Shengfan Zhang

Analyzing factors associated with women’s attitudes and behaviors toward screening mammography using design-based logistic regression

We examined the factors associated with screening mammography adherence behaviors and influencing factors on women’s attitudes toward mammography in non-adherent women. Design-based logistic regression models were developed to characterize the influencing factors, including socio-demographic, health related, behavioral characteristics, and knowledge of breast cancer/mammography, on women’s compliance with and attitudes toward mammography using the 2003 Health Information National Trends Survey data. Findings indicate significant associations among adherence to mammography and marital status, income, health coverage, being advised by a doctor to have a mammogram, having had Pap smear before, perception of chance of getting breast cancer, and knowledge of mammography (frequency of doing mammogram) in both women younger than 65 and women aged 65 and older. However, number of visits to a healthcare provider per year and lifetime number of smoked cigarettes are only significant for women younger than 65. Factors significantly associated with attitudes toward mammography in non-adherent women are age, being advised by a doctor to have a mammogram, and seeking cancer information. To enhance adherence to mammography programs, physicians need to continue to advise their patients to obtain mammograms. In addition, increasing women’s knowledge about the frequency and starting age for screening mammography may improve women’s adherence. Financially related factors such as income and insurance are also shown to be significant factors. Hence, healthcare policies aimed at providing breast cancer screening services to underserved women will likely enhance mammography participation.

Characterizing the impact of mental disorders on HIV patient length of stay and total charges

There are over one million people in the United States living with HIV/AIDS, 20% of whom are undiagnosed, increasing the risk of transmission and the burden on the healthcare system. Those with comorbid diseases may be particularly vulnerable. This paper studies the impact of comorbidities, with a particular focus on mental disorders, on HIV patient outcomes as measured by patient length of stay (LOS) and total charges. Generalized linear models (gamma models) allowing heteroscedasticity are developed to characterize the effects of selected comorbidities on HIV patient outcomes in the adult 2006 National Inpatient Sample. Comorbid HIV patients experience different LOS and total charges. In particular, having mental disorders resulted in a decrease in both LOS (19%) and total charges (15%) for HIV patients. To characterize the role of individual mental disorders, principal component and cluster analyses on ICD-9 codes are used to study the impact of mental disorder, and eight conditions are found to be most strongly associated with HIV. Gamma models with these identified mental disorders as independent variables are then developed. The results have shown different effects on LOS and charges for each condition, and special attention should be given to those mental disorders (e.g., drug dependence) that increased LOS and charges when present.

From Data to Improved Decisions: Operations Research in Healthcare Delivery

Background. The Operations Research Interest Group (ORIG) within the Society of Medical Decision Making (SMDM) is a multidisciplinary interest group of professionals that specializes in taking an analytical approach to medical decision making and healthcare delivery. ORIG is interested in leveraging mathematical methods associated with the field of Operations Research (OR) to obtain data-driven solutions to complex healthcare problems and encourage collaborations across disciplines. This paper introduces OR for the non-expert and draws attention to opportunities where OR can be utilized to facilitate solutions to healthcare problems. Methods. Decision making is the process of choosing between possible solutions to a problem with respect to certain metrics. OR concepts can help systematically improve decision making through efficient modeling techniques while accounting for relevant constraints. Depending on the problem, methods that are part of OR (e.g., linear programming, Markov Decision Processes) or methods that are derived from related fields (e.g., regression from statistics) can be incorporated into the solution approach. This paper highlights the characteristics of different OR methods that have been applied to healthcare decision making and provides examples of emerging research opportunities. Examples. We illustrate OR applications in healthcare using previous studies, including diagnosis and treatment of diseases, organ transplants, and patient flow decisions. Further, we provide a selection of emerging areas for utilizing OR. Conclusions. There is a timely need to inform practitioners and policy makers of the benefits of using OR techniques in solving healthcare problems. OR methods can support the development of sustainable long-term solutions across disease management, service delivery, and health policies by optimizing the performance of system elements and analyzing their interaction while considering relevant constraints.

Strategic level proton therapy patient admission planning: a Markov decision process modeling approach

A relatively new consideration in proton therapy planning is the requirement that the mix of patients treated from different categories satisfy desired mix percentages. Deviations from these percentages and their impacts on operational capabilities are of particular interest to healthcare planners. In this study, we investigate intelligent ways of admitting patients to a proton therapy facility that maximize the total expected number of treatment sessions (fractions) delivered to patients in a planning period with stochastic patient arrivals and penalize the deviation from the patient mix restrictions. We propose a Markov Decision Process (MDP) model that provides very useful insights in determining the best patient admission policies in the case of an unexpected opening in the facility (i.e., no-shows, appointment cancellations, etc.). In order to overcome the curse of dimensionality for larger and more realistic instances, we propose an aggregate MDP model that is able to approximate optimal patient admission policies using the worded weight aggregation technique. Our models are applicable to healthcare treatment facilities throughout the United States, but are motivated by collaboration with the University of Florida Proton Therapy Institute (UFPTI).

Inferring breast cancer concomitant diagnosis and comorbidities from the Nationwide Inpatient Sample using social network analysis

Breast cancer is a complex disease and may be accompanied by other multiple health conditions. The present study investigates associations between diagnosis codes in breast cancer patients using the Nationwide Inpatient Sample data. Concomitant diagnoses codes are identified by statistically significant associations between the diagnoses codes in a given breast cancer patient. These are subsequently represented in the form of a network (Breast Cancer Concomitant Diagnosis Network (BCCDN)). In contrast to more classical approaches, BCCDN provides system-level insights and convenient visualization reflected by the complex wiring patterns between the diagnoses codes. Social network analysis is used to investigate highly connected codes in the BCCDN network, and their variation across three different populations: (i) the deceased breast cancer population (ii) the elderly breast cancer population (age>65 years) and (iii) the adult breast cancer population (age <=65 years). BCCDNs were investigated across years 2005 and 2006 in order to identify associations that are robust to the stratified sampling and population heterogeneity as well as possible errors in documentation characteristic of observational healthcare data. The results presented validate known chronic comorbidities and their persistence across the deceased and elderly breast cancer population. They also provide novel associations and potential comorbidities in breast cancer patients that may warrant a more detailed investigation.

Competing risks analysis in mortality estimation for breast cancer patients from independent risk groups

This study quantifies breast cancer mortality in the presence of competing risks for complex patients. Breast cancer behaves differently in different patient populations, which can have significant implications for patient survival; hence these differences must be considered when making screening and treatment decisions. Mortality estimation for breast cancer patients has been a significant research question. Accurate estimation is critical for clinical decision making, including recommendations. In this study, a competing risks framework is built to analyze the effect of patient risk factors and cancer characteristics on breast cancer and other cause mortality. To estimate mortality probabilities from breast cancer and other causes as a function of not only the patient’s age or race but also biomarkers for estrogen and progesterone receptor status, a nonparametric cumulative incidence function is formulated using data from the community-based Carolina Mammography Registry. Based on the log(−log) transformation, confidence intervals are constructed for mortality estimates over time. To compare mortality probabilities in two independent risk groups at a given time, a method with improved power is formulated using the log(−log) transformation.

Analyzing overdiagnosis risk in cancer screening: A case of screening mammography for breast cancer

ABSTRACT Overdiagnosis is defined as the diagnosis of an asymptotic cancer that would not have presented clinically in a patients lifetime in the absence of screening. Quantifying overdiagnosis is difficult, since it is impossible to distinguish between a cancer that would cause symptoms in the patient lifetime and the ones that would not. In this study, a mathematical framework is developed to estimate the lifetime overdiagnosis and cancer mortality risks associated with cancer screening policies. We also develop an optimization model to extract screening policies with minimum overdiagnosis and lifetime breast cancer mortality risk. The proposed optimization model is highly nonlinear with complex structure. Therefore, we linearize the optimization model by introducing new decision variables and restructuring the equations to solve it optimally. We utilize existing data on breast cancer for average-risk women and evaluated mammography screening policies in terms of their associated lifetime overdiagnosis and breast cancer mortality risk. Optimal policies with minimum overdiagnosis and mortality risks are derived. The optimal policies outperform the existing in-practice policies by recommending more frequent screenings at younger ages, as the cancer is more aggressive and the remaining life expectancy is higher for younger patients.

Optimal Decision Making for Breast Cancer Treatment in the Presence of Cancer Regression and Type II Error in Mammography Results

Breast cancer is the leading cause of cancer death among women worldwide. While breast cancer-screening policies have been widely studied in order to achieve early detection, not much research has been done to optimize treatment decisions once a screening policy is established. In this chapter, we propose a dynamic decision model to determine optimal breast cancer treatment decisions that consider both the impact of overtreatment and the potential delay in cancer detection; these two failures are caused by spontaneous cancer regression and type II error in mammography results, respectively. We measure the impact of medical treatment by means of quality-adjusted life years (QALYs) and our goal is to maximize this metric for a given patient.