Mark P. Becker

Severe Periodontitis and Risk for Poor Glycemic Control in Patients with Non-Insulin-Dependent Diabetes Mellitus

This study tested the hypothesis that severe periodontitis in persons with non-insulin-dependent diabetes mellitus (NIDDM) increases the risk of poor glycemic control. Data from the longitudinal study of residents of the Gila River Indian Community were analyzed for dentate subjects aged 18 to 67, comprising all those: 1) diagnosed at baseline with NIDDM (at least 200 mg/dL plasma glucose after a 2-hour oral glucose tolerance test); 2) with baseline glycosylated hemoglobin (HbA1 ) less than 9%; and 3) who remained dentate during the 2-year follow-up period. Medical and dental examinations were conducted at 2-year intervals. Severe periodontitis was specified two ways for separate analyses: 1) as baseline periodontal attachment loss of 6 mm or more on at least one index tooth; and 2) baseline radiographic bone loss of 50% or more on at least one tooth. Clinical data for loss of periodontal attachment were available for 80 subjects who had at least one follow-up examination, 9 of whom had two follow-up examinations at 2-year intervals after baseline. Radiographic bone loss data were available for 88 subjects who had at least one follow-up examination, 17 of whom had two follow-up examinations. Poor glycemic control was specified as the presence of HbA1 of 9% or more at follow-up. To increase the sample size, observations from baseline to second examination and from second to third examinations were combined. To control for non-independence of observations, generalized estimating equations (GEE) were used for regression modeling. Severe periodontitis at baseline was associated with increased risk of poor glycemic control at follow-up. Other statistically significant covariates in the GEE models were: 1) baseline age; 2) level of glycemic control at baseline; 3) having more severe NIDDM at baseline; 4) duration of NIDDM; and 5) smoking at baseline. These results support considering severe periodontitis as a risk factor for poor glycemic control and suggest that physicians treating patients with NIDDM should be alert to the signs of severe periodontitis in managing NIDDM. J Periodontol 1996;67:1085-1093.

Latent Variable Modeling of Diagnostic Accuracy

Latent class analysis has been applied in medical research to assessing the sensitivity and specificity of diagnostic tests/diagnosticians. In these applications, a dichotomous latent variable corresponding to the unobserved true disease status of the patients is assumed. Associations among multiple diagnostic tests are attributed to the unobserved heterogeneity induced by the latent variable, and inferences for the sensitivities and specificities of the diagnostic tests are made possible even though the true disease status is unknown. However, a shortcoming of this approach to analyses of diagnostic tests is that the standard assumption of conditional independence among the diagnostic tests given a latent class is contraindicated by the data in some applications. In the present paper, models incorporating dependence among the diagnostic tests given a latent class are proposed. The models are parameterized so that the sensitivities and specificities of the diagnostic tests are simple functions of model parameters, and the usual latent class model obtains as a special case. Marginal models are used to account for the dependencies within each latent class. An accelerated EM gradient algorithm is demonstrated to obtain maximum likelihood estimates of the parameters of interest, as well as estimates of the precision of the estimates.

Diabetes and Pregnancy: Factors associated with seeking pre-conception care

OBJECTIVE To define sociodemographic characteristics, medical factors, knowledge, attitudes, and health-related behaviors that distinguish women with established diabetes who seek pre-conception care from those who seek care only after conception. RESEARCH DESIGN AND METHODS A multicenter, case-control study of women with established diabetes making their first pre-conception visit (n = 57) or first prenatal visit without having received pre-conception care (n = 97). RESULTS Pre-conception subjects were significantly more likely to be married (93 vs. 51%), living with their partners (93 vs. 60%), and employed (78 vs. 41%); to have higher levels of education (73% beyond high school vs. 41%) and income (86% >

Analysis of Sets of Two-Way Contingency Tables Using Association Models

20,000 vs. 60%); and to have insulin-dependent diabetes mellims (IDDM) (93 vs. 81%). Pre-conception subjects with IDDM were more likely to have discussed preconception care with their health care providers (98 vs. 51%) and to have been encouraged to get it (77 vs. 43%). In the prenatal group, only 24% of pregnancies were planned. Pre-conception patients were more knowledgeable about diabetes, perceived greater benefits of pre-conception care, and received more instrumental support. CONCLUSIONS Only about one-third of women with established diabetes receive pre-conception care. Interventions must address prevention of unintended pregnancy. Providers must regard every visit with a diabetic woman as a pre-conception visit. Contraception must be explicitly discussed, and pregnancies should be planned. In counseling, the benefits of pre-conception care should be stressed and the support of families and friends should be elicited.

The Role of Organizational Infrastructure in Implementation of Hospitals' Quality Improvement

Abstract A class of models is introduced for the analysis of group differences in the association between two discrete variables. The RC(M) association model for two-way tables is reviewed, and alternative weighting systems for identifying interaction parameters are presented. This model is generalized for the setting where a two-way contingency table is available for two or more groups. Various restricted models can be used to examine possible sources of intergroup heterogeneity in the association. These sources pertain to heterogeneity in the intrinsic association and/or in the scores for the row and column variables. The importance of weights used to identify the row and column scores is emphasized. A classical set of data previously analyzed by many authors is used to illustrate the advantages of the models and methods developed here.

EM algorithms without missing data

Abstract. Quality improvement (QI) is an organized approach to planning and implementing continuous improvement in performance. Although QI holds promise for improving quality of care and patient safety, hospitals that adopt QI often struggle with its implementation. This article examines the role of organizational infrastructure in implementation of quality improvement practices and structures in hospitals. The authors focus specifically on four elements of hospital support and infrastructure for QI—integrated data systems, financial support for QI, clinical integration, and information system capability. These macrolevel factors provide consistent, ongoing support for the QI efforts of clinical teams engaging in direct patient care, thus promoting institutionalization of QI. Results from the multivariate analysis of 1997 survey data on 2,350 hospitals provide strong support for the hypotheses. Results signal that organizations intent upon improving quality must attend to the context in which QI efforts are practiced, and that such efforts are unlikely to be effective unless appropriate support systems are in place to ensure full implementation.

Rural motor vehicle crash mortality: The role of crash severity and medical resources

Most problems in computational statistics involve optimization of an objective function such as a loglikelihood, a sum of squares, or a log posterior function. The EM algorithm is one of the most effective algorithms for maximization because it iteratively transfers maximization from a complex function to a simple, surrogate function. This theoretical perspective clarifies the operation of the EM algorithm and suggests novel generalizations. Besides simplifying maximization, optimization transfer usually leads to highly stable algorithms with well-understood local and global convergence properties. Although convergence can be excruciatingly slow, various devices exist for accelerating it. Beginning with the EM algorithm, we review in this paper several optimization transfer algorithms of substantial utility in medical statistics.

Reliability of nerve conduction studies among active workers.

We did a retrospective case control study to examine the relationship between the risk of dying for Michigan motor vehicle crash (MVC) drivers and the type of county (rural/nonrural) of crash occurrence, while adjusting for crash characteristics, age, sex, and the medical resources in the county of crash occurrence. The 1987 Michigan Accident Census was used to obtain data regarding all MVC driver nonsurvivors (733) and a random sample of all surviving drivers (2,483). County of crash occurrence was defined as rural or nonrural. The crash characteristics analyzed were vehicle deformity, seat belt use, and drivability of the vehicle from the scene. Age and sex of the driver were also analyzed. Medical resource characteristics for the county of crash occurrence were measured as the number of resources per square mile for each of the following: ambulances, emergency medical technicians (EMT), acute care hospital beds, and operating rooms, surgeons and emergency physicians. Also considered were the number and level of emergency rooms in the county of crash occurrence along with the maximum level of prehospital care available (basic life support versus advanced life support) in a county. Before adjusting, the relative risk (RR) for rural MVC drivers dying, compared to their nonrural counterparts, was 1.96. Adjustment for crash characteristics, age, and sex (using logistic regression) decreased the RR to 1.51. An attempt to add medical resource variables to the model resulted in high correlation with the rural/nonrural variable, as well as with each other. This multi-collinearity prevented us from providing a simple explanation of the role of medical resource variables as predictors of survival.(ABSTRACT TRUNCATED AT 250 WORDS)

Marginal modeling of binary cross-over data.

Nerve conduction studies play an important role in clinical practice and research. Given their widespread use, reliability of tests merits careful attention. We assessed interexaminer and intraexaminer reliability of median and ulnar sensory nerve measures of amplitude, onset latency, and peak latency. In a two‐phase cross‐sectional study, two examiners tested 158 workers. Reliability was assessed with intraclass correlations (ICC) and kappa statistics. Median nerve measures were more reliable (ICC range, 0.76 to 0.92) than ulnar measures (ICC range, 0.22 to 0.85). Ulnar‐onset latencies had the worst reliability. The median‐ulnar peak latency difference was a particularly stable measure (ICC range, 0.79 to 0.92). The median‐ulnar peak latency difference had high interexaminer reliability (κ range, 0.71 to 0.79) for normal tests defined by cut points of 0.8 ms and 0.5 ms. Intraexaminer reliability was higher with the 0.8‐ms cut point (κ = 0.90 and κ = 0.85 for examiners 1 and 2, respectively). Rather than absolute cut points to describe normality, a more rational interpretation of results can be made with ordered categories or continuous measures.

Models for the Analysis of Association in Multivariate Contingency Tables

A model specified in terms of linear models for marginal logits and linear models for log-odds ratios is proposed for the analysis of two-period binary cross-over experiments. Hypothesis testing and parameter estimation are facilitated by standard likelihood methodology. Two examples are used to illustrate how the model can be used to analyze two-period binary cross-over experiments. Results from a simulation study demonstrate that this approach to the analysis of binary cross-over data compares favorably with standard procedures, such as the Mainland-Gart test for a treatment difference, Prescotts test for a treatment difference, and the Hills-Armitage test for treatment-by-period interaction.