Nonresponse Weights Research Articles

The 2020 Parliamentary Election in Iran

Introduction: Since its establishment in 1979, the Islamic Republic of Iran has gone through eleven parliamentary elections. The highest participation rate was recorded with 71 percent in 1996. In February 2020, the regime experienced the lowest turnout in its 42-year history. According to Iran's officials, about 42 percent of eligible voters cast their ballots. Independent observers, however, assert that such ordinary claims cannot be held true. The Iranian regime has repeatedly demonstrated a lack of integrity. Objective: This paper aims at assessing the level of participation in the parliamentary election on February 21, 2020. The assessment is expected to shed light on Tehran's approval and election turnout rates. Methods: The author prepared an online survey, which was distributed through widely used social media channels and instant messaging accounts in Iran. An online survey consisting of 17 questions and 21 variables put in the Persian language was publicly distributed on the very same day of the election on February 21, 2020. To ensure that the study represented Iran's social structure, post-stratification or nonresponse weights were applied throughout multiple auxiliary variables. The survey was operationalized based on the snowball effect to reach out to a representative sample of ordinary Iranians. This was accomplished through an effective marketing program. Results: Relying on the snowball effect, with over 20 thousand responses, the author collected a deep sample of 1,060 Iranian citizens. With a probability of 95 percent and 3 percent margin of error, it is fair to assume that that highly likely about 1.9 million citizens elected the new members of Majles. In a nutshell, the majority of eligible voters apparently boycotted the 2020 parliamentary election. Less than 9 million Iranians cast their votes, including seven million invalid ballots. If Iranians could participate in a free referendum, they would overwhelmingly reject the Islamic Republic as a political system. Conclusion: Mounting evidence of an increasing level of dissatisfaction with the Islamic regime combined with an amplifying distrust toward Iran's electoral system has reduced and removed the clerics' legitimacy. Put merely, mullahs in Tehran do not represent the majority of Iranians. Recommendation: The West has two options: (1) partnering with a highly unpopular regime consisting of a septuagenarian group of 2,330 authoritarian religious men or (2) helping freedom-seeking people of Iran who pursue a fundamental change toward democracy. History will judge all decisions and actions.

Identification of Suicide Attempt Risk Factors in a National US Survey Using Machine Learning

Because more than one-third of people making nonfatal suicide attempts do not receive mental health treatment, it is essential to extend suicide attempt risk factors beyond high-risk clinical populations to the general adult population. To identify future suicide attempt risk factors in the general population using a data-driven machine learning approach including more than 2500 questions from a large, nationally representative survey of US adults. Data came from wave 1 (2001 to 2002) and wave 2 (2004 to 2005) of the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC). NESARC is a face-to-face longitudinal survey conducted with a national representative sample of noninstitutionalized civilian population 18 years and older in the US. The cumulative response rate across both waves was 70.2% resulting in 34 653 wave 2 interviews. A balanced random forest was trained using cross-validation to develop a suicide attempt risk model. Out-of-fold model prediction was used to assess model performance, including the area under the receiver operator curve, sensitivity, and specificity. Survey design and nonresponse weights allowed estimates to be representative of the US civilian population based on the 2000 census. Analyses were performed between May 15, 2019, and June 10, 2020. Attempted suicide in the 3 years between wave 1 and wave 2 interviews. Of 34 653 participants, 20 089 were female (weighted proportion, 52.1%). The weighted mean (SD) age was 45.1 (17.3) years at wave 1 and 48.2 (17.3) years at wave 2. Attempted suicide during the 3 years between wave 1 and wave 2 interviews was self-reported by 222 of 34 653 participants (0.6%). Using survey questions measured at wave 1, the suicide attempt risk model yielded a cross-validated area under the receiver operator characteristic curve of 0.857 with a sensitivity of 85.3% (95% CI, 79.8-89.7) and a specificity of 73.3% (95% CI, 72.8-73.8) at an optimized threshold. The model identified 1.8% of the US population to be at a 10% or greater risk of suicide attempt. The most important risk factors were 3 questions about previous suicidal ideation or behavior; 3 items from the 12-Item Short Form Health Survey, namely feeling downhearted, doing activities less carefully, or accomplishing less because of emotional problems; younger age; lower educational achievement; and recent financial crisis. In this study, after searching through more than 2500 survey questions, several well-known risk factors of suicide attempt were confirmed, such as previous suicidal behaviors and ideation, and new risks were identified, including functional impairment resulting from mental disorders and socioeconomic disadvantage. These results may help guide future clinical assessment and the development of new suicide risk scales.

Practice Characteristics and Job Satisfaction of Private Practice and Academic Surgeons

Private practice and academic surgery careers vary significantly in their daily routine, compensation schemes, and definition of productivity. Data are needed regarding the practice characteristics and job satisfaction of these career paths for surgeons and trainees to make informed career decisions and to identify modifiable factors that may be associated with the health of the surgical workforce. To obtain and compare the differences in practice characteristics and career satisfaction measures between academic and private practice surgeons. In this cross-sectional survey performed from June 4 to August 1, 2018, an online survey accommodating smartphone, tablet, and desktop formats was distributed by email to 25 748 surgeons who were actively practicing fellows of the American College of Surgeons; had completed a general surgery residency or categorical fellowship in plastic, cardiothoracic, or vascular surgery; and had an active email address on file. Demographic, training, and current practice characteristics were obtained, and satisfaction measures were measured on a 5-point Likert scale and compared by surgeon type. Nonresponse weights adjusted for respondent sex, age, and presence of subspecialty training between respondents and the total surveyed American College of Surgeons population. There were 3807 responses (15% response rate) from surgeons: 1735 academic surgeons (1390 men [80%]; median age, 53 years [interquartile range (IQR), 44-61 years]) and 1464 private practice surgeons (1276 men [87%]; median age, 56 years [IQR, 48-62 years]); 589 surgeons who reported being neither an academic surgeon nor a private practice surgeon and 19 surgeons who did not respond to questions on their practice type were excluded. Academic surgeons reported working a median of 59 hours weekly (IQR, 38-65 hours) compared with 57 hours weekly (IQR, 45-65 hours) for private practice surgeons. Academic surgeons reported more weekly hours performing nonclinical work than did private practice surgeons (24 hours [IQR, 14-38 hours] vs 9 hours [IQR, 4-17 hours]; P < .001). Academic surgeons were more likely than private practice surgeons to be satisfied with their career as a surgeon (1448 of 1706 [85%] vs 1109 of 1420 [78%]; P < .001) and their financial compensation (997 of 1703 [59%] vs 546 of 1416 [39%]; P < .001). Academic surgeons were less likely than private practice surgeons to feel that competition with other surgeons is a threat to financial security (341 of 1705 [20%] vs 559 of 1422 [39%]; P < .001) and less likely to feel that malpractice experience has decreased job satisfaction (534 of 1703 [31%] vs 686 of 1413 [49%]; P < .001). This study suggests that, although overall surgeon satisfaction was high, academic surgeons reported higher career satisfaction on several measures when compared with private practice surgeons. Advocacy for private practice surgeons is important to encourage career longevity and sustain US surgeon workforce needs.

Mental health, quality of life, self-efficacy, and social support of individuals living with spinal cord injury in Switzerland compared to that of the general population.

Comparative study of survey self-report data. To compare individuals living with spinal cord injury (SCI) in Switzerland to the general population in terms of mental health, quality of life, self-efficacy, and social support. Community, Switzerland. Data from the 2017 community survey of the Swiss Spinal Cord Injury Cohort Study were compared to data from two matched (1:3 nearest neighbor propensity score) general population surveys collected in the same year. Measures of mental health (mental health index, psychological distress item, vitality scale, and energy item), quality of life (WHOQOL-BREF item), self-efficacy (General Self-Efficacy Scale item), and social support (items of relationship satisfaction, living alone, and marital status) were compared across datasets using regression adjusted for non-response correction weights. The analyses were then replicated in subgroups defined by sociodemographic, lesion-related, and secondary health issues factors. Individuals with SCI had significantly higher psychological distress and poorer mental health, vitality, energy, and quality of life than the general population, with medium to large effect sizes (Cohen's d: 0.35-1.08). They also had lower self-efficacy and relationship satisfaction, lived more frequently alone, and were more frequently single. Individuals with less severe secondary health issues reported mental health and quality of life more similar to the general population than those reporting more severe issues. This study highlights a significant long-term impact of SCI on well-being and psychosocial resources, underlining the need for ongoing biopsychosocial care beyond inpatient rehabilitation.

Nurse Effects on Non-response in Survey-Based Biomeasures

ABSTRACT Collecting biological data in representative surveys is becoming more common due to their potential to inform research and policy. Nevertheless, using nurses to collect these data can lead to unintended effects. In this paper, we investigate how nurses influence the non-response process by looking at five waves of data coming from two surveys in the UK: Understanding Society and the English Longitudinal Study of Ageing. We find that nurses explain between 5 and 14% of the variance in non-response to biological data collection. We also find that older nurses are more successful in obtaining cooperation and consent to biological data collection and nurses with more survey experience are more likely to successfully collect blood samples. Finally, we show that including nurse characteristics in non-response weighting models leads to modest changes in population estimates of biological markers.

Insurance Instability and Its Impact on Care of Young Adult Survivors of Pediatric Cancer: A Project Forward Study

By 2020, there will be more than 500,000 childhood cancer survivors (CCS) in the United States. CCS experience disparities in economic, social, and health‐related quality of life outcomes, and these disparities are magnified for CCS who are uninsured. Access to long‐term follow‐up (LTFU) care for surveillance, preventive care, and treatment of late effects for survivors are vital to improve long‐term outcomes. Inadequate insurance coverage, high out‐of‐pocket costs, and lack of perceived need for care have been shown to affect access to LTFU care among CCS. The objectives of this study were to (1) assess insurance instability longitudinally and describe patient factors that correlate with instability and (2) examine whether insurance instability and financial or patient factors influence access to LTFU care.Project Forward was a population‐based, observational study of CCS in Los Angeles County using California Cancer Registry (CCR) data to identify participants who completed a survey. Change in insurance coverage was assessed at diagnosis using CCR data and at survey and its impact on LTFU care. Those who experienced any change in insurance coverage (“insurance instability”) were set equal to one. Multivariable logistic regression models incorporating survey nonresponse weights were used to estimate the change in the marginal predicted probabilities of insurance instability and LTFU care, adjusted for demographic, socioeconomic, and clinical covariates and clustered by treating hospital.Study participants were diagnosed with cancer between the ages of 0 and 19 while living in Los Angeles County between 1996 and 2010 and were older than 21 at the time of survey, from 2015 to 2017 (N = 1106).Almost half (48%; N = 529) of participants experienced insurance instability from diagnosis to survivorship, while 577 did not. After adjusting for demographic, socioeconomic, and clinical covariates, the multivariable model predicting insurance instability indicated that being uninsured at diagnosis or at survey increased the probability of instability by 37% (P &lt; .001) and 58% (P &lt; .001), respectively, in comparison with those with private insurance. The multivariable model predicting LTFU care indicated that those who experienced insurance instability decreased the probability of LTFU care by 5% (P &lt; .05), in comparison with those who did not experience instability. When compared to those with private insurance coverage at diagnosis, participants who were covered by Medicaid, Medicare, or Indian Health Service plans at diagnosis were more likely to participate in LTFU care by 5% (P &lt; .05); however, those who were uninsured at the time of the survey were less likely to participate in LTFU care by 10% (P &lt; .05).CCS who were uninsured at diagnosis or survivorship were more likely to experience insurance instability. Insurance instability and being uninsured during survivorship were negatively associated with access to LTFU care; however, those with public insurance coverage at diagnosis were positively associated with access to LTFU care.Reducing insurance instability for CCS will improve access to LTFU care. This insight is key to improving health, reducing unnecessary and inappropriate health care use, and decreasing costs, while promoting services that can preserve and improve health for CCS.The study was funded by the National Institutes of Health.

The silent epidemic of obesity in The Gambia: evidence from a nationwide, population-based, cross-sectional health examination survey.

ObjectivesNon-communicable diseases account for 70% of global deaths; 80% occur in low-income and middle-income countries. The rapid increase of obesity in sub-Saharan Africa is a concern. We assessed generalised and abdominal obesity and their associated risk factors among adults in The Gambia.DesignNationwide cross-sectional health examination survey using the WHO STEPwise survey methods.SettingThe Gambia.ParticipantsThis study uses secondary analysis of a 2010 nationally representative random sample of adults aged 25–64 years (78% response rate). The target sample size was 5280, and 4111 responded. Analysis was restricted to non-pregnant participants with valid weight and height measurements (n=3533).Primary and secondary outcome measuresThe primary outcome variable was generalised obesity, using WHO body mass index (BMI) thresholds. Analyses used non-response weighting and adjusted for the complex survey design. We conducted multinomial logistic regression analysis to identify factors associated with BMI categories. The secondary outcome variable was abdominal obesity, defined as high waist circumference (using the International Diabetes Federation thresholds for Europeans).ResultsTwo-fifths of adults were overweight/obese, with a higher obesity prevalence in women (17%, 95% CI 14.7 to 19.7; men 8%, 95% CI 6.0 to 11.0). 10% of men and 8% of women were underweight. Urban residence (adjusted relative risk ratio 5.8, 95% CI 2.4 to 14.5), higher education (2.3, 1.2 to 4.5), older age, ethnicity, and low fruit and vegetable intake (2.8, 1.1 to 6.8) were strongly associated with obesity among men. Urban residence (4.7, 2.7 to 8.2), higher education (2.6, 1.1 to 6.4), older age and ethnicity were associated with obesity in women.ConclusionThere is a high burden of overweight/obesity in The Gambia. While obesity rates in rural areas were lower than in urban areas, obesity prevalence was higher among rural residents in this study compared with previous findings. Preventive strategies should be directed at raising awareness, discouraging harmful beliefs on weight, and promoting healthy diets and physical activity.

Abstract A052: Insurance status is a predictor of mode of colon cancer detection but not stage at diagnosis: What this means for early detection

Abstract Background: Colon cancer is among the most incident and one of the most common causes of cancer-related mortality in the United States. Regular screening is the primary method of preventing excess colon cancer-related morbidity and mortality. Self-reports of on-schedule screening through endoscopy were much higher among insured individuals in the US compared to those without insurance (57% versus 24%) in 2015. Privately insured colon cancer patients are more likely to be diagnosed at earlier stages than patients who are uninsured or on Medicaid. Insurance status may influence whether colon cancer is detected via screening or symptomatic discovery and can impact the likelihood of a late versus early stage diagnosis. Insurance may also influence access to initial follow-up care, which could impact the timeliness of diagnosis and/or treatment. Objective: We examined patterns of health insurance continuity/discontinuity leading up to a colon cancer diagnosis, as well as associations between insurance status and three outcomes: mode of cancer detection (screen-detected versus symptomatic), diagnosis stage, and clinical delay. Methods: The Colon Cancer Patterns of Care in Chicago examined racial and SES disparities in colon cancer screening, care initiation and treatment, and diagnostic stage. Eligible patients were non-Hispanic (NH) Black and NH White, aged 50 and older, with colon cancer, and were recruited from nine urban health care institutions. After consent, participants completed an in-person interview wherein they were asked about health insurance access and timing of coverage during diagnosis and treatment and the 5 years prior, diagnostic pathways, treatment, sociodemographic characteristics, and health care utilization. Logistic regression with marginal standardization was used to model the association between insurance variables and colon cancer outcomes. Nonresponse weights were included in all analyses. Results: After adjusting for age, race, gender and SES, being uninsured any time five years prior to diagnosis or during diagnosis or treatment was associated with a 20-percentage-point (95% CI: 0.08, 0.33) increased prevalence of symptomatic detection. Symptomatic detection in turn was associated with a 15-percentage-point increase in late-stage diagnosis (95% CI: 0.03, 0.27); however, 47% of screen-detected patients were nonetheless diagnosed at a late stage (stage 3 or 4). As a result, insurance status was not associated with stage at diagnosis. Insurance status was also not associated with clinical delay. Discussion: For health insurance to effectively prevent late-stage colon cancer diagnosis, adults must not only utilize their insurance benefit to get screened, but also screening must have a strong association with early-stage detection. Findings highlight the need to further study and intervene on factors contributing to late-stage colon cancer diagnoses despite screen detection. Citation Format: Leslie R. Carnahan, Lindsey Jones, Katherine Brewer, Garth H. Rauscher. Insurance status is a predictor of mode of colon cancer detection but not stage at diagnosis: What this means for early detection [abstract]. In: Proceedings of the Eleventh AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2018 Nov 2-5; New Orleans, LA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2020;29(6 Suppl):Abstract nr A052.

Abstract A051: Race and gender differences in awareness of colorectal cancer screening tests among recently diagnosed colon cancer

Abstract Background: Non-Hispanic Black (NHB) populations, compared to non-Hispanic Whites (NHW), are less likely to receive guideline concordant colorectal cancer (CRC) screening. CRC screening barriers are multifaceted and involve factors including health care access and utilization, sociodemographic characteristics, and individuals' beliefs and awareness about cancer, screening tests, and guidelines. Inability to recall or recognize CRC tests and low knowledge of screening guidelines may contribute to disparate outcomes across the colon cancer continuum. Objective: In the present study, we sought to 1) characterize the prevalence of urban colon cancer patients' awareness of screening tests and guidelines, and 2) examine if awareness and knowledge of guidelines were associated with mode of cancer detection (screen-detected versus symptomatic presentation). Methods: The Colon Cancer Patterns of Care in Chicago study was a descriptive cross-sectional study that examined racial, gender, and SES disparities in CRC screening, care initiation, diagnostic stage, and subsequent treatment. Eligible patients were NHB and NHW, aged 45-79, with first primary invasive colon cancer, and were recruited from nine diverse, urban health care institutions. After consent, participants completed an in-person interview wherein they responded to questions related to the recall and recognition of colon cancer stool, sigmoidoscopy, and colonoscopy screening tests and knowledge of screening guidelines, diagnostic pathways and treatment, sociodemographic characteristics, and health care access and utilization. They received $100 for completing the interview and consenting to medical record abstraction. Logistic regression was used to model the association between knowledge and awareness variables and colon cancer mode of detection (symptomatic versus screen detection). incorporating nonresponse weights created to account for differences in response rate by facility, age, race and gender, and models were, and controlling for age, race, gender and the composite SES variable in all models. Results: Recall of stool testing and sigmoidoscopy was low (13% and 5%); name recognition of these tests was 59% and 30%, respectively. Correct guideline knowledge was low for all three tests (7% for sigmoidoscopy, 14% for FOBT, and 19% for colonoscopy). Recall, recognition, and knowledge were lower for NHB and socioeconomically disadvantaged patients. Inability to name or recall a single test was associated with reduced screen-detection compared with recall of at least one test (36% vs. 22%, p=0.01). Discussion: Our results should help to identify target populations in need of enhanced education and additional prompting by their health care providers to ensure that they obtain the necessary surveillance for colon cancer over the long term. Citation Format: Leslie R. Carnahan, Lindsey Jones, Katherine Brewer, Yamile Molina, Garth Rauscher. Race and gender differences in awareness of colorectal cancer screening tests among recently diagnosed colon cancer [abstract]. In: Proceedings of the Eleventh AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2018 Nov 2-5; New Orleans, LA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2020;29(6 Suppl):Abstract nr A051.

Has working-age morbidity been declining? Changes over time in survey measures of general health, chronic diseases, symptoms and biomarkers in England 1994–2014

ObjectivesAs life expectancy has increased in high-income countries, there has been a global debate about whether additional years of life are free from ill-health/disability. However, little attention has been given...

Survey of the US Surgeon Workforce: Practice Characteristics, Job Satisfaction, and Reasons for Leaving Surgery

An impending surgeon shortage looms in the US due to increasing demand and a stagnant number of surgeons entering practice. We sought to evaluate the state of our surgical workforce by exploring current practice patterns, job satisfaction, and reasons why surgeons consider leaving surgery. In 2018, a link to a detailed survey was distributed by email to Fellows of the American College of Surgeons (ACS) who actively practice in the US and have completed a general surgery residency or integrated cardiothoracic, vascular, or plastic surgery fellowship. Detailed questions regarding practice attributes, surgical training, professional choices, and career satisfaction were included. Nonresponse weights were adjusted for respondent sex, age, and presence of subspecialty training between respondents and the total surveyed ACS population. All estimates and analyses were weighted to account for potential nonresponse bias. There were 3,807 surgeons who responded (15% response rate). Overall, 81% reported career satisfaction and 80% would choose surgery over again. Twenty-six percent were considering leaving surgery within the next 2 years. Stratified by age groups 40 to 49, 50 to 59, and 60 to 69 years, the survey found that 9%, 17%, and 45% of respondents were considering leaving surgery within the next 2 years, respectively. Thirteen percent of surgeons less than 60 years old were considering leaving surgery, and only 49% of these surgeons reported career satisfaction. Among surgeons less than 60 years old who were considering leaving surgery, the top 3 reasons rated as very important were overall stress (79%), work time demands (77%), and personal time requirements (73%). Reported surgeon job satisfaction was high, but a surprising number of surgeons reported that they were considering leaving surgery. Work time requirements and lack of personal time are leading factors contributing to surgeons leaving practice.

The use of Administrative Data to combat non-response

Background The issues arising due to participant non-response are problematic for many surveys. An international trend of falling survey response rates has led to more researchers questioning the extent to which non-response can cause issues with analysis, and what can be done to combat this non-response. Non-response weighting is a common method used by surveys, however the construction of these weights can rely upon untested assumptions and low quality and quantity of data. Administrative data can provide new opportunities in survey design, including improved construction of non-response weights, allowing better survey-based estimates.&#x0D; Main Aim In this paper we explore the potential of using administrative data from the sampling frame of a large social survey to model non-response. The data used for this paper is part of the Healthy Aging in Scotland (HAGIS) Survey.&#x0D; Methods The HAGIS pilot wave collected a sample of approximately 1000 individuals aged over 50 in the Scottish mainland. HAGIS utilised administrative data in the sampling with screening being provided to select households with at least one eligible respondent (over 50). Administrative data were also available for the full sample, and not just respondents to the HAGIS pilot wave. The administrative data consists of individual demographics, hospital admissions, and prescription data held by NHS Scotland.&#x0D; Results Survey response is modelled as a function of explanatory variables from the administrative data in order to examine whether there is systematic demographic and health-based non-response in the survey. We further examine whether this non-response is causing significant biases and then how health data might predict non-response.&#x0D; Conclusion Finally, we discuss how these models could be used in improving conventional weighting techniques to combat non-response in survey design.

Multisite Causal Mediation Analysis in the Presence of Complex Sample and Survey Designs and Non-Random Non-Response

Summary This study provides a template for multisite causal mediation analysis using a comprehensive weighting-based analytic procedure that enhances external and internal validity. The template incorporates a sample weight to adjust for complex sample and survey designs, adopts an inverse probability of treatment weight to adjust for differential treatment assignment probabilities, employs an estimated non-response weight to account for non-random non-response and utilizes a propensity-score-based weighting strategy to decompose flexibly not only the population average but also the between-site heterogeneity of the total programme impact. Because the identification assumptions are not always warranted, a weighting-based balance checking procedure assesses the remaining overt bias, whereas a weighting-based sensitivity analysis further evaluates the potential bias related to omitted confounding or to propensity score model misspecification. We derive the asymptotic variance of the estimators for the causal effects that account for the sampling uncertainty in the estimated weights. The method is applied to a reanalysis of the data from the National Job Corps Study.

Adjusting for non-response in the Finnish Drinking Habits Survey.

Aim: We aim to compare four different weighting methods to adjust for non-response in a survey on drinking habits and to examine whether the problem of under-coverage of survey estimates of alcohol use could be remedied by these methods in comparison to sales statistics. Method: The data from a general population survey of Finns aged 15–79 years in 2016 (n=2285, response rate 60%) were used. Outcome measures were the annual volume of drinking and prevalence of hazardous drinking. A wide range of sociodemographic and regional variables from registers were available to model the non-response. Response propensities were modelled using logistic regression and random forest models to derive two sets of refined weights in addition to design weights and basic post-stratification weights. Results: Estimated annual consumption changed from 2.43 litres of 100% alcohol using design weights to 2.36–2.44 when using the other three weights and the estimated prevalence of hazardous drinkers changed from 11.4% to 11.4–11.8%, correspondingly. The use of weights derived by the random forest method generally provided smaller estimates than use of the logistic regression-based weights. Conclusions: The use of complex non-response weights derived from the logistic regression model or random forest are not likely to provide much added value over more simple weights in surveys on alcohol use. Surveys may not catch heavy drinkers and therefore are prone for under-reporting of alcohol use at the population level. Also, factors other than sociodemographic characteristics are likely to influence participation decisions.

One year after the implementation of 2015 Tobacco Control Regulation on persons aged 15 years and over tobacco use in Beijing

Objective: To understand the current status of smoking and smoking cessation in persons aged 15 years and over in Beijing and evaluate the effect of 2015 Beijing Tobacco Control Regulation. Methods: In 2014 and 2016, based on the principles and methodology of the Global Adult Tobacco Survey. A total of 50 communities or townships were selected from 324 communities or townships in Beijing through multistage cluster sampling, and 2 community (village) committees from each community or township were selected with the method of probability proportional to size (PPS). A total of 100 surveillance sites were set, and 100 households were selected from each surveillance site by using simple random sampling. Data were collected through face-to-face interview from the eligible family members aged 15 years and over with the assistance of a tablet computer. Statistical analyses were conducted by using complex sampling analyses module of SPSS 20.0, with weights as a combination of sampling weights, non-response weights and post- stratification weights, for the calculation of current smoking prevalence, daily smoking prevalence, smoking cessation rate, etc. Results: A total of 8 484 and 9 372 valid questionnaires were obtained, respectively, in 2014 and 2016, with the response rate of 86.5% and 96.5%. The current smoking prevalence in persons aged 15 years and over was 23.4% in 2014, and 22.3% in 2016. According to the 6(th) national census data, the current smoking population decreased by 199 000 in Beijing. The proportion of daily smokers declined from 20.7% in 2014 to 19.2% in 2016. The daily number of cigarettes consumed by current smokers increased from 14.6 in 2014 to 15.4 in 2016. The smoking cessation rate was 14.9% in 2014 and 16.8% in 2016. The proportion of current smokers who had at least one smoking cessation attempt in the past 12 months increased from 22.3% in 2014 to 23.2% in 2016, and the proportion of current smokers who planned to quit smoking increased from 11.6% to 15.5%. Among the current smokers who had visited doctors in the past 12 months, the proportion of those having smoking cessation advice was 58.9% in 2014 and 59.2% in 2016. In 2016, among the current smokers who had attempted to quit in the past 12 months, 36.8% were aware of the smoking cessation clinics, and 29.5%were aware of the quitline. Among those who were aware, only 7.7% had actually visited the cessation clinics, and 5.5% had used the quitline. Conclusions: After the implementation of 2015 Beijing Tobacco Control Regulation for 1 year, the current smoking prevalence in persons aged 15 years and over showed decreasing. It is necessary to further prompt the expansion of smoking cessation service to cover more current smokers.

Integrating electronic health records from different sources across the UK: lessons from a record linkage study

BackgroundHarmonisation of different data sources from various electronic health records (EHRs) across systems enhances the potential scope and granularity of data available to health data research.&#x0D; ObjectiveTo describe data harmonisation of routine electronic healthcare records in Wales and Scotland linked to a UK longitudinal birth cohort, the Millennium Cohort Study (MCS).&#x0D; MethodsComparable secondary care data was linked, with parental consent, to MCS information for 1838 and 1431 children participating in MCS and residing in Wales and Scotland, by assigning, respectively, unique Anonymised Linkage Fields to personbased records in the privacy protecting Secure Anonymised Information Linkage (SAIL) databank at Swansea University, and by the National Health Service (NHS) Information Standards Division. Survey and non-response weights were created to account for the clustered sample, sample attrition and consent to linkage. Heterogeneous variables from the Patient Episode Dataset for Wales, Emergency Department Data Set for Wales, Scottish Medical Record 01 and Accident and Emergency dataset for Scotland were harmonised enabling data to be pooled and standardised for research.&#x0D; FindingsOverall linkage to harmonised health care data was achieved for 98.9% (99.9% for Wales and 97.6% for Scotland) of consented MCS participants. 66% of children experienced at least one hospital admission (total 5747 hospital admissions) up totheir 14th birthday, while 60% attended A&amp;E departments at least once (total 5221 attendances) between their 9th and 14th birthday. We managed date granularity by generating random dates of birth, standardising periods of data collection,identifying inconsistencies and then mapping and bridging differences in definitions of periods of care across countries and datasets.&#x0D; ConclusionsCombining and harmonising data from multiple sources and linking them to information from a longitudinal cohort create useful resources for population health research. These methods are reproducible and can be utilised by other researchersand projects.

The Veterans Metrics Initiative study of US veterans’ experiences during their transition from military service

PurposeEfforts to promote the health and well-being of military veterans have been criticised for being inadequately informed of veterans’ most pressing needs as they separate from military service, as well...

An Introduction to Machine Learning Methods for Survey Researchers

Machine learning techniques comprise an array of computer-intensive methods that aim at discovering patterns in data using flexible, often nonparametric, methods for modeling and variable selection. These methods offer an expansion to the more traditional methods, such as OLS or logistic regression, which have been used by survey researchers and social scientists. Many of the machine learning methods do not require the distributional assumptions of the more traditional methods, and many do not require explicit model specification prior to estimation. Machine learning methods are beginning to be used for various aspects of survey research including responsive/adaptive designs, data processing and nonresponse adjustments and weighting. This special issue aims to familiarize survey researchers and social scientists with the basic concepts in machine learning and highlights five common methods. Specifically, articles in this issue will offer an accessible introduction to: LASSO models, support vector machines, neural networks, and classification and regression trees and random forests. In addition to a detailed description, each article will highlight how the respective method is being used in survey research along with an application of the method to a common example. The introductory article will provide an accessible introduction to some commonly used concepts and terms associated with machine learning modeling and evaluation. The introduction also provides a description of the data set that was used as the common application example for each of the five machine learning methods.

Can we assess representativeness of cross-national surveys using the education variable?

Achieving a representative sample is an important goal for every survey. High response rates are often referred to as an indicator of representativeness in survey methodology research. However, a low response rate does not necessarily imply low representativeness, so that alternative ways of assessing representativeness are needed in times where low response rates are almost ubiquitous. This study asks whether education, a socio-demographic variable covered by virtually every survey of individuals, is a good variable for assessing the representativeness of a realised survey sample. We examine this issue in two steps: Firstly, the distributions of the harmonised education variable in six official and academic cross-national surveys by country-year combination are compared with the respective education distributions in a high-quality reference dataset. Doing so, we identify many substantial inconsistencies. Secondly, we try to identify the sources of these inconsistencies, looking at both measurement errors in the education variables and errors of representation. Since in most instances, inconsistent measurement procedures can probably explain the observed inconsistencies, we conclude that the education variable as currently measured in cross-national surveys is, without further processing, unsuitable for assessing sample representativeness, and constructing nonresponse weights. The paper closes with recommendations for achieving a more comparable measurement of the education variable.

Correcting the Underestimation of Top Incomes: Combining Data from Income Tax Reports and the Brazilian 2010 Census

To deal with the problem of underestimation of top incomes in household surveys, we propose a methodology to combine the income distributions of the Brazilian 2010 Census (survey) and of 2010 DIRPF (personal income tax reports). The method consists in estimating a system of non-response weights that uses as frame the tax register and is applied to the top of the distribution. After applying this calibration methodology, we decompose inequality income sources. Correcting survey distributions with tax data increases the contribution of non-labor income to inequality, as the case of the Brazilian Census shows. Changes in the methodology do not affect the results substantially.