Hosmer-Lemeshow Goodness Of Fit Statistic Research Articles

Use of Caregiving Support Services Among Diverse Dementia Caregivers by Geographic Context.

Little is known about how race/ethnicity and geographic context relate to support service use among dementia caregivers. Our objectives were to investigate (a) whether the use of at least one formal caregiving service-support groups, respite care, and training-differed by race/ethnicity and across metro and nonmetro areas; and (b) whether predisposing, enabling, and need characteristics influenced support service use by race/ethnicity. Data were analyzed from a sample of primary caregivers of care recipients aged 65 years or older with probable dementia (n = 482) in the 2017 National Health and Aging Trends Study and National Study of Caregiving. We calculated weighted prevalence estimates and then used the Hosmer-Lemeshow goodness of fit statistic to find the best-fitting logistic regression models. Among minority dementia caregivers, support service use was higher in metro than nonmetro areas (35% and 15%); the trend was reversed for non-Hispanic White caregivers (47% nonmetro and 29% metro). The best-fitting regression models included predisposing, enabling, and need factors for both minority and non-Hispanic White caregivers. Younger age and more disagreement within the family were consistently associated with more service use in both groups. Among minority caregivers, better caregiver and care recipient health were associated with using support services. Among non-Hispanic White caregivers, nonmetro geographic context and caregiving interfering with valued activities were associated with using support services. Geographic context differently affected support service use and the influence of predisposing, enabling, and need factors varied by race/ethnicity.

Abstract 102: Impact of Using Older Data on the Accuracy of Cardiovascular Risk Scores

Background: Recent scores to predict atherosclerotic cardiovascular disease, ASCVD, have been found inaccurate, with some concern that risk scores become inaccurate with time, as changing demographics may also change ASCVD risk. A partial solution to the timeliness problem could be creating risk scores using electronic health records (EHR). EHR-based risk scores can be easily updated and tested for changes over time. They can also use more variables than traditional risk scores. This could capture clinical change that would otherwise be missed. We hypothesized that ASCVD risk scores change over time, but this could be minimized with more robust risk scores. To test this, we looked at change of two EHR-based risk score over three follow-up periods and using different statistical techniques to design the risk scores. Methods: Data sources: VA national data linked to Medicare and the National Death Index. Population: 3 overlapping cohorts from 2002, 2006, and 2009. Each consisted of all active VA patients aged 45-80 who had no documented history of CVD, clinical heart failure or loop diuretic use at baseline. Prediction models 1. VARS-ASCVD: uses the same variables as traditional risk scores, but all variables were re-calibrated to our population. 2. VARS-EHR: Uses 41 predictor variables and more interaction effects. Outcome variables: First occurrence of fatal or nonfatal ASCVD during 5 years of follow-up. Analysis: We looked at the accuracy of risk scores developed in 2002 on patients in 2006 and 2009. The discrimination of the risk scores (the ability to distinguish between those who do and do not develop an event), was evaluated with C-statistisic. The calibration (how closely the predicted probabilities reflect true risk) was evaluated with the Hosmer-Lemeshow Goodness of Fit statistic (GoF). Results: Each cohort had at least 1.4 million participants. Between the 3 cohorts the rate of diabetes mellitus increased from 201% to 27% and statin use increased from 25% to 45% of the population. The VARS-ASCVD risk scores for men developed in 2002 had the same discrimination of 0.67 in 2006 and 2009, but in women fell from 0.77 to 0.72 then increased to 0.74. The goodness of fit worsened. Using the VARS-EHR model, discrimination stayed similar in men and women. The GOF worsened, but by substantially less. Conclusions: ASCVD risk prediction tools become poorly calibrated over fairly short time periods. For effective use, they must be updated regularly.

A population-based, cross-sectional study of the prevalence and correlates of sedentary behaviour of adults with intellectual disabilities.

High levels of sedentary behaviour have a negative impact on health and well-being. There is limited evidence on the prevalence and correlates of sedentary behaviour of adults with intellectual disabilities (ID). A population-based sample of adults with ID were invited to take part in a comprehensive health check programme. Demographic and health data were collected during a structured interview and physical examination. Screen time was used as a proxy measure of sedentary behaviour. Bivariate and multivariate statistical modelling examined correlates of screen time. Fifty per cent of the 725 participants reported four or more hours of screen time per day. Male gender, higher levels of intellectual ability, mobility problems, obesity, not having hearing impairment and not having epilepsy were all significantly associated with higher screen time in the final multivariate model (R2 =0.16; Hosmer-Lemeshow goodness of fit statistic P=0.36). This is the first study to publish population-based data on the prevalence and correlates of sedentary behaviour in adults with ID. Compared with adults who do not have ID, adults with ID have higher levels, and different correlates, of sedentary behaviour. A better understanding of the social context of sedentary behaviour will inform the design of effective behaviour change programmes for adults with ID.

The Next Generation of Clinical Decision Making Tools: Development of a Real-Time Prediction Tool for Outcome of Prostate Biopsy in Response to a Continuously Evolving Prostate Cancer Landscape

We evaluate whether annual updating of the PCPT Risk Calculator would improve institutional validation compared to static use of the PCPT Risk Calculator alone. Data from 5 international cohorts including SABOR, Cleveland Clinic, ProtecT, Tyrol and Durham VA, comprising 18,400 biopsies, were used to evaluate an institution specific annual recalibration of the PCPT Risk Calculator. Using all prior years as a training set and the current year as the test set, annual recalibrations of the PCPT Risk Calculator were compared to static use of the PCPT Risk Calculator in terms of AUC and the Hosmer-Lemeshow goodness of fit statistic. For predicting high grade disease the median AUC (higher is better) of the recalibrated PCPT Risk Calculator (static PCPT Risk Calculator) across all test years for the 5 cohorts was 67.3 (67.5), 65.0 (60.4), 73.4 (73.4), 73.9 (74.1) and 69.6 (67.2), respectively, and median Hosmer-Lemeshow goodness of fit statistics indicated better fit for recalibration compared to the static PCPT Risk Calculator for Cleveland Clinic, ProtecT and the Durham VA but not for SABOR and Tyrol. For predicting overall cancer median AUC was 63.5 (62.7), 61.0 (57.3), 62.1 (62.5), 66.9 (67.3) and 68.5 (65.5), respectively, and median Hosmer-Lemeshow goodness of fit statistics indicated a better fit for recalibration in all cohorts except for Tyrol. A simple method has been provided to tailor the PCPT Risk Calculator to individual hospitals to optimize its accuracy for the patient population at hand.

Evaluation of IScore validity in a Greek cohort of patients with type 2 diabetes

BackgroundDiabetes constitutes a risk factor for stroke that also aggravates stroke prognosis. Several prognostic models have been developed for the evaluation of neurologic status, severity, short-term functional outcome and mortality of stroke patients. IScore is a novel tool recently developed in order to predict mortality rates within 30 days and 1 year after ischemic stroke and diabetes is not included in the scoring scale of IScore. The aim of the present study was to evaluate and compare IScore validity in ischemic stroke patients with and without diabetes.MethodsThis prospective study included 312 consecutive Caucasian patients with type 2 diabetes and 222 Caucasian patients without diabetes admitted for ischemic stroke in a tertiary Greek hospital. Thirty-day and 1-year IScores were individually calculated for each patient and actual mortality was monitored at the same time intervals. IScore’s predictive ability and calibration was evaluated and compared for ischemic stroke patients with and without diabetes. The performance of IScore for predicting 30 and 1-year mortality between patients with and without diabetes was assessed by determining the calibration and discrimination of the score. The area under the receiver operating characteristic curve was used to evaluate the discriminative ability of IScore for patients with and without diabetes, whereas the calibration of IScore was assessed by the Hosmer–Lemeshow goodness-of fit statistic.ResultsBaseline population characteristics and mortality rates did not differ significantly for both cohorts. IScore values were significantly higher for patients with diabetes at 30 days and 1 year after ischemic stroke and patients with diabetes presented more frequently with lacunar strokes. Based on ROC curves analysis IScore’s predictive ability for 30 day mortality was excellent, without statistically significant difference, for both cohorts. Predictive ability for 1 year mortality was also excellent for both groups with significantly better ability for patients with diabetes especially at high score values. Calibration of the model was good for both groups of patients.ConclusionsIScore accurately predicts mortality in acute ischemic stroke Caucasian patients with and without diabetes with higher efficacy in predicting 1 year mortality in patients with diabetes especially with high scores.

Predictive Model for Earlier Diagnosis of Suspected Age-Related Macular Degeneration Patients

The primary goal of tailored medicine is to presymptomatically identify individuals at high risk for disease using information of each individual's genetic profile and collection of environmental risk factors. Recently, algorithms were given the strong recognition of several replicated risk factors for age-related macular degeneration (AMD), this distant goal is beginning to seem less mysterious. The purpose of the study was to develop a statistical model for AMD. This study includes total 106 subjects. To identify the risk of earlier diagnosis of suspected AMD patients, 22 independent variables were included in the study. Forward stepwise (likelihood ratio) binary logistic regression has been used to find significant variables associated with the risk of AMD. Prediction equation, based on significant risk factors, and model authenticity have been developed. Hosmer-Lemeshow goodness of fit statistic (χ(2)=0.143, df=8, p=1.0), which is nonsignificant, indicates the appropriateness of the logistic regression model to predict AMD. After going through stepwise logistic regression, only 6 variables out of the 22 independent variables, namely, serum complement factor H (CFH), serum chemokine (C-C motif) ligand 2 (CCL2), serum superoxide dismutase 1 (SOD1), polymorphism in CCL2 (rs4586), stress, and comorbidity were found to be significant (p<0.05). The binary logistic regression model is an appropriate tool to predict AMD in the presence of serum CFH, serum CCL2, serum SOD1, polymorphism in CCL2 (rs4586), stress, and comorbidity with high specificity and sensitivity. The area under the receiver operating characteristic curve (0.909, p=0.001) with less standard error of 0.034 and close 95% confidence intervals (0.842-0.976) further validates the model.

PRM190 - Extension of the hosmer-lemeshow goodness of fit statistic to linear models with repeated measurements

PRM190 EXTENSION OF THE HOSMER-LEMESHOW GOODNESS OF FIT STATISTIC TO LINEAR MODELS WITH REPEATED MEASUREMENTS Li R1, Su Z2, Mendelsohn A2, Gemmen E1 1Quintiles Outcome, Rockville, MD, USA, 2Quintiles Outcome, Cambridge, MA, USA OBJECTIVES: Studies of health outcomes commonly involve binary measures, assessed multiple times. Although generalized linear mixed (GLIMMIX) models are well suited for analyzing these data, there does not exist a formal statistic to assess the goodness of fit (GOF) for GLIMMIX models. We developed an extension of the Hosmer-Lemeshow GOF test used for logistic regression that can be applied to GLIMMIX models. METHODS: The correlation among repeated measurements of the binary outcome variable was accounted for by a random effect in the GLIMMIX model. The principles of Hosmer-Lemeshow method were followed. The linear unbiased estimate of dependent variables were transformed to the original probability, sorted from least to largest, and divided into deciles. A Chi-square statistic and corresponding p-value with eight degrees of freedom, was calculated based upon the expected and observed numbers among deciles. The proposed GOF test was validated by a simulation study with 1000 runs generated from logistic regression models with and without random effects. The results were compared with the Hosmer-Lemeshow GOF test in situations where the latter is appropriate. The proposed method was used in the analysis of a comparative effectiveness (CE) study of ophthalmologic treatments for openangle glaucoma patients. RESULTS: When there was no random effect, the proposed GOF test results from the GLIMMIX procedure were almost identical to those of Hosmer-Lemeshow GOF test from the logistic procedure. With a random effect built in a correctly specified model, the goodness of fit rejection rate was 5.1%, which is close to the nominal level 5%. The proposed test did not indicate lack of fit for the regression models in the CE study. CONCLUSIONS: The proposed GOF test provides an assessment of model fit for models with binary outcomes and repeated measurements for predictor variables.

Superoxide Dismutase1 Levels in North Indian Population with Age-Related Macular Degeneration

Aim. The aim of the study was to estimate the levels of superoxide dismutase1 (SOD1) in patients of age-related macular degeneration (AMD) and examine the role of oxidative stress, smoking, hypertension, and other factors involved in the pathogenesis of AMD. Methods. 115 AMD patients and 61 healthy controls were recruited for this study. Serum SOD1 levels were determined by ELISA and were correlated to various risk factors. Logistic regression model of authenticity, by considering SOD1 as independent variable, has been developed along with ROC curve. Results. The SOD1 levels were significantly higher in AMD patients as compared to those of the controls. The difference was not significant for wet and dry AMD. However, the difference was significant between wet AMD subtypes. Nonsignificance of the Hosmer-Lemeshow goodness of fit statistic (χ 2 = 10.516, df = 8, P = 0.231) indicates the appropriateness of logistic regression model to predict AMD. Conclusion. Oxidative stress in AMD patients may mount compensatory response resulting in increased levels of SOD1 in AMD patients. To predict the risk of AMD on the basis of SOD1, a logistic regression model shows authenticity of 78%, and area under the ROC curve (0.827, P = .0001) with less standard error of 0.033 coupled with 95% confidence interval of 0.762–0.891 further validates the model.

Risk-adjustment model in health outcomes evaluation: a contribution to strengthen assessment towards quality improvement in interventional cardiology

The aim of this study was to develop a risk adjustment model for major adverse cardiac and cerebrovascular events following percutaneous coronary intervention (PCI), using data from a national registry, and to highlight the use of the risk adjustment when we evaluate the quality of care in interventional cardiology. The was based on a Coorte study. Bivariate and multivariate logistic regression models were used to identify independent risk factors for these major adverse events. A total of 19 hospitals from the Portuguese National Registry of Interventional Cardiology. Data from 10.641 consecutives procedures collected between June 30, 2003 and June 30, 2006. Build a risk adjustment model for these major adverse events, following percutaneous coronary intervention. Factors that were associated with major adverse cardiac and cerebrovascular events following percutaneous coronary intervention. The rate of in-hospital major adverse cardiac and cerebrovascular events was 1.9%. Factors associated with major adverse cardiac and cerebrovascular events included, among others: age >80 years (adjusted odds ratio = 3.91); female gender (1.72); and cardiogenic shock (6.05). Overall, a good discrimination was achieved with receiver operating characteristics curve = 0.84 and Hosmer-Lemeshow goodness of fit statistic across groups of risk was not significant (P = 0.18) indicating little departure from a perfect fit. These findings will represent an important contribution to quality and safety improvement and should help driving new research and innovative approaches to different subgroups of patients who have higher chances of having an adverse event or poorer outcomes following this intervention.

Validation of Trauma and Injury Severity Score in Blunt Trauma Patients by Using a Canadian Trauma Registry

To compare outcomes in blunt trauma by using Trauma and Injury Severity Score (TRISS) models derived from the Major Trauma Outcome Study (MTOS) and the Ontario Trauma Registry (OTR) as well as to evaluate the role of the Revised Trauma Score within the TRISS model. Consecutive blunt trauma cases from 11 Level I trauma centers over a 4-year period were identified from the OTR. Coefficients of the Revised Trauma Score were modified using the Ontario data and this score was tested by using the Hosmer-Lemeshow Goodness of Fit Test. Two Ontario-specific TRISS models were developed with revised coefficients. The first used the standard Revised Trauma Score and the second used the Revised Trauma Score with regenerated coefficients. The accuracy of mortality predictions for all models were compared by using a Hosmer-Lemeshow Goodness of Fit procedure. Additionally, each TRISS models performance characteristics and receiver operating characteristic (ROC) curves were used to evaluate their discriminative capabilities. A total of 5,436 cases were incorporated in the analysis. Patients with all component TRISS variables had a significantly lower mortality compared to all blunt trauma patients (7.0% vs. 15.5%,p < 0.01). Use of the Revised Trauma Score led to the exclusion of 40% of cases because of absent data necessary to compute the score. The Hosmer-Lemeshow Goodness of Fit statistic for the Revised Trauma Score was 79.45 (p = 0.0001). The Hosmer-Lemeshow Goodness of Fit Statistic ranged from 11.42, p = 0.175 and 13.1, p = 0.125 for the Ontario TRISS models compared to 25.62, p < 0.005 for the MTOS TRISS model. Sensitivity of all three TRISS models ranged from 98% to 99% with specificity ranging from 24% to 35%. ROC curves were identical for all three TRISS models. TRISS demonstrated satisfactory performance in a Canadian blunt trauma population. Although revision of coefficients led to a better fit on the Hosmer-Lemeshow statistic, ROC curves demonstrated virtually identical performance of the MTOS and Ontario-based TRISS models. The poor performance of the Revised Trauma Score and the observation that its use led to the exclusion of 40% of cases with a higher mortality raises concerns regarding its use in the TRISS model.