Bayesian estimation of multivariate Gaussian Markov random fields with constraint.

This article concerns with conditionally formulated multivariate Gaussian Markov random fields (MGMRF) for modeling multivariate local dependencies with unknown dependence parameters subject to positivity constraint. In the context of Bayesian hierarchical modeling of lattice data in general and Bayesian disease mapping in particular, analytic and simulation studies provide new insights into various approaches to posterior estimation of dependence parameters under "hard" or "soft" positivity constraint, including the well-known strictly diagonal dominance criterion and options of hierarchical priors. Hierarchical centering is examined as a means to gain computational efficiency in Bayesian estimation of multivariate generalized linear mixed effects models in the presence of spatial confounding and weakly identified model parameters. Simulated data on irregular or regular lattice, and three datasets from the multivariate and spatiotemporal disease mapping literature, are used for illustration. The present investigation also sheds light on the use of deviance information criterion for model comparison, choice, and interpretation in the context of posterior risk predictions judged by borrowing-information and bias-precision tradeoff. The article concludes with a summary discussion and directions of future work. Potential applications of MGMRF in spatial information fusion and image analysis are briefly mentioned.

Linear models of coregionalization for multivariate lattice data: a general framework for coregionalized multivariate CAR models.

We present a general coregionalization framework for developing coregionalized multivariate Gaussian conditional autoregressive (cMCAR) models for Bayesian analysis of multivariate lattice data in general and multivariate disease mapping data in particular. This framework is inclusive of cMCARs that facilitate flexible modelling of spatially structured symmetric or asymmetric cross-variable local interactions, allowing a wide range of separable or non-separable covariance structures, and symmetric or asymmetric cross-covariances, to be modelled. We present a brief overview of established univariate Gaussian conditional autoregressive (CAR) models for univariate lattice data and develop coregionalized multivariate extensions. Classes of cMCARs are presented by formulating precision structures. The resulting conditional properties of the multivariate spatial models are established, which cast new light on cMCARs with richly structured covariances and cross-covariances of different spatial ranges. The related methods are illustrated via an in-depth Bayesian analysis of a Minnesota county-level cancer data set. We also bring a new dimension to the traditional enterprize of Bayesian disease mapping: estimating and mapping covariances and cross-covariances of the underlying disease risks. Maps of covariances and cross-covariances bring to light spatial characterizations of the cMCARs and inform on spatial risk associations between areas and diseases. Copyright © 2016 John Wiley & Sons, Ltd.

Adaptive Gaussian Markov random field spatiotemporal models for infectious disease mapping and forecasting.

Recent disease mapping literature presents adaptively parameterized spatiotemporal (ST) autoregressive (AR) or conditional autoregressive (CAR) models for Bayesian prediction of COVID-19 infection risks. These models were motivated to capture complex spatiotemporal dynamics and heterogeneities of infection risks. In the present paper, we synthesize, generalize, and unify the ST AR and CAR model constructions for models augmented by adaptive Gaussian Markov random fields, with an emphasis on disease forecasting. A general convolution construction is presented, with illustrative models motivated to (i) characterize local risk dependencies and influences over both spatial and temporal dimensions, (ii) model risk heterogeneities and discontinuities, and (iii) predict and forecast areal-level disease risks and occurrences. The broadened constructions allow rich options of intuitive parameterization for disease mapping and spatial regression. Illustrative parameterizations are presented for Bayesian hierarchical models of Poisson, zero-inflated Poisson, and Bernoulli data models, respectively. They are also discussed in the context of quantifying time-varying or time-invariant effects of (omitted) covariates, with application to prediction and forecasting areal-level COVID-19 infection occurrences and probabilities of zero-infection. The model constructions presented herein have much wider scope in offering a flexible framework for modelling complex spatiotemporal data and for estimation, learning, and forecasting purposes.

Revisiting Gaussian Markov random fields and Bayesian disease mapping.

We revisit several conditionally formulated Gaussian Markov random fields, known as the intrinsic conditional autoregressive model, the proper conditional autoregressive model, and the Leroux et al. conditional autoregressive model, as well as convolution models such as the well known Besag, York and Mollie model, its (adaptive) re-parameterization, and its scaled alternatives, for their roles of modelling underlying spatial risks in Bayesian disease mapping. Analytic and simulation studies, with graphic visualizations, and disease mapping case studies, present insights and critique on these models for their nature and capacities in characterizing spatial dependencies, local influences, and spatial covariance and correlation functions, and in facilitating stabilized and efficient posterior risk prediction and inference. It is illustrated that these models are Gaussian (Markov) random fields of different spatial dependence, local influence, and (covariance) correlation functions and can play different and complementary roles in Bayesian disease mapping applications.

Item response theory analysis of the Dysfunctional Beliefs and Attitudes about Sleep-16 (DBAS-16) scale in a university student sample.

Unhelpful beliefs about sleep have been shown to exacerbate distress associated with sleep-related difficulties. University students are particularly vulnerable to experiencing sleep-related problems. The Dysfunctional Beliefs and Attitudes about Sleep-16 (DBAS-16) scale is a widely used instrument that assesses for sleep-disruptive cognitions. Although psychometric support for the DBAS-16 is available, Item Response Theory (IRT) analysis is needed to examine its properties at the item level. Psychometric investigation in non-clinical samples can help identify people who may be at risk for developing sleep problems. We examined the DBAS-16 using IRT on a sample of 759 university students. Our results identified items and subscales that adequately/inadequately differentiated between students who held unhelpful beliefs about sleep and those who did not. The DBAS-16 is a valuable instrument to assess unhelpful beliefs about sleep. We outline recommendations to improve the discriminatory ability of the instrument. Future investigations should establish cross-validation with a clinical sample.

Unemployment and mortality: a comparative study of Germany and the United States.

OBJECTIVES: We examined the relationship between unemployment and mortality in Germany, a coordinated market economy, and the United States, a liberal market economy. METHODS: We followed 2 working-age cohorts from the German Socio-economic Panel and the US Panel Study of Income Dynamics from 1984 to 2005. We defined unemployment as unemployed at the time of survey. We used discrete-time survival analysis, adjusting for potential confounders. RESULTS: There was an unemployment-mortality association among Americans (relative risk [RR]=2.4; 95% confidence interval [CI]=1.7, 3.4), but not among Germans (RR=1.4; 95% CI=1.0, 2.0). In education-stratified models, there was an association among minimum-skilled (RR=2.6; 95% CI=1.4, 4.7) and medium-skilled (RR=2.4; 95% CI=1.5, 3.8) Americans, but not among minimum- and medium-skilled Germans. There was no association among high-skilled Americans, but an association among high-skilled Germans (RR=3.0; 95% CI=1.3, 7.0), although this was limited to those educated in East Germany. Minimum- and medium-skilled unemployed Americans had the highest absolute risks of dying. CONCLUSIONS: The higher risk of dying for minimum- and medium-skilled unemployed Americans, not found among Germans, suggests that the unemployment-mortality relationship may be mediated by the institutional and economic environment.

Preclosure fluid resuscitation influences outcome in gastroschisis.

Optimal preclosure fluid resuscitation in gastroschisis (GS) is unknown. The purpose of our study was to evaluate effects of preclosure intravenous fluid resuscitation on GS outcome. Cases were accrued from a national GS database. Risk variables analyzed included gestational age (GA), birth weight (BW), neonatal illness severity score, and bolus fluid administration within 6 hours of neonatal intensive care unit admission. Outcomes analyzed included closure success, days of ventilation/total parenteral nutrition (TPN), and bacteremic episodes. Linear and logistic regression analyses were performed. Four hundred seven live-born GS cases were identified (362 with complete resuscitative fluids data). Mean BW, GA, and Score for Neonatal Acute Physiology-II score were 2562 ± 539 g, 36.17 ± 1.95 weeks, and 9.97 ± 12.65, respectively. One hundred sixty-two patients received no supplemental fluid, and 200 patients received a mean of 21.49 (0.81 to 134.81) mL/kg of intravenous fluid. Multivariate outcomes analyses demonstrated a significant, direct relationship between resuscitative volume and days of postclosure ventilation, TPN, length of hospital stay, and bacteremic episodes; specifically, every 17 mL/kg of fluid predicted one additional ventilation day (p = 0.002), TPN day (p = 0.01), and hospital day (p = 0.01) and 0.02 odds increase of an episode of bacteremia (p = 0.03). Judicious, preclosure fluid resuscitation is essential in early GS management. Excessive fluid is associated with several adverse survival outcomes.

Bayesian disease mapping: Past, present, and future.

On the occasion of the Spatial Statistics' 10th Anniversary, I reflect on the past and present of Bayesian disease mapping and look into its future. I focus on some key developments of models, and on recent evolution of multivariate and adaptive Gaussian Markov random fields and their impact and importance in disease mapping. I reflect on Bayesian disease mapping as a subject of spatial statistics that has advanced to date, and continues to grow, in scope and complexity alongside increasing needs of analytic tools for contemporary health science research, such as spatial epidemiology, population and public health, and medicine. I illustrate (potential) utility and impact of some of the disease mapping models and methods for analysing and monitoring communicable disease such as the COVID-19 infection risks during an ongoing pandemic.

On Bayesian shared component disease mapping and ecological regression with errors in covariates.

Recent literature on Bayesian disease mapping presents shared component models (SCMs) for joint spatial modeling of two or more diseases with common risk factors. In this study, Bayesian hierarchical formulations of shared component disease mapping and ecological models are explored and developed in the context of ecological regression, taking into consideration errors in covariates. A review of multivariate disease mapping models (MultiVMs) such as the multivariate conditional autoregressive models that are also part of the more recent Bayesian disease mapping literature is presented. Some insights into the connections and distinctions between the SCM and MultiVM procedures are communicated. Important issues surrounding (appropriate) formulation of shared- and disease-specific components, consideration/choice of spatial or non-spatial random effects priors, and identification of model parameters in SCMs are explored and discussed in the context of spatial and ecological analysis of small area multivariate disease or health outcome rates and associated ecological risk factors. The methods are illustrated through an in-depth analysis of four-variate road traffic accident injury (RTAI) data: gender-specific fatal and non-fatal RTAI rates in 84 local health areas in British Columbia (Canada). Fully Bayesian inference via Markov chain Monte Carlo simulations is presented.

An investigation of the adjustment of retrospective noise exposure for use of hearing protection devices.

OBJECTIVE: To account for use of hearing protection devices (HPDs) in retrospective noise exposure assessment, adjust noise exposure estimates accordingly, and validate the adjusted estimates. BACKGROUND: A previous study in the same working population showed a stronger relation for noise and acute myocardial infarction among those who did not wear HPD. Because accurate noise exposure assessment is complicated by the use of HPD, we previously developed a multilevel model of the likelihood of HPD use for British Columbia (Canada) lumber mill workers. Historical estimates of noise exposure can be adjusted according to models predictions and a reduction in misclassifying workers, exposure is expected. METHODS: Work history and exposure information were obtained for 13,147 lumber mill workers followed from 1909 until 1998. Audiometric data for the cohort, including hearing threshold levels at several pure tone frequencies, were obtained from the local regulatory agency for the period from 1978 to 2003. Following the modeling of HPD use, noise estimates were adjusted according to models predictions and attenuation factors based on existing research and standards. Adjusted and unadjusted noise metrics were compared by investigating their ability to predict noise-induced hearing loss. RESULTS: We showed a 4-fold increase in the noise exposure and hearing loss slope, after adjusting for HPD use, while controlling for gender, age, race, as well as medical and non-occupational confounding variables. CONCLUSION: While the relative difference before and after adjustment for use of HPD is considerable, we observed a subtle absolute magnitude of the effect. Using noise-induced hearing loss as a 'gold standard' for testing the assessment of retrospective noise exposure estimates should continue to be investigated.

Determinants of use of hearing protection devices in Canadian lumber mill workers.

BACKGROUND: In a cohort study of lumber mill workers' exposure to noise and incidence of heart disease, initial noise estimates were likely overestimated because they did not account for reductions afforded by the use of hearing protection. As such information was seldom available for individual workers, modeling was necessary to predict hearing protection use and derive adjusted noise measures. OBJECTIVE: To develop a multilevel model of the likelihood of use of hearing protection devices (HPDs) for British Columbia (Canada) lumber mill workers. METHOD: The study population included 13,147 workers in 14 sawmills for whom we had information on HPD use. Subjects self-reported their use of hearing protectors during routine hearing tests over their work history period. Separate multilevel logistic regression models with increasing complexity were developed for a subcohort of workers with complete information (n = 1493) and for a subcohort comprised subjects with hearing tests coinciding with their jobs (n = 10 203). The models included random intercepts for worker and for sawmill. RESULTS: HPD use was associated in both subcohorts with factors such as noise exposure and age. We also showed that specific jobs (such as sawfiling) and departments (planer, in particular) were strongly associated with the use of HPDs. The model illustrates the quantitative importance of including a hierarchical structure which allows for explaining potential sources of outcome variability. CONCLUSIONS: We developed a hierarchical model to predict hearing protection use to enable correction of exposure assessments for use in retrospective epidemiological studies. We showed that this was feasible even in the absence of complete determinant information.

Water and sewage systems, socio-demographics, and duration of residence associated with endemic intestinal infectious diseases: a cohort study.

BACKGROUND: Studies of water-related gastrointestinal infections are usually directed at outbreaks. Few have examined endemic illness or compared rates across different water supply and sewage disposal systems. We conducted a cohort study of physician visits and hospitalizations for endemic intestinal infectious diseases in a mixed rural and urban community near Vancouver, Canada, with varied and well-characterized water and sewage systems. METHODS: Cohort members and their disease events were defined via universal health insurance data from 1995 through 2003. Environmental data were derived from municipal, provincial, and federal government sources. Logistic regression was used to examine associations between disease events and water and sewage systems, socio-demographic characteristics, and temporal factors. RESULTS: The cohort included 126,499 individuals and approximately 190,000,000 person-days. Crude incidence rates were 1,353 physician visits and 33.8 hospitalizations for intestinal infectious diseases per 100,000 person-years. Water supply chlorination was associated with reduced physician visit incidence (OR: 0.92, 95% CI 0.85-1.0). Two water systems with the highest proportions of surface water had increased incidence (ORs: 1.57, 95% CI 1.39-1.78; and 1.45, 95% CI 1.28-1.64). Private well water and well depth were not associated with increased risk, likely because of residents' awareness of and attention to water quality. There was increased crude incidence with increasing precipitation in the population served by surface water supplies, but this trend did not remain with adjustment for other variables. Municipal sewer systems were associated with increased risk (OR: 1.26, 95% CI 1.14-1.38). Most socio-demographic variables had predicted associations with risk: higher rates in females, in the very young and the elderly, and in residents of low income areas. Increased duration of area residence was associated with reduced risk (OR, duration ≥ 6 years: 0.69, 95% CI 0.60-0.80 vs. < 1 year: 1.16, 95% CI 1.03-1.30). CONCLUSIONS: This large cohort study, with objective data on exposures and outcomes, demonstrated associations between endemic infectious intestinal diseases and factors related to water supply, sewage disposal, socio-demographics, and duration of residency. The results did not always follow prior expectations based on studies examining outbreaks and single systems, and underscore the importance of studying factors associated with endemic disease across water and sewage system types.

Does overnight birth influence treatment or outcome in congenital diaphragmatic hernia?

We examined the effect of overnight birth on treatment and outcome (including initial cardiorespiratory stabilization) in newborns with congenital diaphragmatic hernia (CDH). CDH births between May 2005 and May 2008 were abstracted from a national CDH database. Overnight birth was defined as occurring between 8 PM and 8 AM. Patient characteristics, treatment, and outcomes were compared between birth time groups, including the subset of infants in whom a prenatal diagnosis of CDH had been made. Of 132 cases evaluated, 106 (80%) survived. Forty-nine babies (37%) were born overnight. Eighty-five infants (64%) with a prenatal diagnosis were evenly distributed between the birth time cohorts. Survival to discharge, surgical management, duration of mechanical ventilation, and length of hospital stay were comparable between the risk-matched (SNAP-II score) birth time groups. Overnight birth did predict a lower rate of cesarean section deliveries (P = 0.03). Escalation of ventilator mode, inability to achieve/maintain blood gas targets, development of pneumothorax, and need for extracorporeal membrane oxygenation were all considered "stabilization failures." Overnight birth predicted a failure to maintain or improve pH to within target range (P = 0.008). Overnight birth did not adversely affect outcome in this birth time comparison of newborns with CDH.

Proportional hazards frailty models for recurrent methadone maintenance treatment.

The authors' objective in this study was to identify determinants of time to discontinuation of methadone maintenance treatment (MMT) across multiple treatment episodes. Population-level data on drug dispensations for all patients receiving methadone for opioid maintenance throughout the tenure of the British Columbia, Canada, methadone program to date (1996-2007) were extracted from an administrative database. Proportional hazards frailty models were developed to assess factors associated with time to discontinuation from recurrent MMT episodes. A total of 17,005 patients experienced 32,656 treatment episodes over the 11-year follow-up period. Age, medical comorbidity, and physician patient load, as well as neighborhood-level socioeconomic status indicators, were significant predictors of time to discontinuation of treatment; treatment adherence and average daily doses up to and above 120 mg per day were also associated with longer treatment episodes. Studies have shown that while successfully retained in MMT, clients decrease their illicit drug use and criminal activity, and their risk of mortality is substantially lower; however, the majority of clients relapse. Many reenter treatment. The primary finding of this study was that patients experiencing multiple treatment episodes tended to stay in treatment for progressively longer periods in later episodes.

Improving the quality of care for infants: a cluster randomized controlled trial.

BACKGROUND: We developed and tested a new method, called the Evidence-based Practice for Improving Quality method, for continuous quality improvement. METHODS: We used cluster randomization to assign 6 neonatal intensive care units (ICUs) to reduce nosocomial infection (infection group) and 6 ICUs to reduce bronchopulmonary dysplasia (pulmonary group). We included all infants born at 32 or fewer weeks gestation. We collected baseline data for 1 year. Practice change interventions were implemented using rapid-change cycles for 2 years. RESULTS: The difference in incidence trends (slopes of trend lines) between the ICUs in the infection and pulmonary groups was - 0.0020 (95% confidence interval [CI] - 0.0007 to 0.0004) for nosocomial infection and - 0.0006 (95% CI - 0.0011 to - 0.0001) for bronchopulmonary dysplasia. INTERPRETATION: The results suggest that the Evidence-based Practice for Improving Quality method reduced bronchopulmonary dysplasia in the neonatal ICU and that it may reduce nosocomial infection.

Caregiver fear of falling and functional ability among seniors residing in long-term care facilities.

BACKGROUND: Consistent with fear-avoidance models of falling and pain, past research has demonstrated that, among adults living in the community, excessive fear of falling and fear of pain result in activity restriction and predict functional outcomes including falls (possibly because self-imposed activity restriction, due to fear of pain or falling, can lead to muscular decline and deconditioning). Among seniors with dementia, who rely on others for their care, decisions concerning activity restrictions are made by caregivers. As such, caregivers' fear about the possibility of care recipient falls and pain is important to examine. OBJECTIVE: In this investigation of patients with dementia, our goal was to conduct a longitudinal investigation of the relationship between professional caregivers' fears (about the possibility that care recipients will experience falls and pain) with long-term care (LTC) resident functional ability and falls. METHODS: For the purposes of our 3-month longitudinal study, nurses' and special care aides' fears that specific residents might experience pain and falls were examined. Resident functional ability was assessed, based on an established and well-validated caregiver-administered questionnaire, both before and after the 3-month period. Falls and fall-related injuries sustained by residents were recorded. RESULTS: After controlling for physical risk factors for falling and functional ability at the beginning of the study, caregiver fears that residents might experience pain or falls were found to be predictive of restraint/restriction use. In turn, the use of restraints/restrictions was found to be predictive of future functional ability of residents with dementia (after controlling for functional ability at the beginning of the study) and injurious falls (after controlling for physical risk factors for falling). CONCLUSIONS: This is the first study to apply a modified fear-avoidance model of falls and pain to seniors with dementia who reside in LTC facilities. Our results demonstrate the importance of considering caregiver fears concerning falls and pain, when developing programs designed to optimize the use of physical restrictions (to prevent falls and minimize functional decline) in LTC facilities.

Does routine pain assessment result in better care?

BACKGROUND: Although a variety of national organizations such as the Canadian Pain Society, the American Pain Society and the Joint Commission on Accreditation of Health Care Organizations have advanced the idea that pain should be assessed on a routine basis, there is little evidence that systematic pain assessment information is used routinely by clinicians even when it is readily available. OBJECTIVE: To determine whether systematic pain assessment information alters medical practitioners' clinical practices. METHODS: A population of seniors with complex medical problems who were evaluated by case coordinators was studied. Case coordinators were assigned to either an experimental or control patient assessment condition. Control condition patients were assessed as usual. In the experimental condition, a psychometrically valid pain assessment battery as well as the Geriatric Depression Scale - Short Form (because depression and chronic pain are frequently comorbid) were integrated into the routine case coordination assessment. A summary of the results of the depression and pain assessments was subsequently sent to physicians via mail and fax. Patients were also given copies of the assessment summaries and were asked to discuss these with their physicians. Physicians' medication prescriptions were monitored over time through the database of the provincial ministry of health. RESULTS: At the end of the study, no significant differences between experimental and control patients were found with respect to medications prescribed or patient self-reports of pain. Nonetheless, there was a significant relationship between Geriatric Depression Scale -- Short Form scores and pain medications prescribed for patients in the experimental condition. Moreover, indexes of overall pain intensity did not change significantly over time. CONCLUSIONS: The findings do not support the idea that the availability of systematic pain assessment information leads to change in clinician's medication practices. As such, educational interventions and public policy initiatives are needed to ensure that treatment providers do not only gather but also use pain assessment information.

The Experience of Family Physicians and Home Health Staff Involved in an Intervention to Increase Patient-Related Collaboration.

Les médecins de famille (MF) et le personnel de soins de santé à domicile (PSD) canadiens rencontrent d'importants obstacles lorsqu'ils doivent collaborer pour la prestation de soins aux patients qu'ils ont en commun. Cette étude à méthodologie mixte visait à évaluer la qualité et la viabilité de l'utilisation de l'audioconférence sécurisée dans une optique d'amélioration de la planification des soins pour ces patients. Les données primaires incluaient les résultats d'un sondage réalisé avant et après l'intervention, ainsi que des entretiens semi-structurés et des groupes de discussion post-intervention. Des méthodes statistiques non paramétriques ont été utilisées pour analyser les résultats du sondage, et les données qualitatives ont fait l'objet d'une analyse thématique de contenu. Les résultats des analyses quantitatives et qualitatives ont ensuite été intégrés afin de faire ressortir les inférences reflétant les approches des MF et du PSD relatives aux obstacles et aux avantages de la planification interdisciplinaire des soins. Les MF et le PSD ont montré que des obstacles structurels limitent leur capacité à collaborer. Le PSD et les MF ont également convenu que les rencontres entre les intervenants des deux services étaient bénéfiques pour les patients et que l'utilisation de l'audioconférence constituait une méthode efficiente de planification collaborative des soins. Les limites comprenaient la petite taille de l'échantillon et la courte période d'intervention, compte tenu de l'ampleur des changements attendus.Canadian family physicians (FPs) and home health staff (HHS) experience significant barriers to patient-related collaboration about patients they share. This mixed-methods study sought to determine the quality and sustainability of secure audio conferencing as a way to increase care planning about shared patients. Primary data sources included pre-and post-study administration of a published survey and post-study semi-structured interviews and focus groups. Non-parametric statistical procedures were used to analyze survey results and thematic content analysis was undertaken for qualitative data. Results from both quantitative and qualitative analysis were integrated into the overall analysis, in order to draw inferences reflecting both approaches to barriers and benefits of collaborative care planning for FPs and HHS. Both FPs and HHS provided evidence that structural barriers impede their ability to collaborate. HHS and FPs also agreed that joint conferences were beneficial for patients, and that the use of audio conferencing provided an efficient method of collaborative care planning. Limitations included a small sample size and short timeline for the intervention period, given the magnitude of the expected change.

Instituting surveillance guidelines and adverse outcomes in preeclampsia.

OBJECTIVE: To assess the incidence of combined adverse maternal and perinatal outcomes in women with preeclampsia before and after introducing standardized assessment and surveillance. METHODS: This study was a preintervention (retrospective) compared with a postintervention (prospective) cohort comparison in a single-tertiary, perinatal unit that included women admitted to hospital with preeclampsia. We interrogated an existing retrospective 24-month database and then introduced the guidelines, assessing the incidence of the combined adverse maternal and perinatal outcomes for 41 months (September 2003 through February 2007). Tests of organ (dys)function were performed at least as often as on the day of admission, admission day +1, every Monday and Thursday, day of delivery, and delivery day +1. All data were checked for errors. The combined maternal outcome was maternal death or one or more of hepatic failure, hematoma, or rupture, Glasgow coma score of less than 13, stroke, at least two seizures, cortical blindness, need for positive inotrope support, myocardial infarction, infusion of any third antihypertensive, renal dialysis, renal transplantation, at least 50% FIO(2) for greater than 1 hour, intubation, or transfusion of at least 10 units of blood products. The combined perinatal outcome was perinatal or infant mortality, bronchopulmonary dysplasia, necrotizing enterocolitis, grade III/IV intraventricular hemorrhage, cystic periventricular leukomalacia, or stage 3-5 retinopathy of prematurity. RESULTS: Two hundred ninety-five and 405 women were in the preintervention and postintervention cohorts, respectively. The incidence of adverse maternal outcome fell (5.1% to 0.7%; Fisher P<.001; odds ratio 0.14, 95% confidence interval 0.04-0.49). Perinatal outcomes did not change. CONCLUSION: Standardized surveillance of women with preeclampsia was associated with reduced maternal risk.