Digestive and Liver Disease 42 (2010) 624–628 Contents lists available at ScienceDirect Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld Digestive Endoscopy How to predict a high rate of inappropriateness for upper endoscopy in an endoscopic centre? L. Buri a , G. Bersani b , C. Hassan c,∗ , M. Anti d , M.A. Bianco e , L. Cipolletta e , E. Di Giulio f , G. Di Matteo g , L. Familiari h , L. Ficano i , P. Loriga j , S. Morini c , V. Pietropaolo k , A. Zambelli l , E. Grossi m , M. Intraligi n , F. Tessari o , M. Buscema n , the SIED Appropriateness Working Group1 a Gastroenterology and Digestive Endoscopy Unit, Cattinara Hospital, Trieste, Italy Gastrointestinal Endoscopy Service, Malatesta, Cesena, Italy c Gastroenterology, Nuovo Regina Margherita, Rome, Italy d Gastroenterology Unit, Belcolle Hospital, Viterbo, Italy e Division of Gastroenterology and Digestive Endoscopy ASL NA5-Hospital Agostino Maresca, Torre del Greco, Italy f Digestive and Liver Disease Unit, Second Medical School, University “La Sapienza”, Sant’Andrea Hospital, Rome, Italy g Gastroenterology Unit, “Saverio De Bellis” Hospital, Castellana Grotte, Bari, Italy h Gastroenterology, Policlinico G Martino, Messina, Italy i Surgery and Oncology Department, “Università di Palermo”, Palermo, Italy j Endoscopy Unit, SS Trinità Hospital, Cagliari, Italy k Gastroenterology Unit, Policlinico La Sapienza, Rome, Italy l Gastroenterology Unit, Maggiore Hospital, Crema, Italy m Bracco Imaging S.p.A., Medical Affairs Europe, Milan, Italy n Semeion Research Centre for Sciences of Communication, Rome, Italy o Idea99, Padova, Italy b a r t i c l e i n f o Article history: Received 17 November 2009 Accepted 15 February 2010 Available online 21 March 2010 Keywords: Appropriateness Upper endoscopy a b s t r a c t Background: Inappropriateness of upper endoscopy (EGD) indication causes decreased diagnostic yield. Our aim of was to identify predictors of appropriateness rate for EGD among endoscopic centres. Methods: A post-hoc analysis of two multicentre cross-sectional studies, including 6270 and 8252 patients consecutively referred to EGD in 44 (group A) and 55 (group B) endoscopic Italian centres in 2003 and 2007, respectively, was performed. A multiple forward stepwise regression was applied to group A, and independently validated in group B. A <70% threshold was adopted to deﬁne inadequate appropriateness rate clustered by centre. Results: discrete variability of clustered appropriateness rates among the 44 group A centres was observed (median: 77%; range: 41–97%), and a <70% appropriateness rate was detected in 11 (25%). Independent predictors of centre appropriateness rate were: percentage of patients referred by general practitioners (GP), rate of urgent examinations, prevalence of relevant diseases, and academic status. For group B, sensitivity, speciﬁcity and area under receiver operating characteristic curve of the model in detecting centres with a <70% appropriateness rate were 54%, 93% and 0.72, respectively. Conclusions: A simple predictive rule, based on rate of patients referred by GPs, rate of urgent examinations, prevalence of relevant diseases and academic status, identiﬁed a small subset of centres characterised by a high rate of inappropriateness. These centres may be presumed to obtain the largest beneﬁt from targeted educational programs. © 2010 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved. 1. Introduction Open-access upper endoscopy (EGD) is the most widespread access system to endoscopic examinations in the health systems ∗ Corresponding author at: Ospedale Nuovo Regina Margherita, Via Morosini 30, Rome, Italy. Tel.: +39 06 58446608; fax: +39 06 58446533. E-mail address: [email protected] (C. Hassan). 1 See Appendix A for the list of members. of the western world . This type of service allows physicians to directly schedule elective, common endoscopic procedures for their patients without prior consultation. Unfortunately, this has also resulted in a considerable increase in both overall cost and waiting lists for EGD [2,3]. In order to optimise the use of ﬁnite resources in an open-access system, ofﬁcial guidelines for the appropriate use of EGD have been proposed by the American Society for Gastrointestinal Endoscopy (ASGE) and by the European Panel on the appropriateness of Gastrointestinal Endoscopy (EPAGE) [4,5]. Previous studies based on 1590-8658/$36.00 © 2010 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.dld.2010.02.012 L. Buri et al. / Digestive and Liver Disease 42 (2010) 624–628 these guidelines have generally shown a substantial rate of inappropriate EGD indications, which in turn has been associated with a markedly decreased diagnostic yield for both relevant ﬁndings and cancer, and with an unfavourable cost-effectiveness proﬁle as compared to appropriate procedures [5–13]. The rate of inappropriate indications widely ranged between 10% and 40% in different studies , and the few multicentre studies carried out also showed a moderate degree of variability among different centres [6,14]. To identify the endoscopic centres associated with a higher inappropriateness rate of EGD appears of major importance, since it has been clearly shown that educational programmes directed towards the referring physicians result into a substantial reduction of the inappropriateness rate [15,16]. A widespread implementation of such programmes, however, is unlikely, because these initiatives are time-consuming and costly, also being in competition with similar campaigns for several other medical procedures [17,18]. Therefore, to identify simple variables related with the endoscopic setting able to predict the inappropriateness rate of EGDs would allow to target educational interventions only on a relatively small number of centres. The aim of this study, conceived by the Italian Society of Digestive Endoscopy (SIED), was to identify simple factors able to predict the appropriateness rate for EGD aggregated by centres, in order to detect those centres with a high inappropriateness rate that may need interventional programmes. 2. Materials and methods Two cross-sectional, prospective, multicentre study involving 44 (group A, 6270 patients) and 55 (group B, 8252 patients) open-access Endoscopy Units uniformly distributed throughout the country (SIED APPROPRIATENESS PROJECT) were separately performed in 2003 and 2007, respectively [6,19]. In both studies, the participating centres – which represent 6–8% of all the Italian Endoscopic Units censed by SIED – were also selected to produce a mix of small and large practices, as well as of academic and not academic institutions, in order to be representative of the Italian endoscopy setting. Only a minority of the centres (7 centres) participated in both studies. However, since no educational initiative was promoted between the two studies, we did not exclude these centres in both groups. According to the same study protocol of the two studies, all patients referred to the participating centres for openaccess EGD during 1 month were prospectively enrolled. EGDs were performed according to predeﬁned weekly schedules, the referring physicians being unaware of the purpose of the study. Before endoscopy, the following variables were systematically collected: sex and age of the patient; appropriateness of the EGD indication according to ASGE guidelines; referral physician (primary care physician (GP) or specialist); setting (inpatient or outpatient); EGD timing (urgent or elective); ﬁrst EGD or control/follow-up examination. At endoscopy, the presence of relevant ﬁndings (any involving a modiﬁcation of the treatment and/or management of the patient) was assigned according to predeﬁned diagnoses listed in Appendix 1. The clinical association between the appropriateness of the indication according to ASGE guidelines and the detection of relevant ﬁndings at endoscopy in single patients has been already reported in previous studies for group A  and B . In order to identify eventual predictive variables for the inappropriateness rate of the 44 and 55 individual centres of group A and B, respectively, we aggregated both the appropriateness rate (i.e. all appropriate EGD referrals/(appropriate + inappropriate EGD) for each centre) and the data regarding the following variables for each centre: - 625 rate of patients aged ≥45 years rate of relevant endoscopic diseases detected at endoscopy rate of hospitalised patients rate of patients referred by their GP rate of urgent examinations rate of follow-up examinations type of centre (academic/non-academic) 2.1. Statistical analysis In order to evaluate whether there was a correlation between the rate of appropriateness in the various centres and each of the individual variables in the training sample (group A), the correlation index (Pearson’s r) was calculated. Multiple forward stepwise regression analysis was then used to ﬁnd the minimum set of independent variables that maximises the correlation with the rate of appropriateness aggregated by centre (the dependent variable) in group A; this means to ﬁnd the linear regression equation that combines the independent variables so as to maximise the correlation (measured with R2 , the coefﬁcient of determination) with the dependent variable; the forward selection begins with no predictor in the regression equation, so the independent variable with the highest correlation with the dependent variable is the ﬁrst to be included in the equation; at each subsequent step the variable which most increases the value of R2 is added, while maintaining signiﬁcant F values at 0.05% ; the addition of variables stops when it does not produce a signiﬁcant increase of R2 . The coefﬁcient of determination R2 represents the percentage of variance in the data that is explained by the linear model and therefore the closer it is to 1.0, the greater the representativeness of the linear model adopted. By using the regression model, we computed the predicted value of appropriateness for each of the 44 centres of group A. By comparing the predicted values with the observed values of aggregated appropriateness rates – that served as the reference standard – we measured the accuracy of the model for detecting those centres associated with an inadequate appropriateness rate (i.e. too low). In detail, we adopted 70% as a threshold to deﬁne inadequate the appropriateness rate of a centre, because the mean appropriateness rates in the previous studies was higher than 70% in 6 out of 8 series with an overall median of 74%, suggesting that 70% may be a reliable cut-off to identify a small subset of centres that may need a further intervention . The 55 centres of group B were used for validation of the predictive model, adopting the same 70% threshold. Sensitivity, speciﬁcity, positive and negative predictive value, and area under the receiver operating characteristic curve (AUC), and the 95% conﬁdence intervals, were calculated for both groups . The receiver operating curve represents the relationship between sensitivity and speciﬁcity for the prediction of each of the considered outcomes. For comparison between the two groups, the 2 -test and Mann–Whitney U test were used to compare categorical and continuous variables, respectively. All data analyses were carried out with Excel (Microsoft Corp., Redmond Wash.) and SPSS statistical software. Differences were considered signiﬁcant at a 5% probability level. 3. Results The median rate of the aggregated appropriateness of EGD among the 44 and 55 centres of groups A and B was 77% and 78%, with a considerable variability among the different centres (range: 41–97%, group A; 44–94%, group B). The aggregated appropriate- 626 L. Buri et al. / Digestive and Liver Disease 42 (2010) 624–628 Table 2 Correlation between the aggregated appropriateness rate and the operative variables cumulatively clustered by centre in the group A (training group). Variable Correlation (Pearson’s r) p Rate of patients aged ≥45 years Rate of relevant ﬁndings at EGD Rate of inpatients Rate of referrals by GP Rate of urgent examinations Rate of follow-up EGD Academic status 0.15 0.40 0.41 −0.41 0.35 −0.21 0.15 0.42 0.002 0.0007 0.01 0.02 0.17 0.09 Table 3 Results of the forward multiple regression analysis applied to the training group A (a) and validation group (b). The list of the variables included in the model by the forward stepwise linear multiple regression is shown. Fig. 1. Distribution of the aggregated values of appropriateness rate observed in the 44 centres of the Study A and in the 55 centres of the Study B. The threshold line corresponds to a 70% appropriateness rate. Each circle represents an endoscopic centre, with black circles representing those centres predicted to have a <70% appropriateness rate by the model. Table 1 Operative characteristics of the endoscopy centres participating in the study. Data are reported aggregated by centres, so that the median corresponds to the central value among the 44 and 55 values computed in the individual centres of groups A and B, respectively. Variable Group A Median (range) Group B Median (range) Rate of patients aged ≥45 years Rate of relevant ﬁndings at EGD Rate of referrals by GP Rate of inpatients Rate of urgent examinations Rate of follow-up EGD 73% (60–85%) 44% (24–89%) 62% (29–88%) 28% (0–58%) 8% (0–28%) 35% (2–66%) 73% (56–88%)a 46% (26–81%)a 56% (11–84%)a 25% (7–55%)a 7% (1–16%)a 35% (10–63%)a a No statistically signiﬁcant difference between group A and B, respectively. Coefﬁcient Std. err. (a) Costant % referred by GP in centre Ci % urgent examinations in centre Ci % relevant disease in centre Ci Ci academic centre or not 0.68 −0.24 0.70 0.34 0.07 0.067 0.126 0.264 0.143 0.04 (b) Costant % referred by GP in centre Ci % urgent examinations in centre Ci % relevant disease in centre Ci 0.88 −0.3 0.43 0.19 0.07 0.21 0.28 0.11 t p 7.023 2.454 2.303 2.299 2.01 13.34 2.02 2.05 2.28 0.0001 0.019 0.027 0.027 0.04 0.0001 0.04 0.03 0.028 Forward stepwise multiple regression selected as independently associated with the aggregated appropriateness rate of the individual centres in group A the following 4 variables (Table 3): aggregated percentage of patients referred by their GP, aggregated rate of urgent examinations, aggregated prevalence of relevant diseases and the academic status of a centre. These variables were combined in a predictive model, producing the following regression equation: Appropriateness rate of centre Ci = 0.68 − 0.24 (% referred by GP in centre Ci ) ness rate was lower than the 70% adopted threshold in 11 (25%) centres in group A, and in 11 (20%) in group B (Fig. 1). Regarding the operative characteristics of the different centres in both groups, median values and ranges aggregated by centres are reported in Table 1. Overall, 8 (18%) centres were academic in group A, and 11 (20%) in the group B. The correlation between the appropriateness rate of the individual centres and each variable aggregated by centre, expressed as Pearson’s r, is shown in Table 2 (group A). The correlation with the aggregated appropriateness rate appeared to be higher for the percentage of patients referred by GP in the individual centres, the rate of hospitalised patients, the prevalence of relevant endoscopic diseases, and the percentage of urgent examinations, whilst it was lower for the rate of follow-up examinations, the percentage of patients aged ≥45 years, and the academic proﬁle, all aggregated by centre. + 0.70 (% urgent examinations in centre Ci ) + 0.34 (% relevant disease in centre Ci ) + 0.07(Ci academic centre or not) This model appeared to explain 41% of the variability of the aggregated appropriateness rates among the different centres (R2 : 0.405, F4:36:6.468, p < 0.001), and the corresponding AUC was 0.76. The regression model predicted 9 (16%) centres in group B to have an appropriateness rate lower than the 70% adopted threshold (Fig. 1). In detail, sensitivity, speciﬁcity, and AUC were 54%, 93% and 0.72 (Table 4). There was no statistically signiﬁcant difference between the AUCs (Fig. 2) calculated in the training (group A) and validating samples (group B). The corresponding predicted values of appropriateness rate computed for each of the 55 centres have Table 4 Accuracy values of the predictive model for correctly identifying those centres with an aggregated appropriateness rate lower than the adopted threshold (70%) in the validating group B. The false negatives represent the centres with an observed (aggregated) appropriateness rate below the threshold, but with a predicted value higher than the same threshold. PPV: positive predictive value; NPV: negative predictive value; FP: false positives; TP: true positives. Rate of centres below the threshold All positives (FP + TP) False negatives Sensitivity (95%CI) Speciﬁcity (95%CI) PPV (95%CI) NPV (95%CI) AUC–ROC (95%CI) 20%a 16%b 9%c 54% (25–84%) 93% (85–100%) 67% (36–97%) 89% (80–98%) 0.72 (0.58–0.82) a b c Corresponding to 11 centres. Corresponding to 9 centres. Corresponding to 5 centres. L. Buri et al. / Digestive and Liver Disease 42 (2010) 624–628 Fig. 2. Receiver operating curve (ROC) representing the accuracy of the regression model in identifying those endoscopic centres associated with an aggregated appropriateness rate below the adopted threshold of 70% for the training (group A, squares) and the validation (group B, circles) samples. The ROC represents the relationship between sensitivity and speciﬁcity for the prediction of each of the considered outcomes. No statistically signiﬁcant difference emerged when comparing the area under the ROC of groups A (0.76) and group B (0.72). been compared with the original values – that served as reference criteria to assess the overall accuracy – in Fig. 1. 4. Discussion Our study identiﬁed several variables – rates of patients referred by GP and urgent examinations, prevalence of relevant diseases and academic status – able to predict the aggregated appropriateness rate of individual endoscopic centres. Of note, most of the selected variables – i.e. rate of prescriptions by GP or urgent EGDs – are simple to be assessed and promptly available in most GI units. The only exception may be represented by the prevalence of relevant disease. However, the relatively high prevalence of relevant ﬁndings – nearly 50% – allows an accurate estimate with a relatively small number of patients (i.e. 100–200), also facilitated by a widely accepted deﬁnition of the relevant endoscopic ﬁndings [5–13]. The inverse association between the aggregated appropriateness rate of a single centre and the overall rate of prescription by GPs conﬁrms, at a different level, the same association shown in clinical studies dealing with individual patients [5–13]. Analogously, the direct association between the aggregated appropriateness rate and the rate of urgent examination is well in line with the proposal of prioritisation of urgent examinations with a dedicated triage . Regarding the association with the prevalence of relevant ﬁndings, it may be speculated that physicians practicing in a setting with more prevalent disease are more used to appropriately select patients for EGD, and a similar assumption could be put forward for the academic centres. The importance of this predictive rule is strictly related with the very high variability of the aggregated appropriateness rates among the different centres, widely ranging between 41% and 97% and between 44% and 94% in the two included series, so that it may be useful to discriminate those centres characterised by a very poor performance. In our series, for instance, we showed a clearly inadequate aggregated appropriateness rate – lower than 70% – in 20–25% of the centres. The importance of predicting a poor appropriateness rate in individual centres is also strengthened by the possibility to effec- 627 tively improve it with dedicated interventions. A 1-day training course to all GPs referring to an Italian GI Unit reduced the pre-interventional inappropriateness rate from 23% to 7%, also decreasing the waiting lists by 15% . Similarly, when referring GPs received training with respect to indications during the ﬁrst 2 years of the program, the rate of inappropriate referrals for EGD was only 3% . When taking into consideration the very low prevalence of relevant ﬁndings and cancer in inappropriate procedures, and the unfavourable cost-effectiveness proﬁle of inappropriate EGDs [5–13], the possibility to target educational interventions only to those centres that may gain a substantial beneﬁt appears relevant. Our analysis showed that, assuming as acceptable an aggregated appropriateness rate ≥70%, only 16–20% of all the endoscopic centres would be predicted to deserve a dedicated program. This would be enough to substantially reduce the poor-performing centres from 20–25% to 9%. To focus health interventions only on a relatively small number of centres would appear as a more realistic and practical approach than a widespread implementation of the educational programs. Not only the number of endoscopic centres in any western country is countless, but also each endoscopic centre has several referring physicians who would need to be addressed by the educational initiative. It is extremely unlikely that, in a period of resource and budget constraint, a widespread implementation of educational programs only to improve the appropriateness rate of EGD will take place. On the other hand, to limit these programmes to 16–20% of the centres sounds as a more rational and reasonable health strategy. According to our analysis, the impact of any educational program would be superior, when addressing GPs in a setting of low prevalence of relevant ﬁndings. There are limitations to the present analysis. Although we aggregated all the available variables to predict the appropriateness rate of individual centres, we failed to explain a substantial part of the study variability. This may be related with the existence of other variables that are difﬁcult to be identiﬁed, or with the complex social and psychological dynamics involving the patient and the prescribing doctor, that may not be explained by means of a linear model. For instance, we cannot exclude that other characteristics of the centres, such as the length of the waiting lists, diverse availability of radiology or breath test for H. pylori infection, and different costs among the competing strategies, may have some inﬂuence on appropriateness, suggesting the need for further research in this ﬁeld. However, such uncertainty appeared to affect more the sensitivity than the speciﬁcity at the adopted thresholds, marginalising the risk of a waste of resources, when implementing educational programmes. Secondly, the degree of correlation among the operative variables and the aggregated appropriateness was rather poor. This mirrors the results from clinical studies in which no single variable appeared to be strongly related with the appropriateness of the request. For instance, endoscopic relevant ﬁndings were detected in a substantial rate of inappropriate or follow-up EGD [5–12]. Similarly, although the rate of inappropriateness among GPs’ was signiﬁcantly higher than that among specialists, more than half of the GP’s prescriptions were still appropriate [5–12]. Thirdly, It could also be argued that some of the factors identiﬁed by the present analysis – such as GP prescription or follow-up examination – were already shown to be clinical predictors of inappropriateness for single patients [5–12]. However, this is the ﬁrst study, as far as we know, that showed that some of these variables are also effective in discriminating among individual centres, when aggregated. This means that ASGE guidelines, although originally intended only to be a help in the clinical decision regarding an individual patient, should not be restricted to a clinical use, whilst they may also be applied as a health policy indicator. Fourthly, despite no ofﬁcial initiative by SIED was concealed after the publication of the ﬁrst study, we cannot exclude that the dissemination of such 628 L. Buri et al. / Digestive and Liver Disease 42 (2010) 624–628 information may have changed physician attitude to prescribe EGD. However, the similar level of inappropriateness observed between the two studies seems to marginalise this possibility [6,19]. Finally, the suboptimal sensitivity of the predictive rule is likely to misclassify some centres with a high rate of inappropriateness as centres that do not need further educational initiatives. However, this risk would appear to be compensated by a substantial reduction of the implementation cost to only a 16–20% of the centres. In conclusion, we showed that simple aggregated variables, related with the different settings in which GI-units operate, may be able to predict the appropriateness rate of individual endoscopic centres, and, especially, to identify a small subset of centres with a high rate of inappropriateness that require further intervention. This should reduce wasting of health and economic resources when implementing dedicated educational programmes. Conﬂict of interest Enzo Grossi is an employee for BRACCO SPA. Appendix A. SIED Appropriateness Working Group Costa G (Ospedale di Tortona – ASL 20, Piemonte); Grassini M, Niola P (Ospedale “Cardinal Massaia”, Asti); Del Piano M, Carmagnola S (Ospedale Maggiore della Carità, Novara); Allegretti A, Vallarino E (E.O. Ospedali Galliera, Genova); Leoci C, Popovic A (Ospedale di Manerbio – A.O. Desenzano G., Brescia); Snider L, Bellini O (Azienda Ospedaliera “S. Anna”, Como); Manes G (Azienda Ospedaliera “L. Sacco”, Milano); Lancini GP, Cestari R (Spedali Civili di Brescia, Brescia); Centenaro R, Boni F (Ospedale Civile di Melegnano, Milano); Gebbia C, Iollo P (Ospedale “Città di Sesto S. Giovanni”, Milano); Bertozzo A, Piazzi L (Ospedale Civile, Bolzano); Bottona E, Pantalena M (Ospedale “Cazzavillan”, Arzignano, Vicenza); Battaglia L, Marino S (Ospedale Piove di Sacco ULSS 14 – Chioggia, Padova); Ederle A, Ntakirutimana E (Ospedale “G. Fracastoro” – S. Bonifacio – ULSS 20, Verona); Marcon V (Ospedale Civile Agordo, Belluno); Marin R, Cervellin E (Ospedale Di Dolo, Venezia); Okolicsanyi L, Monica F (Ospedale Regionale – ULSS 9, Treviso); Cappiello R, Sablich R (Ospedale S. Maria degli Angeli, Pordenone); Simeth C (Ospedale Cattinara, Trieste); Zilli M (Ospedale Santa Maria della Misericordia, Udine); Trande P, De Martinis E (Nuovo Ospedale Civile S. Agostino Estense, Modena); Merighi A, Boarino V (Azienda Ospedaliera Universitaria Policlinico, Modena); Cortini C, Maltoni S (Ospedale G.B. Morgagni, Forli’); Rossi A (Casa di Cura “Malatesta – Novello”, Cesena); Frosini G, Longobardi L (Azienda Ospedaliera Universitaria Senese “S. Maria delle Scotte”, Siena); Macri’ G, Dabizzi E (Azienda Ospedaliera Universitaria “Careggi”, Firenze); Pincione F, Widmayer C (ASL 1 – Ospedale Civico, Massa); Tatali M (AUSL Marche – Zona Territoriale 2 Urbino, Pesaro); Ferrini G (Ospedale “F. Renzetti”, Chieti); Neri M, Laterza F (Ospedale Clinicizzato SS. Annunziata, Chieti); Spadaccini A, Silla M (Ospedale “San Pio da Pietrelcina”, Vasto, Chieti); Pastorelli A (Ospedale “Belcolle”, Viterbo); Spada C, Costamagna G. Petruzziello L (Università Cattolica del Sacro Cuore “Policlinico A. Gemelli”, Roma); Gabbrielli A, Di Matteo F (Università Campus “Bio Medico”, Roma); Zullo A (Ospedale “Nuovo Regina Margherita”, Roma); Onorato M (Policlinico Umberto I, Roma); Stroppa I, Andrei F (Policlinico Tor Vergata, Roma); Grimaldi E, De Filippo FR (Ospedale Civile – A.O. Caserta, Caserta); Di Giorgio P, Giannattasio F (Ospedale “S.Maria di Loreto – Mare”, Napoli); Piscopo R (Ospedale Evangelico “Villa Betania”, Napoli); Ingrosso M, Spera G (Università Cattolica del Sacro Cuore, Campobasso); Fregola G, Quatraro F (Ente Ecclesiastico Osp. Reg. “F.Miulli”, Bari); De Maio G, Corazza L (Ospedale “Madonna delle Grazie”, Matera); Giglio A, Rodino’ S (Ospedale “Pugliese – Ciaccio”, Catanzaro); Iaquinta L. (Ospedale di San Giovanni in Fiore, Cosenza); Ciliberto E, Cavaliere C (Ospedale S. Giovanni di Dio, Crotone); Fatta MF, Stiriti A., Naim G (Ospedale Civile, Scilla, Reggio Calabria); Assenza G (Ospedale “E. Muscatello”, Siracusa); Brancato FP (Ospedale di Alcamo, Trapani); Tortora A, Giacobbe G (Policlinico Universitario “G. Martino”, Messina); Italy; Cannizzaro R (Centro di riferimento Oncologico, Aviano, Pordenone); Di Napoli A, Recchia S (Ospedale “San Giovanni Bosco”, Torino); Marino M (Osp. Civ. “G.Bernabeo”, Ortona, Chieti). Appendix B. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.dld.2010.02.012. References  Charles RJ, Chak A, Cooper GS, et al. Use of open access in GI endoscopy at an academic medical center. Gastrointest Endosc 1999;50:480–5.  Froehlich F, Burnand B, Pache I, et al. Overuse of upper gastrointestinal endoscopy in a country with open-access endoscopy: a prospective study in primary care. Gastrointest Endosc 1997;45:13–9.  Caselli M, Parente F, Palli D, et al. Guidelines on the diagnosis and treatment of Helicobacter pylori infection. Dig Liver Dis 2001;33:75–80.  American Society for Gastrointestinal Endoscopy. Appropriate use of gastrointestinal endoscopy. Gastrointest Endosc 2000;52:831–7.  Froehlich F, Repond C, Mullhaupt B, et al. Is the diagnostic yield of upper GI endoscopy improved by the use of explicit panel-based appropriateness criteria? Gastrointest Endosc 2000;52:333–41.  Hassan C, Bersani G, Buri L, et al. Appropriateness of upper-GI endoscopy: an Italian survey on behalf of the Italian Society of Digestive Endoscopy. Gastrointest Endosc 2007;65:767–74.  Rossi A, Bersani G, Ricci G, et al. ASGE guidelines for the appropriate use of upper endoscopy: association with endoscopic ﬁndings. Gastrointest Endosc 2002;56:714–9.  Chan YM, Goh KL. Appropriateness and diagnostic yield of EGD: a prospective study in a large Asian hospital. Gastrointest Endosc 2004;59:517–24.  Bersani G, Rossi A, Suzzi A, et al. Comparison between the two systems to evaluate the appropriateness of endoscopy of the upper digestive tract. Am J Gastroenterol 2004;99:2128–35.  Al-Romaih WR, Al-Shehri AM. Appropriateness of upper gastrointestinal endoscopy referrals from primary health care. Ann Saudi Med 2006;26:224–7.  Kaliszan B, Soule JC, Vallot T, et al. Applicability and efﬁcacy of qualifying criteria for an appropriate use of diagnostic upper gastrointestinal endoscopy. Gastroenterol Clin Biol 2006;30:673–80.  Gonvers JJ, Burnand B, Froehlich F, et al. Appropriateness and diagnostic yield of upper gastrointestinal endoscopy in an open-access endoscopy unit. Endoscopy 1996;28:661–6.  Di Giulio E, Hassan C, Pickhardt PJ, et al. Cost-effectiveness of upper gastrointestinal endoscopy according to the appropriateness of the indication. Scand J Gastroenterol 2008:1–8.  Froehlich F, Pache I, Burnand B, et al. Underutilization of upper gastrointestinal endoscopy. Gastroenterology 1997;112:690–7.  Grassini M, Verna C, Battaglia E, et al. Education improves colonoscopy appropriateness. Gastrointest Endosc 2008;67:88–93.  Mahajan RJ, Barthel JS, Marshall JB. Appropriateness of referrals for open-access endoscopy. How do physicians in different medical specialties do? Arch Intern Med 1996;156:2065–9.  Brook RH, Park RE, Chassin MR, et al. Predicting the appropriate use of carotid endarterectomy, upper gastrointestinal endoscopy, and coronary angiography. N Engl J Med 1990;323:1173–7.  Andriulli A, Annese V, Terruzzi V, et al. “Appropriateness” or “prioritization” for GI endoscopic procedures? Gastrointest Endosc 2006;63:1034–6.  Buri L, Hassan C, Bersani G, et al. Appropriateness guidelines and predictive rules to select patients for upper endoscopy: A nationwide, multicenter study. Am J Gastroenterol 2009. PMID: 20029414.  Wonnacot TH, Wonnacot RJ. Multiple regression and multiple regression extensions. In: Regression: A Second Course in Statistics. Wiley New York; 1981. pp. 75–150.  Hanley JA, Mcneil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143:29–36.
© Copyright 2018