Biomedical subjects

R D Gibbons

Publications and source records attributed to R D Gibbons.

At least 19 recordsLinked to original sources

PubMed1996

MIXOR: a computer program for mixed-effects ordinal regression analysis.

MIXOR provides maximum marginal likelihood estimates for mixed-effects ordinal probit, logistic, and complementary log-log regression models. These models can be used for analysis of dichotomous and ordinal outcomes from either a clustered or longitudinal design. For clustered data, the mixed-effects model assumes that data within clusters are dependent. The degree of dependency is jointly estimated with the usual model parameters, thus adjusting for dependence resulting from clustering of the data. Similarly, for longitudinal data, the mixed-effects approach can allow for individual-varying intercepts and slopes across time, and can estimate the degree to which these time-related effects vary in the population of individuals. MIXOR uses marginal maximum likelihood estimation, utilizing a Fisher-scoring solution. For the scoring solution, the Cholesky factor of the random-effects variance-covariance matrix is estimated, along with the effects of model covariates. Examples illustrating usage and features of MIXOR are provided.

Antipsychotic Agents↗

PubMed1996

MIXREG: a computer program for mixed-effects regression analysis with autocorrelated errors.

MIXREG is a program that provides estimates for a mixed-effects regression model (MRM) for normally-distributed response data including autocorrelated errors. This model can be used for analysis of unbalanced longitudinal data, where individuals may be measured at a different number of timepoints, or even at different timepoints. Autocorrelated errors of a general form or following an AR(1), MA(1), or ARMA(1,1) form are allowable. This model can also be used for analysis of clustered data, where the mixed-effects model assumes data within clusters are dependent. The degree of dependency is estimated jointly with estimates of the usual model parameters, thus adjusting for clustering. MIXREG uses maximum marginal likelihood estimation, utilizing both the EM algorithm and a Fisher-scoring solution. For the scoring solution, the covariance matrix of the random effects is expressed in its Gaussian decomposition, and the diagonal matrix reparameterized using the exponential transformation. Estimation of the individual random effects is accomplished using an empirical Bayes approach. Examples illustrating usage and features of MIXREG are provided.

Adolescent↗

PubMed1996

Prevalence rates and correlates of psychiatric disorders among preschool children.

OBJECTIVE: To determine the prevalence and correlates of psychiatric disorders among preschool children in a primary care pediatric sample. METHOD: In a two-stage design, 3,860 preschool children were screened; 510 received fuller evaluations. RESULTS: For quantitative assessment of disorder (> or = 90th percentile), prevalence of behavior problems was 8.3%. "Probable" occurrence of an Axis I DSM-III-R disorder was 21.4% (9.1%, severe). Logistic regression analyses indicated significant demographic correlates for quantitative outcomes (older age, minority status, male sex, low socio-economic status, father absence, small family size) but not for DSM-III-R diagnoses. Maternal and family characteristics were generally not significant. Child correlates included activity level, timidity, persistence, and IQ. CONCLUSIONS: Overall prevalence of disorder was consistent with rates for older children; correlates varied by approach used for classification.

Age Factors↗

PubMed1996

Length of inpatient stay and recidivism among patients with schizophrenia.

OBJECTIVE: The study examined whether length of hospital stay is related to recidivism among psychiatric patients. A quasi-experimental approach was used to address limitations of controlled and epidemiological research. METHODS: Three matched groups, each consisting of 55 inpatients with schizophrenia, were selected from public psychiatric units with different mean lengths of stay. Regression models were used to compare the groups on three variables: time to first readmission (survival analysis), number of readmissions (ordinal logit regression), and total time in the community in the postdischarge year (multiple linear regression). RESULTS: An analysis based on the units with different lengths of stay, which was similar to that typically used in controlled studies, found no differences in the three outcome measures. However, a second analysis that examined data for all patients irrespective of their unit assignment found that inpatients treated for 30 days or less relapsed sooner than those with stays longer than 30 days. The disparity in results was largely due to overlapping quasi-experimental conditions: many patients on the short-stay units had a long lengths of stay, and vice versa. The first analysis supports an administrative policy of short stays. The second reinforces previous findings that a group of patients, primarily young males with onset of illness at an early age and multiple previous hospitalizations, is at greater risk of relapse with short-term treatment. CONCLUSIONS: The apparent contradiction between a unit- or patient-based analysis suggests that unit-based results should be interpreted with caution when used to make clinical or utilization review decisions.

Adult↗

PubMed1996

High-dimensional multivariate probit analysis.

A computationally practical form of probit analysis for multiple response variables based on an assumed common factor model for the latent tolerances is proposed. Numerical integration over the factor space provides maximum likelihood estimation of the probit regression parameters and of the probabilities of response combinations under the model. The procedure is applied to five variables from the Pneumoconiosis Field Trial, two variables of which were previously analyzed by Ashford and Sowden (1970, Biometrics 26, 535-546).

Algorithms↗

PubMed1995

Initial severity and differential treatment outcome in the National Institute of Mental Health Treatment of Depression Collaborative Research Program.

Random regression models (RRMs) were used to investigate the role of initial severity in the outcome of 4 treatments (cognitive-behavior therapy [CBT], interpersonal psychotherapy [IPT], imipramine plus clinical management [IMI-CM], and placebo plus clinical management [PLA-CM]) for outpatients with major depressive disorder seen in the National Institute of Mental Health Treatment of Depression Collaborative Research Program. Initial severity of depression and impairment of functioning significantly predicted differential treatment effects. A larger number of differences than previously reported were found among the active treatments for the more severely ill patients; this was due, in large part, to the greater power of the present statistical analyses.

Activities of Daily Living↗

PubMed1995

Role of previous claims and specialty on the effectiveness of risk-management education for office-based physicians.

We analyzed the medical malpractice claims data of 1,903 physicians between 1981 and 1990 to assess the efficacy--a reduced incidence of future claims and decreased payout in the event of a claim--of risk-management education for office-based physicians. Physicians were participants in the Oregon Medical Association's medical liability program and represented all recognized specialties and all geographic areas of the state. Each physician's claim and payout history before and after 4 sequential risk-management education programs was entered into a random-effects probit model that allowed for a longitudinal rather than a cross-sectional analysis. For most physicians, there was increased claim vulnerability following 1 or 2 risk-management education courses but decreased vulnerability after additional courses. Among all physicians, having a previous claim substantially increased the risk for a future claim. Risk for an additional claim doubled (from 7% to 14%) for physicians who had a claim in the previous year. Of all specialists who have had claims, anesthesiologists (reduction in claims incidence from 18.8% to 9.1% and in payout from 14.6% to 5%) and obstetrician-gynecologists (reduction in claims incidence from 23.3% to 15.2% and in payout from 11.6% to 4.2%) benefit most from cumulative risk-management education.

Humans↗

PubMed1994

Application of random-effects probit regression models.

A random-effects probit model is developed for the case in which the outcome of interest is a series of correlated binary responses. These responses can be obtained as the product of a longitudinal response process where an individual is repeatedly classified on a binary outcome variable (e.g., sick or well on occasion t), or in "multilevel" or "clustered" problems in which individuals within groups (e.g., firms, classes, families, or clinics) are considered to share characteristics that produce similar responses. Both examples produce potentially correlated binary responses and modeling these person- or cluster-specific effects is required. The general model permits analysis at both the level of the individual and cluster and at the level at which experimental manipulations are applied (e.g., treatment group). The model provides maximum likelihood estimates for time-varying and time-invariant covariates in the longitudinal case and covariates which vary at the level of the individual and at the cluster level for multilevel problems. A similar number of individuals within clusters or number of measurement occasions within individuals is not required. Empirical Bayesian estimates of person-specific trends or cluster-specific effects are provided. Models are illustrated with data from mental health research.

Cluster Analysis↗

PubMed1994

Random-effects regression models for clustered data with an example from smoking prevention research.

A random-effects regression model is proposed for analysis of clustered data. Unlike ordinary regression analysis of clustered data, random-effects regression models do not assume that each observation is independent but do assume that data within clusters are dependent to some degree. The degree of this dependency is estimated along with estimates of the usual model parameters, thus adjusting these effects for the dependency resulting from the clustering of the data. A maximum marginal likelihood solution is described, and available statistical software for the model is discussed. An analysis of a dataset in which students are clustered within classrooms and schools is used to illustrate features of random-effects regression analysis, relative to both individual-level analysis that ignores the clustering of the data, and classroom-level analysis that aggregates the individual data.

Attitude to Health↗

PubMed1994

A computerized image analysis system for estimating Tanner-Whitehouse 2 bone age.

A method for assigning Tanner-Whitehouse 2 skeletal maturity scores (or bone ages) to hand-wrist X-rays by an image analysis computer system is described. An operator positions the relevant area of the X-ray on a light box beneath a video camera. Correct positioning is assured by computer templates of each bone stage. Thereafter the process is automatic; the computer, not the operator, rates the bones. The system produces continuous stage scores, not discrete ones such as B, C or D. Data are given which show that the computer-assisted skeletal age score is more repeatable than the usual manual (or unassisted) rating. The absolute difference between duplicates averaged 0.25 stages; differences of as much as 1.0 stage occurred in only 3% of duplicates compared with 15% obtained in manual ratings.

Age Determination by Skeleton↗

PubMed1994

A random-effects ordinal regression model for multilevel analysis.

A random-effects ordinal regression model is proposed for analysis of clustered or longitudinal ordinal response data. This model is developed for both the probit and logistic response functions. The threshold concept is used, in which it is assumed that the observed ordered category is determined by the value of a latent unobservable continuous response that follows a linear regression model incorporating random effects. A maximum marginal likelihood (MML) solution is described using Gauss-Hermite quadrature to numerically integrate over the distribution of random effects. An analysis of a dataset where students are clustered or nested within classrooms is used to illustrate features of random-effects analysis of clustered ordinal data, while an analysis of a longitudinal dataset where psychiatric patients are repeatedly rated as to their severity is used to illustrate features of the random-effects approach for longitudinal ordinal data.

Adolescent↗

PubMed1993

Some conceptual and statistical issues in analysis of longitudinal psychiatric data. Application to the NIMH treatment of Depression Collaborative Research Program dataset.

Longitudinal studies have a prominent role in psychiatric research; however, statistical methods for analyzing these data are rarely commensurate with the effort involved in their acquisition. Frequently the majority of data are discarded and a simple end-point analysis is performed. In other cases, so called repeated-measures analysis of variance procedures are used with little regard to their restrictive and often unrealistic assumptions and the effect of missing data on the statistical properties of their estimates. We explored the unique features of longitudinal psychiatric data from both statistical and conceptual perspectives. We used a family of statistical models termed random regression models that provide a more realistic approach to analysis of longitudinal psychiatric data. Random regression models provide solutions to commonly observed problems of missing data, serial correlation, time-varying covariates, and irregular measurement occasions, and they accommodate systematic person-specific deviations from the average time trend. Properties of these models were compared with traditional approaches at a conceptual level. The approach was then illustrated in a new analysis of the National Institute of Mental Health Treatment of Depression Collaborative Research Program dataset, which investigated two forms of psychotherapy, pharmacotherapy with clinical management, and a placebo with clinical management control. Results indicated that both person-specific effects and serial correlation play major roles in the longitudinal psychiatric response process. Ignoring either of these effects produces misleading estimates of uncertainty that form the basis of statistical tests of hypotheses.

Analysis of Variance↗

PubMed1993

Length of stay and recidivism in schizophrenia: a study of public psychiatric hospital patients.

OBJECTIVE: Psychiatric beds in public hospitals have decreased 80% since 1955, but admissions have risen correspondingly, largely because of high recidivism rates. Decreases in numbers of beds have been partly achieved by shortening the length of stay, which lessened by half between 1970 and 1980. This study was undertaken to determine whether duration of hospital treatment affects the rate and rapidity of relapse among schizophrenic patients. METHOD: Data on 1,500 patients from 10 state hospitals were gathered for 18 months after initial discharge. Predictor variables included age, sex, marital status, race, number of previous admissions, location of the facility, and length of stay. Data were analyzed by survival analysis with a Cox regression model for two times to initial relapse: 30 days and 18 months (outcome). RESULTS: Length of stay was significantly related to each time to relapse after the effects of number of previous admissions and age were partialed out. Facility location was not predictive, but intrahospital effects were tested by examining the data on the largest facility; again, length of stay significantly predicted relapse. CONCLUSIONS: Although the magnitude of the effect was small, the clinical significance of the findings is the greater likelihood that brief-stay patients will be rehospitalized within 30 days after discharge than will patients treated for longer periods. Brief hospitalization seems generally applicable to psychiatric populations, but there may be a small but important group of seriously mentally ill patients for whom other alternatives are possibly more appropriate and should be explored.

Adult↗

PubMed1993

Assessing the severity of depressive states in recently detoxified alcoholics.

The problem of how to assess "unbiased" symptoms of depression meaningfully in the context of alcoholism bedevils studies of concomitant alcoholism and affective disorder. This study summarizes this controversy and discusses its implications for developing improved measures of depression severity among alcoholics. The Beck Depression Inventory (BDI) responses of 130 alcoholic applicants applying for inpatient care were evaluated using a two-parameter normal item response model. This study demonstrates that a single dimension of depression severity accounts for patient responses well, but that seven BDI items were relatively poor markers of syndrome severity for these alcoholics. The study documents growing consensus among investigators as to which BDI items constitute a fair scale of depression severity among alcoholic patients The availability of "unbiased criteria" for assessing the severity of depression among alcoholics applying for inpatient treatment will enable investigators and clinicians to recognize patients with concomitant alcoholism and affective disorder for special attention and/or treatment. The modified BDI scale proposed here is a candidate for a clinical definition of depression severity among alcoholic patients by virtue of this study's demonstration that it defines a distinct syndrome in this patient sample; but the modified scale must undergo independent tests of validity before its clinical utility can be established.

Adult↗

PubMed1992

An application of item response theory to alexithymia assessment among abstinent alcoholics.

An item response theory (IRT) model identified three dimensions assessed by the Toronto Alexithymia Scale (TAS) in a sample of 130 male applicants for inpatient care at a Veterans Administration (VA) medical center alcoholism treatment program. A unidimensional solution did not capture all of alexithymia's theoretical features. Subjects with lower alexithymia scores gave positive responses to items tapping emotional awareness deficits; only those with higher alexithymia scores gave positive responses to items tapping external, operative cognitive style. Thus, a total TAS score may not represent alexithymia accurately in substance-abusing patient populations.

Adult↗

PubMed1992

Predicting risk for medical malpractice claims using quality-of-care characteristics.

The current fault-based tort system assumes that claims made against physicians are inversely related to the quality of care they provide. In this study we identified physician characteristics associated with elements of medical care that make physicians vulnerable to malpractice claims. A sample of physicians (n = 248) thought to be at high or low risk for claims was surveyed on various personal and professional characteristics. Statistical analysis showed that 9 characteristics predicted risk group. High risk was associated with increased age, surgical specialty, emergency department coverage, increased days away from practice, and the feeling that the litigation climate was "unfair." Low risk was associated with scheduling enough time to talk with patients, answering patients' telephone calls directly, feeling "satisfied" with practice arrangements, and acknowledging greater emotional distress. Prediction was more accurate for physicians in practice 15 years or less. We conclude that a relationship exists between a history of malpractice claims and selected physician characteristics.

Clinical Medicine↗

PubMed1992

The efficacy of psychotropic drugs: implications for power analysis.

The estimation of the correct sample size to successfully test a hypothesis has become critical. A common approach to this problem is for the investigating team to complete a pilot study of a few patients to establish the "active drug-placebo" difference, using this "effect size" to perform the power analysis for sample size estimation. Given the variability evident in the effect size from completed and published studies, the pilot study approach may not be entirely dependable. The authors propose a method to obtain this initial "active drug-placebo" difference, in the field of psychotropic drug research. They apply meta-analysis to statistically summarize effect sizes obtained from an exhaustive review of the literature for a specific psychotropic drug in a given clinical condition. All double-blind, random assignment studies are used to calculate the effect size; therefore, no selection bias exists. These literature-based effect sizes are then used to perform the traditional power analysis for sample size estimation. The authors propose these estimations as a convenient reference source for future clinical investigators.

Humans↗

PubMed1991

Lack of a bimodal distribution of ventricular size in schizophrenia: a Gaussian mixture analysis of 1056 cases and controls.

The finding of clinical and laboratory differences between schizophrenic patients with large and small cerebral ventricles has led to the widespread assumption that large ventricles are a marker that characterizes a subgroup of patients with schizophrenia. We reviewed all published English language ventricle-to-brain ratio (VBR) studies in which individual data points were available (schizophrenics: n = 691, medical controls; n = 205, normal volunteers: n = 160). Using a univariate normal mixture model to examine the distribution of ventricular size in each group, we found no evidence of a mixture of Gaussian distributions (i.e., "bimodality") within any of the three groups. The same analysis was then performed on the combined sample of schizophrenic patients and normal and medical controls, respectively. In each case the improvement in fit of a mixture of normal distributions compared to a single component normal distribution was significant. The data do not support the notion that ventricular enlargement is a discontinuous marker of a subtype of schizophrenia.

Analysis of Variance↗