Age heaping among individuals in selected South Asian countries: evidence from Demographic and Health Surveys

Age misreporting is a common phenomenon in Demographic and Health Surveys, and there are numerous reasons for this. The trend and pattern of disparity in age heaping vary between countries. The present study assesses age heaping in the selected South Asian countries of Afghanistan, India, Nepal, Bangladesh and Pakistan using data from the most recent round of the Demographic and Health Survey. The respondent sample sizes were 203,703 for Afghanistan, 2,869,043 for India, 49,064 for Nepal, 81,618 for Bangladesh and 100,868 for Pakistan. Age heaping was assessed by respondent’s age, education level, sex and level of education. Whipple’s index was calculated to assess systematic heaping on certain ages as a result of digit preference. Bangladesh, Afghanistan and India showed stronger preference for ages ending with the digits ‘0’ and ‘5’ compared with Pakistan and Nepal among uneducated respondents. On the other hand, strong avoidance of ages ending in the digits ‘1’, ‘4’ and ‘9’ was observed in Bangladesh, Afghanistan and India. However, urban–rural place of residence was not found to be associated with digit preference in the study countries. Among males, age misreporting with the final digits ‘0’ and ‘5’ was highest in Bangladesh, followed by Afghanistan and India, and Nepal showed the least displacement. Strong digit preference and avoidance, and upper age displacement, were witnessed in the surveys conducted in Bangladesh, Afghanistan and India on the parameters of sex and education level. Innovative methods of data collection with the measurement and minimization of errors using statistical techniques should be used to ensure accuracy of age data.

Digit Preference and Biased Conclusions in Cardiac Arrest Studies

IntroductionIn cardiac arrest (CA), time is directly predictive of patient prognosis. The increase in mortality resulting from delayed cardiopulmonary resuscitation has been quantified minute by minute. Times reported in such a situation could reflect a bias referred to as “digit preference”. This phenomenon leads to privilege certain numerical values (like 2, 5, or 10) over others (like 13). We investigated this bias in times reported during cardiac arrest management in a national register. MethodsWe analyzed data from the French National Electronic Registry of Cardiac Arrests. All the data, including the twelve times corresponding to the main steps of the management of cardiac arrest are prospectively reported by the emergency physician who managed the patients in prehospital settings. The frequency of times as multiples of 15 (0, 15, 30, and 45) was our primary end-point.Results 47,211 times related to 6,131 cardiac arrests were analyzed. The most overrepresented numbers were: 0, with 3,737 occurrences (8% vs 2% expected, p<0,0001) and 30, with 2,807 occurrences (6% vs 2% expected, p<0,0001). Times as multiples of 5 and 10 were overrepresented (52% vs 20% and 10% expected, p<0,0001).ConclusionProspectively collected time were considerably influenced by digit preference phenomenon. Studies that are not based on automatic recording of times as well as studies that have not evaluated and considered this bias should be interpretated with caution.

Measurement of Head Circumference: Implications for Microcephaly Surveillance in Zika-Affected Areas

Worldwide recognition of the Zika virus outbreak in the Americas was triggered by an unexplained increase in the frequency of microcephaly. While severe microcephaly is readily identifiable at birth, diagnosing less severe cases requires comparison of head circumference (HC) measurement to a growth chart. We examine measured values of HC and digit preference in those values, and, by extension, the prevalence of microcephaly at birth in two data sources: a research study in Honduras and routine surveillance data in Uruguay. The Zika in Pregnancy in Honduras study enrolled pregnant women prenatally and followed them until delivery. Head circumference was measured with insertion tapes (SECA 212), and instructions including consistent placement of the tape and a request to record HC to the millimeter were posted where newborns were examined. Three indicators of microcephaly were calculated: (1) HC more than 2 standard deviations (SD) below the mean, (2) HC more than 3 SD below the mean (referred to as “severe microcephaly”) and (3) HC less than the 3rd percentile for sex and gestational age, using the INTERGROWTH-21st growth standards. We compared these results from those from a previous analysis of surveillance HC data from the Uruguay Perinatal Information System (Sistema Informático Perinatal (SIP). Valid data on HC were available on 579 infants, 578 with gestational age data. Nine babies (1.56%, 95% CI 0.71–2.93) had HC < 2SD, including two (0.35%, 95% CI 0.04–1.24) with HC < 3SD, and 11 (1.9%, 95% CI, 0.79–3.02) were below the 3rd percentile. The distribution of HC showed strong digit preference: 72% of measures were to the whole centimeter (cm) and 19% to the half-cm. Training and use of insertion tapes had little effect on digit preference, nor were overall HC curves sufficient to detect an increase in microcephaly during the Zika epidemic in Honduras. When microcephaly prevalence needs to be carefully analyzed, such as during the Zika epidemic, researchers may need to interpret HC data with caution.

Retrospective Study of Investigation of Possible Predictors for Total Fertility Rate in India

Total fertility rate (TFR) is the most acceptable and widely used measure of current fertility. Since TFR is based on age-specific fertility rate which required the total number of births in different age groups as well as age of female. When the population is illiterate or older then the information on age may have some recall bias, misreporting digit preference etc., thus in this situation TFR may departed from the actual. Therefore, need some indirect methodology which enables us to have an idea about the estimation of TFR. In this study an attempt has been made to identify some predictors that the explain TFR and try to suggest the best combination of predictors to get estimate of TFR. The methodology used in this study is essentially based on the regression technique. The identification and acceptance of possible predictors are based on the coefficient of determination. The data for the major states of India from National Family Health Survey (NFHS 4) is used for the analysis.

Digit Preference in Office Blood Pressure Measurements, United States 2015–2019

Background Blood pressure (BP) measurement error may lead to under- or overtreatment of hypertension. One common source of error is terminal digit preference, most often a terminal digit of ‘0’. The objective was to evaluate national trends in terminal digit preference in office BP measurements among adults with treated hypertension. Methods Data were from IQVIA’s National Disease and Therapeutic Index, a nationally representative, serial cross-sectional survey of office-based physicians. The analysis included office visits from 2015 to 2019 among adults aged ≥18 years receiving antihypertensive treatment. Annual trends were examined in the percent of systolic and diastolic BP measurements ending in zero by patient sex, age, and race/ethnicity, physician specialty, and first or subsequent hypertension treatment visit. Results From 2015 to 2019, there were ~60 million hypertension treatment visits annually (unweighted N: 5,585-9,085). There was a decrease in the percent of visits with systolic (41.7% to 37.7%) or diastolic (42.7% to 37.8%) BP recordings ending in zero. Trends were similar by patient characteristics. However, a greater proportion of measurements ended in zero among patients aged ≥80 (vs 15-59 or 60-79) years, first (vs subsequent) treatment visits, visits to cardiologists (vs primary care physicians), and visits with systolic BP ≥140 or diastolic BP ≥90 (vs &lt;140/90) mmHg. Conclusions Despite modest improvement, terminal digit preference remains a common problem in office BP measurement in the U.S. Without bias, 10-20% of measurements are expected to end in zero. Reducing digit preference is a priority for improving BP measurement accuracy and hypertension management.

Comparison of anthropometric data quality in children aged 6-23 and 24-59 months: lessons from population-representative surveys from humanitarian settings

Background Ensuring the quality of anthropometry data is paramount for getting accurate estimates of malnutrition prevalence among children aged 6–59 months in humanitarian and refugee settings. Previous reports based on data from Demographic and Health Surveys suggested systematic differences in anthropometric data quality between the younger and older groups of preschool children. Methods We analyzed 712 anthropometric population-representative field surveys from humanitarian and refugee settings conducted during 2011–2018. We examined and compared the quality of five anthropometric indicators in children aged 6–23 months and children aged 24–59 months: weight for height, weight for age, height for age, body mass index for age and mid-upper arm circumference (MUAC) for age. Using the z-score distribution of each indicator, we calculated the following parameters: standard deviation (SD), percentage of outliers, and measures of distribution normality. We also examined and compared the quality of height, weight, MUAC and age measurements using missing data and rounding criteria. Results Both SD and percentage of flags were significantly smaller on average in older than in younger age group for all five anthropometric indicators. Differences in SD between age groups did not change meaningfully depending on overall survey quality or on the quality of age ascertainment. Over 50% of surveys overall did not deviate significantly from normality. The percentage of non-normal surveys was higher in older than in the younger age groups. Digit preference score for weight, height and MUAC was slightly higher in younger age group, and for age slightly higher in the older age group. Children with reported exact date of birth (DOB) had much lower digit preference for age than those without exact DOB. SD, percentage flags and digit preference scores were positively correlated between the two age groups at the survey level, such as those surveys showing higher anthropometry data quality in younger age group also tended to show higher quality in older age group. Conclusions There should be an emphasis on increased rigor of training survey measurers in taking anthropometric measurements in the youngest children. Standardization test, a mandatory component of the pre-survey measurer training and evaluation, of 10 children should include at least 4–5 children below 2 years of age.

Using 3D Imaging Technology to Accurately Collect Postmortem Anthropometric Measurements in Children Under 5 Years of Age

Objectives The Child Health and Mortality Prevention Surveillance Network (CHAMPS) aims to identify causes of under-5 mortality in sub-Saharan African and South Asian surveillance sites. To address challenges in postmortem nutritional assessment, we evaluated anthropometry training and 3D imaging in the CHAMPS Kenya site. Methods Staff were trained using World Health Organization (WHO) recommended standard anthropometry equipment as well as 3D imaging to collect postmortem measurements. Following the training, 76 cases were measured in duplicate using standard anthropometry and 3D imaging and were compared to 75 pre-intervention cases. Outcomes included data quality metrics [standard deviations (SD), digit preference, % biologically implausible values (BIV, Length-for-age z-score (LAZ) BIV = ± 6 SD), measurement reliability (technical errors of measurement, TEM), and accuracy (correlation coefficients and Bland Altman plots of standard vs. 3D scan measurements). We used both WHO growth standard and internal standardization to produce sex and age-specific LAZ. Results Standard anthropometry data quality improved as indicated by digit preference (all measures rounded to 0.0 or 0.5 pre-intervention vs. no preference post-intervention). When using the WHO growth standards, we observed increases between pre- and post-training LAZ SD (2.55 vs. 2.92) and % BIV (5.33 vs. 15.13). Internal standardization eliminated the % BIV, with pre-intervention LAZ ranging from-1.78 to 2.27, and post intervention LAZ: −2.27 to 2.04, falling within the WHO ranges for biologically plausible values (−6 SD &lt; LAZ &lt; 6 SD). Reliability of length measurements post-intervention was high as indicated by low relative TEM of 0.53%. Accuracy of 3D imaging was high (R = 0.99) comparing post-training vs. 3D imaging for length; however, examination of Bland Altman plots revealed that on average 3D scans overestimated length by 3.87 centimeters. Conclusions Training on standard anthropometry improved data quality. 3D imaging may be an accurate alternative to standard anthropometry, but adjustment of the technology is needed to avoid overestimation of length. Future research on the appropriate use of reference standards to define malnutrition in this severely ill population is needed. Funding Sources Bill & Melinda Gates Foundation.

Detecting fabrication in large-scale molecular omics data

Background: Fraud is a pervasive problem and can occur as fabrication, falsification, plagiarism or theft. The scientific community is not exempt from this universal problem and several studies have recently been caught manipulating or fabricating data. Current measures to prevent and deter scientific misconduct come in the form of the peer-review process and on-site clinical trial auditors. As recent advances in high-throughput omics technologies have moved biology into the realm of big-data, fraud detection methods must be updated for sophisticated computational fraud. In the financial sector, machine learning and digit-preference are successfully used to detect fraud. Results: Drawing from these sources, we develop methods of fabrication detection in biomedical research and show that machine learning can be used to detect fraud in large-scale omic experiments. Using the raw data as input, the best machine learning models correctly predicted fraud with 84-95% accuracy. With digit frequency as input features, the best models detected fraud with 98%-100% accuracy. All of the data and analysis scripts used in this project are available at https://github.com/MSBradshaw/FakeData . Conclusions: Using digit frequencies as a generalized representation of the data, multiple machine learning methods were able to identify fabricated data with near perfect accuracy.

Abstract P333: National Trends in the Quality of Office Blood Pressure Measurements, 2014-2019

Introduction: Clinical practice guidelines emphasize the importance of accurate blood pressure measurement and recording to diagnose and treat hypertension. Trends in terminal digit preference (typically manifest by a terminal digit of ‘0’) have not been examined nationally. The growing use of automated blood pressure devices may have reduced terminal digit preference and improved accuracy over time. Objective: To evaluate trends in terminal digit preference in office blood pressure measurements among adults with hypertension by patient and provider characteristics. Methods: We used IQVIA National Disease and Therapeutic Index (NDTI) data from January 2014 through June 2019. The NDTI is designed to be nationally-representative of all patient visits to office-based physicians and uses a two-stage stratified sampling design to sample ~4,000 physicians per quarter who report information on all patient visits on 2 random workdays. We included all hypertension treatment visits (~60M/year) among adults aged ≥18. We examined trends in the proportion of hypertension treatment visits with recorded systolic (SBP) and diastolic (DBP) blood pressure measurements with a terminal digit ‘0’. The expected percent of blood pressures with ‘0’ is 10% for automated and 20% for manual readings. Results: There was a decrease in the percent of visits with SBP (43.0% to 37.4%) or DBP (44.3% to 38.1%) recordings ending in zero ( Table ). The decrease in percent of SBPs with a terminal zero was similar by patient and provider characteristics, though the percentage of SBPs with a terminal ‘0’ was consistently higher among patients aged ≥60, when SBP ≥140 mmHg, and among cardiologists. Conclusions: Terminal digit preference is common indicating systematic error in blood pressure measurement and recording, despite some improvement over time. This may lead to under- and overtreatment of patients with hypertension. Improving the quality of blood pressure measurement is central to improving hypertension diagnosis and control in clinical practice.