 91.6% (95% CI, 0.90-0.93) vs 88.0% (95% CI, 0.85-0.91) at 15 years. On multivariable analysis accounting for confounding factors, the boost remained significantly associated with reduced IBTR (HR compared with no boost, 0.68; 95% CI, 0.50-0.91; P = .01) independent of age and tamoxifen citrate use.

Conclusions and Relevance

This patient-level analysis suggests that the RT boost confers a statistically significant benefit in decreasing IBTR across all DCIS age groups, similar to that seen in patients with invasive breast cancer. These findings suggest that a DCIS RT boost to the tumor bed could be considered to provide an added incremental benefit in decreasing IBTR after a shared discussion between the patient and her radiation oncologist.

Development and validation of clinical prediction models for mortality, functional outcome and cognitive impairment after stroke: a study protocol

INTRODUCTION

Stroke is a leading cause of adult disability and death worldwide. The neurological impairments associated with stroke prevent patients from performing basic daily activities and have enormous impact on families and caregivers. Practical and accurate tools to assist in predicting outcome after stroke at patient level can provide significant aid for patient management. Furthermore, prediction models of this kind can be useful for clinical research, health economics, policymaking and clinical decision support.

METHODS

2869 patients with first-ever stroke from South London Stroke Register (SLSR) (1995-2004) will be included in the development cohort. We will use information captured after baseline to construct multilevel models and a Cox proportional hazard model to predict cognitive impairment, functional outcome and mortality up to 5 years after stroke. Repeated random subsampling validation (Monte Carlo cross-validation) will be evaluated in model development. Data from participants recruited to the stroke register (2005-2014) will be used for temporal validation of the models. Data from participants recruited to the Dijon Stroke Register (1985-2015) will be used for external validation. Discrimination, calibration and clinical utility of the models will be presented.

ETHICS

Patients, or for patients who cannot consent their relatives, gave written informed consent to participate in stroke-related studies within the SLSR. The SLSR design was approved by the ethics committees of Guy's and St Thomas' NHS Foundation Trust, Kings College Hospital, Queens Square and Westminster Hospitals (London). The Dijon Stroke Registry was approved by the Comité National des Registres and the InVS and has authorisation of the Commission Nationale de l'Informatique et des Libertés.

Quality of life among perinatally HIV-affected and HIV-unaffected school-aged and adolescent Ugandan children: a multi-dimensional assessment of wellbeing in the post-HAART era

OBJECTIVE

To examine quality of life (QOL) in perinatally HIV-infected (PHIV) or HIV-exposed uninfected (PHEU) vs. healthy HIV-unexposed uninfected (HUU) children during school-age/adolescence.

METHODS

PHIV infection was diagnosed via DNA PCR. Current HIV status was confirmed by HIV rapid diagnostic test. Three HIV groups were defined: PHIV, PHEU, and HUU. QOL was assessed with proxy and self-report versions of the PedsQL™ 4.0 instrument at 6-18 years of age. QOL scores ranged from zero (least QOL) to 100 (highest QOL) in the following dimensions: combined QOL inventory (CQOLI), multi-dimensional vigor (MDV), general wellbeing (GWB), present functioning, and general cognitive functioning (CF). Multivariable linear regression models estimated HIV-related percent differences (β) in QOL scores and 95% confidence intervals (CI).

FINDINGS

Compared to HUU CQOLI deficits ranged from 6.5 to 9.2% (95% CI -15.4, -1.6), GWB deficit ranged from 6.5 to 10.5% (95% CI -16.0, -1.3), MDV deficit ranged from 6.8 to 11.6% (95% CI -14.5, 0.9), and CF deficit ranged from 9.7 to 13.1% for PHIV children. QOL deficits of similar magnitude and direction in most domains were observed for PHIV compared to PHEU. However, self-reported indicators of GWB (β = -3.5; 95% CI -9.0, 2.0) and present functioning (β = 4.0; 95% CI -4.6, 12.5) were similar for PHIV compared to PHEU. QOL scores were generally similar for PHEU compared to HUU.

CONCLUSION

PHEU and HUU had similar QOL profile but PHIV predicted sustained deficits in multiple QOL domains. PHIV and PHEU children were similar with respect to general wellbeing and present functioning. Psychosocial and scholastic interventions in combination with HIV care are likely to improve QOL in PHIV.

Influence of pre-pregnancy leisure time physical activity on gestational and postpartum weight gain and birth weight - a cohort study

In order to examine the association between pre-pregnancy leisure time physical activities and gestational weight gain, postpartum weight gain and birth weight, we analysed prospectively collected data from 1827 women with singleton term pregnancies. Women were categorised in groups of sedentary women, light exercisers, moderate exercisers and competitive athletes. The results showed that sedentary women on average gained 14.1 kg during pregnancy, whereas light exercisers gained 13.7 kg, moderate exercisers gained 14.3 kg and competitive athletes 16.1 kg. Competitive athletes had an increased risk of having a gestational weight gain above Institute of Medicine (IOM) recommendations with an odds ratio of 2.60 (1.32-5.15) compared to light exercisers. However, birth weight and one year postpartum weight was similar for all four groups. Thus, although competitive athletes gain more weight than recommended during pregnancy, this may not affect birth weight or postpartum weight. Impact statement What is already known on this subjectPrevious studies have found that increased pre-pregnancy physical activity is associated with lower gestational weight gain during the last trimester, but showed no association between the pre-pregnancy level of physical activity and mean birth weight. What the results of this study addWe found that women classified as competitive exercisers had a 2.6-fold increased risk of gaining more weight than recommended compared to light exercisers. Nearly 6 out of 10 women among the competitive exercisers gained more weight than recommended by IOM. Surprisingly, this did not appear to increase birth weight or post-partum weight gain, but other adverse effects cannot be excluded. What the implications are of these findings for clinical practice and/or further researchIn the clinical practice it may be relevant to focus on and advise pre-pregnancy competitive exercisers in order to prevent excessive gestational weight gain.

The Effect of Partly Missing Covariates on Statistical Power in Randomized Controlled Trials With Discrete-Time Survival Endpoints

Abstract. In randomized controlled trials (RCTs), a common strategy to increase power to detect a treatment effect is adjustment for baseline covariates. However, adjustment with partly missing covariates, where complete cases are only used, is inefficient. We consider different alternatives in trials with discrete-time survival data, where subjects are measured in discrete-time intervals while they may experience an event at any point in time. The results of a Monte Carlo simulation study, as well as a case study of randomized trials in smokers with attention deficit hyperactivity disorder (ADHD), indicated that single and multiple imputation methods outperform the other methods and increase precision in estimating the treatment effect. Missing indicator method, which uses a dummy variable in the statistical model to indicate whether the value for that variable is missing and sets the same value to all missing values, is comparable to imputation methods. Nevertheless, the power level to detect the treatment effect based on missing indicator method is marginally lower than the imputation methods, particularly when the missingness depends on the outcome. In conclusion, it appears that imputation of partly missing (baseline) covariates should be preferred in the analysis of discrete-time survival data.

Mortality and morbidity after spinal surgery in patients with Parkinson's disease: a retrospective matched-pair cohort study

BACKGROUND CONTEXT

There is a lack of information about postoperative outcomes and related risk factors associated with spinal surgery in patients with Parkinson's disease (PD).

PURPOSE

This study aimed to investigate the postoperative morbidity and mortality associated with spinal surgery for patients with PD, and the risk factors for poor outcomes.

STUDY DESIGN

This is a retrospective matched-pair cohort study.

PATIENT SAMPLE

Data of patients who underwent elective spinal surgery between July 2010 and March 2013 were extracted from the Diagnosis Procedure Combination database, a nationwide inpatient database in Japan.

OUTCOME MEASURES

In-hospital mortality and occurrence of postoperative complications.

METHODS

For each patient with PD, we randomly selected up to four age- and sex-matched controls in the same hospital in the same year. The differences in in-hospital mortality and occurrence of postoperative complications were compared between patients with PD and controls. A multivariable logistic regression model fitted with a generalized estimation equation was used to identify significant predictors of major complications (surgical site infection, sepsis, pulmonary embolism, respiratory complications, cardiac events, stroke, and renal failure). Multiple imputation was used for missing data.

RESULTS

Among 154,278 patients undergoing spinal surgery, 1,423 patients with PD and 5,498 matched controls were identified. Crude in-hospital mortality was higher in patients with PD than in controls (0.8% vs. 0.3%, respectively). The crude proportion of major complications was also higher in patients with PD (9.8% vs. 5.1% in controls). Postoperative delirium was more common in patients with PD (30.3%) than in controls (4.3%). Parkinson's disease was a significant predictor of major postoperative complications, even after adjusting for other risk factors (odds ratio, 1.74; 95% confidence intervals, 1.37-2.22; p<.001).

CONCLUSIONS

Patients with PD had a significantly increased risk of postoperative complications following spinal surgery. Postoperative delirium was the most frequently observed complication.

Leveraging electronic health records for predictive modeling of post-surgical complications

Hospital-specific electronic health record systems are used to inform clinical practice about best practices and quality improvements. Many surgical centers have developed deterministic clinical decision rules to discover adverse events (e.g. postoperative complications) using electronic health record data. However, these data provide opportunities to use probabilistic methods for early prediction of adverse health events, which may be more informative than deterministic algorithms. Electronic health record data from a set of 9598 colorectal surgery cases from 2010 to 2014 were used to predict the occurrence of selected complications including surgical site infection, ileus, and bleeding. Consistent with previous studies, we find a high rate of missing values for both covariates and complication information (4-90%). Several machine learning classification methods are trained on an 80% random sample of cases and tested on a remaining holdout set. Predictive performance varies by complication, although an area under the receiver operating characteristic curve as high as 0.86 on testing data was achieved for bleeding complications, and accuracy for all complications compares favorably to existing clinical decision rules. Our results confirm that electronic health records provide opportunities for improved risk prediction of surgical complications; however, consideration of data quality and consistency standards is an important step in predictive modeling with such data.

A psychological intervention to promote acceptance and adherence to non-invasive ventilation in people with chronic obstructive pulmonary disease: study protocol of a randomised controlled trial

BACKGROUND

People with chronic obstructive pulmonary disease (COPD) sometimes experience anxiety, depression and comorbid cognitive deficits. Rather than being merely a consequence of symptom-related physical impairments these additional problems may be part of the clinical course of the condition. The relationship between the physical and psychological aspects of the condition is illustrated by the patterns of use of non-invasive ventilation (NIV); NIV is often rejected or used inappropriately, resulting in clinical deterioration and an increase in health care costs. The study aims to analyse the effects of psychological support on the acceptance of, and adherence to, NIV. The primary outcome will be a latent variable related to indices of use of NIV equipment and adherence to treatment regime; while survival rates and psychological variables will constitute the secondary outcomes.

METHODS

A two-arm randomised controlled trial will be conducted. We aim to recruit 150 COPD patients for whom NIV is indicated. The experimental group will receive a brief course of psychological support that will include counselling, relaxation and mindfulness-based exercises. In some cases, it will also include neuropsychological rehabilitation exercises. Support will be delivered via four to eight meetings at the HD Respiratory Rehabilitation Unit, at home or via telemedicine. Controls will receive standard care and watch educational videos related to the management of their disease.

DISCUSSION

This investigation will gain insight about the role of a psychological intervention as part of a treatment plan during the process of adaptation to NIV in COPD patients.

TRIAL REGISTRATION

ClinicalTrials.gov, ID: NCT02499653 . Registered on 14 July 2015.

Should multiple imputation be the method of choice for handling missing data in randomized trials?

The use of multiple imputation has increased markedly in recent years, and journal reviewers may expect to see multiple imputation used to handle missing data. However in randomized trials, where treatment group is always observed and independent of baseline covariates, other approaches may be preferable. Using data simulation we evaluated multiple imputation, performed both overall and separately by randomized group, across a range of commonly encountered scenarios. We considered both missing outcome and missing baseline data, with missing outcome data induced under missing at random mechanisms. Provided the analysis model was correctly specified, multiple imputation produced unbiased treatment effect estimates, but alternative unbiased approaches were often more efficient. When the analysis model overlooked an interaction effect involving randomized group, multiple imputation produced biased estimates of the average treatment effect when applied to missing outcome data, unless imputation was performed separately by randomized group. Based on these results, we conclude that multiple imputation should not be seen as the only acceptable way to handle missing data in randomized trials. In settings where multiple imputation is adopted, we recommend that imputation is carried out separately by randomized group.

Physical Activity, Diet Quality, and Mortality among Community-Dwelling Prefrail and Frail Older Adults

It is unknown if physical activity and diet quality are associated with the risk of poor outcomes, such as mortality, among prefrail and frail older adults. This was a population-based cohort study among 1487 prefrail and frail older-adults from the Third National Health and Nutrition Survey. Compared to participants who were sedentary (0 bouts of physical activity per week), those who were physically inactive (1-4 bouts of physical activity per week) were 24% less likely to die [HR: 0.76 (95% CI: 0.58-0.98)], and those who were physically active (≥5 bouts of physical activity per week) were 34% less likely to die [HR: 0.66 (95% CI: 0.51-0.86); Ptrend = 0.002]. Compared to participants with poor diet quality, those with fair diet quality were 26% less likely to die [HR: 0.74 (95% CI: 0.52-0.98)], and those with good diet quality were 33% less likely to die [HR: 0.67 (95% CI: 0.55-1.00); Ptrend = 0.050]. There was a synergistic interaction between physical activity and diet quality on the risk of mortality (Pinteraction = 0.058). Participation in physical activity and consumption of a healthy diet is associated with a lower risk of mortality among prefrail and frail older adults.

The Interplay Between Continuity of Care, Multimorbidity, and Adverse Events in Patients With Diabetes

OBJECTIVES

To evaluate the impact of continuity of care and multimorbidity on health outcomes in patients with diabetes.

RESEARCH DESIGN

Using a US claims database of insured patients, we identified those with incident diabetes between 2004 and 2008 and followed them until death, disenrollment, or December 31, 2010. Continuity of care was defined using Breslau's Usual Provider of Continuity (UPC; proportion of visits to the usual or predominant provider within 2 y of diabetes diagnosis). Multivariable logistic regression was used to determine the association between UPC in the first 2 years after diabetes diagnosis and subsequent 1-year composite primary outcome of all-cause hospitalization or death in year 3 in patients with/without multimorbidity.

RESULTS

Of the 285,231 patients with incident diabetes, 74% had multimorbidity; their average age was 53 years (SD=10.5) and 49% were female. A total of 77,270 (27%) individuals had a mean UPC≥75% in the first 2 years. During year 3 of follow-up, 33,632 (12%) patients died or were hospitalized for any cause. Greater continuity of care (UPC≥75%) was associated with reduced risk of subsequent death or hospitalization [7.2% vs. 13.5%; adjusted odds ratio (aOR)=0.72; 95% CI, 0.70-0.75]. Although multimorbidity was independently associated with an increased risk of our primary composite endpoint (13.4% vs. 7.2%; aOR=1.26; 95% CI, 1.21-1.30), the association between greater continuity and better outcomes was similar in those with multimorbidity (aOR=0.71; 95% CI, 0.69-0.71) as in those without (aOR=0.75; 95% CI, 0.71-0.80).

CONCLUSIONS

In patients with incident diabetes, greater continuity of care is associated with improved outcomes, irrespective of whether or not they have multimorbidity.

Birth by Caesarean Section and the Risk of Adult Psychosis: A Population-Based Cohort Study

Despite the biological plausibility of an association between obstetric mode of delivery and psychosis in later life, studies to date have been inconclusive. We assessed the association between mode of delivery and later onset of psychosis in the offspring. A population-based cohort including data from the Swedish National Registers was used. All singleton live births between 1982 and 1995 were identified (n= 1,345,210) and followed-up to diagnosis at age 16 or later. Mode of delivery was categorized as: unassisted vaginal delivery (VD), assisted VD, elective Caesarean section (CS) (before onset of labor), and emergency CS (after onset of labor). Outcomes included any psychosis; nonaffective psychoses (including schizophrenia only) and affective psychoses (including bipolar disorder only and depression with psychosis only). Cox regression analysis was used reporting partially and fully adjusted hazard ratios (HR) with 95% confidence intervals (CI). Sibling-matched Cox regression was performed to adjust for familial confounding factors. In the fully adjusted analyses, elective CS was significantly associated with any psychosis (HR 1.13, 95% CI 1.03, 1.24). Similar findings were found for nonaffective psychoses (HR 1.13, 95% CI 0.99, 1.29) and affective psychoses (HR 1.17, 95% CI 1.05, 1.31) (χ(2)for heterogeneityP= .69). In the sibling-matched Cox regression, this association disappeared (HR 1.03, 95% CI 0.78, 1.37). No association was found between assisted VD or emergency CS and psychosis. This study found that elective CS is associated with an increase in offspring psychosis. However, the association did not persist in the sibling-matched analysis, implying the association is likely due to familial confounding by unmeasured factors such as genetics or environment.

Associations between ADHD symptoms and smoking outcome expectancies in a non-clinical sample of daily cigarette smokers

BACKGROUND AND OBJECTIVES

Smoking outcome expectancies for positive reinforcement (PR: beliefs that smoking produces desirable outcomes) and negative reinforcement (NR: beliefs that smoking alleviates negative affect) are modifiable cognitive manifestations of affect-mediated smoking motivation. Based on prior data and theory, we hypothesized that NR and PR expectancies are associated with ADHD symptom levels in a non-clinical sample of cigarette smokers. (Am J Addict 2016; XX:XX -XX) METHODS: Daily cigarette smokers (N = 256) completed self-report measures of ADHD symptoms and smoking outcome expectancies. Cross-sectional associations of overall ADHD symptomatology and the ADHD symptom dimensions of inattention (IN: difficulty concentrating and distractibility) and hyperactivity impulsivity (HI: poor inhibitory control and motor activity restlessness) with PR and NR smoking outcome expectancies were examined.

RESULTS

Higher levels of overall, IN and HI ADHD symptoms were positively associated with NR smoking expectancies after statistically controlling for anxiety, depression, alcohol/drug use problems, nicotine dependence, and other smoking expectancies. Although neither HI nor IN symptom dimensions exhibited empirically unique relations to NR expectancies over and above one another, the collective variance across IN and HI was associated with NR expectancies. PR expectancies were not associated with ADHD symptoms.

DISCUSSION AND CONCLUSIONS

Although PR and NR expectancies may be important etiological influences in the overall population of smokers, NR outcome expectancies appear to be disproportionately expressed in smokers with elevated ADHD symptoms. Cognitive manifestations of NR motivation, which may be modifiable via intervention, are prominent in smokers with elevated ADHD symptoms.

SCIENTIFIC SIGNIFICANCE

Beliefs that smoking alleviates negative affect may underlie ADHD-smoking comorbidity.

Cost-effectiveness of abatacept, rituximab, and TNFi treatment after previous failure with TNFi treatment in rheumatoid arthritis: a pragmatic multi-centre randomised trial

Introduction

For patients with rheumatoid arthritis (RA) whose treatment with a tumour necrosis factor inhibitor (TNFi) is failing, several biological treatment options are available. Often, another TNFi or a biological with another mode of action is prescribed. The objective of this study was to compare the effectiveness and cost-effectiveness of three biologic treatments with different modes of action in patients with RA whose TNFi therapy is failing.

Methods

We conducted a pragmatic, 1-year randomised trial in a multicentre setting. Patients with active RA despite previous TNFi treatment were randomised to receive abatacept, rituximab or a different TNFi. The primary outcome (Disease Activity Score in 28 joints) and the secondary outcomes (Health Assessment Questionnaire Disability Index and 36-item Short Form Health Survey scores) were analysed using linear mixed models. Cost-effectiveness was analysed on the basis of incremental net monetary benefit, which was based on quality-adjusted life-years (calculated using EQ-5D scores), and all medication expenditures consumed in 1 year. All analyses were also corrected for possible confounders.

Results

Of 144 randomised patients, 5 were excluded and 139 started taking abatacept (43 patients), rituximab (46 patients) or a different TNFi (50 patients). There were no significant differences between the three groups with respect to multiple measures of RA outcomes. However, our analysis revealed that rituximab therapy is significantly more cost-effective than both abatacept and TNFi over a willingness-to-pay range of 0 to 80,000 euros.

Conclusions

All three treatment options were similarly effective; however, when costs were factored into the treatment decision, rituximab was the best option available to patients whose first TNFi treatment failed. However, generalization of these costs to other countries should be undertaken carefully.

Trial registration

Netherlands Trial Register number NTR1605. Registered 24 December 2008.

Electronic supplementary material

The online version of this article (doi:10.1186/s13075-015-0630-5) contains supplementary material, which is available to authorized users.

Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): explanation and elaboration

The TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) Statement includes a 22-item checklist, which aims to improve the reporting of studies developing, validating, or updating a prediction model, whether for diagnostic or prognostic purposes. The TRIPOD Statement aims to improve the transparency of the reporting of a prediction model study regardless of the study methods used. This explanation and elaboration document describes the rationale; clarifies the meaning of each item; and discusses why transparent reporting is important, with a view to assessing risk of bias and clinical usefulness of the prediction model. Each checklist item of the TRIPOD Statement is explained in detail and accompanied by published examples of good reporting. The document also provides a valuable reference of issues to consider when designing, conducting, and analyzing prediction model studies. To aid the editorial process and help peer reviewers and, ultimately, readers and systematic reviewers of prediction model studies, it is recommended that authors include a completed checklist in their submission. The TRIPOD checklist can also be downloaded from www.tripod-statement.org.

Predictors of re-entry into the child protection system in Singapore: a cumulative ecological-transactional risk model

This study examines the risk factors of re-entry for 1,750 child protection cases in Singapore using a cumulative ecological-transactional risk model. Using administrative data, the present study found that the overall percentage of Child Protection Service (CPS) re-entry in Singapore is 10.5% based on 1,750 cases, with a range from 3.9% (within 1 year) to 16.5% (within 8 years after case closure). One quarter of the re-entry cases were observed to occur within 9 months from case closure. Seventeen risk factors, as identified from the extant literature, were tested for their utility to predict CPS re-entry in this study using a series of Cox regression analyses. A final list of seven risk factors (i.e., children's age at entry, case type, case closure result, duration of case, household income, family size, and mother's employment status) was used to create a cumulative risk score. The results supported the cumulative risk model in that higher risk score is related to higher risk of CPS re-entry. Understanding the prevalence of CPS re-entry and the risk factors associated with re-entry is the key to informing practice and policy in a culturally relevant way. The results from this study could then be used to facilitate critical case management decisions in order to enhance positive outcomes of families and children in Singapore's care system.

Patient attrition from the HIV antiretroviral therapy program at two hospitals in Haiti

OBJECTIVE

To identify factors associated with antiretroviral therapy (ART) attrition among patients initiating therapy in 2005-2011 at two large, public-sector department-level hospitals, and to inform interventions to improve ART retention.

METHODS

This retrospective cohort study used data from the iSanté electronic medical record (EMR) system. The study characterized ART attrition levels and explored the patient demographic, clinical, temporal, and service utilization factors associated with ART attrition, using time-to-event analysis methods.

RESULTS

Among the 2 023 patients in the study, ART attrition on average was 17.0 per 100 person-years (95% confidence interval (CI): 15.8-18.3). In adjusted analyses, risk of ART attrition was up to 89% higher for patients living in distant communes compared to patients living in the same commune as the hospital (hazard ratio: 1.89, 95%CI: 1.54-2.33; P < 0.001). Hospital site, earlier year of ART start, spending less time enrolled in HIV care prior to ART initiation, receiving a non-standard ART regimen, lacking counseling prior to ART initiation, and having a higher body mass index were also associated with attrition risk.

CONCLUSIONS

The findings suggest quality improvement interventions at the two hospitals, including: enhanced retention support and transportation subsidies for patients accessing care from remote areas; counseling for all patients prior to ART initiation; timely outreach to patients who miss ART pick-ups; "bridging services" for patients transferring care to alternative facilities; routine screening for anticipated interruptions in future ART pick-ups; and medical case review for patients placed on non-standard ART regimens. The findings are also relevant for policymaking on decentralization of ART services in Haiti.