Population-based denominators matter: Bias in U.S. Virgin Islands COVID-19 vaccination coverage under changing population counts.

PURPOSE: The U.S. Virgin Islands (USVI) receives an updated population count once every 10 years and used 2010 decennial census population counts to estimate COVID-19 vaccination coverage during the COVID-19 emergency response. We investigated whether using outdated (2010) or modeled (2020 international database [IDB]) population counts biased vaccination coverage estimates used to inform public health priorities during the 2020-2022 COVID-19 response. METHODS: We estimated percentage of USVI residents with a completed primary COVID-19 vaccination series during December 16, 2020-September 20, 2022. Vaccination coverage was calculated as number of persons who completed the vaccination series divided by 2010 and 2020 decennial census population counts and 2020 IDB intercensal estimate. RESULTS: COVID-19 vaccination coverage using the 2020 population count was 12 % points higher than coverage using 2010 denominator (2010 denominator: 51 %; 2020 denominator: 63 %). Vaccination coverage estimated using 2020 IDB was approximately equal with the 2010 decennial census estimate (52 %). CONCLUSIONS: Using 2010 and modeled population counts underestimated 2020 USVI COVID-19 vaccination coverage given the 18 % population decline during 2010-2020, potentially limiting USVI's ability to assess vaccination progress. Identifying mechanisms for more reliable population enumeration or improved estimate modeling are essential for accurately guiding USVI public health decision-making.

Severe Maternal Morbidity in Georgia, 2009-2020.

BACKGROUND: The increasing focus of population surveillance and research on maternal-and not only fetal and infant-health outcomes is long overdue. The United States maternal mortality rate is higher than any other high-income country, and Georgia is among the highest rates in the country. Severe maternal morbidity (SMM) is conceived of as a "near miss" for maternal mortality, is 50 times more common than maternal death, and efforts to systematically monitor SMM rates in populations have increased in recent years. Much of the current population-based research on SMM has occurred in coastal states or large cities, despite substantial geographical variation with higher maternal and infant health burdens in the Southeast and rural regions. METHODS: This population-based study uses hospital discharge records linked to vital statistics to describe the epidemiology of SMM in Georgia between 2009 and 2020. RESULTS: Georgia had a higher SMM rate than the United States overall (189.2 vs. 144 per 10,000 deliveries in Georgia in 2014, the most recent year with US estimates). SMM was higher among racially minoritized pregnant persons and those at the extremes of age, of lower socioeconomic status, and with comorbid chronic conditions. SMM rates were 5 to 6 times greater for pregnant people delivering infants <1500 grams or <32 weeks' gestation as compared with those delivering normal weight or term infants. Since 2015, SMM has increased in Georgia. CONCLUSION: SMM represents a collection of life-threatening emergencies that are unevenly distributed in the population and require increased attention. This descriptive analysis provides initial guidance for programmatic interventions intending to reduce the burden of SMM and, subsequently, maternal mortality in the US South.

Epidemiology beyond its limits.

In 1995, journalist Gary Taubes published an article in Science titled "Epidemiology faces its limits," which questioned the utility of nonrandomized epidemiologic research and has since been cited more than 1000 times. He highlighted numerous examples of research topics he viewed as having questionable merit. Studies have since accumulated for these associations. We systematically evaluated current evidence of 53 example associations discussed in the article. Approximately one-quarter of those presented as doubtful are now widely viewed as causal based on current evaluations of the public health consensus. They include associations between alcohol consumption and breast cancer, residential radon exposure and lung cancer, and the use of tanning devices and melanoma. This history should inform current debates about the reproducibility of epidemiologic research results.

On selection bias in comparison measures of smartphone-generated population mobility: an illustration of no-bias conditions with a commercial data source.

PURPOSE: Passively generated cell-phone location ("mobility") data originally intended for commercial use has become frequently used in epidemiologic research, notably during the COVID-19 pandemic to study the impact of physical-distancing recommendations on aggregate population behavior (e.g., average daily mobility). Given the opaque nature of how individuals are selected into these datasets, researchers have cautioned that their use may give rise to selection bias, yet little guidance exists for assessing this potential threat to validity in mobility-data research. Through an example analysis of cell-phone-derived mobility data, we present a set of conditions to guide the assessment of selection bias in measures comparing aggregate mobility patterns over time and between groups. METHODS: We specifically consider bias in measures comparing group-level mobility in the same group (difference, ratio, percent difference) and between groups (difference in differences, ratio of ratios, ratio of percent differences). We illustrate no-bias conditions in these measures through an example comparing block-group-level mobility between income groups in United States metro areas before (January 1st-March 10, 2020) and after (March 11th-April 19th, 2020) the day COVID-19 was declared a pandemic. RESULTS: Within-group contrasts describing mobility over time, especially for the higher-income decile, were expected to be most resistant to bias during the example study period. CONCLUSIONS: The presented conditions can be used to assess the susceptibility to selection bias of group-level measures comparing mobility. Importantly, they can be used even without knowledge of the degree of bias in each group at each time point. We further highlight links between no-bias principles originating in epidemiology and economics, showing that certain assumptions (e.g., parallel trends) can apply to biases beyond their original application.

Misclassification of Neonatal Abstinence Syndrome Surveillance Estimates: Is Considering the Positive Predictive Value Enough?

BACKGROUND: Validation studies estimating the positive predictive value (PPV) of neonatal abstinence syndrome (NAS) have consistently suggested overreporting in hospital discharge records. However, few studies estimate the negative predictive value (NPV). Even slightly imperfect NPVs have the potential to bias estimated prevalences of rare outcomes like NAS. Given the challenges in estimating NPV, our objective was to evaluate whether the PPV was sufficient to understand the influence of NAS misclassification bias on conclusions of the NAS prevalence in surveillance research. METHODS: We used hospital discharge data from the 2016 New Jersey State Inpatient Databases, Healthcare Cost and Utilization Project. We adjusted surveillance data for misclassification using quantitative bias analysis models to estimate the expected NAS prevalence under a range of PPV and NPV bias scenarios. RESULTS: The 2016 observed NAS prevalence was 0.61%. The misclassification-adjusted prevalence estimates ranged from 0.31% to 0.91%. When PPV was assumed to be ≥90%, the misclassification-adjusted prevalence was typically greater than the observed prevalence but the reverse was true for PPV ≤70%. Under PPV 80%, the misclassification-adjusted prevalence was less than the observed prevalence for NPV >99.9% but flipped for NPV <99.9%. CONCLUSIONS: When we varied the NPV below 100%, our results suggested that the direction of bias (over or underestimation) was dependent on the PPV, and sometimes dependent on the NPV. However, NPV was important for understanding the magnitude of bias. This study serves as an example of how quantitative bias analysis methods can be applied in NAS surveillance to supplement existing validation data when NPV estimates are unavailable.

Validation of Hypertensive Disorders During Pregnancy: ICD-10 Codes in a High-burden Southeastern United States Hospital.

BACKGROUND: Identification of hypertensive disorders in pregnancy research often uses hospital International Classification of Diseases v. 10 (ICD-10) codes meant for billing purposes, which may introduce misclassification error relative to medical records. We estimated the validity of ICD-10 codes for hypertensive disorders during pregnancy overall and by subdiagnosis, compared with medical record diagnosis, in a Southeastern United States high disease burden hospital. METHODS: We linked medical record data with hospital discharge records for deliveries between 1 July 2016, and 30 June 2018, in an Atlanta, Georgia, public hospital. For any hypertensive disorder (with and without unspecified codes) and each subdiagnosis (hemolysis, elevated liver enzymes, and low platelet count [HELLP] syndrome, eclampsia, preeclampsia with and without severe features, chronic hypertension, superimposed preeclampsia, and gestational hypertension), we calculated positive predictive value (PPV), negative predictive value (NPV) sensitivity, and specificity for ICD-10 codes compared with medical record diagnoses (gold standard). RESULTS: Thirty-seven percent of 3,654 eligible pregnancies had a clinical diagnosis of any hypertensive disorder during pregnancy. Overall, ICD-10 codes identified medical record diagnoses well (PPV, NPV, specificity >90%; sensitivity >80%). PPV, NPV, and specificity were high for all subindicators (>80%). Sensitivity estimates were high for superimposed preeclampsia, chronic hypertension, and gestational hypertension (>80%); moderate for eclampsia (66.7%; 95% confidence interval [CI] = 22.3%, 95.7%), HELLP (75.0%; 95% CI = 50.9%, 91.3%), and preeclampsia with severe features (58.3%; 95% CI = 52.6%, 63.8%); and low for preeclampsia without severe features (3.2%; 95% CI, 1.4%, 6.2%). CONCLUSIONS: We provide bias parameters for future US-based studies of hypertensive outcomes during pregnancy in high-burden populations using hospital ICD-10 codes.

Sociodemographic and Service Use Characteristics of Abortion Fund Cases from Six States in the U.S. Southeast.

Abortion funds are key actors in mitigating barriers to abortion access, particularly in contexts where state-level abortion access restrictions are concentrated. Using 2017-2019 case management data from a regional abortion fund in the southeastern U.S., we described the sociodemographic and service use characteristics of cases overall (n = 9585) and stratified by state of residence (Alabama, Florida, Georgia, Mississippi, South Carolina, and Tennessee). Overall, cases represented people seeking abortion fund assistance who predominately identified as non-Hispanic Black (81%), 18-34 years of age (84%), publicly or uninsured (87%), having completed a high school degree or some college (70%), having one or more children (77%), and as Christian (58%). Most cases involved an in-state clinic (81%), clinic travel distance under 50 miles (63%), surgical abortion (66%), and pregnancy under 13 weeks' gestation (73%), with variation across states. The median abortion fund contribution pledge was $75 (interquartile range (IQR): 60-100), supplementing median caller contributions of $200 (IQR: 40-300). These data provide a unique snapshot of a population navigating disproportionate, intersecting barriers to abortion access, and abortion fund capacity for social care and science. Findings can inform abortion fund development, data quality improvement efforts, as well as reproductive health, rights and justice advocacy, policy, and research.

Validation of ICD-10 Codes for Gestational and Pregestational Diabetes During Pregnancy in a Large, Public Hospital.

BACKGROUND: The use of billing codes (ICD-10) to identify and track cases of gestational and pregestational diabetes during pregnancy is common in clinical quality improvement, research, and surveillance. However, specific diagnoses may be misclassified using ICD-10 codes, potentially biasing estimates. The goal of this study is to provide estimates of validation parameters (sensitivity, specificity, positive predictive value, and negative predictive value) for pregestational and gestational diabetes diagnosis using ICD-10 diagnosis codes compared with medical record abstraction at a large public hospital in Atlanta, Georgia. METHODS: This study includes 3,654 deliveries to Emory physicians at Grady Memorial Hospital in Atlanta, Georgia, between 2016 and 2018. We linked information abstracted from the medical record to ICD-10 diagnosis codes for gestational and pregestational diabetes during the delivery hospitalization. Using the medical record as the gold standard, we calculated sensitivity, specificity, positive predictive value, and negative predictive value for each. RESULTS: For both pregestational and gestational diabetes, ICD-10 codes had a high-negative predictive value (>99%, Table 3) and specificity (>99%). For pregestational diabetes, the sensitivity was 85.9% (95% CI = 78.8, 93.0) and positive predictive value 90.8% (95% CI = 85, 97). For gestational diabetes, the sensitivity was 95% (95% CI = 92, 98) and positive predictive value 86% (95% CI = 81, 90). CONCLUSIONS: In a large public hospital, ICD-10 codes accurately identified cases of pregestational and gestational diabetes with low numbers of false positives.

Estimating the Unknown: Greater Racial and Ethnic Disparities in COVID-19 Burden After Accounting for Missing Race and Ethnicity Data.

BACKGROUND: Black, Hispanic, and Indigenous persons in the United States have an increased risk of SARS-CoV-2 infection and death from COVID-19, due to persistent social inequities. However, the magnitude of the disparity is unclear because race/ethnicity information is often missing in surveillance data. METHODS: We quantified the burden of SARS-CoV-2 notification, hospitalization, and case fatality rates in an urban county by racial/ethnic group using combined race/ethnicity imputation and quantitative bias analysis for misclassification. RESULTS: The ratio of the absolute racial/ethnic disparity in notification rates after bias adjustment, compared with the complete case analysis, increased 1.3-fold for persons classified Black and 1.6-fold for those classified Hispanic, in reference to classified White persons. CONCLUSIONS: These results highlight that complete case analyses may underestimate absolute disparities in notification rates. Complete reporting of race/ethnicity information is necessary for health equity. When data are missing, quantitative bias analysis methods may improve estimates of racial/ethnic disparities in the COVID-19 burden.

Measuring the missing: greater racial and ethnic disparities in COVID-19 burden after accounting for missing race/ethnicity data.

Black, Hispanic, and Indigenous persons in the United States have an increased risk of SARS-CoV-2 infection and death from COVID-19, due to persistent social inequities. The magnitude of the disparity is unclear, however, because race/ethnicity information is often missing in surveillance data. In this study, we quantified the burden of SARS-CoV-2 infection, hospitalization, and case fatality rates in an urban county by racial/ethnic group using combined race/ethnicity imputation and quantitative bias-adjustment for misclassification. After bias-adjustment, the magnitude of the absolute racial/ethnic disparity, measured as the difference in infection rates between classified Black and Hispanic persons compared to classified White persons, increased 1.3-fold and 1.6-fold respectively. These results highlight that complete case analyses may underestimate absolute disparities in infection rates. Collecting race/ethnicity information at time of testing is optimal. However, when data are missing, combined imputation and bias-adjustment improves estimates of the racial/ethnic disparities in the COVID-19 burden.

Comparison of diameter-based and image-based measures of surface area from gross placental pathology for use in epidemiologic studies.

Placental surface area is often estimated using diameter measurements. However, as many placentas are not elliptical, we were interested in the validity of these estimates. We compared placental surface area from images for 491 singletons from the Stillbirth Collaborative Research Network (SCRN) Study (416 live births, 75 stillbirths) to estimates obtained using diameter measurements. Placental images and diameters were obtained from pathologic assessments conducted for the SCRN Study and images were analyzed using ImageJ software. On average, diameter-based measures underestimated surface area by -5.58% (95% confidence interval: -30.23, 19.07); results were consistent for normal and abnormal shapes. The association between surface area and birthweight was similar for both measures. Thus, diameter-based surface area can be used to estimate placental surface area.