PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 9364651

Logistic regression models with missing covariate values for complex survey data.

S Gao·S L Hui
View original source ↗Export metadata ↓

Abstract

Maximum likelihood methods are used to incorporate partially observed covariate values in fitting logistic regression models. We extend these methods to data collected through complex surveys using the pseudo-likelihood approach. One can obtain parameter estimates of the logistic regression model using standard statistical software and their standard errors by Taylor series expansion or the jackknife method. We apply the approach to data from a two-phase survey screening for dementia in a community sample of African Americans age 65 and older living in Indianapolis. The binary response variable is dementia and the covariate with missing values is a daily functioning score collected from interviews with a relative of the study subject.

Explore related subjects

Black or African AmericanAgedCognitionDementiaFemaleHealth SurveysHumansIndianaInterviews as TopicLikelihood FunctionsLogistic ModelsMalePrevalenceSoftwareComparative StudyJournal ArticleResearch Support, Non-U.S. Gov'tResearch Support, U.S. Gov't, P.H.S.

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

S Gao, S L Hui. 1997-11-15. Logistic regression models with missing covariate values for complex survey data.. https://doi.org/10.1002/(sici)1097-0258(19971115)16%3A21%3C2419%3A%3Aaid-sim686%3E3.0.co%3B2-e

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2010

Common charge-shift mutation Glu65Lys in K+ channel β₁-Subunit KCNMB1: pleiotropic consequences for glomerular filtration rate and progressive renal disease.

BACKGROUND: Glomerular filtration rate (GFR) is a heritable trait, and hyperfiltration (GFR increment in remnant nephrons) may accelerate renal functional decline in chronic kidney disease (CKD). Mesangial and vascular smooth myocytes control GFR by contraction, dependent on voltage-gated Ca(2+) influx, which is controlled by the regulatory β₁-subunit (KCNMB1) of large-conductance heteromeric K+ ('BK') channels. KCNMB1 gain-of-function variant Glu65Lys results in generalized vasorelaxation and thus protection against systemic hypertension. Here we asked whether the Glu65Lys variant influences GFR, in the basal state or during progressive renal decline. METHODS: We explored Glu65Lys effects on GFR in three populations spanning two ethnicities and two diseases (hypertension and nephrosclerosis). GFR was either estimated (eGFR from serum creatinine) or directly measured (iothalamate clearance). RESULTS: The 65Lys variant was relatively common, occurring on ∼5-10% of chromosomes in different biogeographic ancestry groups, and 65Lys carriers exhibited higher eGFR in two primary care populations: extreme BP values in Kaiser clinics (p = 0.029, accounting for ∼0.2% of trait variance), or treated hypertensives in VA clinics (p = 0.017, accounting for ∼0.9% of trait variance). In blacks with progressive renal disease (NIDDK AASK), 65Lys carriers displayed a steeper slope in GFR chronic decline (p = 0.030, accounting for ∼0.4% of trait variance), and Glu65Lys genotype also predicted time of onset of renal failure (log rank p = 0.019). CONCLUSIONS: Common KCNMB1 gain-of-function variant Glu65Lys influences GFR, and 65Lys carriers exhibit not only elevated baseline GFR, but also more rapid GFR decline (and consequent development of renal failure) in CKD. The results suggest that profiling patients at Glu65Lys can assist in gauging renal prognosis as well as selection of rational therapy in hypertension with progressive renal disease.

Black or African American↗
PubMed2003

What does "a gene for heart disease" mean? A focus group study of public understandings of genetic risk factors.

There is growing concern in the medical community about potential genetic determinism in the patient population. Limited information about the public understanding of genetic factors in disease formation is available. To access public perceptions of potentially deterministic phrasing of genetic risk factors, we sought to establish interpretations of the phrase, "a gene for heart disease." Focus groups in urban, suburban, and rural communities were conducted from July through October, 2001 in Georgia. A total of 108 participants were recruited. Participants were recruited to balance sex and racial representation. We used three outcome measures for participants understandings of the phrase: (1) participants' statements of the meaning of the phrase; (2) the level of determinism assigned to genetic factors by participants; and (3) participant reports of the health consequences of having "a gene for heart disease." Participants did not report a single interpretation of the phrase. There were dominant participant interpretations under each outcome measure: (1) "a gene for heart disease" was interpreted as meaning genetic and environmental factors both played roles in disease formation; (2) genetic predisposition was perceived as heightened, not absolute, risk; (3) the perceived health impact was a greater risk of becoming sick. Minority interpretations were found under each measure. Overall, naming "a gene for heart disease" does not appear to have a deterministic impact on a plurality of participants' perceptions of risks associated with genetic factors. Genetic fatalism in patient populations may be confined to a sizable minority. Important considerations for provider intervention and patient education are indicated.

Black or African American↗
PubMed2003

Mortality and apolipoprotein E in African-American, and White elders: an attempted replication.

We have tried, with only partial success, to confirm findings in a recently reported study in this journal on the relationship of APOE genotype to mortality in community representative Hispanics (n = 659), Whites (n = 272), and African-Americans (n = 450), aged 65 and over, living in Northern Manhattan, New York. That study found that using proportional hazards models adjusted for sex and lipid levels, Hispanics and Whites with the E2/E3 genotype, but not African-Americans, had the lowest mortality risk. Those under age 75 had risks comparable to those over age 75, suggesting minimal survivor bias. Nearly 50% of the mortality risk associated with the APOE genotype appeared to act through heart disease, diabetes, and stroke. The current study of African-Americans (n = 1,083) and Whites (n = 915) aged 71 and over living in the more rural Southeastern US, found no protective effect of the E2/E3 genotype for either African-Americans or Whites. Among younger Whites (age 71-75), point estimates suggested that the E2/E3 genotype might be protective, but at a nonsignificant level; self-reported African-American race, but not genotype, was a risk factor for mortality in this age group. Neither lipid level nor health condition attenuated the effect of APOE genotype. Differences in findings may reflect issues of sampling, age, the relative distribution of the APOE alleles, or some other factor. Until such time as studies use truly representative samples and include younger ages, findings in this area must be treated with caution.

Black or African American↗