Anna Rieger

Risk calculators—methods, development, implementation, and validation

IntroductionA surgical risk calculator (SRC) estimates the probabilities of unfavorable outcomes such as complications or death after a specific surgery. The risk estimates are based on information regarding the patient’s medical history and his current status. They are calculated using risk models derived from the analysis of data from a large number of previous patients in a similar clinical situation.MethodsThis paper discusses several aspects of the SRC development and its implementation into clinical practice: the development of the statistical risk models, their validation and software implementation, the use of the SRC output for shared decision making in clinical settings, and the evaluation of the SRC’s impact on individual patient outcomes as well as on the institution’s quality of care of the clinical institution.ResultsProbably the most elaborate SRC is the ACS NSQIP SRC. A comparable project was started by the German Society for Visceral and General Surgery (DGAV) in the framework of its Study, Documentation, and Quality Center (StuDoQ). It is relevant to consider that the transportability of a SRC from a US American to a German setting is not straightforward.ConclusionsRisk calculators are important instruments for shared decision making between patients and doctor. Their implementation into clinical practice has to solve technical issues, and it is related to appropriate training of clinicians. There are specific study designs to evaluate the clinical impact of a SCR.

The DGAV risk calculator: development and validation of statistical models for a web-based instrument predicting complications of colorectal cancer surgery

PurposeThe purpose of this study is to provide a web-based calculator predicting complication probabilities of patients undergoing colorectal cancer (CRC) surgery in Germany.MethodsAnalyses were based on records of first-time CRC surgery between 2010 and February 2017, documented in the database of the Study, Documentation, and Quality Center (StuDoQ) of the Deutsche Gesellschaft für Allgemein- und Viszeralchirurgie (DGAV), a registry of CRC surgery in hospitals throughout Germany, covering demography, medical history, tumor features, comorbidity, behavioral risk factors, surgical procedures, and outcomes. Using logistic ridge regression, separate models were developed in learning samples of 6729 colon and 4381 rectum cancer patients and evaluated in validation samples of sizes 2407 and 1287. Discrimination was assessed using c statistics. Calibration was examined graphically by plotting observed versus predicted complication probabilities and numerically using Brier scores.ResultsWe report validation results regarding 15 outcomes such as any major complication, surgical site infection, anastomotic leakage, bladder voiding disturbance after rectal surgery, abdominal wall dehiscence, various internistic complications, 30-day readmission, 30-day reoperation rate, and 30-day mortality. When applied to the validation samples, c statistics ranged between 0.60 for anastomosis leakage and 0.85 for mortality after rectum cancer surgery. Brier scores ranged from 0.003 to 0.127.ConclusionsWhile most models showed satisfactory discrimination and calibration, this does not preclude overly optimistic or pessimistic individual predictions. To avoid misinterpretation, one has to understand the basic principles of risk calculation and risk communication. An e-learning tool outlining the appropriate use of the risk calculator is provided.

Efficient Maximum Likelihood Estimation for Pedigree Data with the Sum-Product Algorithm

Objective: We analyze data sets consisting of pedigrees with age at onset of colorectal cancer (CRC) as phenotype. The occurrence of familial clusters of CRC suggests the existence of a latent, inheritable risk factor. We aimed to compute the probability of a family possessing this risk factor as well as the hazard rate increase for these risk factor carriers. Due to the inheritability of this risk factor, the estimation necessitates a costly marginalization of the likelihood. Methods: We propose an improved EM algorithm by applying factor graphs and the sum-product algorithm in the E-step. This reduces the computational complexity from exponential to linear in the number of family members. Results: Our algorithm is as precise as a direct likelihood maximization in a simulation study and a real family study on CRC risk. For 250 simulated families of size 19 and 21, the runtime of our algorithm is faster by a factor of 4 and 29, respectively. On the largest family (23 members) in the real data, our algorithm is 6 times faster. Conclusion: We introduce a flexible and runtime-efficient tool for statistical inference in biomedical event data with latent variables that opens the door for advanced analyses of pedigree data.

Investigation of three outbreaks of Porcine Epidemic Diarrhea in Germany in 2016 demonstrates age dependent differences in the development of humoral immune response

Porcine epidemic diarrhea (PED) has reemerged in Europe since 2014. Characterized by a rapid onset of diarrhea in pigs of all ages, morbidity can reach up to 100% whereas mortality is variable. The virus strains involved in the recent European outbreaks all cluster together with US strains (S INDEL) that lead to less severe clinical signs. In this study, fattening pigs and suckling piglets (n = 105) on farms with no prior PED history were monitored after an acute outbreak of the disease, caused by an S INDEL strain of PED virus (PEDV). For diagnostic investigations in the affected farms, real time RT-PCR was performed to detect PEDV RNA in individually taken fecal samples, and two commercial ELISA kits, both based on the N protein of PEDV, were used to detect IgG in serum samples of pigs experiencing acute signs of the disease. PEDV RNA could be detected in fecal samples up to 14 days after initial sampling. Comparing both ELISAs by Cohens Kappa showed substantial agreement (κ = 0,771). Antibodies were detectable in all fattening pigs (100%) within 10 days after the occurrence of first clinical signs and remained detectable for about two months at least in 20.6% (farm 1) and 45.7% (farm 2) of the animals, respectively. In contrast, only 18 of 34 (52.9%) suckling piglets seroconverted. Although, PEDV RNA was found in fecal samples of all piglets, 13 piglets did not demonstrate antibodies at any sampling day. PCR to detect PEDV RNA in fecal samples seems to be a reliable diagnostic tool during and after the acute outbreak. In the present study, IgG ELISA kits proved to be a feasible diagnostic tool, but age dependent differences in detection rate and persistence of antibodies need to be considered.

Impact of robot scrapers on clinical mastitis and somatic cell count in lactating cows

ABSTRACT The objective of the present study was to explore the impact of robotic walkway cleaning on clinical mastitis and the somatic cell count in lactating cows. Data collection was carried out on a large dairy farm for two six-month periods in 2012 and 2013. Walkway cleaning with five robot scrapers was performed only in 2013. The incidence of clinical mastitis was analysed using the chi-square test. A linear mixed-effects model was applied for the analysis of the somatic cell count. Results indicated that the proportion of incidences of clinical mastitis decreased between 2012 and 2013 by 2.42 percent points. On the other hand, the somatic cell count of the cows slightly rose between both investigation periods and thus increased the likelihood of intramammary infection. This contrary development between clinical mastitis and somatic cell count also occurred in previous studies in which it was attributed to a pathogen-specific effect owing to farm management. An investigation over a longer period can help to clarify the influence of robot scrapers on udder health in dairy cows.

A Bayesian scoring rule on clustered event data for familial risk assessment - an example from colorectal cancer screening

Colorectal cancer screening is well established. The identification of high risk populations is the key to implement effective risk-adjusted screening. Good statistical approaches for risk prediction do not exist. The familys colorectal cancer history is used for identification of high risk families and usually assessed by a questionnaire. This paper introduces a prediction algorithm to designate a family for colorectal cancer risk and discusses its statistical properties. The new algorithm uses Bayesian reasoning and a detailed family history illustrated by a pedigree and a Lexis diagram. The algorithm is able to integrate different hereditary mechanisms that define complex latent class or random factor structures. They are generic and do not reflect specific genetic models. This is comparable to strategies in complex segregation analysis. Furthermore, the algorithm can integrate different statistical penetrance models for right censored event data. Computational challenges related to the handling of the likelihood are discussed. Simulation studies assess the predictive quality of the new algorithm in terms of ROC curves and corresponding AUCs. The algorithm is applied to data of a recent study on familial colorectal cancer risk. Its predictive performance is compared to that of a questionnaire currently used in screening for familial colorectal cancer. The results of the proposed algorithm are robust against different inheritance models. Using the simplest hereditary mechanism, the simulation study provides evidence that the algorithm improves detection of families with high cancer risk in comparison to the currently used questionnaire. The applicability of the algorithm goes beyond the field of colorectal cancer.