Evaluating data quality for blended data using a data quality framework.

In 2020 the U.S. Federal Committee on Statistical Methodology (FCSM) released "A Framework for Data Quality", organized by 11 dimensions of data quality grouped among three domains of quality (utility, objectivity, integrity). This paper addresses the use of the FCSM Framework for data quality assessments of blended data. The FCSM Framework applies to all types of data, however best practices for implementation have not been documented. We applied the FCSM Framework for three health-research related case studies. For each case study, assessments of data quality dimensions were performed to identify threats to quality, possible mitigations of those threats, and trade-offs among them. From these assessments the authors concluded: 1) data quality assessments are more complex in practice than anticipated and expert guidance and documentation are important; 2) each dimension may not be equally important for different data uses; 3) data quality assessments can be subjective and having a quantitative tool could help explain the results, however, quantitative assessments may be closely tied to the intended use of the dataset; 4) there are common trade-offs and mitigations for some threats to quality among dimensions. This paper is one of the first to apply the FCSM Framework to specific use-cases and illustrates a process for similar data uses.

A Comparison of the Mortality Experience of U.S. Adults Estimated With the 2006-2018 National Health Interview Survey Linked Mortality Files and the Annual U.S. Life Tables.

Objective-Linking data is a powerful mechanism to provide policy-relevant information. The National Center for Health Statistics' Data Linkage Program produces linked mortality files (LMFs) for research by linking data from the National Center for Health Statistics' surveys, including the National Health Interview Survey (NHIS), to mortality data from the National Death Index. Assessing the accuracy of the linked data is an important step in ensuring its analytic use. This report compares the cumulative survival probabilities estimated with the 2006-2018 NHIS LMFs to those from the annual U.S. life tables.

A methodological assessment of privacy preserving record linkage using survey and administrative data.

BACKGROUND: The National Center for Health Statistics (NCHS) links data from surveys to administrative data sources, but privacy concerns make accessing new data sources difficult. Privacy-preserving record linkage (PPRL) is an alternative to traditional linkage approaches that may overcome this barrier. However, prior to implementing PPRL techniques it is important to understand their effect on data quality. METHODS: Results from PPRL were compared to results from an established linkage method, which uses unencrypted (plain text) identifiers and both deterministic and probabilistic techniques. The established method was used as the gold standard. Links performed with PPRL were evaluated for precision and recall. An initial assessment and a refined approach were implemented. The impact of PPRL on secondary data analysis, including match and mortality rates, was assessed. RESULTS: The match rates for all approaches were similar, 5.1% for the gold standard, 5.4% for the initial PPRL and 5.0% for the refined PPRL approach. Precision ranged from 93.8% to 98.9% and recall ranged from 98.7% to 97.8%, depending on the selection of tokens from PPRL. The impact of PPRL on secondary data analysis was minimal. DISCUSSION: The findings suggest PPRL works well to link patient records to the National Death Index (NDI) since both sources have a high level of non-missing personally identifiable information, especially among adults 65 and older who may also have a higher likelihood of linking to the NDI. CONCLUSION: The results from this study are encouraging for first steps for a statistical agency in the implementation of PPRL approaches, however, future research is still needed.

Using Synthetic Data to Replace Linkage Derived Elements: A Case Study.

While record linkage can expand analyses performable from survey microdata, it also incurs greater risk of privacy-encroaching disclosure. One way to mitigate this risk is to replace some of the information added through linkage with synthetic data elements. This paper describes a case study using the National Hospital Care Survey (NHCS), which collects patient records under a pledge of protecting patient privacy from a sample of U.S. hospitals for statistical analysis purposes. The NHCS data were linked to the National Death Index (NDI) to enhance the survey with mortality information. The added information from NDI linkage enables survival analyses related to hospitalization, but as the death information includes dates of death and detailed causes of death, having it joined with the patient records increases the risk of patient re-identification (albeit only for deceased persons). For this reason, an approach was tested to develop synthetic data that uses models from survival analysis to replace vital status and actual dates-of-death with synthetic values and uses classification tree analysis to replace actual causes of death with synthesized causes of death. The degree to which analyses performed on the synthetic data replicate results from analysis on the actual data is measured by comparing survival analysis parameter estimates from both data files. Because synthetic data only have value to the degree that they can be used to produce statistical estimates that are like those based on the actual data, this evaluation is an essential first step in assessing the potential utility of synthetic mortality data.

Using supervised machine learning to identify efficient blocking schemes for record linkage.

Record linkage enables survey data to be integrated with other data sources, expanding the analytic potential of both sources. However, depending on the number of records being linked, the processing time can be prohibitive. This paper describes a case study using a supervised machine learning algorithm, known as the Sequential Coverage Algorithm (SCA). The SCA was used to develop the join strategy for two data sources, the National Center for Health Statistics' (NCHS) 2016 National Hospital Care Survey (NHCS) and the Center for Medicare & Medicaid Services (CMS) Enrollment Database (EDB), during record linkage. Due to the size of the CMS data, common record joining methods (i.e. blocking) were used to reduce the number of pairs that need to be evaluated to identify the vast majority of matches. NCHS conducted a case study examining how the SCA improved the efficiency of blocking. This paper describes how the SCA was used to design the blocking used in this linkage.

Assessing Linkage Eligibility Bias in the National Health Interview Survey.

Background Linking health survey data to administrative records expands the analytic utility of survey participant responses, but also creates the potential for new sources of bias when not all participants are eligible for linkage. Residual differences-bias-can occur between estimates made using the full survey sample and the subset eligible for linkage. Objective To assess linkage eligibility bias and provide examples of how bias may be reduced by changes in questionnaire design and adjustment of survey weights for linkage eligibility. Methods Linkage eligibility bias was estimated for various sociodemographic groups and health-related variables for the 2000-2013 National Health Interview Surveys. Conclusions Analysts using the linked data should consider the potential for linkage eligibility bias when planning their analyses and use approaches to reduce bias, such as survey weight adjustments, when appropriate.

Comparative Analysis of the National Health Interview Survey Public-use and Restricted-use Linked Mortality Files.

Linking national survey data with administrative data sources enables researchers to conduct analyses that would not be possible with each data source alone. Recently, the Data Linkage Program at the National Center for Health Statistics (NCHS) released updated linked mortality files, including the National Health Interview Survey data linked to the National Death Index mortality files. Two versions of the files were released: restricted-use files available through NCHS and Federal Statistical Research Data Centers and public-use files. To reduce the reidentification risk, statistical disclosure limitation methods were applied to the public-use files before they were released. This included limiting the amount of mortality information available and perturbing cause of death and follow-up time for select records. To assess the comparability of the restricted-use and public-use files, relative hazard ratios for all-cause and cause-specific mortality using Cox proportional hazards models were estimated and compared. The comparative analysis found that the two data files yielded very similar descriptive and model results.

Evaluating Survey Report of Social Security Disability Benefit Receipt Using Linked National Health Interview Survey and Social Security Administration Data.

Linking nationally representative population health survey data with Social Security Administration (SSA) disability program data provides a rich source of information on program recipients. Survey participant data from the 1998-2005 National Health Interview Survey (NHIS) were linked to SSA administrative records from 1997 through 2005. The goal of this study was to assess agreement between the actual benefit receipt based on the SSA administrative records and the survey report of benefit receipt in the linked NHIS and SSA file for the U.S. civilian noninstitutionalized population. This evaluation provides information on the expected accuracy of survey report of Social Security Disability Insurance (SSDI) and Supplemental Security Income (SSI) benefit receipt, including how participant characteristics may be associated with reporting misclassification. The results indicate that there is some underreporting of SSA disability benefit receipt based on the NHIS responses compared with the SSA administrative records. The analysis identified some differences between the concordant and discordant groups for selected characteristics, but there were no clear patterns among the different survey questions or the different survey participant characteristics.

Linkage of 1999-2012 National Health Interview Survey and National Health and Nutrition Examination Survey Data to U.S. Department of Housing and Urban Development Administrative Records.

Objectives This report presents the development, plan, and operation of the 2011-2012 National Survey of Children's Health, a module of the State and Local Area Integrated Telephone Survey, conducted by the National Center for Health Statistics. Funding was provided by the Maternal and Child Health Bureau, Health Resources and Services Administration. The survey was designed to produce national and state prevalence estimates of the physical and emotional health of children aged 0-17 years, as well as factors that may relate to child well-being including medical homes, family interactions, parental health, school and after-school experiences, and neighborhood characteristics.

Concordance between survey report of Medicaid enrollment and linked Medicaid administrative records in two national studies.

The National Health and Nutrition Examination Survey (NHANES) and the National Health Interview Survey (NHIS) are population-based surveys that have each been linked to administrative data from the Centers for Medicare and Medicaid Services (CMS): the Medicaid Analytic eXtract (MAX) files. These linked data were used to examine, among children under age 18 years, respondent-level concordance between Medicaid or the Children's Health Insurance Program (CHIP) enrollment as reported in each survey (NHANES and NHIS) and as indicated by administrative data from the MAX files. Concordance was defined as having Medicaid/CHIP reported as a health insurance source in the survey questionnaire and having a CMS Medicaid/CHIP administrative record in the same month and year as the interview. Records were also considered concordant if there was no report of Medicaid/CHIP coverage based on the interview response and no match to the CMS administrative records for Medicaid enrollment. Between NHANES and MAX, 88% of observations were concordant with respect to Medicaid or CHIP enrollment, yielding a Kappa of 0.71. Between NHIS and MAX, 89% of observations were concordant with respect to Medicaid or CHIP enrollment, yielding a Kappa of 0.73. These concordance rates provide support for the use of both administrative and NHANES and NHIS as a valuable tool for public health researchers and survey methodologists.

The prevalence of using iodine-containing supplements is low among reproductive-age women, NHANES 1999-2006.

During pregnancy, the iodine requirement rises to meet demands for neurological development and fetal growth. If these requirements are not met, irreversible pathological cognitive and behavioral changes to the fetus may ensue. This study estimated the prevalence of iodine-containing dietary supplement (DS) use and intakes of iodine from DSs among pregnant women and nonpregnant women of reproductive age (15-39 y) who were interviewed and examined in NHANES 1999-2006 (n = 6404). Although 77.5% of pregnant women reported taking one or more DSs in the past 30 d, only 22.3% consumed an iodine-containing supplement. Most pregnant women reported using one DS and reported taking this product daily. The vast majority of iodine-containing DSs reported by pregnant women claimed an iodine content of 150 µg iodine/serving on the label. Pregnant women using at least one DS containing iodine had a mean daily iodine intake of 122 µg/d from supplements; the median value was 144 µg/d. Median urinary iodine concentrations (UICs) were similar for pregnant and nonpregnant women in the population aged 15-39 y. The median UIC was 148 µg/L for pregnant women and 133 µg/L for nonpregnant women. The WHO has established a cutoff for insufficient iodine intake at <150 µg/L for pregnant women and <100 mg/L for those who are not pregnant. This suggests that as a population, we may not be meeting adequate intakes of iodine for pregnant women. More research is needed on the iodine intakes of pregnant women and women of reproductive age on their total iodine intake from all sources, not just DSs.

National Health and Nutrition Examination Survey: sample design, 2007-2010.

BACKGROUND: Data collection for the National Health and Nutrition Examination Survey (NHANES) comprises three levels: a household screener, an interview, and a physical examination. The primary objective of the screener is to determine whether any household members are eligible for the interview an dexamination. Eligibility is determined by preset selection probabilities for the desired demographic subdomains. After an eligible sample person is selected, the interview collects person-level demographic, health, and nutrition information, as well as information about the household. The examination includes physical measurements, tests such as hearing and dental examinations, and the collection of blood and urine specimens for laboratory testing. OBJECTIVES: This report provides some background on the NHANES program, beginning with the first survey cycle in the 1970s and highlighting significant changes since its inception. The report then describes the broad design specifications for the 2007-2010 survey cycle, including survey objectives, domain and precision specifications, and operational requirements unique to NHANES. In addition, the report describes the details of the survey design, including the calculation of sampling rates and sample selection methods. Documentation of survey content, data collection procedures, estimation methods, and methods to assess nonsampling errors are reported elsewhere.

National Health and Nutrition Examination Survey: estimation procedures, 2007-2010.

BACKGROUND: Data collection for the National Health and Nutrition Examination Survey (NHANES), comprises three levels: an initial household screening interview (or ''screener''), an in-home personal interview, and a physical examination. The primary objective of the screener is to determine whether any household members are eligible for the interview and examination. Eligibility is determined by preset selection probabilities for the desired demographic subdomains. After an eligible sample person is selected, the in-home interview collects person-level demographic, health, and nutrition information, as well as information about the household. The examination includes physical measurements such as blood pressure, a dental examination, and the collection of blood and urine specimens for laboratory testing. OBJECTIVES: This report provides background for the NHANES program and summarizes the sample design specifications for the 2007-2010 survey cycle. Estimation procedures are then presented, including the methods used to calculate survey weights for the full sample and for examination subsamples, as well as guidelines for combining 2-year weights for the analysis of multiyear data. Finally, the appropriate variance estimation methods are described. The sample selection methods, survey content, data collection procedures, and methods for assessing nonsampling errors are documented elsewhere.

Health characteristics of Medicare traditional fee-for-service and Medicare Advantage enrollees: 1999-2004 National Health and Nutrition Examination Survey linked to 2007 Medicare data.

BACKGROUND: National Health and Nutrition Examination Survey (NHANES) records have been linked to health care use and expenditure information from Medicare records. Claims data are generally available only for traditional fee-for-service (FFS) enrollees and not for Medicare Advantage enrollees. Differences in health characteristics between Medicare Advantage and traditional FFS enrollees could affect conclusions from analyses of the linked files that are restricted to traditional FFS enrollees. METHODS: Data from the 1999-2004 NHANES linked to the 2007 Medicare enrollment data were analyzed. Using examination and interview data collected in NHANES, we compared health characteristics of Medicare beneficiaries aged 65 and over at the NHANES interview by their type of Medicare enrollment in 2007. RESULTS: We found that the overall percentage of Medicare beneficiaries who had hypertension at the time of the NHANES medical examination was lower for Medicare Advantage enrollees compared with traditional FFS enrollees; this was found for the non-Hispanic white population but not for other race and ethnicity groups. We found no statistically significant differences between Medicare Advantage and traditional FFS enrollees overall or within race and ethnicity groups for other health characteristics that were measured or reported at the time of the NHANES interview or medical examination in 1999-2004. CONCLUSIONS: Despite finding no large differences in health characteristics by Medicare enrollment in this analysis, users of the National Center for Health Statistics Medicare linked files should consider potential differences in health characteristics between Medicare Advantage and traditional FFS enrollees that could influence results limited to traditional FFS beneficiaries only.

The National Health and Nutrition Examination Survey: Sample Design, 1999-2006.

BACKGROUND: Data collection for the National Health and Nutrition Examination Survey (NHANES) comprises three levels: a household screener, an interview, and a physical examination. The primary objective of the screener is to determine whether any household members are eligible for the interview and examination. Eligibility is determined by the preset selection probabilities for the desired demographic subdomains. After selection as an eligible sample person, the interview collects person-level demographic, health, and nutrition information as well as information about the household. The examination includes physical measurements, tests such as eye and dental examinations, and the collection of blood and urine specimens for laboratory testing. OBJECTIVES: This report will first describe the broad design specifications for the 1999-2006 survey including survey objectives, domain and precision specifications, operational requirements, sample design, and estimations procedures. Details of the sample design are divided into two sections. The first section (NHANES 1999-2001 Sample Design) broadly describes the sample design and various design changes during the first three years of the continuous NHANES (1999-2001). The second section (NHANES 2002-2006 Sample Design) describes the final sample design developed and applied for 2002-2006. Weighting and variance estimation procedures are presented in the same manner; however, to correspond to the public data release cycles, the weighting and variance sections are separated into those used for 1999-2002, and those used for 2003-2006. Much of this report is based on survey operations documents and sample design reports prepared by Westat. Documentation of the survey content, procedures, and methods to assess nonsampling errors are reported elsewhere.

Serum soluble transferrin receptor concentrations in US preschool children and non-pregnant women of childbearing age from the National Health and Nutrition Examination Survey 2003-2010.

BACKGROUND: Serum soluble transferrin receptor (sTfR) is recommended as a sensitive and accurate measure of iron deficiency (ID) in populations when only a single indicator can be used. The lack of assay standardization and of representative data on the distribution of sTfR in at-risk populations currently limits its utility. METHODS: Using data from NHANES 2003-2010, we examined the distribution of sTfR and developed assay-specific cutoff values for defining elevated sTfR in 2 US populations groups: children aged 1-5 y (n=2820) and non-pregnant women aged 15-49 y (n=6575). RESULTS: On average, children had higher geometric mean sTfR concentrations (4.09 mg/l; 95% CI: 4.04-4.14) than non-pregnant women (3.31 mg/l; 95% CI: 3.26-3.35) (p<0.001). Among children, those aged 1-2 y (compared to those aged 3-5 y), boys (compared to girls), and non-Hispanic black (NHB) children (compared to non-Hispanic white (NHW) and Mexican-American (MA) children) had higher sTfR concentrations. Among non-pregnant women, adolescents (15-19 y) had higher sTfR concentrations than adults aged 20-34 y but not compared to adults aged 35-49 y; NHB women (compared to NHW and MA women) and multiparous women (compared to nulliparous women) had higher sTfR concentrations. The derived cutoff values (97.5th percentile in a defined healthy reference population) for defining elevated sTfR in the US were 6.00 mg/l for children 1-5 y and 5.33 mg/l for non-pregnant women 15-49 y. CONCLUSIONS: A different sTfR cutoff value may be needed in children and non-pregnant women to define ID.

Limited effectiveness of antiviral treatment for hepatitis C in an urban HIV clinic.

OBJECTIVE: To evaluate predictors and trends of referral for hepatitis C virus (HCV) care, clinic attendance and treatment in an urban HIV clinic. DESIGN AND METHODS: A retrospective cohort analysis in which 845 of 1318 co-infected adults who attended the Johns Hopkins HIV clinic between 1998 and 2003 after an on-site viral hepatitis clinic was opened, attended regularly (>/= 1 visit/year for >/= 2 years). Logistic regression was used to examine predictors of referral. RESULTS: A total of 277 (33%) of 845 were referred for HCV care. Independent predictors of referral included percentage elevated alanine aminotransferase levels [adjusted odds ratio (AOR) for 10% increase,1.16; 95% confidence interval (CI), 1.10-1.22] and CD4 cell count > 350 cells/microl (AOR, 3.20; 95% CI, 2.10-4.90), while injection drug use was a barrier to referral (AOR, 0.26; 95% CI, 0.11-0.64). Overall referral rate increased from < 1% in 1998 to 28% in 2003; however, even in 2003, 65% of those with CD4 cell count > 200 cells/microl were not referred. One hundred and eighty-five (67%) of 277 referred kept their appointment, of whom 32% failed to complete a pre-treatment evaluation. Of the remaining 125, only 69 (55%) were medically eligible for treatment, and 29 (42%) underwent HCV treatment. Ninety percent of 29 were infected with genotype 1 and 70% were African American; six (21%) achieved sustained virologic response (SVR). Only 0.7% of the full cohort achieved SVR. CONCLUSIONS: Although the potential for SVR and the recent marked increase in access to HCV care are encouraging, overall effectiveness of anti-HCV treatment in this urban, chiefly African American, HCV genotype 1 HIV clinic is extremely low. New therapies and treatment strategies are an urgent medical need.

Blood lead levels and death from all causes, cardiovascular disease, and cancer: results from the NHANES III mortality study.

BACKGROUND: Analyses of mortality data for participants examined in 1976-1980 in the second National Health and Nutrition Examination Survey (NHANES II) suggested an increased risk of mortality at blood lead levels > 20 microg/dL. Blood lead levels have decreased markedly since the late 1970s. In NHANES III, conducted during 1988-1994, few adults had levels > 20 microg/dL. OBJECTIVE: Our objective in this study was to determine the risk of mortality in relation to lower blood lead levels observed for adult participants of NHANES III. METHODS: We analyzed mortality information for 9,757 participants who had a blood lead measurement and who were > or = 40 years of age at the baseline examination. Using blood lead levels categorized as < 5, 5 to < 10, and > or = 10 microg/dL, we determined the relative risk of mortality from all causes, cancer, and cardiovascular disease through Cox proportional hazard regression analysis. RESULTS: Using blood lead levels < 5 microg/dL as the referent, we determined that the relative risk of mortality from all causes was 1.24 [95% confidence interval (CI), 1.05-1.48] for those with blood levels of 5-9 microg/dL and 1.59 (95% CI, 1.28-1.98) for those with blood levels > or = 10 microg/dL (p for trend < 0.001). The magnitude of risk was similar for deaths due to cardiovascular disease and cancer, and tests for trend were statistically significant (p < 0.01) for both causes of death. CONCLUSION: In a nationally representative sample of the U.S. population, blood lead levels as low as 5-9 mug/dL were associated with an increased risk of death from all causes, cardiovascular disease, and cancer.