MyVA Access: An Evaluation of Changes in Access for a System-Wide Program Implemented in the Veterans Health Administration.

The current study evaluates changes in access as a result of the MyVA Access program-a system-wide effort to improve patient access in the Veterans Health Administration. Data on 20 different measures were collected, and changes were analyzed using t tests and Chow tests. Additionally, organizational health-how able a system is to create health care practice change-was evaluated for a sample of medical centers (n = 36) via phone interviews and surveys conducted with facility staff and technical assistance providers. An organizational health variable was created and correlated with the access measures. Results showed that, nationally, average wait times for urgent consults, new patient wait times for mental health and specialty care, and slot utilization for primary and specialty care patients improved. Patient satisfaction measures also improved, and patient complaints decreased. Better organizational health was associated with improvements in patient access.

A Personalized Approach of Patient-Health Care Provider Communication Regarding Colorectal Cancer Screening Options.

BACKGROUND: Current colorectal cancer screening guidelines by the US Preventive Services Task Force endorse multiple options for average-risk patients and recommend that screening choices should be guided by individual patient preferences. Implementing these recommendations in practice is challenging because they depend on accurate and efficient elicitation and assessment of preferences from patients who are facing a novel task. OBJECTIVE: To present a methodology for analyzing the sensitivity and stability of a patient's preferences regarding colorectal cancer screening options and to provide a starting point for a personalized discussion between the patient and the health care provider about the selection of the appropriate screening option. METHODS: This research is a secondary analysis of patient preference data collected as part of a previous study. We propose new measures of preference sensitivity and stability that can be used to determine if additional information provided would result in a change to the initially most preferred colorectal cancer screening option. RESULTS: Illustrative results of applying the methodology to the preferences of 2 patients, of different ages, are provided. The results show that different combinations of screening options are viable for each patient and that the health care provider should emphasize different information during the medical decision-making process. CONCLUSION: Sensitivity and stability analysis can supply health care providers with key topics to focus on when communicating with a patient and the degree of emphasis to place on each of them to accomplish specific goals. The insights provided by the analysis can be used by health care providers to approach communication with patients in a more personalized way, by taking into consideration patients' preferences before adding their own expertise to the discussion.

Modeling Patient No-Show History and Predicting Future Outpatient Appointment Behavior in the Veterans Health Administration.

BACKGROUND: Missed appointments reduce the efficiency of the health care system and negatively impact access to care for all patients. Identifying patients at risk for missing an appointment could help health care systems and providers better target interventions to reduce patient no-shows. OBJECTIVES: Our aim was to develop and test a predictive model that identifies patients that have a high probability of missing their outpatient appointments. METHODS: Demographic information, appointment characteristics, and attendance history were drawn from the existing data sets from four Veterans Affairs health care facilities within six separate service areas. Past attendance behavior was modeled using an empirical Markov model based on up to 10 previous appointments. Using logistic regression, we developed 24 unique predictive models. We implemented the models and tested an intervention strategy using live reminder calls placed 24, 48, and 72 hours ahead of time. The pilot study targeted 1,754 high-risk patients, whose probability of missing an appointment was predicted to be at least 0.2. RESULTS: Our results indicate that three variables were consistently related to a patient's no-show probability in all 24 models: past attendance behavior, the age of the appointment, and having multiple appointments scheduled on that day. After the intervention was implemented, the no-show rate in the pilot group was reduced from the expected value of 35% to 12.16% (p value < 0.0001). CONCLUSIONS: The predictive model accurately identified patients who were more likely to miss their appointments. Applying the model in practice enables clinics to apply more intensive intervention measures to high-risk patients.

A mixed-ensemble model for hospital readmission.

OBJECTIVE: A hospital readmission is defined as an admission to a hospital within a certain time frame, typically thirty days, following a previous discharge, either to the same or to a different hospital. Because most patients are not readmitted, the readmission classification problem is highly imbalanced. MATERIALS AND METHODS: We developed a hospital readmission predictive model, which enables controlling the tradeoff between reasoning transparency and predictive accuracy, by taking into account the unique characteristics of the learned database. A boosted C5.0 tree, as the base classifier, was ensembled with a support vector machine (SVM), as a secondary classifier. The models were induced and validated using anonymized administrative records of 20,321 inpatient admissions, of 4840 Congestive Heart Failure (CHF) patients, at the Veterans Health Administration (VHA) hospitals in Pittsburgh, from fiscal years (FY) 2006 through 2014. RESULTS: The SVM predictions are characterized by greater sensitivity values (true positive rates) than are the C5.0 predictions, for a wider range of cut off values of the ROC curve, depending on a predefined confidence threshold for the base C5.0 classifier. The total accuracy for the ensemble ranges from 81% to 85%. Different predictors, including comorbidities, lab values, and vitals, play different roles in the two models. CONCLUSIONS: The mixed-ensemble model enables easy and fast exploratory knowledge discovery of the database, and a control of the classification error for positive readmission instances. Implementation of this ensembling method for predicting all-cause hospital readmissions of CHF patients allows overcoming some of the limitations of the classifiers considered individually, and of other traditional ensembling methods. It also increases the classification accuracy for positive readmission instances, particularly when strong predictors are not available.

Large-Scale No-Show Patterns and Distributions for Clinic Operational Research.

Patient no-shows for scheduled primary care appointments are common. Unused appointment slots reduce patient quality of care, access to services and provider productivity while increasing loss to follow-up and medical costs. This paper describes patterns of no-show variation by patient age, gender, appointment age, and type of appointment request for six individual service lines in the United States Veterans Health Administration (VHA). This retrospective observational descriptive project examined 25,050,479 VHA appointments contained in individual-level records for eight years (FY07-FY14) for 555,183 patients. Multifactor analysis of variance (ANOVA) was performed, with no-show rate as the dependent variable, and gender, age group, appointment age, new patient status, and service line as factors. The analyses revealed that males had higher no-show rates than females to age 65, at which point males and females exhibited similar rates. The average no-show rates decreased with age until 75-79, whereupon rates increased. As appointment age increased, males and new patients had increasing no-show rates. Younger patients are especially prone to no-show as appointment age increases. These findings provide novel information to healthcare practitioners and management scientists to more accurately characterize no-show and attendance rates and the impact of certain patient factors. Future general population data could determine whether findings from VHA data generalize to others.

Classifying highly imbalanced ICU data.

Highly imbalanced data sets are those where the class of interest is rare. In this paper, we compare the performance of several common data mining methods, logistic regression, discriminant analysis, Classification and Regression Tree (CART) models, C5, and Support Vector Machines (SVM) in predicting the discharge status (alive or deceased, with "deceased" being the class of interest) of patients from an Intensive Care Unit (ICU). Using a variety of misclassification cost ratio (MCR) values and using specificity, recall, precision, the F-measure, and confusion entropy (CEN) as criteria for evaluating each method's performance, C5 and SVM performed better than the other methods. At a MCR of 100, C5 had the highest recall and SVM the highest specificity and lowest CEN. We also used Hand's measure to compare the five methods. According to Hand's measure, logistic regression performed the best. This article makes several contributions. We show how the use of MCR for analyzing imbalanced medical data significantly improves the method's classification performance. We also found that the F-measure and precision did not improve as the MCR was increased.

A data mining approach to characterizing medical code usage patterns.

This research describes a synthetic data mining approach to identifying diagnostic (ICD-9) and procedure (CPT) code usage patterns in two US. hospitals, with the goal of determining the adequacy and effectiveness of the current coding classification systems. We combine relative frequency measurements with measures of industry concentration borrowed from industrial economics in order to (1) ascertain the extent to which physicians utilize the available codes in classifying patients and (2) discover the factors that impinge on code usage. Our results partition the domain into areas for which the coding systems perform well and those areas for which the systems perform relatively poorly. The goal is to use this approach to understand how coding systems are used and to highlight areas for targeted improvement of the current coding

Estimating procedure times for surgeries by determining location parameters for the lognormal model.

We present an empirical study of methods for estimating the location parameter of the lognormal distribution. Our results identify the best order statistic to use, and indicate that using the best order statistic instead of the median may lead to less frequent incorrect rejection of the lognormal model, more accurate critical value estimates, and higher goodness-of-fit. Using simulation data, we constructed and compared two models for identifying the best order statistic, one based on conventional nonlinear regression and the other using a data mining/machine learning technique. Better surgical procedure time estimates may lead to improved surgical operations.

Estimating times of surgeries with two component procedures: comparison of the lognormal and normal models.

BACKGROUND: Variability inherent in the duration of surgical procedures complicates surgical scheduling. Modeling the duration and variability of surgeries might improve time estimates. Accurate time estimates are important operationally to improve utilization, reduce costs, and identify surgeries that might be considered outliers. Surgeries with multiple procedures are difficult to model because they are difficult to segment into homogenous groups and because they are performed less frequently than single-procedure surgeries. METHODS: The authors studied, retrospectively, 10,740 surgeries each with exactly two CPTs and 46,322 surgical cases with only one CPT from a large teaching hospital to determine if the distribution of dual-procedure surgery times fit more closely a lognormal or a normal model. The authors tested model goodness of fit to their data using Shapiro-Wilk tests, studied factors affecting the variability of time estimates, and examined the impact of coding permutations (ordered combinations) on modeling. RESULTS: The Shapiro-Wilk tests indicated that the lognormal model is statistically superior to the normal model for modeling dual-procedure surgeries. Permutations of component codes did not appear to differ significantly with respect to total procedure time and surgical time. To improve individual models for infrequent dual-procedure surgeries, permutations may be reduced and estimates may be based on the longest component procedure and type of anesthesia. CONCLUSIONS: The authors recommend use of the lognormal model for estimating surgical times for surgeries with two component procedures. Their results help legitimize the use of log transforms to normalize surgical procedure times prior to hypothesis testing using linear statistical models. Multiple-procedure surgeries may be modeled using the longest (statistically most important) component procedure and type of anesthesia.