Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Selection Bias”

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 37 records · Page 2Linked to original sources
PubMed1995

Who says yes? Identifying selection biases in a psychosocial intervention study of multiple sclerosis.

The purpose of this work is to examine whether patients' sociodemographic and medical characteristics are associated with participation in a randomized controlled trial of two psychosocial interventions. After elimination of basic ineligibles from a registry of multiple sclerosis patients, 325 patients were sent a letter inviting participation in the randomized trial. Among those invited, 29% expressed an interest in participating, 19% scheduled an intake interview and 13% were actually randomized. chi 2 and logistic regression analyses were used to determine what factors were associated with successive stages of participation or non-participation in the study. Differential referral and participation rates could be traced in part to physician factors. With respect to patient factors, people were more likely to participate in the successive stages of recruitment if they had a higher median family income, lived a moderate distance from the hospital, and were disabled from working. Multivariate analyses of patient factors revealed that having a higher income and being disabled from work were the strongest predictors of participation, after adjusting for the effects of course of disease, disability status, and other sociodemographic predictors. Implications of these findings are discussed in terms of potential sources of bias in the selection of eligible patients, as well as the role the investigator may play in highlighting or underplaying this selection bias. Suggestions are made to minimize the bias-generating barriers.

Adult↗
PubMed2004

The intervention selection bias: an underrecognized confound in intervention research.

Selection bias can be the most important threat to internal validity in intervention research, but is often insufficiently recognized and controlled. The bias is illustrated in research on parental interventions (punishment, homework assistance); medical interventions (hospitalization); and psychological interventions for suicide risk, sex offending, and juvenile delinquency. The intervention selection bias is most adequately controlled in randomized studies or strong quasi-experimental designs, although recent statistical innovations can enhance weaker designs. The most important points are to increase awareness of the intervention selection bias and to systematically evaluate plausible alternative explanations of data before making causal conclusions.

Child↗
PubMedDate not supplied

Outcomes from stroke rehabilitation in Veterans Affairs rehabilitation units: detecting and correcting for selection bias.

This paper addresses the issue of statistical selection bias in multivariate models of functional gain estimated from observational data. Stroke patients from 20 high-volume Veterans Affairs Medical Centers (VAMCs) with acute and subacute inpatient rehabilitation treatment units were observed. Their gains in overall, motor, and cognitive functional status were measured with the use of the Functional Independence Measure (FIM). In estimating multivariate models of FIM gain during rehabilitation using these observational data, we found statistically significant evidence of selection bias, along with considerable differences in inferences between standard multivariate analyses and our selectivity-corrected models. Our results demonstrate the importance of detecting and correcting for statistical selection bias when one uses nonexperimental data to study gains in functional status.

Activities of Daily Living↗
PubMed2001

Selection bias and confounding in case-crossover analyses of environmental time-series data.

The case-crossover study design is a popular analytic tool for estimating the effects of triggers of acute outcomes by environmental exposures. Although this approach controls for time-invariant confounders by design, it may allow for selection bias and confounding by time-varying factors. We conducted a simulation study of the sensitivity of the symmetric bidirectional case-crossover design to time-varying patterns in exposure and outcome. We identified the effects of selection bias and confounding on symmetric bidirectional case-crossover results and offer strategies to eliminate or reduce these biases. Selection bias results when exposure in the reference periods is not identically representative of exposure in the hazard periods, even when the distribution of exposure is stationary. This bias can be estimated and removed. Selection bias also occurs when the distribution of exposure is nonstationary, but the adjusted symmetric bidirectional case-crossover methodology substantially controls for this. Confounding results from a common temporal pattern in the exposure and the outcome time series, but can also be the result of patterns in exposure and outcome that, although asymptotically uncorrelated, are correlated at finite series lengths. All three biases are reduced by choosing shorter referent-spacing lengths. This effect is illustrated using data on air pollution and daily deaths in Chicago.

Air Pollution↗
PubMed2003

Selection bias in prenatal care use by Medicaid recipients.

OBJECTIVE: This study examines whether four types of selection bias in estimates of the effectiveness of prenatal care utilization for improving birthweight occur in a population of economically disadvantaged women. METHODS: We categorized adequacy of prenatal care use using the Adequacy of Prenatal Care Utilization Index (APNCU) and the Revised-GINDEX for 142,381 Medicaid recipients who gave birth to a live, singleton infant in Washington State (1994-1998). Multinomial logistic regression was used to model categories of adequacy of prenatal care use as functions of variables chosen to indicate high- or low-risk status. A series of linear regression models were estimated to quantify the magnitude and direction of any bias in the effects of prenatal care on birthweight that could be attributed to accounting for each risk covariate. Results were examined for patterns of risk, prenatal care use, and estimation bias equated with the four selection processes. RESULTS: We found modest evidence of adverse, favorable, confidence, and estrangement selection biases. The overriding effect, relative to low prenatal care use, was overestimation of the adequate care coefficient by 8.68 g with the APNCU, and underestimation by 3.36 g with the R-GINDEX because of competing confidence and estrangement selection biases. Relative to intensive use, the effect of adequate care on birthweight was underestimated (17.58 g with the APNCU; 13.34 g with the R-GINDEX) because of adverse selection bias and a small countervailing favorable selection process. CONCLUSIONS: The underestimation of birthweight associated with prenatal care noted in prior studies appears to result from multiple selection processes working in different directions. Understanding selection processes can help the assessment of the contribution of prenatal care to birth outcomes and development of appropriate programs and policies.

Adult↗
PubMed2001

Matched case-control data analysis with selection bias.

Case-control studies offer a rapid and efficient way to evaluate hypotheses. On the other hand, proper selection of the controls is challenging, and the potential for selection bias is a major weakness. Valid inferences about parameters of interest cannot be drawn if selection bias exists. Furthermore, the selection bias is difficult to evaluate. Even in situations where selection bias can be estimated, few methods are available. In the matched case-control Northern Manhattan Stroke Study (NOMASS), stroke-free controls are sampled in two stages. First, a telephone survey ascertains demographic and exposure status from a large random sample. Then, in an in-person interview, detailed information is collected for the selected controls to be used in a matched case-control study. The telephone survey data provides information about the selection probability and the potential selection bias. In this article, we propose bias-corrected estimators in a case-control study using a joint estimating equation approach. The proposed bias-corrected estimate and its standard error can be easily obtained by standard statistical software.

Bias↗
PubMed1998

Controlling for selection bias in the evaluation of Alcoholics Anonymous as aftercare treatment.

OBJECTIVE: The purpose of this research was to control for self-selection bias in the evaluation of Alcoholics Anonymous (AA) as aftercare treatment. Observational studies of alcoholism aftercare treatment are subject to selection bias whenever the self-selection process results in important differences in unobserved casemix dimensions between treatment groups. METHOD: The sample included 118 male veterans discharged from inpatient alcohol treatment, 85% of whom were followed-up at 3 months. Drinking outcomes were measured by self-reported abstinence in the third month after discharge. The aftercare treatment effect of AA was estimated using standard logistic regression analysis and instrumental variables analysis. Instruments included the subject's ability to drive oneself to AA meetings and the presence/absence of an AA meeting in the subject's town of residence. A Hausman exogeniety test was used to determine whether the standard logistic regression results were subject to self-selection bias. RESULTS: Estimates from the standard logistic regression yielded a positive (OR = 3.7) and significant (p = .018) treatment effect for AA attendance. However, the instrumental variables analysis yielded a smaller (OR* = 1.7) and insignificant treatment effect estimate (p = .782). The Hausman exogeniety test indicated that the treatment effect estimate from the standard logistic regression was subject to significant self-selection bias (chi2 = 83.9, 1 df, p <.01). CONCLUSIONS: The AA aftercare treatment effect observed in this sample was due to differences in unobserved casemix factors between the treatment groups. Results suggest that previous AA aftercare research may have also been subject to self-selection bias. Researchers of substance abuse outcomes should consider analyzing nonexperimental data using instrumental variables methodologies.

Adult↗
PubMedDate not supplied

Personal exposure assessment studies may suffer from exposure-relevant selection bias.

We evaluated exposure-relevant selection bias within the framework of a study on personal air pollution exposure, using traffic data as exposure proxy. Based on random samples of 3000 (Basel) and 2532 (Helsinki) persons, 50 and 250 subjects, respectively, were recruited for direct monitoring and 250 (Basel, Helsinki) for indirect monitoring. In Basel, participants of direct monitoring as compared to non-participants were more likely to live at streets with low traffic volume (49% below 1st quartile vs. 27%). Adjusted for sex, age and nationality, an increase of 100 cars per hour was associated with 14% less participation (odds ratio (OR): 0.861; 95% CI: 0.731, 1.007). Although in Helsinki, traffic volume was neither significantly related to participation in direct nor indirect monitoring, the point estimates indicate a tendency to decreased participation with increasing traffic intensity at home. We conclude that selection bias regarding exposure-relevant characteristics is likely to occur when recruiting participants for studies including demanding personal exposure assessment. Correction for factors routinely collected may not fully account for exposure-relevant bias. This is of particular importance when using exposure data for modelling population exposure distributions, whereas in epidemiological studies, a reduced range of exposure must not a priori distort the exposure-response relationship.

Adult↗
PubMed2004

Low all-cause mortality in the volunteer-based Rotterdam section of the European randomised study of screening for prostate cancer: self-selection bias?

OBJECTIVES: The likelihood of self-selection bias in a cohort of men that entered the volunteer-based randomised prostate cancer screening trial was assessed. In addition, the accuracy of the randomisation was evaluated by comparison amongst the trial arms. SETTING: The Rotterdam section of the European Randomised Study of Screening for Prostate Cancer (ERSPC). Participants were recruited from population registries and randomised after receipt of written informed consent (49% participation rate). METHODS: Data from 37,614 men who entered the Rotterdam trial between June 1994 and July 1999 were linked to the Causes of Death Registry of Statistics Netherlands. For the assessment of self-selection bias, mortality in the trial cohort and the area of Rotterdam was compared by calculating standardised mortality ratios (SMRs, 95% confidence interval [CI]). Comparisons between the screening and the control arm were made by means of relative risk (RR). RESULTS: There were a total of 1661 deaths (4.4%) in the trial cohort, during an average follow-up time of 2.8 years. The observed number of deaths in the two arms combined was lower than expected (SMR 86.7%, 95% CI 81.6-92.1), which was reflected in the major cause of death categories, especially in all malignant neoplasms. No difference was observed in the all-cause mortality rate among the screening and control arm, aged 55-69 years at entry (RR 1.00, 95% CI 0.89-1.12), nor when all causes, excluding prostate cancer mortality, were considered. CONCLUSIONS: The preliminary results show that the screening and the control arm were comparable, but men recruited for the trial seemed healthier than men in the target population.

Aged↗
PubMedDate not supplied

Affirmative actions: can the discriminant accuracy of a test be determined in the face of selection bias?

Clinical estimates of test efficacy can be distorted by the differential referral of positive and negative test responders for outcome verification. Accordingly, a series of computer simulations was performed to quantify the effects of various degrees of this selection bias on the observed true-positive rate, false-positive rate, and discriminant accuracy of a hypothetical test. The error in observed true- and false-positive rates was positive with respect to diagnosis, and negative with respect to prognosis. The magnitude of error was highly correlated with the magnitude of bias associated with the test response (primary selection bias), but not with the magnitude of bias associated with additional independent factors (secondary selection bias). Mathematical correction for preferential referral based on the test response using a previously published algorithm completely removed the correlation with primary selection bias for both diagnosis and prognosis. Although a significant correlation with secondary selection bias persisted at intermediate base rates, its magnitude was small. Discriminant accuracy was assessed in terms of area under a receiver operating characteristic (ROC) curve. Biased values of true- and false-positive rates were distributed along the curve defined by the actual true- and false-positive rates of the test for both diagnosis and prognosis. As a result, the areas under ROC curves calculated from biased true- and false-positive rates were within 2% of the areas calculated from the actual rates. Only when the primary and secondary observations were independent with respect to one outcome and dependent with respect to the other outcome did a systematic error appear in ROC area.(ABSTRACT TRUNCATED AT 250 WORDS)

Computer Simulation↗
PubMed2003

Selection bias in the assessment of gene-environment interaction in case-control studies.

Selection bias is a common concern in epidemiologic studies, particularly case-control studies. Selection bias in the odds ratio occurs when participation depends jointly on exposure and disease status. General results for understanding when selection bias may affect studies involving gene-environment interactions have not yet been developed. In this paper, the authors show that the assessment of gene-environment interactions will not be subject to selection bias under the assumption that genotype does not influence participation conditional on exposure and disease status. This is true even when selection, including self-selection of subjects, is jointly influenced by exposure and disease and regardless of whether the genotype is related to exposure, disease, or both. The authors present an example to illustrate this concept.

Bias↗
PubMed1998

The effects of sample selection bias on racial differences in child abuse reporting.

OBJECTIVE: The aim was to examine whether design features of Wave 1, 1980 National Incidence Study (NIS) data resulted in sample selection bias when certain victims of maltreatment were excluded. METHOD: Logistic regression models for the probability of child abuse reports to the child protective services (CPS) were estimated using maximum likelihood methods for Black (n = 511) and White (n = 2499) child abuse cases. The models were estimated with and without correction for selection bias using a two-step procedure proposed by Heckman. RESULTS: Substantial differences were found in the characteristics of Black and White victims by source of report and by type of maltreatment. Also found were sizeable differences within each racial group between sampled agencies and nonsampled agencies. Sample selection bias affected the estimation of both White and Black child abuse reporting rates. In the Black sample, however, the effect of sample selection bias was to reduce the statistical significance of the impacts of reporting agency and physical and sexual abuse on report rates. In the White sample, most significant factors in the basic model remained statistically significant with correction for selection bias. CONCLUSIONS: Selection bias was found to be caused by the exclusion of family, friends, and neighbors in the NIS sample design. Such exclusion has the effect of altering the interpretation of the determinants of child abuse reporting among Blacks, but not among Whites. Thus, conclusions about racial differences in child maltreatment must be reached cautiously, given the NIS study design.

Child↗
PubMed2004

Power, selection bias and predictive performance of the Population Pharmacokinetic Covariate Model.

Identification and quantification of covariate relations is often an important part of population pharmacokinetic/pharmacodynamic (PK/PD) modelling. The covariate model is regularly built in a stepwise manner. With such methods, selection bias may be a problem if only statistically significant covariates are accepted into the model. Competition between multiple covariates may further increase selection bias, especially when there is a moderate to high correlation between the covariates. This can also result in a loss of power to find the true covariates. The aim of this simulation study was to investigate the effect on power, selection bias and predictive performance of the covariate model, when altering study design and system-related quantities. Data sets with 20-1000 subjects were investigated. Five covariates were created by sampling from a multivariate standard normal distribution. The true covariate was set up to have no, low, moderate and high correlation to the other four covariates, respectively. Data sets, in which each individual had two or three PK observations, were simulated using a one-compartment i.v. bolus model. The true covariate influenced clearance according to one of several magnitudes. Different magnitudes of residual error and inter-individual variability in the structural model parameters were also introduced to the simulation model. A total of 7400 replicate data sets were simulated independently for each combination of the above conditions. Models with one of the five simulated covariates influencing clearance and the model without any covariate were fitted to the data. The probability of selecting (according to a pre-specified P-value) the different covariates, along with the estimated covariate coefficient, was recorded. The results show that selection bias is very high for small data sets (< or = 50 subjects) simulated with a weak covariate effect. If selected under these circumstances, the covariate coefficient is on average estimated to be more than twice its true value, making the covariate model useless for predictive purposes. Surprisingly, even though competition from false covariates caused substantial loss in the power of selecting the true covariate, the already high selection bias increased only marginally. This means that the bias due to competition is negligible if statistical significance is also required for covariate selection. Bias and predictive performance are direct functions of power, only indirectly affected by study design and system-related quantities. Mainly because of selection bias, low-powered covariates can be expected to harm the predictive performance when selected. For the same reason these low-powered covariates may falsely appear to be clinically relevant when selected. If the aim of an analysis is predictive modelling, we do not recommend stepwise selection or significance testing of covariates to be performed on small or moderately sized data sets (<50-100 subjects).

Computer Simulation↗
PubMed2006

Country development and manuscript selection bias: a review of published studies.

BACKGROUND: Manuscript selection bias is the selective publication of manuscripts based on study characteristics other than quality indicators. One reason may be a perceived editorial bias against the researches from less-developed world. We aimed to compare the methodological quality and statistical appeal of trials from countries with different development status and to determine their association with the journal impact factors and language of publication. METHODS SELECTION CRITERIA: Based on the World Bank income criteria countries were divided into four groups. All records of clinical trials conducted in each income group during 1993 and 2003 were included if they contained abstract and study sample size. DATA SOURCES: Cochrane Controlled Trials Register was searched and 50 articles selected from each income group using a systematic random sampling method in years 1993 and 2003 separately. DATA EXTRACTION: Data were extracted by two reviewers on the language of publication, use of randomization, blinding, intention to treat analysis, study sample size and statistical significance. Disagreement was dealt with by consensus. Journal impact factors were obtained from the institute for scientific information. RESULTS: Four hundred records were explored. Country income had an inverse linear association with the presence of randomization (chi2 for trend = 5.6, p = 0.02) and a direct association with the use of blinding (chi2 for trend = 6.9, p = 0.008); although in low income countries the probability of blinding was increased from 36% in 1993 to 46% in 2003. In 1993 the results of 68% of high income trials and 64.7% of other groups were statistically significant; but in 2003 they were 66% and 82% respectively. Study sample size and income were the only significant predictors of journal impact factor. CONCLUSION: The impact of country development on manuscript selection bias is considerable and may be increasing over time. It seems that one reason may be more stringent implementation of the guidelines for improving the reporting quality of trials on developing world researchers. Another reason may be the presumptions of the researchers from developing world about the editorial bias against their nationality.

Developing Countries↗
PubMed1994

Measuring second-order selection bias in a work site health program.

Magnitude and direction of second-order self-selection bias were assessed in a sample of 93,807 IBM employees who participated in the company's Voluntary Health Assessment (VHA) Program, by comparing repeat participants with one-time participants and by simulating selection into the repeat sample. One-time and repeat VHA participants differed systematically but not uniformly in several health characteristics. Repeat participants improved significantly in risk-relevant behaviors and health risk measures. Simulation of selection bias by excluding healthier or less healthy participants from the repeat VHA sample showed findings of gain to be robust. In studies of gain, second-order selection bias cannot automatically be assumed to inflate gain nor to be of sufficient magnitude to affect conclusions about program effects. Simulation is a useful tool for gauging direction and magnitude of selection bias.

Adult↗
PubMed1999

Detecting selection bias in randomized clinical trials.

Lack of concealment of allocation in randomized clinical trials can invite selection bias, which is the preferential enrollment of specific patients into one treatment group over another. For example, patients more likely to respond may be enrolled only when the next treatment to be assigned is known to be the active treatment, and patients less likely to respond may be enrolled only when the next treatment to be assigned is known to be the control. Despite the fact that selection bias can compromise both the internal and external validity of trials, little methodology has been developed for its detection. An investigator may test the success of the randomization by comparing baseline characteristics across treatment groups, but such test is limited by the potential inability of the measured baseline variables to predict response. A new method for detecting selections bias, based on response data only, is developed for the case in which a small block size, and either unmasking of treatment codes or an open-label design, have compromised the concealment of allocation. This new method complements baseline comparisons, and is sensitive to detect selection bias even in situations in which baseline comparisons are not.

Computer Simulation↗
PubMed2000

Studies on natural recovery from alcohol dependence: sample selection bias by media solicitation?

AIMS: To assess the selection bias of recruiting participants in studies on natural recovery from alcohol dependence through media solicitation. DESIGN: Two samples with different recruitment strategies are compared. SETTING: Media solicitation and general population. PARTICIPANTS: Sample 1 consists of 176 alcohol-dependent individuals remitted without formal help and recruited through media solicitation, sample 2 consists of 32 natural remitters derived from a representative general population study with a sample size of 4075 respondents and a response rate of 70.2%. MEASUREMENTS: Several triggering mechanisms and maintenance factors of remission were assessed in a personal interview using standardized questionnaires. FINDINGS: Results of logistic regression analyses show that media-solicited subjects were more often abstinent in the last 12 months, were more severely dependent, were less satisfied with eight life domains prior to remission and showed higher scores in a coping behaviour measure. Besides these major differences from the multivariate analysis, media subjects revealed more health problems prior to remission, experienced more social pressure to change drinking behaviour, and showed differences in reasons for not seeking help. CONCLUSIONS: Media solicitation leads to a sample selection bias in research on natural recovery from alcohol dependence. When measures to foster self-change are derived from such studies, findings from representative samples have to be considered.

Female↗
PubMed2006

Predictors of follow-up and assessment of selection bias from dropouts using inverse probability weighting in a cohort of university graduates.

Dropouts in cohort studies can introduce selection bias. In this paper, we aimed (i) to assess predictors of retention in a cohort study (the SUN Project) where participants are followed-up through biennial mailed questionnaires, and (ii) to evaluate whether differential follow-up introduced selection bias in rate ratio (RR) estimates. The SUN Study recruited 9907 participants from December 1999 to January 2002. Among them, 8647 (87%) participants answered the 2-year follow-up questionnaire. The presence of missing information in key variables at baseline, being younger, smoker, a marital status different of married, being obese/overweight and a history of motor vehicle injury were associated with being lost to follow-up, while a self-reported history of cardiovascular disease predicted a higher retention proportion. To assess whether differential follow-up affected RR estimates, we studied the association between body mass index and the risk of hypertension, using inverse probability weighting (IPW) to adjust for confounding and selection bias. Obese individuals had a higher crude rate of hypertension compared with normoweight participants (RR=6.4, 95% confidence interval (CI): 3.9-10.5). Adjustment for confounding using IPW attenuated the risk of hypertension associated to obesity (RR=2.4, 95% CI: 1.1-5.3). Additional adjustment for selection bias did not modify the estimations. In conclusion, we show that the follow-up through mailed questionnaires of a geographically disperse cohort in Spain is possible. Furthermore, we show that despite existing differences between retained or lost to follow-up participants this may not necessarily have an important impact on the RR estimates of hypertension associated to obesity.

Adult↗