E Barbini

Importance of sociodemographic and morbidity aspects in measuring health-related quality of life: performances of three tools

BackgroundSince health-related quality of life (HRQL) measures are numerous, comparisons have been suggested.AimTo compare three HRQL measures: SF6D, HUI3 and EQ5D.MethodsThree questionnaires (SF36, HUI3, EQ5D) were administered to 1,011 patients attending 16 general practices in two Italian cities. Information about patients’ gender, age, education, marital status, smoking, body mass index (BMI) and chronic diseases (hypertension, diabetes, cardiovascular and musculoskeletal diseases) were also collected. Questionnaires scores were calculated using the appropriate algorithms; in particular SF6D scores were obtained from SF36 items. Agreement and correlation between questionnaires scores were investigated using Bland and Altman method and Spearman coefficient. The influence of socio-demographic and morbidity indicators on scores was analysed using the nonparametric quantile regression.ResultsThe Spearman coefficient was about 0.6 for all questionnaires. The 95% limits of agreement of the scores were approximately from −0.5 to 0.3 except for SF6D and EQ5D when they were from −0.4 to 0.2. The measures were influenced by socio-demographic and clinical variables in a similar way, especially SF6D (the index obtained from SF36) and EQ5D, which appeared to be influenced by the same pattern of factors, including gender, chronic diseases, smoking and BMI.ConclusionsOverall, the agreement between questionnaires scores was quite low, whilst the correlation level was good. Questionnaire scores were influenced by socio-demographic and clinical variables in a similar way, especially SF6D and EQ5D. Therefore, the descriptive capacity of SF6D and EQ5D was found to be similar.

A comparative analysis of predictive models of morbidity in intensive care unit after cardiac surgery – Part I: model planning

BackgroundDifferent methods have recently been proposed for predicting morbidity in intensive care units (ICU). The aim of the present study was to critically review a number of approaches for developing models capable of estimating the probability of morbidity in ICU after heart surgery. The study is divided into two parts. In this first part, popular models used to estimate the probability of class membership are grouped into distinct categories according to their underlying mathematical principles. Modelling techniques and intrinsic strengths and weaknesses of each model are analysed and discussed from a theoretical point of view, in consideration of clinical applications.MethodsModels based on Bayes rule, k- nearest neighbour algorithm, logistic regression, scoring systems and artificial neural networks are investigated. Key issues for model design are described. The mathematical treatment of some aspects of model structure is also included for readers interested in developing models, though a full understanding of mathematical relationships is not necessary if the reader is only interested in perceiving the practical meaning of model assumptions, weaknesses and strengths from a user point of view.ResultsScoring systems are very attractive due to their simplicity of use, although this may undermine their predictive capacity. Logistic regression models are trustworthy tools, although they suffer from the principal limitations of most regression procedures. Bayesian models seem to be a good compromise between complexity and predictive performance, but model recalibration is generally necessary. k-nearest neighbour may be a valid non parametric technique, though computational cost and the need for large data storage are major weaknesses of this approach. Artificial neural networks have intrinsic advantages with respect to common statistical models, though the training process may be problematical.ConclusionKnowledge of model assumptions and the theoretical strengths and weaknesses of different approaches are fundamental for designing models for estimating the probability of morbidity after heart surgery. However, a rational choice also requires evaluation and comparison of actual performances of locally-developed competitive models in the clinical scenario to obtain satisfactory agreement between local needs and model response. In the second part of this study the above predictive models will therefore be tested on real data acquired in a specialized ICU.

A comparative analysis of predictive models of morbidity in intensive care unit after cardiac surgery – Part II: an illustrative example

BackgroundPopular predictive models for estimating morbidity probability after heart surgery are compared critically in a unitary framework. The study is divided into two parts. In the first part modelling techniques and intrinsic strengths and weaknesses of different approaches were discussed from a theoretical point of view. In this second part the performances of the same models are evaluated in an illustrative example.MethodsEight models were developed: Bayes linear and quadratic models, k-nearest neighbour model, logistic regression model, Higgins and direct scoring systems and two feed-forward artificial neural networks with one and two layers. Cardiovascular, respiratory, neurological, renal, infectious and hemorrhagic complications were defined as morbidity. Training and testing sets each of 545 cases were used. The optimal set of predictors was chosen among a collection of 78 preoperative, intraoperative and postoperative variables by a stepwise procedure. Discrimination and calibration were evaluated by the area under the receiver operating characteristic curve and Hosmer-Lemeshow goodness-of-fit test, respectively.ResultsScoring systems and the logistic regression model required the largest set of predictors, while Bayesian and k-nearest neighbour models were much more parsimonious. In testing data, all models showed acceptable discrimination capacities, however the Bayes quadratic model, using only three predictors, provided the best performance. All models showed satisfactory generalization ability: again the Bayes quadratic model exhibited the best generalization, while artificial neural networks and scoring systems gave the worst results. Finally, poor calibration was obtained when using scoring systems, k-nearest neighbour model and artificial neural networks, while Bayes (after recalibration) and logistic regression models gave adequate results.ConclusionAlthough all the predictive models showed acceptable discrimination performance in the example considered, the Bayes and logistic regression models seemed better than the others, because they also had good generalization and calibration. The Bayes quadratic model seemed to be a convincing alternative to the much more usual Bayes linear and logistic regression models. It showed its capacity to identify a minimum core of predictors generally recognized as essential to pragmatically evaluate the risk of developing morbidity after heart surgery.

A naïve approach for deriving scoring systems to support clinical decision making.

RATIONALE, AIMS AND OBJECTIVES Scoring systems are frequently proposed in medicine to summarize a set of qualitative and quantitative items by means of a numeric score. Their design often requires modelling ability and subjective judgments. This can make it difficult to adapt a scoring system to a clinical setting different from that in which the system was developed. The objective of this study was to discuss an approach to derive scoring systems, which can be easily modified and matched to any scenario. METHODS A naïve Bayes approach was used to develop a scoring system that is completely defined by descriptive tables obtained by frequency counts from the training set. The approach was implemented to build a locally customized scoring system for planning transfusion requirements after cardiac surgery. The performance of this system was evaluated and compared with that of a logistic regression model designed using the same predictors. The working sample was a set of 3182 consecutive patients undergoing cardiac surgery at the University Hospital of Siena, Italy. RESULTS The area under the receiver operating characteristic curve was equal to 0.811 and 0.824 for the scoring system and for the logistic regression model, respectively. This result proves that this global index of discrimination capacity was virtually identical and very good for both models. The values of sensitivity, specificity and overall correct-classification percentage obtained by the leave-one-out method were practically the same for the two models (73.9% versus 75.3%). CONCLUSIONS An easy-fitting and trustworthy scoring system can be directly developed using a naïve Bayes approach. The simplicity of its design allows the system to be customized to any specific institution and updated regularly. This aspect has important practical implications because it can encourage the use of scoring systems among clinicians, enabling their performance to be properly assessed in a wider clinical context.Rationale, aims and objectives Scoring systems are frequently proposed in medicine to summarize a set of qualitative and quantitative items by means of a numeric score. Their design often requires modelling ability and subjective judgments. This can make it difficult to adapt a scoring system to a clinical setting different from that in which the system was developed. The objective of this study was to discuss an approach to derive scoring systems, which can be easily modified and matched to any scenario. Methods A naive Bayes approach was used to develop a scoring system that is completely defined by descriptive tables obtained by frequency counts from the training set. The approach was implemented to build a locally customized scoring system for planning transfusion requirements after cardiac surgery. The performance of this system was evaluated and compared with that of a logistic regression model designed using the same predictors. The working sample was a set of 3182 consecutive patients undergoing cardiac surgery at the University Hospital of Siena, Italy. Results The area under the receiver operating characteristic curve was equal to 0.811 and 0.824 for the scoring system and for the logistic regression model, respectively. This result proves that this global index of discrimination capacity was virtually identical and very good for both models. The values of sensitivity, specificity and overall correct-classification percentage obtained by the leave-one-out method were practically the same for the two models (73.9% versus 75.3%). Conclusions An easy-fitting and trustworthy scoring system can be directly developed using a naive Bayes approach. The simplicity of its design allows the system to be customized to any specific institution and updated regularly. This aspect has important practical implications because it can encourage the use of scoring systems among clinicians, enabling their performance to be properly assessed in a wider clinical context.

A naïve Bayes classifier for planning transfusion requirements in heart surgery

RATIONALE, AIMS AND OBJECTIVES Transfusion of allogeneic blood products is a key issue in cardiac surgery. Although blood conservation and standard transfusion guidelines have been published by different medical groups, actual transfusion practices after cardiac surgery vary widely among institutions. Models can be a useful support for decision making and may reduce the total cost of care. The objective of this study was to propose and evaluate a procedure to develop a simple locally customized decision-support system. METHODS We analysed 3182 consecutive patients undergoing cardiac surgery at the University Hospital of Siena, Italy. Univariate statistical tests were performed to identify a set of preoperative and intraoperative variables as likely independent features for planning transfusion quantities. These features were utilized to design a naïve Bayes classifier. Model performance was evaluated using the leave-one-out cross-validation approach. All computations were done using spss and matlab code. RESULTS The overall correct classification percentage was not particularly high if several classes of patients were to be identified. Model performance improved appreciably when the patient sample was divided into two classes (transfused and non-transfused patients). In this case the naïve Bayes model correctly classified about three quarters of patients with 71.2% sensitivity and 78.4% specificity, thus providing useful information for recognizing patients with transfusion requirements in the specific scenario considered. CONCLUSIONS Although the classifier is customized to a particular setting and cannot be generalized to other scenarios, the simplicity of its development and the results obtained make it a promising approach for designing a simple model for different heart surgery centres needing a customized decision-support system for planning transfusion requirements in intensive care unit.

A straightforward approach to designing a scoring system for predicting length-of-stay of cardiac surgery patients

BackgroundLength-of-stay prediction for cardiac surgery patients is a key point for medical management issues, such as optimization of resources in intensive care units and operating room scheduling. Scoring systems are a very attractive family of predictive models, but their retraining and updating are generally critical. The present approach to designing a scoring system for predicting length of stay in intensive care aims to overcome these difficulties, so that a model designed in a given scenario can easily be adjusted over time or for internal purposes.MethodsA naïve Bayes approach was used to develop a simple scoring system. A set of 36 preoperative, intraoperative and postoperative variables collected in a sample of 3256 consecutive adult patients undergoing heart surgery were considered as likely risk predictors. The number of variables was reduced by selecting an optimal subset of features. Scoring system performance was assessed by cross-validation.ResultsAfter the selection process, seven variables were entered in the prediction model, which showed excellent discrimination, good generalization power and suitable sensitivity and specificity. No significant difference was found between AUC of the training and testing sets. The 95% confidence interval for AUC estimated by the BCa bootstrap method was [0.841, 0.883] and [0.837, 0.880] in the training and testing sets, respectively. Chronic dialysis, low postoperative cardiac output and acute myocardial infarction proved to be the major risk factors.ConclusionsThe proposed approach produced a simple and trustworthy scoring system, which is easy to update regularly and to customize for other centers. This is a crucial point when scoring systems are used as predictive models in clinical practice.

Bayesian Approach in Medicine and Health Management

The Bayesian procedure is a particular way of formulating and dealing with these type of problems. It has great promise in putting health-related decision making on a more rational basis, thus making the assumptions more obvious, and making the decisions easier to ex‐ plain and defend [1]. This approach can be used to support the decision-making process in many application fields, as, for example, diagnosis and prognosis [2], risk assessment [3] and health technology assessment [4]. A wide-ranging collection of applications of Bayesian statistics in the biomedical field can be found in thematic books [5-7].

La valutazione dello stato di salute percepita: strumenti psicometrici ed econometrici

Introduzione : per la messa a punto di modelli decisionali in Sanita Pubblica appare sempre piu indispensabile collegare i tradizionali riscontri “oggettivi” (spesso insufficientemente sensibili perche molto dipendenti dal versante dell’offerta) della statistica sanitaria, alle attese/percezioni soggettive degli utenti ed alla valutazione del bene salute. Perseguendo la nostra ricerca dell’ “anello mancante” tra Epidemiologia, Sociologia ed Economia, abbiamo voluto saggiare le correlazioni tra uno strumento tipicamente psicometrico ed uno econometrico di valutazione soggettiva dello stato di salute. Obiettivo : confrontare profili e valori di salute ottenuti con uno strumento psicometrico (SF36) ed uno econometrico (Health Utility Index-HUI). Materiali e metodi: SF36 descrive lo stato di salute attraverso 8 scale: Salute Generale, Attivita Fisica, Ruolo Fisico, Ruolo Emotivo, Attivita Sociali, Dolore Fisico, Vitalita e Salute Mentale. L’HUI produce sia un profilo di salute con 8 scale (Vista, Udito, Parola, Cognitivita, Deambulazione, Uso delle mani, Emotivita e Dolore), sia uno score riassuntivo del livello di salute. I questionari SF36 e HUI sono stati somministrati a 98 studenti dell’Universita di Siena. I punteggi delle otto scale dei due questionari e lo score sono stati confrontanti mediante il Coefficiente di Spearman. Risultati : i punteggi raggiunti nelle 8 scale dell’HUI rientrano nel 95% dei casi nella fascia piu alta (0,9-1), mentre quelli dell’SF36 rientrano nella fascia 90-100 solo per il 36%. Piu discriminante sembra essere lo score (solo il 44% dei valori compresi nella fascia 0,9-1). La correlazione piu forte e quella tra scala Salute Generale e score (r=0,458). La scala Attivita Fisica e correlata con Deambulazione (r=0,382) e Dolore (0,312); Ruolo Fisico con Cognitivita (r=0,425), Deambulazione (r=0,332), Dolore (r=0,383) e score (r= 0,347); Ruolo Emotivo con Cognitivita (r=0,398); Attivita Sociali con Cognitivita (r=0,324); Vitalita con Cognitivita (r=0,382), Emotivita (r=0,432) e score (r=0,438); Salute Mentale con Cognitivita (r=0,426) e score (r=0,383). Conclusioni : l’HUI appare, alle risultanze preliminari, meno discriminante rispetto all’ SF36. Le scale dell’Area Fisica dei due strumenti sembrano correlare, cosi come quelle dell’Area Emotiva. La scala “Salute Generale” e correlata con lo score.