Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Imputation”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 469 records · Page 26Linked to original sources
PubMed2000

Assessing research outcomes by postal questionnaire with telephone follow-up. TOTAL Study Group. Trial of Occupational Therapy and Leisure.

BACKGROUND: Face-to-face assessment of research outcomes is expensive and may introduce bias. Postal questionnaires offer a cheaper alternative which avoids observer bias, but non-response and incomplete response reduce the effective sample size and may be equally serious sources of bias. This study examines the extent and potential effects of missing data in the postal collection of outcomes for a large rehabilitation trial. METHODS: Questionnaires containing a number of established scales were posted to participants in a trial of occupational therapy after stroke. Response was maximized by telephone and postal reminders, and incomplete questionnaires were followed up by telephone. Scale scores obtained by imputing values to questionnaire items missing on return were compared with those achieved by telephone follow-up. FINDINGS: Response to the initial posting was 60%, rising to 85% after reminders. Participants receiving the experimental treatment were more likely to respond without a reminder. There were no significant differences on any known factors between eventual responders and non-responders. Of the questionnaires, 43% were incomplete on return: partial responders were significantly different to complete responders on baseline disability and home circumstances. Of the incomplete questionnaires, 71% were resolved by telephone follow-up. In these, the scale scores achieved by telephone were generally higher than those derived by conventional imputation. CONCLUSION: Postal outcome assessment achieved a good response rate, but considerable effort was needed to minimize non-response and incomplete response, both of which could have been serious sources of bias.

Bias↗
PubMed2004

The systemic lupus erythematosus Tri-nation Study: absence of a link between health resource use and health outcome.

OBJECTIVE: Health consumption and health status in SLE in three countries with different health funding structures were compared. METHODS: Seven hundred and fifteen SLE patients (Canada 231, USA 269, UK 215) were surveyed semi-annually over 4 yr for health resource utilization and health status. Cross-country comparisons of (i) cumulative health expenditure (calculated by applying 2002 Canadian prices to resources in all countries) and (ii) disease damage (Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index, SLICC/ACR DI) at study conclusion were performed after adjustment. Missing expenditure and damage data were managed through multiple imputation using best predictive regressions with all available data from all patients as potential covariates. RESULTS: Four hundred and eighty-five patients provided data at study entry and conclusion and at least four resource questionnaires (Canada 162, USA 157, UK 166); 41 died (Canada 13, USA 18, UK 10); 189 withdrew, were lost to follow-up or provided data at entry and conclusion but fewer than four resource questionnaires (Canada 56, USA 94, UK 39). At conclusion, after imputation, in Canada, the USA and the UK respectively, mean cumulative costs per patient over 4 yr [95% confidence interval (CI)] were $15,845 (13,509, 18,182), $20,244 (17,764, 22,724) and $17,647 (15,557, 19,737) and mean changes in SLICC/ACR DI were 0.49 (0.39, 0.60), 0.63 (0.52, 0.74) and 0.48 (0.39, 0.57). After adjustment for baseline differences, on average (95% CI), Canadian and British patients utilized 20% (8%, 32%) and 13% (1%, 24%) less resources than patients in the USA respectively, but experienced similar health outcomes. CONCLUSION: Despite patients in the USA incurring higher health expenditures, they did not experience superior health outcomes.

Adult↗
PubMed1986

Rapid estimation of hospitalization charges from a brief medical record review. Evaluation of a multivariate prediction model.

In settings where an itemized hospital bill is not generated, estimation of hospitalization charges for research or administrative purposes can be a laborious task. This article examines the extent to which the number of hospital days spent outside an intensive care unit (ICU), number of days in an ICU, number of laboratory tests performed, number of x-rays, and number of surgeries can be used in a multiple regression equation to impute inpatient charges for a sample of 103 hospitalizations at a Veterans Administration hospital. These predictor variables, all of which are readily ascertained in a brief medical record review, accounted for about 97% of the variance in imputed hospital charges. The bootstrap method was applied for validation of the prediction equation. Application of the method described here may be of value to researchers concerned with hospital charge estimation in non-fee-for-service settings.

Aged↗
PubMed2006

Psychological job demands increase the risk of ischaemic heart disease: a 14-year cohort study of employed Danish men.

AIMS: To test the relationship between job strain and the incidence of ischaemic heart disease (IHD) prospectively in the Danish working population. METHODS AND RESULTS: In 1986, a clinical examination was undertaken of 659 men, all employed and without known IHD, together with a questionnaire-based evaluation of living conditions and psychosocial factors at work, including items identified in the job strain model. This study was part of the World Health Organization-initiated MONICA II study. In the job strain model, job strain is defined as the combination of high psychological demands and a low degree of control in the work situation. An objective classification of the components in the job strain model was made by imputation by utilizing the participants' job title and the principles guiding the payment of their salaries/wages. In addition, a questionnaire-based subjective classification was undertaken. All participants were followed until the end of 1999 with regard to hospitalization and death as a result of IHD. Stepwise analyses were made, adjusting for age, social class, social network and established behavioural and physiological coronary risk factors. Self-reported job strain was significantly associated with IHD independently of standard coronary risk factors. Of the two components in the job strain model only high demands contributed significantly to this result. The study did not support the job strain hypothesis when an imputed, objective classification of the components in the job strain model was applied. This is in accordance with the majority of other studies in this area. An unexpected finding was that the incidence of IHD was highest among employers and managers. CONCLUSION: High psychological demands at work are a risk factor for IHD, a fact that should affect the primary and secondary prevention of IHD.

Adult↗
PubMed2005

Methods for addressing missing data in psychiatric and developmental research.

OBJECTIVE: First, to provide information about best practices in handling missing data so that readers can judge the quality of research studies. Second, to provide more detailed information about missing data analysis techniques and software on the Journal's Web site at www.jaacap.com. METHOD: We focus our review of techniques on those that are based on the "Missing at Random" assumption and are either extremely popular because of their convenience or that are harder to employ but yield more precise inferences. RESULTS: The literature regarding missing data indicates that deletion of observations with missing data can yield biased findings. Other popular methods for handling missing data, notably replacing missing values with means, can lead to confidence intervals that are too narrow as well as false identifications of significant differences (type I statistical errors). Methods such as multiple imputation and direct maximum likelihood estimation are often superior to deleting observations and other popular methods for handling missing data problems. CONCLUSIONS: Psychiatric and developmental researchers should consider using multiple imputation and direct maximum likelihood estimation rather than deleting observations with missing values.

Adolescent↗
PubMed2005

Reducing misclassification in assignment of timing of events during pregnancy.

BACKGROUND: Perinatal epidemiology studies often collect only the calendar month in which an event occurs in early pregnancy because it is difficult for women to recall a specific day when queried later in pregnancy or postpartum. Lack of day information may result in incorrect assignment of completed gestational month because calendar months and pregnancy months are not aligned. METHODS: To examine the direction and magnitude of misclassification, we compared 3 methods for assignment of completed gestational month: 1) calendar month difference, 2) conditional month difference, and 3) imputed month midpoint. We used data from the Pregnancy, Infection, and Nutrition Study for simulations. RESULTS: Calendar month difference misclassified 54% of events as 1 month later in pregnancy compared with the actual completed month of gestation. Each of the other 2 methods misclassified approximately 12% of events to 1 month earlier and 12% to 1 month later. CONCLUSIONS: Calendar month difference, a common method, has the greatest misclassification. Conditional month difference and imputed month midpoint, which require little effort to implement, are superior to calendar month difference for reducing misclassification.

Cohort Studies↗
PubMed2003

Accounting for deaths in longitudinal studies using the SF-36: the performance of the Physical Component Scale of the Short Form 36-item health survey and the PCTD.

BACKGROUND: Commonly used measures such as the Physical Component Scale of the Short Form 36-item health survey (PCS) are undefined at death, limiting longitudinal analyses to survivors, a healthier cohort that cannot be identified prospectively, and that might have had little change in health. One proposed approach is to transform the PCS into the Physical Component Transformed, with Deaths included (PCTD), which is the probability of being healthy 1 year later and for which deaths logically have a value of zero. Data missing for other reasons than death have not been considered. OBJECTIVE: To examine the performance of the PCTD, to determine the influence of including deaths, the additional effects of imputing missing values and adjusting for covariates, and the calibration of the PCTD in different populations. METHODS: We imputed missing values of the PCTD, calling the new variable the PCTDI. We compared the distributions of the PCS, PCTD, and PCTDI cross-sectionally and over time. In 3 different populations, we determined whether the PCTD accurately predicted the probability of being healthy 1 year later. RESULTS: The patients who died did not have extreme values on the PCTD. The experience of the cohort was best described by the PCTDI. The calibration of the PCTD was surprisingly good in all the populations examined. Results were similar for the physical function index. CONCLUSION: The PCTDI is an improvement over the PCS, in which patients who had died have no influence, and over the PCTD, where they might have too much influence. We recommend the PCTDI for longitudinal analyses of physical health when deaths occur, for primary or secondary analysis.

Aged↗
PubMed2000

Assessing overall risk in reproductive experiments.

Toxicologists are often interested in assessing the joint effect of an exposure on multiple reproductive endpoints, including early loss, fetal death, and malformation. Exposures that occur prior to mating or extremely early in development can adversely affect the number of implantation sites or fetuses that form within each dam and may even prevent pregnancy. A simple approach for assessing overall adverse effects in such studies is to consider fetuses or implants that fail to develop due to exposure as missing data. The missing data can be imputed, and standard methods for the analysis of quantal response data can then be used for quantitative risk assessment or testing. In this article, a new bias-corrected imputation procedure is proposed and evaluated. The procedure is straightforward to implement in standard statistical packages and has excellent operating characteristics when used in combination with a marginal model fit with generalized estimating equations. The methods are applied to data from a reproductive toxicity study of Nitrofurazone conducted by the National Toxicology Program.

Abnormalities, Drug-Induced↗
PubMedDate not supplied

An assessment of the impact of informative dropout and nonresponse in measuring health-related quality of life using the EuroQol (EQ-5D) descriptive system.

OBJECTIVES: To investigate the impact of imputing EQ-5D values to allow for informative dropout and nonresponse in a longitudinal assessment of the health-related quality of life (HRQL) of liver transplant recipients. METHODS: The EQ-5D was administered at defined time intervals pre- and post-transplantation to all adults who were listed to receive liver transplants as National Health Service (NHS) treatment at each of the six Department of Health designated centers in England and Wales over a time-period of 36 months (12 month recruitment period and 24 month follow-up period). During the course of the study missing data arose for two main reasons, informative dropout and nonresponse. Informative dropout was accounted for by giving those patients who died an EQ-5D score of 0 and those patients who were too ill to respond to an EQ-5D score equivalent to the 5th percentile of respondents for each time point pretransplantation. Nonresponse was accounted for using relatively naive approaches (last value carried forward, and upper/lower 95% confidence interval around the mean) and contrasted with a more sophisticated multiple imputation method. RESULTS: Adjusting for informative dropout in isolation resulted in a marked deterioration in mean scores over time pretransplant relative to the base case situation in which no such adjustments were made. Nevertheless, adjusting for informative dropout and/or nonresponders did not alter the base case conclusion of no statistically significant differences in mean EQ-5D scores over time pretransplant. In contrast, post-transplant data indicated highly statistically significant improvements in quality of life over time for the base case (P < 0.001) whereas no statistically significant improvements over time were found when informative dropout was allowed for in isolation (P = 0.402) or when informative dropout and nonresponse were allowed for simultaneously (P = 0.105-0.185). CONCLUSIONS: It is important that future studies which purport to assess the HRQL over time of patients, such as these with end-stage liver disease, include an allowance for informative dropout and nonresponse within the analysis.

Bias↗
PubMed2004

A new method for computing the multipoint posterior probability of linkage.

The posterior probability of linkage (PPL) is a Bayesian statistic which directly measures the probability of linkage between a trait locus and a marker (in the 2-point case) or a genomic region (in the multipoint case). It has several benefits, including ease of interpretation, the ability to incorporate prior genomic information, and a mathematically rigorous and robust procedure for accumulating linkage information across multiple heterogeneous datasets. To date, the majority of work on the PPL has focused on the development of the 2-point statistic, with only preliminary attempts at the development of an equivalent multipoint version. In this paper we present a new way of computing of the multipoint PPL. This new version imputes to each genomic point an estimate of the 2-point PPL we would have obtained from a fully informative marker giving similar evidence for linkage. This version, which we call the imputed PPL, is shown to be superior to previously developed versions.

Bayes Theorem↗
PubMedDate not supplied

Options for handling missing data in the Health Utilities Index Mark 3.

BACKGROUND: The Health Utilities Index Mark 3 (HUI3) is a tool composed of 41 questions, covering 8 attributes: vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain. Responses to these questions can define more than 972,000 health situations. This tool allows respondents to answer "Don't Know," for which there is no scoring instruction, to any given question. This situation creates a break in the scoring algorithm and leads to considerable amounts of missing data. The goal of this study is to develop strategies to deal with HUI3 scores for participants who have missing data. METHODS: The authors used data from 248 individuals enrolled in the Cataract Management Trial, focusing on the HUI3 vision and ambulation attributes, which had 19% and 10% of attribute levels missing, respectively. Inspection and deduction were used to fill in values independent of the value of the missing data, then alternative analytic techniques were compared, including mean substitution, model scoring, hot deck, multiple imputation, and regression imputation. RESULTS: Inspection and logical deduction reduced the percentage of missing information in the HUI3 by 49% to 87%. A comparison of analytic techniques used for the remaining HUI3 vision data missing demonstrated the value of building models based on internal response patterns and that simple analytic techniques fare as well as more complicated ones when the number of missing cases is small. CONCLUSION: Analyzing the pattern of responses in cases where the attribute level score is missing reduces the amount of missing data and can simplify the analytic process for the remaining missing data.

Activities of Daily Living↗
PubMed2001

Multilevel analysis with messy data.

This paper reviews applications of the method of multiple imputation to dealing with multilevel data that have several kinds of imperfections. These are classified into two broad categories: missing values and imprecise measurement (corrupted recording). The role of the model describing the data imperfections is emphasized. With multiple imputation, these imperfections and information about the processes underlying them can be taken into account. The inferences drawn exploit all the collected information and appropriately reflect the information contained in the data.

Algorithms↗
PubMed1996

Handling missing data in survey research.

Missing data occur in survey research because an element in the target population is not included on the survey's sampling frame (noncoverage), because a sampled element does not participate in the survey (total nonresponse) and because a responding sampled element fails to provide acceptable responses to one or more of the survey items (item nonresponse). A variety of methods have been developed to attempt to compensate for missing survey data in a general purpose way that enables the survey's data file to be analysed without regard for the missing data. Weighting adjustments are often used to compensate for noncoverage and total nonresponse. Imputation methods that assign values for missing responses are used to compensate for item nonresponses. This paper describes the various weighting and imputation methods that have been developed, and discusses their benefits and limitations.

Data Interpretation, Statistical↗
PubMed2003

Comparison of missing data approaches in linkage analysis.

BACKGROUND: Observational cohort studies have been little used in linkage analyses due to their general lack of large, disease-specific pedigrees. Nevertheless, the longitudinal nature of such studies makes them potentially valuable for assessing the linkage between genotypes and temporal trends in phenotypes. The repeated phenotype measures in cohort studies (i.e., across time), however, can have extensive missing information. Existing methods for handling missing data in observational studies may decrease efficiency, introduce biases, and give spurious results. The impact of such methods when undertaking linkage analysis of cohort studies is unclear. Therefore, we compare here six methods of imputing missing repeated phenotypes on results from genome-wide linkage analyses of four quantitative traits from the Framingham Heart Study cohort. RESULTS: We found that simply deleting observations with missing values gave many more nominally statistically significant linkages than the other five approaches. Among the latter, those with similar underlying methodology (i.e., imputation- versus model-based) gave the most consistent results, although some discrepancies remained. CONCLUSION: Different methods for addressing missing values in linkage analyses of cohort studies can give substantially diverse results, and must be carefully considered to protect against biases and spurious findings.

Algorithms↗
PubMed2001

A functional update of the Escherichia coli K-12 genome.

BACKGROUND: Since the genome of Escherichia coli K-12 was initially annotated in 1997, additional functional information based on biological characterization and functions of sequence-similar proteins has become available. On the basis of this new information, an updated version of the annotated chromosome has been generated. RESULTS: The E. coli K-12 chromosome is currently represented by 4,401 genes encoding 116 RNAs and 4,285 proteins. The boundaries of the genes identified in the GenBank Accession U00096 were used. Some protein-coding sequences are compound and encode multimodular proteins. The coding sequences (CDSs) are represented by modules (protein elements of at least 100 amino acids with biological activity and independent evolutionary history). There are 4,616 identified modules in the 4,285 proteins. Of these, 48.9% have been characterized, 29.5% have an imputed function, 2.1% have a phenotype and 19.5% have no function assignment. Only 7% of the modules appear unique to E. coli, and this number is expected to be reduced as more genome data becomes available. The imputed functions were assigned on the basis of manual evaluation of functions predicted by BLAST and DARWIN analyses and by the MAGPIE genome annotation system. CONCLUSIONS: Much knowledge has been gained about functions encoded by the E. coli K-12 genome since the 1997 annotation was published. The data presented here should be useful for analysis of E. coli gene products as well as gene products encoded by other genomes.

Bacterial Proteins↗
PubMed2000

Long-term health effects of particulate and other ambient air pollution: research can progress faster if we want it to.

There is need for the assessment of long-term effects of outdoor air pollution. In fact, a considerable part of the large amount of U.S. research money that has been dedicated to investigate effects of ambient particulate pollution should be invested to address long-term effects. Studies that follow the health status of large numbers of subjects across long periods of time (i.e., cohort studies) should be considered the key research approach to address these questions. However, these studies are time consuming and expensive. We propose efficient strategies to address these questions in less time. Apart from long-term continuation of the few ongoing air pollution cohort studies in the United States, data from large cohorts that were established decades ago may be efficiently used to assess cardiorespiratory effects and to target research on detection of the most susceptible subgroups in the population, which may be related to genetic, molecular, behavioral, societal, and/or environmental factors. This approach will be efficient only if the available air pollution monitoring data will be used to spatially model long-term outdoor pollution concentrations across a given country for each year with available pollution data. Such concentration maps will allow researchers to impute outdoor air pollution levels at any residential location, independent of the location of monitors. Exposure imputation may be based on residential location(s) of participants in long-standing cardiorespiratory cohort studies, which can be matched to pollutant levels using geographic information systems. As shown in European impact assessment studies, such maps may be derived relatively quickly.

Air Pollutants↗
PubMed2006

Does celecoxib potentiate the anticoagulant effect of warfarin? A randomized, double-blind, controlled trial.

BACKGROUND: The management of patients who are receiving warfarin therapy and have musculoskeletal problems that require treatment with a nonsteroidal antiinflammatory drug (NSAID) is problematic because NSAID use may increase the risk for bleeding. Cyclooxygenase-2 selective NSAIDs such as celecoxib may be less likely to promote gastrointestinal bleeding; however, there are concerns that they could potentiate the anticoagulation effect of warfarin. OBJECTIVE: To determine whether celecoxib potentiates the anticoagulant effect of warfarin, as measured by the international normalized ratio (INR). METHODS: We performed a randomized, controlled, crossover trial to assess the effect on INR of celecoxib versus codeine (control treatment) in 15 patients who were receiving warfarin therapy and required analgesic treatment for osteoarthritis. During Phase 1 of the study, patients were randomly allocated to receive celecoxib 200 mg/day or codeine phosphate 7-15 mg 3-4 times daily for 5 weeks. During Phase 2 of the study, patients stopped the first study medication and started the other study medication; there was no drug-free interval between phases. Weekly INR testing was performed during the 10 week study period. Adopting the intent-to-treat principle, we used generalized estimating equations to analyze the data. RESULTS: There was no significant difference in the mean INR values during each 5 week treatment period when patients received either celecoxib or codeine. There was, therefore, insufficient evidence to reject the hypothesis that these 2 treatments had an equal effect on the INR (mean difference [95% CI] 0.10 [-0.04 to 0.24]; p = 0.16) based on mean imputation. This finding was confirmed after we repeated the analysis with multiple imputations (mean difference [95% CI] 0.093 [-0.16 to 0.35]; p = 0.47). CONCLUSIONS: Our results suggest that treatment with celecoxib does not potentiate the INR when taken with warfarin. Larger randomized trials are warranted to address the effects of coadministered warfarin and celecoxib on clinical outcomes.

Aged↗
PubMed2000

Direct expenditures related to otitis media diagnoses: extrapolations from a pediatric medicaid cohort.

BACKGROUND: Treatment of otitis media in children is associated with substantial expenditures because of its high frequency during childhood. Vaccines against respiratory pathogens causing otitis media are now being developed. Information about otitis media-related medical expenditures will be needed to determine the cost-effectiveness of these preventive interventions. METHODS: This study used utilization data from claims to impute otitis media-related expenditures for medical visits, pharmaceuticals, and surgical procedures for 87 057 children 13 years of age and younger who were continuously enrolled in Colorado's fee-for-service Medicaid program during 1992. International Classification of Disease, Ninth Revision diagnostic codes were used to identify visits for otitis media. An antibiotic was considered to have been prescribed to treat otitis media if it was dispensed up to 24 hours before or within 48 hours after a physician encounter showing a diagnosis of otitis media. All tympanostomies, mastoidectomies, and adenoidectomies were assumed to be related to otitis media. Expenditures were imputed from utilization using a Medicaid fee schedule. National expenditures for 1992 to treat otitis media were extrapolated from Colorado's Medicaid data. We adjusted for differences between Colorado and the United States as a whole in terms of price, number, and intensity of services; for differences in reimbursement rates by service between Medicaid and private insurance; and for differences in utilization between Medicaid enrollees and the uninsured. To provide a more current expression of medical expenditures for otitis media, we inflated the 1992 expenditure estimates to 1998 dollars using the Consumer Price Index published by the US Bureau of Labor Statistics. RESULTS: Twenty-eight percent of children experienced at least 1 episode of diagnosed otitis media. The proportion of children with a diagnosis of otitis media was highest (42%-60%) in the 7-month to 36-month age range. The proportion was also higher among white (34.5%) and Hispanic (25.3%) children than among black children (18.5%), as well as among rural (34.5%) compared with urban children (27.2%). Children 19 to 24 months of age incurred the highest total annual expenditures per child with otitis media ($239.68). Expenditures for drugs, visits, and procedures were all highest for this group. The per-patient cost to Medicaid was greater for visits than for drugs or procedures across all age groups. Total per-patient expenditures were higher for males ($174.67) than for females ($154.47) and higher for white children ($176.59) than for Hispanic ($154.12) or black children ($134.44). The differences among the ethnic groups can be attributed almost entirely to differences in expenditures for procedures and drugs. Although mean expenditures per patient varied substantially by some patient characteristics (eg, race), these differences accounted for only a small fraction of the enormous variation in costs per patient. Including children with and without otitis media, age-specific estimated expenditures per child peaked among children 1 ($132.94) and 2 years of age ($88.72). Children 3 to 6 years of age incurred expenditures only one third as great as those incurred by children 1 year of age. Total national expenditures were estimated to have been approximately $4.1 billion in 1992 dollars and $5.3 billion in 1998 dollars. Over 40% of national expenditures to treat otitis media in children younger than 14 years of age were incurred for children between 1 and 3 years of age ($453 per capita in 1992 dollars over these 2 years vs $1027 for all years of age from 2 to 13). Nationally, expenditures for visits remained the largest component of expenditures. LIMITATIONS: This study assessed expenditures from the point of view of the health care system; that is, no social costs, such as lost work time, or expenditures not normally covered by insurance, such as those for transportation, we

Adolescent↗