@article {13033, title = {Validation of Self-Reported Cancer Diagnoses by Respondent Cognitive Status in the U.S. Health and Retirement Study.}, journal = {The Journals of Gerontology. Series A }, year = {Forthcoming}, abstract = {

BACKGROUND: Cancer and dementia are becoming increasingly common co-occurring conditions among older adults. Yet, the influence of participant cognitive status on the validity of self-reported data among older adults in population-based cohorts is unknown. We thus compared self-reported cancer diagnoses in the US Health and Retirement Study (HRS) against claims from linked Medicare records to ascertain the validity of self-reported diagnoses by participant cognitive and proxy interview status.

METHODS: Using data from HRS participants aged >=67 who had at least 90\% continuous enrollment in fee-for-service Medicare, we examined the validity of self-reported first incident cancer diagnoses from biennial HRS interviews against diagnostic claim records in linked Medicare data (reference standard) for interviews from 2000-2016. Cognitive status was classified as normal, cognitive impairment no dementia (CIND), or dementia using the Langa-Weir method. We calculated the sensitivity, specificity, and κfor cancer diagnosis.

RESULTS: Of the 8,280 included participants, 23.6\% had cognitive impairment without dementia (CIND) or dementia ,and 10.7\% had a proxy respondent due to an impairment. Self-reports of first incident cancer diagnoses for participants with normal cognition had 70.2\% sensitivity and 99.8\% specificity (κ=0.79). Sensitivity declined substantially with cognitive impairment and proxy response (56.7\% for CIND, 53.0\% for dementia, 60.0\% for proxy respondents), indicating poor validity for study participants with CIND, dementia, or a proxy respondent.

CONCLUSION: Self-reported cancer diagnoses in the US HRS have poor validity for participants with cognitive impairment, dementia, or a proxy respondent. Population-based cancer research among older adults will be strengthened with linkage to Medicare claims.

}, keywords = {Cognition, Dementia, self-reported diagnoses, sensitivity, specificity, Validation}, issn = {1758-535X}, doi = {10.1093/gerona/glac248}, author = {Mullins, Megan A and Kabeto, Mohammed and Wallner, Lauren P and Lindsay C Kobayashi} } @article {10817, title = {Comprehensive review of ICD-9 code accuracies to measure multimorbidity in administrative data}, journal = {BMC Health Services Research}, volume = {20}, year = {2020}, chapter = {489}, abstract = {Background Quantifying the burden of multimorbidity for healthcare research using administrative data has been constrained. Existing measures incompletely capture chronic conditions of relevance and are narrowly focused on risk-adjustment for mortality, healthcare cost or utilization. Moreover, the measures have not undergone a rigorous review for how accurately the components, specifically the International Classification of Diseases, Ninth Revision (ICD-9) codes, represent the chronic conditions that comprise the measures. We performed a comprehensive, structured literature review of research studies on the accuracy of ICD-9 codes validated using external sources across an inventory of 81 chronic conditions. The conditions as a weighted measure set have previously been demonstrated to impact not only mortality but also physical and mental health-related quality of life. Methods For each of 81 conditions we performed a structured literature search with the goal to identify 1) studies that externally validate ICD-9 codes mapped to each chronic condition against an external source of data, and 2) the accuracy of ICD-9 codes reported in the identified validation studies. The primary measure of accuracy was the positive predictive value (PPV). We also reported negative predictive value (NPV), sensitivity, specificity, and kappa statistics when available. We searched PubMed and Google Scholar for studies published before June 2019. Results We identified studies with validation statistics of ICD-9 codes for 51 (64\%) of 81 conditions. Most of the studies (47/51 or 92\%) used medical chart review as the external reference standard. Of the validated using medical chart review, the median (range) of mean PPVs was 85\% (39{\textendash}100\%) and NPVs was 91\% (41{\textendash}100\%). Most conditions had at least one validation study reporting PPV >=70\%. Conclusions To help facilitate the use of patient-centered measures of multimorbidity in administrative data, this review provides the accuracy of ICD-9 codes for chronic conditions that impact a universally valued patient-centered outcome: health-related quality of life. These findings will assist health services studies that measure chronic disease burden and risk-adjust for comorbidity and multimorbidity using patient-centered outcomes in administrative data.}, keywords = {ICD-9, Literature Review, multimorbidity, Validation}, isbn = {1472-6963}, doi = {10.1186/s12913-020-05207-4}, author = {Melissa Y Wei and Jamie E Luster and Chan, Chiao-Li and Lillian C. Min} } @article {9365, title = {Eliciting Stock Market Expectations: The Effects of Question Wording on Survey Experience and Response Validity}, journal = {Journal of Behavioral Finance}, volume = {19}, year = {2018}, pages = {101-110}, abstract = {Expectations about stock market movements are an important factor in models of investments and savings. To better understand consumers{\textquoteright} financial behavior, economic surveys such as the Health and Retirement Study (HRS) ask participants to report expectations about the stock market. However, readability statistics suggest that the HRS{\textquoteright} stock market expectations questions use relatively complex wording, which may contribute to their relatively high rates of missing responses. Here, the authors build on survey design research to improve the readability of these questions. In 2 experiments using national online panels, they test whether revising stock market expectation questions to reduce their difficulty affects respondents{\textquoteright} (1) survey experience, as measured by percent of missing answers and ratings of question clarity and difficulty, and (2) response validity, as assessed by respondents{\textquoteright} confidence in their answer and comparisons between expectations and stock market outcomes. In both studies, the authors{\textquoteright} revisions improved survey experience. Unfortunately, revisions also decreased the perceived (Study 1) and actual (Study 2) validity of responses. The authors discuss implications of question revisions for the design of economic surveys.}, keywords = {Financial literacy, Meta-analyses, Survey Methodology, Validation}, issn = {1542-7560}, doi = {10.1080/15427560.2017.1373353}, url = {https://www.tandfonline.com/doi/full/10.1080/15427560.2017.1373353https://www.tandfonline.com/doi/pdf/10.1080/15427560.2017.1373353}, author = {Chin, Alycia and Bruine De Bruin, W{\"a}ndi} } @article {9351, title = {A machine-learning heuristic to improve gene score prediction of polygenic traits.}, journal = {Scientific Reports}, volume = {7}, year = {2017}, pages = {12665}, abstract = {Machine-learning techniques have helped solve a broad range of prediction problems, yet are not widely used to build polygenic risk scores for the prediction of complex traits. We propose a novel heuristic based on machine-learning techniques (GraBLD) to boost the predictive performance of polygenic risk scores. Gradient boosted regression trees were first used to optimize the weights of SNPs included in the score, followed by a novel regional adjustment for linkage disequilibrium. A calibration set with sample size of ~200 individuals was sufficient for optimal performance. GraBLD yielded prediction R (2) of 0.239 and 0.082 using GIANT summary association statistics for height and BMI in the UK Biobank study (N = 130 K; 1.98 M SNPs), explaining 46.9\% and 32.7\% of the overall polygenic variance, respectively. For diabetes status, the area under the receiver operating characteristic curve was 0.602 in the UK Biobank study using summary-level association statistics from the DIAGRAM consortium. GraBLD outperformed other polygenic score heuristics for the prediction of height (p < 2.2 {\texttimes} 10(-16)) and BMI (p < 1.57 {\texttimes} 10(-4)), and was equivalent to LDpred for diabetes. Results were independently validated in the Health and Retirement Study (N = 8,292; 688,398 SNPs). Our report demonstrates the use of machine-learning techniques, coupled with summary-level data from large genome-wide meta-analyses to improve the prediction of polygenic traits.}, keywords = {Genetics, Machine learning, Meta-analyses, Validation}, issn = {2045-2322}, doi = {10.1038/s41598-017-13056-1}, author = {Pare, Guillaume and Mao, Shihong and Wei Q. Deng} } @article {9348, title = {Validation of a theoretically motivated approach to measuring childhood socioeconomic circumstances in the Health and Retirement Study.}, journal = {PLoS One}, volume = {12}, year = {2017}, month = {2017}, pages = {e0185898}, abstract = {Childhood socioeconomic status (cSES) is a powerful predictor of adult health, but its operationalization and measurement varies across studies. Using Health and Retirement Study data (HRS, which is nationally representative of community-residing United States adults aged 50+ years), we specified theoretically-motivated cSES measures, evaluated their reliability and validity, and compared their performance to other cSES indices. HRS respondent data (N = 31,169, interviewed 1992-2010) were used to construct a cSES index reflecting childhood social capital (cSC), childhood financial capital (cFC), and childhood human capital (cHC), using retrospective reports from when the respondent was <16 years (at least 34 years prior). We assessed internal consistency reliability (Cronbach{\textquoteright}s alpha) for the scales (cSC and cFC), and construct validity, and predictive validity for all measures. Validity was assessed with hypothesized correlates of cSES (educational attainment, measured adult height, self-reported childhood health, childhood learning problems, childhood drug and alcohol problems). We then compared the performance of our validated measures with other indices used in HRS in predicting self-rated health and number of depressive symptoms, measured in 2010. Internal consistency reliability was acceptable (cSC = 0.63, cFC = 0.61). Most measures were associated with hypothesized correlates (for example, the association between educational attainment and cSC was 0.01, p < 0.0001), with the exception that measured height was not associated with cFC (p = 0.19) and childhood drug and alcohol problems (p = 0.41), and childhood learning problems (p = 0.12) were not associated with cHC. Our measures explained slightly more variability in self-rated health (adjusted R2 = 0.07 vs. <0.06) and number of depressive symptoms (adjusted R2 > 0.05 vs. < 0.04) than alternative indices. Our cSES measures use latent variable models to handle item-missingness, thereby increasing the sample size available for analysis compared to complete case approaches (N = 15,345 vs. 8,248). Adopting this type of theoretically motivated operationalization of cSES may strengthen the quality of research on the effects of cSES on health outcomes.}, keywords = {Childhood adversity, Meta-analyses, Socioeconomic factors, Validation}, issn = {1932-6203}, doi = {10.1371/journal.pone.0185898}, author = {Anusha M Vable and Paola Gilsanz and Thu T Nguyen and Ichiro Kawachi and M. Maria Glymour} }