A hypothetical intervention to reduce inequities in anxiety for Multiracial people: simulating an intervention on childhood adversity.

Multiracial people report higher mean Adverse Childhood Experiences (ACEs) scores and prevalence of anxiety than other racial groups. Studies using statistical interactions to estimate racial differences in ACEs-anxiety associations do not show stronger associations for Multiracial people. Using data from Waves 1 (1995-97) through 4 (2008-09) of the National Longitudinal Study of Adolescent to Adult Health (Add Health), we simulated a stochastic intervention over 1,000 resampled datasets to estimate the race-specific cases averted per 1,000 of anxiety if all racial groups had the same exposure distribution of ACEs as Whites. Simulated cases averted were greatest for the Multiracial group, (median = -4.17 cases per 1,000, 95% CI: -7.42, -1.86). The model also predicted smaller risk reductions for Black participants (-0.76, 95% CI: -1.53, -0.19). CIs around estimates for other racial groups included the null. An intervention to reduce racial disparities in exposure to ACEs could help reduce the inequitable burden of anxiety on the Multiracial population. Stochastic methods support consequentialist approaches to racial health equity, and can encourage greater dialogue between public health researchers, policymakers, and practitioners.

Neighborhood-level Fatal Police Violence and Severe Maternal Morbidity in California.

Police violence is a pervasive issue that may have adverse implications for severe maternal morbidity (SMM). We assessed how the occurrence of fatal police violence (FPV) in one's neighborhood before/during pregnancy may influence SMM risk. Hospital discharge records from California between 2002-2018 were linked with the Fatal Encounters database (N=2,608,682). We identified 2,184 neighborhoods (census-tracts) with at least one FPV incident during the study period and used neighborhood fixed-effects models adjusting for individual sociodemographic characteristics to estimate odds of SMM associated with experiencing FPV in one's neighborhood anytime within the 24-months before childbirth. We did not find conclusive evidence on the link between FPV occurrence before delivery and SMM. However, estimates show that birthing people residing in neighborhoods where one or more FPV events had occurred within the preceding 24-months of giving birth may have a mildly elevated odds of SMM than those residing in the same neighborhoods with no FPV occurrence during the 24-months preceding childbirth (Odds Ratio (OR)=1.02; 95% Confidence Interval (CI): 0.99-1.05), particularly among those living in neighborhoods with fewer (1-2) FPV incidents throughout the study period (OR=1.03; 95% CI:1.00-1.06). Our findings provide evidence for the need to continue to examine the health consequences of police violence.

Risk of Adverse Perinatal Outcomes Among African-born Black Women in California, 2011-2020.

BACKGROUND: African-born women have a lower risk of preterm birth and small for gestational age (SGA) birth compared with United States-born Black women, however variation by country of origin is overlooked. Additionally, the extent that nativity disparities in adverse perinatal outcomes to Black women are explained by individual-level factors remains unclear. METHODS: We conducted a population-based study of nonanomalous singleton live births to United States- and African-born Black women in California from 2011 to 2020 (n = 194,320). We used age-adjusted Poisson regression models to estimate the risk of preterm birth and SGA and reported risk ratios (RR) and 95% confidence intervals (CI). Decomposition using Monte Carlo integration of the g-formula computed the percentage of disparities in adverse outcomes between United States- and African-born women explained by individual-level factors. RESULTS: Eritrean women (RR = 0.4; 95% CI = 0.3, 0.5) had the largest differences in risk of preterm birth and Cameroonian women (RR = 0.5; 95% CI = 0.3, 0.6) in SGA birth, compared with United States-born Black women. Ghanaian women had smaller differences in risk of preterm birth (RR = 0.8; 95% CI = 0.7, 1.0) and SGA (RR = 0.9; 95% CI = 0.8, 1.1) compared with United States-born women. Overall, we estimate that absolute differences in socio-demographic and clinical factors contributed to 32% of nativity-based disparities in the risk of preterm birth and 26% of disparities in SGA. CONCLUSIONS: We observed heterogeneity in risk of adverse perinatal outcomes for African- compared with United States-born Black women, suggesting that nativity disparities in adverse perinatal outcomes were not fully explained by differences in individual-level factors.

Structural racism, nativity and risk of adverse perinatal outcomes among Black women.

BACKGROUND: Black women in the United States (US) have the highest risk of preterm birth (PTB) and small for gestational age (SGA) births, compared to women of other racial groups. Among Black women, there are disparities by nativity whereby foreign-born women have a lower risk of PTB and SGA compared to US-born women. Differential exposure to racism may confer nativity-based differences in adverse perinatal outcomes between US- and foreign-born Black women. This remains unexplored among US- and African-born women in California. OBJECTIVES: Evaluate the relationship between structural racism, nativity, PTB and SGA among US- and African-born Black women in California. METHODS: We conducted a population-based study of singleton births to US- and African-born Black women in California from 2011 to 2017 (n = 131,424). We examined the risk of PTB and SGA by nativity and neighbourhoods with differing levels of structural racism, as measured by the Index of Concentration at the Extremes. We fit crude and age-adjusted Poisson regression models, estimated using generalized estimating equations, with risk ratios (RR) and 95% confidence intervals (CI) as the effect measure. RESULTS: The proportions of PTB and SGA were 9.7% and 14.5%, respectively, for US-born women, while 5.6% and 8.3% for African-born women. US-born women (n = 24,782; 20.8%) were more likely to live in neighbourhoods with high structural racism compared to African-born women (n = 1474; 11.6%). Structural racism was associated with an elevated risk of PTB (RR 1.19, 95% CI 1.12, 1.26) and SGA (RR 1.19, 95% CI 1.13, 1.25) for all Black women, however, there was heterogeneity by nativity, with US-born women experiencing a higher magnitude of effect than African-born women. CONCLUSIONS: Among Black women in California, exposure to structural racism and the impacts of structural racism on the risk of PTB and SGA varied by nativity.

Harnessing Google Health Trends Data for Epidemiologic Research.

Interest in using internet search data, such as that from the Google Health Trends Application Programming Interface (GHT-API), to measure epidemiologically relevant exposures or health outcomes is growing due to their accessibility and timeliness. Researchers enter search term(s), geography, and time period, and the GHT-API returns a scaled probability of that search term, given all searches within the specified geographic-time period. In this study, we detailed a method for using these data to measure a construct of interest in 5 iterative steps: first, identify phrases the target population may use to search for the construct of interest; second, refine candidate search phrases with incognito Google searches to improve sensitivity and specificity; third, craft the GHT-API search term(s) by combining the refined phrases; fourth, test search volume and choose geographic and temporal scales; and fifth, retrieve and average multiple samples to stabilize estimates and address missingness. An optional sixth step involves accounting for changes in total search volume by normalizing. We present a case study examining weekly state-level child abuse searches in the United States during the coronavirus disease 2019 pandemic (January 2018 to August 2020) as an application of this method and describe limitations.

Comparing Two-way Fixed Effects and New Estimators for Difference-in-Differences: A Simulation Study and Empirical Example.

BACKGROUND: Two-way fixed effects methods have been used to estimate effects of policies adopted in different places over time, but they can provide misleading results when effects are heterogeneous or dynamic, and alternate methods have been proposed. METHODS: We compared methods for estimating the average treatment effect on the treated (ATT) under staggered adoption of policies, including two-way fixed effects, group-time ATT, cohort ATT, and target-trial approaches. We applied each method to assess the impact of Medicaid expansion on preterm birth using the National Center for Health Statistics' birth records. We compared each estimator's performance in a simulation parameterized to mimic the empirical example. We generated constant, heterogeneous, and dynamic effects and calculated bias, mean squared error, and confidence interval coverage of each estimator across 1000 iterations. RESULTS: Two-way fixed effects estimated that Medicaid expansion increased the risk of preterm birth (risk difference [RD], 0.12; 95% CI = 0.02, 0.22), while the group-time ATT, cohort ATT, and target-trial approaches estimated protective or null effects (group-time RD, -0.16; 95% CI = -0.58, 0.26; cohort RD, -0.02; 95% CI = -0.46, 0.41; target trial RD, -0.16; 95% CI = -0.59, 0.26). In simulations, two-way fixed effects performed well when treatment effects were constant and less well under heterogeneous and dynamic effects. CONCLUSIONS: We demonstrated why new approaches perform better than two-way fixed effects when treatment effects are heterogeneous or dynamic under a staggered policy adoption design, and created simulation and analysis code to promote understanding and wider use of these methods in the epidemiologic literature.

US shelter in place policies and child abuse Google search volume during the COVID-19 pandemic.

The COVID-19 pandemic has led to unemployment, school closures, movement restrictions, and social isolation, all of which are child abuse risk factors. Our objective was to estimate the effect of COVID-19 shelter in place (SIP) policies on child abuse as captured by Google searches. We applied a differences-in-differences design to estimate the effect of SIP on child abuse search volume. We linked state-level SIP policies to outcome data from the Google Health Trends Application Programming Interface. The outcome was searches for child abuse-related phrases as a scaled proportion of total searches for each state-week between December 31, 2017 and June 14, 2020. Between 914 and 1512 phrases were included for each abuse subdomain (physical, sexual, and emotional). Eight states and DC were excluded because of suppressed outcome data. Of the remaining states, 38 introduced a SIP policy between March 19, 2020 and April 7, 2020 and 4 states did not. The introduction of SIP generally led to no change, except for a slight reduction in child abuse search volume in weeks 8-10 post-SIP introduction, net of changes experienced by states that did not introduce SIP at the same time. We did not find strong evidence for an effect of SIP on child abuse searches. However, an increase in total search volume during the pandemic that may be differential between states with and without SIP policies could have biased these findings. Future work should examine the effect of SIP at the individual and population level using other data sources.

Excess Google Searches for Child Abuse and Intimate Partner Violence During the COVID-19 Pandemic: Infoveillance Approach.

BACKGROUND: The COVID-19 pandemic has created environments with increased risk factors for household violence, such as unemployment and financial uncertainty. At the same time, it led to the introduction of policies to mitigate financial uncertainty. Further, it hindered traditional measurements of household violence. OBJECTIVE: Using an infoveillance approach, our goal was to determine if there were excess Google searches related to exposure to child abuse, intimate partner violence (IPV), and child-witnessed IPV during the COVID-19 pandemic and if any excesses are temporally related to shelter-in-place and economic policies. METHODS: Data on relative search volume for each violence measure was extracted using the Google Health Trends application programming interface for each week from 2017 to 2020 for the United States. Using linear regression with restricted cubic splines, we analyzed data from 2017 to 2019 to characterize the seasonal variation shared across prepandemic years. Parameters from prepandemic years were used to predict the expected number of Google searches and 95% prediction intervals (PI) for each week in 2020. Weeks with searches above the upper bound of the PI are in excess of the model's prediction. RESULTS: Relative search volume for exposure to child abuse was greater than expected in 2020, with 19% (10/52) of the weeks falling above the upper bound of the PI. These excesses in searches began a month after the Pandemic Unemployment Compensation program ended. Relative search volume was also heightened in 2020 for child-witnessed IPV, with 33% (17/52) of the weeks falling above the upper bound of the PI. This increase occurred after the introduction of shelter-in-place policies. CONCLUSIONS: Social and financial disruptions, which are common consequences of major disasters such as the COVID-19 pandemic, may increase risks for child abuse and child-witnessed IPV.

Declining Life Expectancy in the United States: Missing the Trees for the Forest.

In recent years, life expectancy in the United States has stagnated, followed by three consecutive years of decline. The decline is small in absolute terms but is unprecedented and has generated considerable research interest and theorizing about potential causes. Recent trends show that the decline has affected nearly all race/ethnic and gender groups, and the proximate causes of the decline are increases in opioid overdose deaths, suicide, homicide, and Alzheimer's disease. A slowdown in the long-term decline in mortality from cardiovascular diseases has also prevented life expectancy from improving further. Although a popular explanation for the decline is the cumulative decline in living standards across generations, recent trends suggest that distinct mechanisms for specific causes of death are more plausible explanations. Interventions to stem the increase in overdose deaths, reduce access to mechanisms that contribute to violent deaths, and decrease cardiovascular risk over the life course are urgently needed to improve mortality in the United States.

A New Approach for Classifying Fetal Growth Restriction.

BACKGROUND: Fetal growth restriction is commonly defined using small for gestational age (SGA) birth (birthweight < 10th percentile) as a proxy, but this approach is problematic because most SGA infants are small but healthy. In this proof-of-concept study, we sought to develop a new approach for identifying fetal growth restriction at birth that combines information on multiple, imperfect measures of fetal growth restriction in a probabilistic manner. METHODS: We combined information on birthweight, placental weight, placental malperfusion lesions, maternal disease, and fetal acidemia using latent profile analysis to classify fetal growth in births at the Royal Victoria Hospital in Montreal, Canada, 2001-2009. We examined the clinical characteristics and health outcomes of infants classified as growth-restricted and nongrowth-restricted by our model, and among the subgroup of growth-restricted infants who had a birthweight ≥10th percentile (i.e., would have been missed by the conventional SGA proxy). RESULTS: Among 26,077 births, 345 (1.3%) were classified as growth-restricted by our latent profile model. Growth-restricted infants were more likely than nongrowth-restricted infants to have an Apgar score <7 (10% vs. 2%), have hypoglycemia at birth (17% vs. 3%), require neonatal intensive care unit admission (59% vs. 6%), die in the perinatal period (3.8% vs. 0.2%), and require an emergency cesarean delivery (42% vs. 15%). Risks remained elevated in growth-restricted infants who were not SGA, suggesting our model identified at-risk infants not detected using the SGA proxy. CONCLUSIONS: Latent profile analysis is a promising strategy for classifying growth restriction at birth in fetal growth restriction research.

Hyperlocalized Measures of Air Pollution and Preeclampsia in Oakland, California.

Exposure to nitrogen dioxide (NO2), black carbon (BC), and ultrafine particles (UFPs) during pregnancy may increase the risk of preeclampsia, but previous studies have not assessed hyperlocalized differences in pollutant levels, which may cause exposure misclassification. We used data from Google Street View cars with mobile air monitors that repeatedly sampled NO2, BC, and UFPs every 30 m in Downtown and West Oakland neighborhoods during 2015-2017. Data were linked to electronic health records of pregnant women in the 2014-2016 Sutter Health population, who resided within 120 m of monitoring data (N = 1095), to identify preeclampsia cases. We used G-computation with log-binomial regression to estimate risk differences (RDs) associated with a hypothetical intervention reducing pollutant levels to the 25th percentile observed in our sample on preeclampsia risk, overall and stratified by race/ethnicity. Prevalence of preeclampsia was 6.8%. Median (interquartile range) levels of NO2, BC, and UFPs were 10.8 ppb (9.0, 13.0), 0.34 µg/m3 (0.27, 0.42), and 29.2 # × 103/cm3 (26.6, 32.6), respectively. Changes in the risk of preeclampsia achievable by limiting each pollutant to the 25th percentile were NO2 RD = -1.5 per 100 women (95% confidence interval (CI): -2.5, -0.5); BC RD = -1.0 (95% CI: -2.2, 0.02); and UFP RD = -0.5 (95% CI: -1.8, 0.7). Estimated effects were the largest for non-Latina Black mothers: NO2 RD = -2.8 (95% CI: -5.2, -0.3) and BC RD = -3.0 (95% CI: -6.4, 0.4).

Using change in a seat belt law to study racially-biased policing in South Carolina.

Seat belt laws have increased seat belt use in the US and contributed to reduced fatalities and injuries. However, these policies provide the potential for increased discrimination. The objective of this study is to determine whether a change in seat belt use enforcement led to a differential change in the number of stops, arrests, and searches to White, Black and Hispanic drivers in one US state. We used data on 1,091,424 traffic stops conducted by state troopers in South Carolina in 2005 and 2006 to examine how the change from secondary to primary enforcement of seat belt use in December 2005 affected the number of stops, arrests, and searches to White, Black, and Hispanic drivers using quasi-Poisson and logistic regressions. We found that the policy led to a 50% increase in the number of non-speeding stops for White drivers, and that this increase was 5% larger among Black drivers [RR (95% CI) = 1.05 (1.00, 1.10)], but not larger among Hispanic drivers [1.00 (0.93, 1.08)]. The policy decreased arrests and searches among non-speeding stops, with larger decreases for Black vs. White drivers [RR searches = 0.86 (0.81, 0.91) and RR arrests = 0.90 (0.85, 0.96)]. For Hispanic drivers, effects of the policy change were also found among stops for speeding, which failed the falsification test and suggested that other changes likely affected this group. These findings may support the hypothesis of differential enforcement of seat belt policy in South Carolina for Black and White drivers.

Comparison of Rates of Firearm and Nonfirearm Homicide and Suicide in Black and White Non-Hispanic Men, by U.S. State.

Background: The extent to which differences in homicide and suicide rates in black versus white men vary by U.S. state is unknown. Objective: To compare the rates of firearm and nonfirearm homicide and suicide in black and white non-Hispanic men by U.S. state and to examine whether these deaths are associated with state prevalence of gun ownership. Design: Surveillance study. Setting: 50 states and the District of Columbia, 2008 to 2016. Cause-of-death data were abstracted by using the Centers for Disease Control and Prevention's WONDER (Wide-ranging Online Data for Epidemiologic Research) database. Participants: Non-Hispanic black and non-Hispanic white males, all ages. Measurements: Absolute rates of and rate differences in firearm and nonfirearm homicide and suicide in black and white men. Results: During the 9-year study period, 84 113 homicides and 251 772 suicides occurred. Black-white differences in rates of firearm homicide and suicide varied widely across states. Relative to white men, black men had between 9 and 57 additional firearm homicides per 100 000 per year, with black men in Missouri, Michigan, Illinois, Indiana, and Pennsylvania having more than 40 additional firearm homicides per 100 000 per year. White men had between 2 fewer and 16 more firearm suicides per 100 000 per year, with the largest inequalities observed in southern and western states and the smallest in the District of Columbia and densely populated northeastern states. Limitations: Some homicides and suicides may have been misclassified as deaths due to unintentional injury. Survey data on state household gun ownership were collected in 2004 and may have shifted during the past decade. Conclusion: The large state-to-state variation in firearm homicide and suicide rates, as well as the racial inequalities in these numbers, highlights states where policies may be most beneficial in reducing homicide and suicide deaths and the racial disparities in their rates. Primary Funding Source: McGill University and the National Institutes of Health.

Safety of labour and delivery following closures of obstetric services in small community hospitals.

BACKGROUND: In recent decades, many smaller hospitals in British Columbia, Canada, have stopped providing planned obstetric services. We examined the effect of these service closures on the labour and delivery outcomes of pregnant women living in affected communities. METHODS: We used maternal postal codes to identify delivery records (1998-2014) of women residing in a community affected by service closure. The records were obtained from the British Columbia Perinatal Data Registry. We examined the effect of the closures using a within-communities fixed-effects framework and included similar-sized communities without service closures to control for underlying time trends. The primary outcome was a previously published composite measure of labour and delivery safety, the Adverse Outcome Index, which includes adverse events such as birth injury and unanticipated operative procedures, and includes weights for severity of adverse events. Secondary outcomes included maternal or newborn transfer, and use of obstetric interventions. RESULTS: We found little evidence that closure of planned obstetric services affected the risk of composite adverse maternal-newborn outcome (-0.4 excess adverse events per 100 deliveries, 95% confidence interval [CI] -2.0 to 1.1), or most other secondary outcomes. The severity of composite outcome events decreased following the closures (rate ratio 0.58, 95% CI 0.36 to 0.89). Closures were associated with increases in use of epidural analgesia (3.4 excess events per 100 deliveries, 95% CI 0.4 to 6.3) and length of antepartum stay (0.6 h, 95% CI 0.1 to 1.0 h). INTERPRETATION: Closure of planned obstetric services in low-volume hospitals was not associated with an increase or decrease in frequency of adverse events during labour and delivery.

Classifying Gestational Weight Gain Trajectories Using the SITAR Growth Model.

BACKGROUND: Gestational weight gain is often characterized by the total amount of weight gained during pregnancy, however, the pattern of gain may be an important determinant of health outcomes. The SITAR (Super Imposition by Translation And Rotation) model has been used to describe childhood growth trajectories and has appeal because of the biological interpretability of its parameters. The objective of this study was to determine the feasibility of applying this model to gestational weight gain trajectories. METHODS: The study cohort included 3470 normal-weight, overweight, and obese women delivering at Magee-Womens Hospital in Pittsburgh, Pennsylvania, 1998 to 2010. We applied the SITAR model, a non-linear mixed effects model, to serial prenatal weight gain measurements in each pre-pregnancy body mass index (BMI) category. We fit models of varying complexity, and chose the best-fitting model to describe the pattern of weight gain (by its absolute amount, timing, and acceleration) for each BMI group. RESULTS: The most complex SITAR models failed to converge, but reduced models could successfully be fit by specifying fewer random effects and simplifying the modelling of gestational age. Best-fitting models for each BMI group explained between 95% and 97% of the variation in weight gain trajectories. Peak rates of weight gain were reached between the 20th and 22nd weeks, and were higher for normal and overweight women (0.59 kg/week and 0.57 kg/week, respectively) than obese women (0.46 kg/week). CONCLUSIONS: Following some modifications, the SITAR model can be used to characterize pregnancy weight gain patterns.

Inter-institutional Variation in Use of Caesarean Delivery for Labour Dystocia.

OBJECTIVE: To establish the degree of variation across hospitals in the use of Caesarean delivery for the indication of labour dystocia before and after accounting for maternal, fetal, and hospital characteristics. METHODS: This study was a retrospective, population-based cohort study of nulliparous women delivering term singletons in cephalic position following labour. Delivery visits were extracted from three provincial perinatal registries in the Canadian provinces of Ontario, Alberta, and British Columbia, from 2008-2012. Crude hospital-specific rates of Caesarean delivery for labour dystocia were reported, and these rates were then stabilized to account for hospitals with low delivery volumes. Rates were then adjusted for maternal, fetal, and hospital characteristics using hierarchical logistic regression. RESULTS: Among 403 205 women delivering at 170 hospitals, the overall Caesarean delivery rate was 21.0%, and the rate of Caesarean delivery for labour dystocia was 12.7%, indicating that 60% of all Caesarean deliveries were performed in part for this indication. The middle 95% of hospitals had Caesarean delivery rates for labour dystocia ranging from 4.5% to 24.7%. Differences in maternal case mix and hospital characteristics explained only a small proportion of this variation (95% central range 6.3%-21.7%). CONCLUSION: Considerable inter-hospital variation in rates of Caesarean delivery for labour dystocia remained after accounting for differences in maternal and hospital factors. Reporting systems that monitor variation in inter-institutional rates should incorporate stabilization and adjustment for case-mix differences and consider indication-specific rates of Caesarean delivery to more fairly compare hospital performance and better target interventions to reduce Caesarean delivery for specific indications.

Differences in obstetric care among nulliparous First Nations and non-First Nations women in British Columbia, Canada.

BACKGROUND: Canada's Aboriginal population faces significantly higher rates of stillbirth and neonatal and postnatal death than those seen in the general population. The objective of this study was to compare indicators of obstetric care quality and use of obstetric interventions between First Nations and non-First Nations mothers in British Columbia, Canada. METHODS: We linked obstetrical medical records with the First Nations Client File for all nulliparous women who delivered single infants in British Columbia from 1999 to 2011. Using logistic regression models, we examined differences in the proportion of women who received services aligned with best practice guidelines, as well as the overall use of obstetric interventions among First Nations mothers compared with the general population, controlling for geographic barriers (distance to hospital) and other relevant confounders. RESULTS: During the study period, 215,993 single births occurred in nulliparous women in British Columbia, 9152 of which were to members of our First Nations cohort. First Nations mothers were less likely to have early ultrasonography (adjusted risk difference = 10.2 fewer women per 100 deliveries [95% confidence interval {CI} -11.3 to -9.3]), to have at least 4 antenatal care visits (3.6 fewer women per 100 deliveries [95% CI -4.6 to -2.6]), and to undergo labour induction after prolonged (> 24 hours) prelabour rupture of membranes (-5.9 [95% CI -11.8 to 0.1]) or at post-dates gestation (-10.6 [95% CI -13.8 to -7.5]). Obstetric interventions including epidural, labour induction, instrumental delivery and cesarean delivery were used less often in First Nations mothers. INTERPRETATION: We identified differences in the obstetric care received by First Nations mothers compared with the general population. Such differences warrant further investigation, given increases in perinatal mortality that are consistently shown and that may be a downstream consequence of differences in care.

Beyond intention to treat: what is the right question?

BACKGROUND: Most methodologists recommend intention-to-treat (ITT) analysis in order to minimize bias. Although ITT analysis provides an unbiased estimate for the effect of treatment assignment on the outcome, the estimate is biased for the actual effect of receiving treatment (active treatment) compared to some comparison group (control). Other common analyses include measuring effects in (1) participants who follow their assigned treatment (Per Protocol), (2) participants according to treatment received (As Treated), and (3) those who would comply with recommended treatment (Complier Average Causal Effect (CACE) as estimated by Principal Stratification or Instrumental Variable Analyses). As each of these analyses compares different study subpopulations, they address different research questions. PURPOSE: For each type of analysis, we review and explain (1) the terminology being used, (2) the main underlying concepts, (3) the questions that are answered and whether the method provides valid causal estimates, and (4) the situations when the analysis should be conducted. METHODS: We first review the major concepts in relation to four nuances of the clinical question, 'Does treatment improve health?' After reviewing these concepts, we compare the results of the different analyses using data from two published randomized controlled trials (RCTs). Each analysis has particular underlying assumptions and all require dichotomizing adherence into Yes or No. We apply sensitivity analyses so that intermediate adherence is considered (1) as adherence and (2) as non-adherence. RESULTS: The ITT approach provides an unbiased estimate for how active treatment will improve (1) health in the population if a policy or program is enacted or (2) health of patients if a clinician changes treatment practice. The CACE approach generally provides an unbiased estimate of the effect of active treatment on health of patients who would follow the clinician's advice to take active treatment. Unfortunately, there is no current analysis for clinicians and patients who want to know whether active treatment will improve the patient's health if taken, which is different from the effect in patients who would follow the clinician's advice to take active treatment. Sensitivity analysis for the CACE using two published data sets suggests that the underlying assumptions appeared to be violated. LIMITATIONS: There are several methods within each analytical approach we describe. Our analyses are based on a subset of these approaches. CONCLUSIONS: Although adherence-based analyses may provide meaningful information, the analytical method should match the clinical question, and investigators should clearly outline why they believe assumptions hold and should provide empirical tests of the assumptions where possible.

Suicide prevention-related Google searches and subsequent emergency department visits in California and Arizona, 2007-2015.

Introduction: United States emergency departments (ED) visit rates for nonfatal self-harm increased by 42% from 2001 to 2016. Previous suicide mortality research has provided conflicting evidence on the use of suicide-related Internet searches as a surveillance tool for self-harm and suicidal ideation. However, few have used rigorous approaches to account for autocorrelation at the aggregate level, and none have focused on Internet searches related to suicide prevention. Methods and results: Over a 9-year study period (2007-2015), suicidality-related search data were extracted using the Google Health Application Programming Interface (API) for Arizona and California - states, chosen for their differing age distributions and rigorous ED injury coding policies. We examined several combined suicide prevention-related search queries. Using autoregressive integration moving average (ARIMA) models and a Box-Jenkins approach, we assessed whether increased prevention-related Internet searches related to suicidality are predictive of lower subsequent ED visits related to suicidal ideation with or without self-harm injury. In both states, greater prevention-related queries were associated with lower ED visits approximately four to six weeks later. Conclusions: Our results indicate that Internet-based search volumes related to suicide prevention may have the potential to monitor suicidality and online suicide prevention resources offer meaningful opportunities for mental health support.