Misclassification Bias Research Articles

Accuracy of Administrative Database Algorithms for Hospitalized Pneumonia in Adults: a Systematic Review.

Administrative data algorithms (ADAs) to identify pneumonia cases are commonly used in the analysis of pneumonia burden, trends, etiology, processes of care, outcomes, health care utilization, cost, and response to preventative and therapeutic interventions. However, without a good understanding of the validity of ADAs for pneumonia case identification, an adequate appreciation of this literature is difficult. We systematically reviewed the quality and accuracy of published ADAs to identify adult hospitalized pneumonia cases. We reviewed the Medline, EMBase, and Cochrane Central databases through May 2020. All studies describing ADAs for adult hospitalized pneumonia and at least one accuracy statistic were included. Investigators independently extracted information about the sampling frame, reference standard, ADA composition, and ADA accuracy. Thirteen studies involving 24 ADAs were analyzed. Compliance with a 38-item study-quality assessment tool ranged from 17 to 29 (median, 23; interquartile range [IQR], 20 to 25). Study setting, design, and ADA composition varied extensively. Inclusion criteria of most studies selected for high-risk populations and/or increased pneumonia likelihood. Reference standards with explicit criteria (clinical, laboratorial, and/or radiographic) were used in only 4 ADAs. Only 2 ADAs were validated (one internally and one externally). ADA positive predictive values ranged from 35.0 to 96.5% (median, 84.8%; IQR, 65.3 to 89.1%). However, these values are exaggerated for an unselected patient population because pneumonia prevalences in the study cohorts were very high (median, 66%; IQR, 46 to 86%). ADA sensitivities ranged from 31.3 to 97.8% (median, 65.1%; IQR 52.5-72.4). ADAs for identification of adult pneumonia hospitalizations are highly heterogeneous, poorly validated, and at risk for misclassification bias. Greater standardization in reporting ADA accuracy is required in studies using pneumonia ADA for case identification so that results can be properly interpreted.

A Case-Control Study of Peripartum Cardiomyopathy Using the Rochester Epidemiology Project

BackgroundThe incidence of peripartum cardiomyopathy (PPCM) is known through referral center databases that may be affected by referral, misclassification, and other biases. We sought to determine the community-based incidence and natural history of PPCM using the Rochester Epidemiology Project. Methods and ResultsIncident cases of PPCM occurring between January 1, 1970, and December 31, 2014, were identified in Olmsted County, Minnesota. A total of 15 PPCM cases were confirmed yielding an incidence of 20.3 cases per 100,000 live births in Olmsted County, Minnesota. Clinical information, disease characteristics, and outcomes were extracted from medical records in a 27-county region of the Rochester Epidemiology Project including Olmsted County and matched in a 1:2 ratio with pregnant women without PPCM. A total of 48 women were identified with PPCM in the expanded 27-county region. There was 1 death and no transplants over a median of 7.3 years of follow-up. Six of the 23 women with subsequent pregnancies developed recurrent PPCM, all of whom recovered. Migraine and anxiety were identified as novel possible risk factors for PPCM. ConclusionsThe population-based incidence of PPCM was 20.3 cases per 100,000 live births in Olmsted County, Minnesota. Cardiovascular outcomes were generally excellent in this community cohort.

Experimenting first with e-cigarettes versus first with cigarettes and transition to daily cigarette use among adolescents: the crucial effect of age at first experiment.

Most studies in English-speaking countries have found a positive association between e-cigarette experimentation and subsequent daily tobacco smoking among adolescents. However, this result may not be valid in other cultural contexts; in addition, few studies have assessed whether this association varies with the subject' age at the time of e-cigarette experimentation. This study aimed to estimate the association between experimenting first with e-cigarette (rather than tobacco) and subsequent daily smoking according to age at the time of experimentation. Secondary analysis; risk ratios (RRs) computed using modified Poisson regressions with inverse probability weighting. A cross-sectional nation-wide representative survey performed in 2017 in France. French adolescents (n=24 111), aged 17 to 18.5years, who had previously experimented with either e-cigarettes or tobacco. Exposure was defined as the experimentation with e-cigarettes first (whether or not followed by experimentation with tobacco); the outcome as daily tobacco smoking at the time of data collection. Gender, age, literacy, socio-economic status, pre-exposure repeat school years and experimentation with drunkeness, 3 licit and 8 illicit drugs were adjusted for. Uncertainties about the sequence of events defining exposure were handled by the definition of three patterns of exposure, to avoid a misclassification bias. Exposure reduced the risk of transition to daily smoking: RR=0.58, 95% confidence interval (CI)=0.54, 0.62. This effect increased in a linear manner with age at exposure (RR=0.87, 95% CI=0.78; 0.98 for 1year, P<0.001): from RR=1.30, 95% CI=1.09; 1.54 at age 9 to RR=0.38, 95% CI=0.32; 0.45 at age 17. Experimenting with e-cigarettes first (as opposed to tobacco first) appears to be associated with a reduction in the risk of daily tobacco smoking among French adolescents aged 17-18.5, but this risk varies negatively with age at experimentation, and early e-cigarette experimenters are at higher risk.

Systematic Literature Review of the Epidemiology of Advanced Prostate Cancer and Associated Homologous Recombination Repair Gene Alterations.

To characterize the global epidemiology of metastatic castration-sensitive prostate cancer (mCSPC), nonmetastatic castration-resistant prostate cancer (nmCRPC) and metastatic castration-resistant prostate cancer (mCRPC). Additionally, to assess the prevalence of homologous recombination repair gene alterations (HRRm) and their prognostic impact in advanced disease setting. A systematic literature review of real-world evidence published from January 2009 through May 2019 was conducted to assess global epidemiology and clinical practice trends for mCSPC, nmCRPC, mCRPC and HRRm; 4,732 papers were systematically screened for inclusion. Ten conference proceedings from 2014 through 2019 were reviewed. Of the screened articles 22 relevant publications were identified for this paper. Six publications reported global epidemiology of advanced prostate cancer. The prevalence of nmCRPC was estimated as 1.1% to 12.3% of prostate cancer cases and for mCRPC 1.2% to 2.1% of prostate cancer cases. No mCSPC prevalence was captured. Sixteen publications investigated HRRm prevalence in advanced prostate cancer with the majority conducted in mCRPC assessed using next-generation sequencing of tissue and germline samples. In mCRPC, the highest prevalence HRRm in both germline (3.3%-6.0%) and somatic (5.0%-15.1%) was BRCA2. Five publications reported the prognostic impact of HRRm in advanced prostate cancer. Published real-world evidence quantifying the prevalence of advanced prostate cancer and HRRm beyond mCRPC is sparse. Published data on HRRm, specifically BRCA2, are consistent with published clinical trial data for poly (ADP-ribose) polymerase inhibitors in mCRPC. In mCRPC, real-world evidence suggests that patients with HRRm have different clinical outcomes to noncarriers. More data are needed to better understand real-world patient segmentation and clinical outcomes for biomarkers given increasing interest in profiling.

Pharmacoepidemiology for nephrologists (part 2): potential biases and how to overcome them.

Observational pharmacoepidemiological studies using routinely collected healthcare data are increasingly being used in the field of nephrology to answer questions on the effectiveness and safety of medications. This review discusses a number of biases that may arise in such studies and proposes solutions to minimize them during the design or statistical analysis phase. We first describe designs to handle confounding by indication (e.g. active comparator design) and methods to investigate the influence of unmeasured confounding, such as the E-value, the use of negative control outcomes and control cohorts. We next discuss prevalent user and immortal time biases in pharmacoepidemiology research and how these can be prevented by focussing on incident users and applying either landmarking, using a time-varying exposure, or the cloning, censoring and weighting method. Lastly, we briefly discuss the common issues with missing data and misclassification bias. When these biases are properly accounted for, pharmacoepidemiological observational studies can provide valuable information for clinical practice.

Clinical pathways and outcomes of patients with Barrett's esophagus in tertiary care settings: a prospective longitudinal cohort study in Australia, 2008-2016.

Clinical services for Barrett's esophagus have been rising worldwide including Australia, but little is known of the long-term outcomes of such patients. Retrospective studies using data at baseline are prone to both selection and misclassification bias. We investigated the clinical characteristics and outcomes of Barrett's esophagus patients in a prospective cohort. We recruited patients diagnosed with Barrett's esophagus in tertiary settings across Australia between 2008 and 2016. We compared baseline and follow-up epidemiological and clinical data between Barrett's patients with and without dysplasia. We calculated age-adjusted incidence rates and estimated minimally and fully adjusted hazard ratios (HR) to identify those clinical factors related to disease progression. The cohort comprised 268 patients with Barrett's esophagus (median follow-up 5years). At recruitment, 224 (84%) had no dysplasia, 44 (16%) had low-grade or indefinite dysplasia (LGD/IND). The age-adjusted incidence of esophageal adenocarcinoma (EAC) was 0.5% per year in LGD/IND compared with 0.1% per year in those with no dysplasia. Risk of progression to high-grade dysplasia/EAC was associated with prior LGD/IND (fully adjusted HR 6.55, 95% confidence interval [CI] 1.96-21.8) but not long-segment disease (HR 1.03, 95%CI 0.29-3.58). These prospective data suggest presence of dysplasia is a stronger predictor of progression to cancer than segment length in patients with Barrett's esophagus.

Unsterile injection equipment associated with HIV outbreak and an extremely high prevalence of HCV-A case-control investigation from Unnao, India.

The integrated counseling and testing center (ICTC) located in the district hospital, Unnao in the northern state of Uttar Pradesh (UP), India witnessed an increased detection of HIV among its attendees in July 2017. Subsequently, health camps were organized by the UP State AIDS Control Society in the villages and townships contributing to such detection. We conducted a case-control study to identify factors associated with this increased detection; 33 cases and 125 controls were enrolled. Cases were individuals, detected HIV sero-reactive during November 2017-April 2018 from three locations namely Premganj, Karimuddinpur and Chakmeerapur in the Bangarmau block of the district of Unnao. Controls hailed from the same geographical setting and tested HIV sero-nonreactive either in health camps or at ICTC centers from where the cases were detected. Misclassification bias was avoided by confirming HIV sero-status of both cases as well as controls prior to final analysis. Study participants were interviewed on various risk practices and invasive treatment procedures. They were also tested for HIV and other bio-markers reflecting unsafe injecting and sexual exposures such as hepatitis B surface antigen (HBsAg), anti-HCV antibody (HCV Ab), anti-herpes simplex-2 Immunoglobulin G (HSV-2 IgG) and rapid plasma regain (RPR) test for syphilis. Secondary data analysis on three time points during 2015 through 2018 revealed a rising trend of HIV among attendees of the ICTCs (ICTC-Hasanganj, ICTC-Unnao district hospital and ICTC- Nawabganj) catering to the entire district of Unnao. While there was a seven fold rise of HIV among ICTC attendees of Hasanganj (χ2 value for trend 23.83; p < 0.001), the rise in Unnao district hospital was twofold (χ2 value for trend 4.37; p < 0.05) and was tenfold at ICTC-Nawabganj (χ2 value for trend 5.23; p < 0.05) indicating risk of infection prevailing throughout the district. Primary data was generated through interviews and laboratory investigations as mentioned above. The median age of cases and controls was 50 year (minimum 18 -maximum 68; IQR 31-57) and 38 year (minimum 18 -maximum 78; IQR 29-50) respectively. Thirty six percent of the cases and 47% of controls were male. A significantly higher proportion of cases (85%) had HCV Ab compared to controls (56%; OR 4.4, 95% CI 1.5-12.1); none reported injection drug use. However, cases and controls did not differ significantly regarding presence of HSV-2 IgG (6% versus 8% respectively). Neither any significant difference existed between cases and controls in terms of receiving blood transfusion, undergoing invasive surgical procedures, tattooing, tonsuring of head or skin piercing. In multivariate logistic regression model, 'unsafe injection exposure during treatment-seeking'(AOR 6.61, 95% CI 1.80-24.18) and 'receipt of intramuscular injection in last five years' (AOR 7.20, 95% CI 1.48-34.88) were independently associated with HIV sero-reactive status. The monophyletic clustering of HIV sequences from 14 cases (HIV-1 pol gene amplified) indicated a common ancestry. Availability of auto-disabled syringes and needles, empowerment of the local communities and effective regulatory practices across care settings would serve as important intervention measures in this context.

Separating the genetics of childhood and adult obesity: a validation study of genetic scores for body mass index in adolescence and adulthood in the HUNT Study.

From a life-course perspective, genetic and environmental factors driving childhood obesity may have a lasting influence on health later in life. However, how obesity trajectories vary throughout the life-course remains unknown. Recently, Richardson et al. created powerful early life and adult gene scores for body mass index (BMI) in a comprehensive attempt to separate childhood and adult obesity. The childhood score was derived using questionnaire-based data administered to adults aged 40–69 regarding their relative body size at age 10, making it prone to recall and misclassification bias. We therefore attempted to validate the childhood and adult scores using measured BMI data in adolescence and adulthood among 66 963 individuals from the HUNT Study in Norway from 1963 to 2019. The predictive performance of the childhood score was better in adolescence and early adulthood, whereas the predictive performance of the adult score was better in adulthood. In the age group 12–15.9 years, the variance explained by the childhood polygenic risk score (PRS) was 6.7% versus 2.4% for the adult PRS. In the age group 24–29.9 years, the variance explained by the adult PRS was 3.9% versus 3.6% for the childhood PRS. Our findings support that genetic factors driving BMI differ at young age and in adulthood. Within the framework of multivariable Mendelian randomization, the validated childhood gene score can now be used to determine the consequence of childhood obesity on later disease.

Statistical inference for association studies using electronic health records: handling both selection bias and outcome misclassification.

Health research using electronic health records (EHR) has gained popularity, but misclassification of EHR-derived disease status and lack of representativeness of the study sample can result in substantial bias in effect estimates and can impact power and type I error. In this paper, we develop new strategies for handling disease status misclassification and selection bias in EHR-based association studies. We first focus on each type of bias separately. For misclassification, we propose three novel likelihood-based bias correction strategies. A distinguishing feature of the EHR setting is that misclassification may be related to patient-varying factors, and the proposed methods leverage data in the EHR to estimate misclassification rates without gold standard labels. For addressing selection bias, we describe how calibration and inverse probability weighting methods from the survey sampling literature can be extended and applied to the EHRsetting. Addressing misclassification and selection biases simultaneously is a more challenging problem than dealing with each on its own, and we propose several new strategies. For all methods proposed, we derive valid standard error estimators and provide software for implementation. We provide a new suite of statistical estimation and inference strategies for addressing misclassification and selection bias simultaneously that is tailored to problems arising in EHR data analysis. We apply these methods to data from The Michigan Genomics Initiative, a longitudinal EHR-linkedbiorepository.

Systemic adverse effects from inhaled corticosteroid use in asthma: a systematic review

BackgroundOral corticosteroid use increases the risk of systemic adverse effects including osteoporosis, bone fractures, diabetes, ocular disorders and respiratory infections. We sought to understand if inhaled corticosteroid (ICS) use in...

Gestational Weight Gain and Long-term Maternal Obesity Risk: A Multiple-Bias Analysis

Lifecourse research provides an important framework for chronic disease epidemiology. However, data collection to observe health characteristics over long periods is vulnerable to systematic error and statistical bias. We present a multiple-bias analysis using real-world data to estimate associations between excessive gestational weight gain and mid-life obesity, accounting for confounding, selection, and misclassification biases. Participants were from the multiethnic Study of Women's Health Across the Nation. Obesity was defined by waist circumference measured in 1996-1997 when women were age 42-53. Gestational weight gain was measured retrospectively by self-recall and was missing for over 40% of participants. We estimated relative risk (RR) and 95% confidence intervals (CI) of obesity at mid-life for presence versus absence of excessive gestational weight gain in any pregnancy. We imputed missing data via multiple imputation and used weighted regression to account for misclassification. Among the 2,339 women in this analysis, 937 (40%) experienced obesity in mid-life. In complete case analysis, women with excessive gestational weight gain had an estimated 39% greater risk of obesity (RR = 1.4, CI = 1.1, 1.7), covariate-adjusted. Imputing data, then weighting estimates at the guidepost values of sensitivity = 80% and specificity = 75%, increased the RR (95% CI) for obesity to 2.3 (2.0, 2.6). Only models assuming a 20-point difference in specificity between those with and without obesity decreased the RR. The inference of a positive association between excessive gestational weight gain and mid-life obesity is robust to methods accounting for selection and misclassification bias.

The impact of periodontitis exposure misclassification bias from partial-mouth measurements on association with diabetes and cardiovascular disease.

To quantify exposure misclassification bias arising from use of partial-mouth protocols in studies of periodontitis-systemic disease associations. Using data from 10,134 adults participating in the National Health and Nutrition Examination Survey, we classified periodontal status based on full-mouth clinical examinations and three commonly used partial-mouth protocols. Associations between periodontitis and self-reported diabetes and cardiovascular disease were evaluated under each protocol using adjusted logistic regression. Percent relative bias was calculated to evaluate magnitude and direction of bias. Misclassification primarily resulted in underestimation of associations, the extent of which depended on both the outcome under study and exposure severity. Bias due to misclassification of severe periodontitis was negligible for cardiovascular disease (0%-4.1%) compared to diabetes (177.7%-234.1%). In contrast, bias in moderate periodontitis associations was comparable across each outcome-diabetes (28.4%-39.5%) and cardiovascular disease (8.9%-46.7%). Results did not meaningfully change based on the partial-mouth protocol implemented. Stratified analyses showed increased bias among those with ≤15 teeth. Use of mean attachment loss as a continuous exposure resulted in minimal-to-no bias. Exposure misclassification bias due to use of partial-mouth protocols can yield inaccurate conclusions about periodontitis-systemic disease associations, the extent of which may depend on periodontitis classification and the association under study.

Venous Thromboembolism Incidence and Impact on Survival in Patients with Brain Metastases

Background: Patients with brain metastases are considered especially high risk for venous thromboembolism (VTE). The incidence of VTE following a diagnosis of brain metastases and the impact of VTE on overall survival among patients with brain metastases have not been well characterized. Methods: Using the linked Surveillance, Epidemiology and End Results (SEER) and Medicare databases between 2008-2015, we identified all patients ≥66 years with a primary diagnosis of lung, renal cell, breast cancer or melanoma and brain metastases (ICD-9-CM 198.3 present anytime in Medicare claims). Patients were required to have continuous Medicare part A and B enrollment from 12 months before and 12 months after diagnosis or until death. The primary outcome was incidence of VTE (deep vein thrombosis or pulmonary embolism or both) in the 12- months following a diagnosis of brain metastases. Patients with a VTE diagnosis prior to brain metastases were excluded. Incidence rates were calculated as VTE events per 100 patient years. Tumor specific multivariable Cox proportional hazards model adjusting for age, sex, race, time from primary cancer diagnosis to brain metastasis, comorbidities and VTE as a time varying covariate was used to assess the association of VTE and 12-month overall survival following a diagnosis of brain metastasis. Results: Among 154,918 individuals with a primary cancer diagnosis of interest, 21,623 (14.0%) had a brain metastasis diagnosis. Majority of patients with brain metastases had a primary diagnosis of lung cancer (N=19,141 (88.5%)), followed by melanoma (N=1,607 (7.4%)), renal cell cancer (N=576 (2.7%)) and breast cancer (299 (1.4%)). The baseline characteristics are highlighted in Table 1. Among these patients, VTE was diagnosed in 3,432 (15.9%). In the year following a diagnosis of brain metastasis, the incidence of VTE per 100 patient-years was highest among those with a diagnosis of renal cell cancer (19.4), followed by breast (16.3), lung (16.0) and melanoma (12.5). The median time to VTE from a diagnosis of brain metastasis is 1 month (IQR 0-3 months) and the cumulative incidence of VTE stratified by tumor type and accounting for the competing risk of death over 12 months is illustrated in Figure 1.The median overall survival following a diagnosis of brain metastasis was 2.7 months (95% CI 2.1-3.1 months). In a multivariable Cox proportional hazards model, a diagnosis of VTE following brain metastasis was associated with increased risk of death in patients with lung cancer (HR 1.23 (95% CI 1.18-1.28)), renal cell cancer (HR 1.36 (95% CI 1.06-1.75)) and melanoma (HR 1.32 (95% CI 1.12-1.55)), but not breast cancer (HR 1.16 (95% CI 0.80-1.69)). Conclusions: This large real-world retrospective cohort analysis demonstrates and quantifies the high tumor specific incidences of VTE among individuals with brain metastases. The VTE incidence rates observed are comparable to metastatic pancreatic cancer and highest in the initial 3 months following a diagnosis of brain metastasis. Furthermore, a diagnosis of VTE is associated with a reduced 12-month overall survival in this patient population. We are mindful to highlight the inherent limitations of claims data in identifying clinical events such as brain metastases and VTE, thus possible misclassification and selection biases should be considered when interpreting these findings. However, considering the complexity of managing VTE in this high-risk patient population and the high thrombotic burden demonstrated, these findings warrant further prospective validation. Disclosures Key: Takeda: Research Funding; Grifols: Research Funding; Uniqure: Consultancy; Novo Nordisk: Other: Chair of Grants Committee. Lund:GlaxoSmithKline: Current Employment, Current equity holder in publicly-traded company; AbbVie: Research Funding.

Analysis of hidradenitis suppurativa hospitalizations: A report from the National Inpatient Sample database

Analysis of hidradenitis suppurativa hospitalizations: A report from the National Inpatient Sample database

The Association between Alcohol Drinking Patterns and Health-Related Quality of Life in the Korean Adult Population: Effects of Misclassification Error on Estimation of Association.

The purpose of this study was to investigate the association between drinking patterns and health-related quality of life (HRQoL) in the Korean general population and to validate the estimation of the association of alcohol use on HRQoL when former drinkers are separated from never drinkers and low-risk drinkers depending on gender. Data were collected from 23,055 adults (over 19 years old) who completed the Korean National Health and Nutritional Examination Survey (2010–2013). Multivariate logistic and linear regression analyses were performed to investigate the association between drinking patterns and HRQoL. When former drinkers were separated from never drinkers and low-risk drinkers to control for misclassification bias, there were gender differences in the associations between alcohol use and HRQoL. Although the estimation of the association of alcohol use was not valid in men, the estimation of association was valid in women, as low-risk women drinkers showed better HRQoL than nondrinkers. Therefore, when conducting research related to alcohol and health, analyses should correct for the various confounding variables and minimize the misclassification bias of drinking patterns. It is also necessary to consider gender characteristics when intervening to improve HRQoL related to drinking.

Ondansetron use in early pregnancy and the risk of miscarriage

Ondansetron is commonly used to treat nausea and vomiting in pregnancy despite inconclusive evidence of its safety. Previous studies have reported no increase in risk of miscarriage but relied on methods that failed to account for gestational weeks at risk and non-user comparators, which may increase the potential for unmeasured confounding. Our objective was to estimate the risk of miscarriage among women prescribed ondansetron vs alternative antiemetics during the first 20 weeks of pregnancy. A pregnancy cohort was created using electronic health record data from a health care system in North Carolina. Women were classified as exposed to either ondansetron or comparator antiemetics (metoclopramide or promethazine) based on the first antiemetic prescription received in the first 20 weeks of gestation. Cumulative incidence of miscarriage at 20 weeks was estimated in each antiemetic group. Hazard ratios (HR) were estimated with 95% confidence intervals and measured confounding was controlled using inverse probability of treatment weights. Sensitivity analyses assessed the potential impact of exposure misclassification, latency period, and selection bias. We identified 2620 eligible pregnancies with antiemetic orders; 65% had a first ondansetron order and 35% had a first comparator antiemetic order. In total, 95 women had a miscarriage. After adjustment, there was no difference in risk of miscarriage (HR 1.21, 95% CI 0.77, 1.90). Results from the per-protocol and other sensitivity analyses were similar to the main analysis. We did not observe an increase in the risk of miscarriage for pregnancies exposed to ondansetron vs comparator antiemetics.

Embracing uncertainty to reduce bias in hydroacoustic species apportionment

Species-specific abundance estimates lend insight into fish ecology, inform sustainable fisheries management, and remain the goal of hydroacoustic fisheries sampling. The apportionment process, assigning aggregate hydroacoustic data to individual species, is affected by uncertainties in both hydroacoustic and species composition data. These uncertainties have associated biases that can propagate through the apportionment process and degrade abundance estimates. We developed an apportionment procedure that reduces the influence of sampling, threshold, and misclassification biases leading to more accurate species-specific abundance estimates. We applied our method to Lake Erie walleye, using paired hydroacoustic and gillnet sampling data, and generated distribution and abundance estimates supported by known ecological patterns. Flexibility in this approach performed better than traditional threshold methods that ignore uncertainty in catch composition, threshold choice, and target-strength uncertainty.

Business Data Categorization and Refinement for Application in Longitudinal Neighborhood Health Research: a Methodology.

Retail environments, such as healthcare locations, food stores, and recreation facilities, may be relevant to many health behaviors and outcomes. However, minimal guidance on how to collect, process, aggregate, and link these data results in inconsistent or incomplete measurement that can introduce misclassification bias and limit replication of existing research. We describe the following steps to leverage business data for longitudinal neighborhood health research: re-geolocating establishment addresses, preliminary classification using standard industrial codes, systematic checks to refine classifications, incorporation and integration of complementary data sources, documentation of a flexible hierarchical classification system and variable naming conventions, and linking to neighborhoods and participant residences. We show results of this classification from a dataset of locations (over 77 million establishment locations) across the contiguous U.S. from 1990 to 2014. By incorporating complementary data sources, through manual spot checks in Google StreetView and word and name searches, we enhanced a basic classification using only standard industrial codes. Ultimately, providing these enhanced longitudinal data and supplying detailed methods for researchers to replicate our work promotes consistency, replicability, and new opportunities in neighborhood health research.

Bias correction in self-reported high blood pressure prevalence based on objectively measured data

Abstract Background Reliable and precise estimates of high blood pressure (HBP) prevalence are essential to inform decision-making and policies evaluation. Self-reported HBP may be underestimated by surveys due to misclassification of health status by participants. Misclassification may lead to inaccurate inference. We aimed to assess a feasibility of correcting misclassification bias in self-reported HBP in the Portuguese component of the European Health Interview Survey (INS2014) using data on objective blood pressure measurements from a smaller health examination survey (INSEF). Methods We assumed that “true” measured HBP status was missing at random for INS2014 participants (n = 13937). Using a multiple imputation method of logistic regression for monotone missing data patterns, HBP was imputed for INS2014 sample. Auxiliary data on measured HBP from INSEF (n = 4910) was used. Crude and corrected HBP prevalence rates stratified by sex and age group were calculated. Results HBP prevalence based on INS2014 increased significantly after multiple imputation correction, from 22.0% [CI95%:20.6-23.5%] to 38.8% [35.4-41.9%] for men and from 26.8% [25.3-28.3%] to 33.0% [30.5-35.4%] for women. Corrected estimates were similar to the objectively measured from INSEF, 39.6% for men and 32.7% for women. Corrected HBP prevalence in INS2014 increased from 4.1% [2.9-5.9%] to 7.9% [4.0-11.8%] in 25-34 years old (yo); from 9.3% [7.8-11.0%] to 17.6% [14.0-21.3%] in 35-44 yo; from 22.5% [20.3-24.8%] to 38.7% [34.2-43.2%] in 45-54 yo; from 39.8% [37.2-42.4%] to 53.8% [49.1-58.5%] in 55-64 yo and from 54.3% [51.5-57.1%] to 67.7% [63.1-72.3%] in 65-74 yo. Corrected rates were similar to the obtained by INSEF (5.7%, 17.0%, 35.8%, 54.8% and 71.3%, respectively). Conclusions Our results highlight the importance of bias analysis when using self-reported data on HBP. Multiple imputation is a feasible approach to adjust misclassification bias in health outcomes collected by population surveys. Key messages The magnitude and direction of misclassification bias in self-reported health outcomes should be investigated. Multiple imputation is a feasible approach to misclassification bias correction in self-reported survey data.

Estimating diet costs: Bridging the gap between food supply price databases and dietary intake data

Abstract Introduction The cost of food is a key influence on diet. The majority of diet cost studies match intake data from population-based surveys to a single source of food supply prices such as the Consumer Price Index (CPI). Our aim was to examine the nutritional significance of using food supply data to price dietary intakes in Canada. Methods We examined food groups and nutrients in dietary intakes captured by the CPI. For prices, we used 2015 Canadian CPI average monthly item prices. For dietary intakes, we used reported intakes from the 2015 Canadian Community Health Survey (CCHS)-Nutrition, 1st 24-hour recall (n = 20,487). i) 2015 CPI item prices ($/g) were matched to the 156 food items from the 2015 CCHS-Nutrition as full, partial, or non-match; ii) CPI capture (full or partial match) per total intake (g), without water, was calculated for each respondent; iii) descriptive statistics and quantile regression (α = 0.05) were used to compare intakes of Canadian Nutrient File food groups and nutrients by quantile of CPI capture. Results The CPI captured on average 74% of total dietary intake (g) without water. A greater proportion of protein and fat intake was captured by the CPI as compared to carbohydrate, sodium, fibre, and sugar intake. Intakes of beef, poultry, sausages, pork, and breakfast foods had among the best match; snack foods, nuts, veal, and alcoholic beverages had among the worst. Individuals in the poorest CPI capture quantile consumed the greatest fibre (g), carbohydrates (g), total sugar (g), fat (g), protein (g), and energy (kcal) as compared to those with best CPI capture. Conclusions The poorest quantile of CPI capture reflects individuals with high intakes of nutrients of concern including fat, carbohydrates, and sugar; potential bias in estimating fibre and protein intake was also detected. Researchers and decision makers should attend to differential misclassification bias and opportunities for tailored datasets to price dietary intakes. Key messages Given the proliferation of diet cost studies using food supply prices, this novel study highlights the importance of understanding the biases in using food supply data to price dietary intakes. Nutrition researchers and decision makers can use these findings to strengthen food supply price data to support the monitoring of diet costs in relation to diet quality and health outcomes.