PHASE-TYPE DISTRIBUTIONS FOR CLAIM SEVERITY REGRESSION MODELING

This paper addresses the task of modeling severity losses using segmentation when the data distribution does not fall into the usual regression frameworks. This situation is not uncommon in lines of business such as third-party liability insurance, where heavy-tails and multimodality often hamper a direct statistical analysis. We propose to use regression models based on phase-type distributions, regressing on their underlying inhomogeneous Markov intensity and using an extension of the expectation–maximization algorithm. These models are interpretable and tractable in terms of multistate processes and generalize the proportional hazards specification when the dimension of the state space is larger than 1. We show that the combination of matrix parameters, inhomogeneity transforms, and covariate information provides flexible regression models that effectively capture the entire distribution of loss severities.

Intake of Flavonoids and Odds of Frailty Onset in Adults in the Framingham Offspring Cohort

Polyphenols (antioxidants derived from plant-foods) could play a role in inhibition of oxidative stress and frailty reduction, yet data on the polyphenol subclass of dietary flavonoids is limited. This study sought to determine the association between dietary flavonoids and frailty onset in middle-aged and older adults. This prospective cohort study included non-frail individuals from the Framingham Offspring Cohort (FOC) with total flavonoid intake (mg/day; defined as sum flavonols, flavan-3-ols, flavonones, flavones, and anthocyanins via Harvard Food Frequency Questionnaire), frailty (via Fried phenotype), and covariate information measured at baseline (1998-2001). Follow-up frailty was evaluated in 2011-2014. Logistic regression estimated odds ratio (OR) and 95% confidence intervals (95% CI) adjusting for relevant confounders. Participants (n=1,701; 55.5% female) had a mean age of 58.4 years (SD ± 8.3). Mean flavonoid intake was 309 mg/d (SD ± 266). After 12.4 years (SD ± 0.8), 224 (13.2%) individuals exhibited frailty. In age and sex adjusted models, every 50 mg/day of higher total flavonoid intake was associated with 3% reduced odds of frailty [OR (95%CI): 0.97 (0.94-1.00), p-value: 0.05). Further adjustment for smoking, energy and protein intake, and disease indicators did not appreciably change the association, and associations became non-significant (p-value=0.12). Thus, there was no association between flavonoid intake and odds of frailty onset in adults in the FOC. This could be due to participants' higher intake of flavonoids compared to average intake of ~200 mg/d in Americans.

Distributionally robust stochastic programs with side information based on trimmings

We consider stochastic programs conditional on some covariate information, where the only knowledge of the possible relationship between the uncertain parameters and the covariates is reduced to a finite data sample of their joint distribution. By exploiting the close link between the notion of trimmings of a probability measure and the partial mass transportation problem, we construct a data-driven Distributionally Robust Optimization (DRO) framework to hedge the decision against the intrinsic error in the process of inferring conditional information from limited joint data. We show that our approach is computationally as tractable as the standard (without side information) Wasserstein-metric-based DRO and enjoys performance guarantees. Furthermore, our DRO framework can be conveniently used to address data-driven decision-making problems under contaminated samples. Finally, the theoretical results are illustrated using a single-item newsvendor problem and a portfolio allocation problem with side information.

Learning social networks from text data using covariate information

Accurately describing the lives of historical figures can be challenging, but unraveling their social structures perhaps is even more so. Historical social network analysis methods can help in this regard and may even illuminate individuals who have been overlooked by historians, but turn out to be influential social connection points. Text data, such as biographies, are a useful source of information for learning historical social networks but the identifcation of links based on text data can be challenging. The Local Poisson Graphical Lasso model models social networks by conditional independence structures, and leverages the number of name co-mentions in the text to infer relationships. However, this method does not take into account the abundance of covariate information that is often available in text data. Conditional independence structure like Poisson Graphical Model, which makes use name mention counts in the text can be useful tools to avoid false positive links due to the co-mentions but given historical tendency of frequently used or common names, without additional distinguishing information, we may introduce incorrect connections. In this work, we therefore extend the Local Poisson Graphical Lasso model with a (multiple) penalty structure that incorporates covariates, opening up the opportunity for similar individuals to have a higher probability of being connected. We propose both greedy and Bayesian approaches to estimate the penalty parameters. We present results on data simulated with characteristics of historical networks and show that this type of penalty structure can improve network recovery as measured by precision and recall. We also illustrate the approach on biographical data of individuals who lived in early modern Britain between 1500 and 1575. We will show how these covariates affect the statistical model’s performance using simulations, discuss how it helps to better identify links for the people with common names and those who are traditionally underrepresented in the biography text data.

POS0122 THE ASSOCIATIONS BETWEEN DISEASE MODIFYING ANTI-RHEUMATIC DRUGS AND INCIDENT, AND PROGRESSION OF, RADIOGRAPHIC HAND OSTEOARTHRITIS IN RHEUMATOID ARTHRITIS PATIENTS

Background:Hand osteoarthritis (OA) is characterized by bone erosions, joint space remodeling, and new bone formation mainly in distal interphalangeal (DIP) joints and thereby differs from hand manifestations in rheumatoid arthritis (RA). There are conflicting data about the benefit of treatment with conventional synthetic (cs) and biologic (b) disease modifying anti-rheumatic treatment (DMARD) on DIP OA.Objectives:To assess the associations between DMARDs and incident, and progression of, radiographic DIP OA in RA patients.Methods:We performed two observational cohort studies in the longitudinal Swiss Clinical Quality Management in Rheumatic Diseases registry (SCQM) between 1997 and 2014. RA patients who had ≥2 eligible hand radiographs were included at their first eligible radiograph (baseline) and were followed until the outcome or their last eligible radiograph. Radiographs were eligible if all 8 DIP joints could be scored. Modified Kellgren-Lawrence scores (KLS) were obtained by evaluating DIP joints for severity of osteophytes, joint space narrowing, subchondral sclerosis, and erosions. Incident/existing DIP OA was defined as KLS ≥2 in ≥1 DIP joint. Progression of existing DIP OA was defined as an increase of ≥1 in KLS in ≥1 DIP joint. We divided the study population into two cohorts based on whether DIP OA was present or absent at cohort entry (cohorts 1 and 2, respectively). Exposure status was defined time-dependently into mutually exclusive exposure groups: csDMARD monotherapy, bDMARD monotherapy, bDMARD/csDMARD combination therapy, past bDMARD/csDMARD therapy, or never DMARD use. Cox time-varying proportional hazard regression analyses were used to estimate hazard ratios (HRs) with 95% confidence intervals (CI) of DIP OA progression (cohort 1) or DIP OA incidence (cohort 2) associated with DMARD exposure categories (csDMARD monotherapy was the reference group because it was the largest group). Exposure and covariate information were extracted at every radiograph and other visit date. Missing covariate information was imputed using multiple imputation by chained equations. In sensitivity analyses, we repeated all analyses using generalised estimation equations (GEE).Results:Among 2234 RA patients with 5928 eligible radiographs, 1340 patients had radiographic DIP OA at cohort entry (cohort 1) and 894 were DIP OA naïve (cohort 2). In cohort 1, radiographic progression of existing DIP OA was characterized by new osteophyte formation (666, 52.4%), followed by joint space narrowing (379, 27.5%), subchondral sclerosis (238, 17.8%), and erosion (62, 4.3%). bDMARD monotherapy was associated with an increased risk of radiographic DIP OA progression compared to csDMARD monotherapy (adjusted HR 1.34, 95% CI 1.07–1.69). The risk of DIP OA progression was not significant in csDMARD/bDMARD combination therapy users (adjusted HR 1.12, 95% CI 0.96–1.31), absent in past DMARD users (adjusted HR 0.96, 95% CI 0.66–1.41), and significantly lower among never DMARD users (adjusted HR 0.54, 95% CI 0.33–0.90), compared to csDMARD monotherapy users. In cohort 2, the risk of incident OA did not differ materially between treatment groups. Results from GEE analyses corroborated all findings.Conclusion:Our results from this real-world RA cohort suggest that monotherapy with bDMARDs is not associated with incident DIP OA but may increase the risk of radiographic progression of existing DIP OA when compared to csDMARDs.Acknowledgements:We thank all patients and rheumatologists involved for their contribution to the SCQM RA cohort. A list of rheumatology offices and hospitals that contribute to the SCQM registry can be found at http://www.scqm.ch/institutions. The SCQM is financially supported by pharmaceutical industries and donors. A list of financial supporters can be found at http://www.scqm.ch/sponsors.Disclosure of Interests:Theresa Burkard: None declared, Christian Lechtenboehmer: None declared, Stephan Reichenbach: None declared, Monika Hebeisen: None declared, Ulrich Walker: None declared, Andrea Michelle Burden: None declared, Thomas Hügle Consultant of: Pfizer, Abbvie, Novartis, Grant/research support from: GSK, Jansen, Pfizer, Abbvie, Novartis, Roche, MSD, Sanofi, BMS, Eli Lilly, UCB

Common Methodological Challenges Encountered With Multiple Systems Estimation Studies

Multiple systems estimation refers to a class of inference procedures that are commonly used to estimate the size of hidden populations based on administrative lists. In this paper we discuss some of the common challenges encountered in such studies. In particular, we summarize theoretical issues relating to the existence of maximum likelihood estimators, model identifiability, and parameter redundancy when there is sparse overlap among the lists. We also discuss techniques for matching records when there are no unique identifiers, exploiting covariate information to improve estimation, and addressing missing data. We offer suggestions for remedial actions when these issues/challenges manifest. The corresponding R coding packages that can assist with the analyses of multiple systems estimation data sets are also discussed.