A risk score system for identification of patients with upper-GI bleeding suitable for outpatient management

A risk score system for identification of patients with upper-GI bleeding suitable for outpatient management
of 10
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A risk score system for identification of patients withupper-GI bleeding suitable for outpatient management Pedro Almela, MD, PhD, Adolfo Benages, MD, PhD, Salvador Peiro´ , MD, PhD, Ramo´ n An ˜ o´ n, MD, PhD,Miguel Minguez Pe´ rez, MD, PhD, Andre´ s Pen ˜ a, MD, PhD, Isabel Pascual, MD, PhD,Francisco Mora, MD, PhD Valencia, Spain  Background:   The aim of this study was to develop a risk score system for identification of patientswith upper-GI hemorrhage who are suitable for outpatient management. Methods:   From a prospective cohort of 983 consecutive patients with upper-GI hemorrhage notassociated with portal hypertension, 581 cases that did not meet pre-established criteria foradmission were selected, and a logistic regression analysis was performed to identify factorsassociated with two adverse outcomes: recurrent bleeding and/or the need for emergency surgery.The risk score system was developed by using the beta coefficients of the logistic model, and itsperformance was evaluated. The results of this model were combined with pre-established criteriafor admission to build a simplified scoring system for identification of patients who can bemanaged safely on an outpatient basis. Results:   Chronic alcoholism, active malignancy, prior upper digestive tract surgery, wastingsyndrome, hemodynamic compromise, duodenal ulcer as the cause of upper-GI hemorrhage, andhemorrhage of unknown cause were independently associated with a greater risk of unfavorableoutcomes in the group that did not meet pre-established criteria for admission. The logistic modelshowed a high capacity for discrimination (C statistic: 0.87) and good calibration ( p   value forHosmer-Lemeshow goodness-of-fit test, 0.62), with a sensitivity of 100% and specificity of 64%.The simplified score had a sensitivity of 100% and specificity of 29% for adverse outcomes, andsensitivity of 78% and specificity of 38% for mortality. Conclusions:   The score system developed in this study may be helpful in deciding betweenhospitalization and outpatient management for patients with upper-GI hemorrhage, but it remainsto be validated in patient groups other than those used for its development. (Gastrointest Endosc2004;59:772-81.)  Acuteupper-GIhemorrhage(UGIH)isaclinicallysignificant and costly health care problem, with anestimated incidence at 50 to 160 cases per 100,000 adultsperyear. 1,2 In80%to85%ofepisodes, 3-5 UGIHnot associated with portal hypertension is self-limited, and, in some groups, such as young patientswith no concomitant diseases, the mortality rate isless than 1%. 1,6,7  Accumulating evidence suggeststhateffectiveandsafeoutpatientmanagementcouldbe implemented for 20% to 30% of the patients withUGIH admitted to the hospital. 8-16 Efforts are un-derway to develop reliable and valid risk scoringsystems as well as clinical guidelines for the identi-fication of patients at low risk of adverse outcomeswho may be suitable candidates f or outpatient man- agement 3,17-28 or early discharge. 29-31 These scoringsystems vary depending on the severity of thecondition in the cohort of patients from which theyare derived, the outcomes selected as end points(death,recurrentbleeding,surgery,orcompositeendpoints), the risk factors incorporated into the model,andtheweightgiveneachfactor;nosystemhasbeenadopted widely.The aim of this study was to develop a risk scoresystem that can be used to identify low-risk patientswho would be suitable candidates for outpatientmanagement. The scoring system, as presented inthis study, should be regarded as preliminary; it hasnot been prospectively validated in an internal or anexternal patient population. PATIENTS AND METHODS The model developed was based on findings in a pro-spective cohort study of 983 consecutive patients seen in  Received July 30, 2003. For revision December 4, 2003. Accepted February 11, 2004.Current affiliations: Servicio de Gastroenterologı´ a, HospitalClı´ nico Universitario, Universitat de Valencia, Valencia, Spain, Fundacio´ n Instituto de Investigacio´ n en Servicios de Salud,Valencia, Spain.Supported in part by the Fundacio´ n Instituto de Investigacio´ n enServicios de Salud and a grant from the Instituto de Salud Carlos III (C02/03). Reprint requests: Adolfo Benages, MD, Hospital Clı´ nico Uni-versitario, Servicio de Gastroenterologı´ a, Avda. Blasco Iban˜ez 17, 46010, Valencia, Spain.Copyright    2004 by the American Society for Gastrointestinal Endoscopy 0016-5107/$30.00 PII: S0016-5107(04)00362-1 772  GASTROINTESTINAL ENDOSCOPY VOLUME 59, NO. 7, 2004  the emergency department of a university hospital in Valencia, Spain, for UGIH not associated with portalhypertension over a 45-month period (January 1, 1994 toSeptember 30, 1997). The clinical course of the patientswas followed to assess the safety of outpatient manage-ment. Criteria for inclusion, study protocols, patientcharacteristics, and outcomes are fully described else-where. 15 Briefly, UGIH was diagnosed if hematemesis,melena, blood in the nasogastric tube, or bleeding at EGDwas observed. All patients were managed according toa protocol that included immediate examination andhemodynamic evaluation, a detailed medical history,laboratory tests, and EGD (performed in all cases within24 hours of arrival in the emergency department).Pre-established criteria for hospitalization included thepresence of severe or massive hemodynamic compromise;stigmataofactivebleeding,avisiblevessel,oradherentclotat the EGD; circumstances that made the performanceof EGD impossible; and any comorbid condition listedin Table 1. Mild to moderate hemodynamic compromise,a stigmata of black base at endoscopy, absence of anendoscopic lesion, and some clinical conditions (Table 1)were relative criteria for hospitalization, with the decisionfor admission left to the clinical judgment of the gastroen-terologistoncallintheemergencydepartment.Definitionsused in the study are given in the Appendix. The studywas observational in design, and the patients were notrandomized; participation in the study did not entail anytest or treatment procedure that differed from those of theestablished protocols for management of patients withUGIH in our hospital.Ofthe983patientsincluded,387(39.4%)wereadmittedbecausetheymetatleastoneofthepre-establishedcriteriafor hospitalization. The remaining 596 were either admit-ted (380, 38.7%) or discharged from the emergency de-partment and referred for outpatient follow-up (216,22.0%), based on the judgment of the attending gastroen-terologist. In the latter group, 15 patients either did notreturn for examination or could not be contacted, leaving581 patients for analysis. Outcomes measured were re-currentbleedingduringthe10daysaftertheinitialepisodeandtheneedforemergencysurgeryduringthefollowing15days. Mortality for any cause during the 30 days after theinitial episode of UGIH also was recorded. A logistic regression analysis was performed by usingonly the 581 cases in which the decision to admit or todischarge was based on the clinical judgment of theattending physician. Patients who met any of the pre-established admission criteria were hospitalized automat-ically, and these criteria could be used for inclusion in thequalitative aspects of the risk score system. The outcomeused was the composite end point of recurrent bleedingand/or the need for emergency surgery. A full model thenwas constructed with all clinically relevant variablesestablished in the medical publications as factors associ-ated with poor outcome, and the factors that werestatistically significant in the bivariable analysis in a pre-vious study. 15  All variables considered for inclusion in themodel were introduced as dichotomous variables. A back-ward stepwise method was used to exclude from the fullmodel the variables that were not significant. The Cstatistic (area under the receiver operating characteristiccurve) was used to evaluate the capacity of the model todiscriminate between patients who had good and pooroutcomes (a C statistic of 0.50 indicates the model does notdiscriminate better than random chance; a C statistic of 1indicatesthatitdiscriminatesperfectly) 32,33 ;patientsweregrouped in deciles according to the probability of theirexperiencing adverse outcomes as foreseen by the model,andtheHosmer-Lemeshowgoodness-of-fittestwasappliedto examine whether there were differences between thecases actually observed and those expected by themodel. 34 The beta coefficients of the variables associated withadverse outcomes in the logistic regression model wereusedtoderivetheexpectedprobabilityofadverseoutcomesfor each patient. This can be calculated for each individualpatient with the formula e logit  /(1 + e logit ), where the logit isequaltothesumoftheconstanttermobtainedandthebetacoefficients of each of the factors present. The predictioncapacity of the model also was assessed by comparing Table 1. Absolute and relative pre-established criteria supporting, indication for hospitalization  Absolute criteria Comorbidity dependent Severe heart failureRecent acute myocardial infarctionRecent strokeSevere coagulopathy (INR $ 1.5)Socio-family Unsuitable socio-family conditionsUGIH dependent Hemodynamic compromise severe or massive Active bleeding, visible vessel, or adherent clot (endoscopy)Impossibility of performing endoscopyRelative criteria Comorbidity dependent Mild/moderate heart failureMild/moderate respiratory failureChronic kidney failureOld myocardial infarction or strokeModerate coagulopathy (INR<1.5)Psychologic alterationsUGIH dependent Black rests (endoscopy)Bleeding of unknown cause (endoscopy)  INR , International normalization ratio;  UGIH,  upper GI hemorrhage. (See definitions in Appendix.) VOLUME 59, NO. 7, 2004 GASTROINTESTINAL ENDOSCOPY   773 Upper bleeding outpatient management: risk score system P Almela, A Benages, S Peiro´  , et al.  expected results with actual results and by using a cut-off pointthatmaximizedsensitivity.Finally,asimplifiedscorewas created by combining the results generated with thismodel and the pre-established criteria for admission. Allcalculationsweremadewithastatisticalsoftwareprogram(STATA; Stata Corp., College Station, Tex.). RESULTS Thegroupofpatientswhometnoneoftheexplicit,predefined criteria for hospitalization had the fol-lowing characteristics when compared with thegroup that was automatically hospitalized (Table 2): Table 2. Clinical, laboratory, and endoscopic characteristics of the patients with and without absolutepre-established criteria for hospitalization Patients without absolutepre-established criteria forhospitalization (n = 581)Patients with absolutepre-established criteria forhospitalization (n = 387)n % or mean 95% CI n % or mean 95% CI Age Y (mean) 581 61.29 59.76 62.83 387 62.44 60.59 64.30Sex Men 374 64.37 60.32 68.27 288 74.42 69.76 78.69Women 207 35.63 31.72 39.67 99 25.58 21.30 30.23Smoking Non-smokers 317 54.56 50.41 58.66 99 25.58 21.30 30.23Smokers 129 22.20 18.88 25.80 198 51.16 46.06 56.24Ex-smokers 135 23.24 19.85 26.88 90 23.26 19.13 27.79 Alcohol Alcoholics 54 9.29 7.05 11.95 51 13.17 9.97 16.96Nonalcoholics 527 90.71 88.04 92.94 336 86.82 83.03 90.02NSAIDs Yes 302 51.98 43.89 52.17 214 55.29 50.18 60.32No 279 48.02 47.82 56.10 173 44.70 39.67 49.81Comorbidity Yes 332 57.14 53.00 61.20 221 57.11 52.00 62.09No 249 42.86 38.79 46.99 166 42.89 37.90 47.99History of GI pathology Yes 308 53.01 48.94 57.08 206 53.23 48.12 58.28No 273 46.99 42.51 51.05 181 46.77 41.71 51.87GI symptoms Yes 308 53.01 48.94 57.08 188 48.58 43.49 53.68No 273 46.99 42.91 51.05 199 51.42 46.31 56.50Manifestationof UGIHMelena 307 52.84 48.68 56.96 182 47.03 41.96 52.13Hematemesis 180 30.98 27.23 34.91 98 25.32 21.06 29.96Both 94 16.18 13.27 19.43 107 27.65 23.24 32.39Time since onset Hr (mean) 581 47.17 42.09 52.25 387 28.38 23.16 33.60HemodynamiccompromiseMild 223 38.38 34.40 42.47 70 18.09 14.38 22.29Moderate 358 61.62 57.52 65.59 274 70.80 65.99 75.28Severe/massive 0 0.00 — — 43 11.11 8.15 14.67Hb g/dL (mean) 581 11.02 10.79 11.25 387 10.19 9.91 10.47Urea mg/dL (mean) 581 69.72 65.44 74.01 387 90.29 83.63 96.95Cause of UGIH Duodenal ulcer 193 33.22 29.39 37.21 169 43.67 38.66 48.77Gastric ulcer 130 22.38 19.04 25.98 107 27.65 23.24 32.39Erosive duodenitis 55 9.47 7.21 12.14 2 0.52 0.06 1.85Erosive gastritis 90 15.49 12.64 18.69 10 2.58 1.24 4.70Mallory-Weisssyndrome28 4.82 3.22 6.89 15 3.88 2.18 6.31Peptic esophagitis 53 9.12 6.90 11.76 13 3.36 1.80 5.67Malignancy 8 1.38 0.59 2.69 10 2.58 1.24 4.70Undetermined 61 10.50 8.12 13.28 2 0.52 0.06 1.85Others 3 0.52 0.10 1.50 10 2.58 1.24 4.70No endoscopy 0 0.00 — — 51 13.18 9.97 16.96Endoscopic stigmataof recent bleeding Active bleeding/ spurting0 0.00 — — 8 2.07 0.89 4.03 Active bleeding/oozing 0 0.00 — — 52 13.44 10.20 17.24 Visible vessel 0 0.00 — — 55 14.21 10.88 18.09 Adherent clot 0 0.00 — — 214 55.30 50.18 60.32Black base 231 39.76 35.75 43.86 4 1.03 0.28 2.62None 289 49.74 45.60 53.88 1 0.26 0.01 1.43Undetermined orno endoscopy61 10.50 8.12 13.28 53 13.70 10.42 17.52 95% CI,  Confidence interval, 95% estimated by the binomial method;  NSAID,  non-steroid anti-inflammatory drug,  UGIH  , upper GIhemorrhage. (See definitions in Appendix.) 774  GASTROINTESTINAL ENDOSCOPY VOLUME 59, NO. 7, 2004 P Almela, A Benages, S Peiro´  , et al. Upper bleeding outpatient management: risk score system  there were more women than men; more non-smokers than smokers; more time had elapsedbetween the onset of hemorrhage and arrival at thehospital; fewer patients presented with both hema-temesis and melena; hemodynamic compromise wasless severe, Hb level was higher and serum ureanitrogen lower; and more patients had erosive le-sions, peptic esophagitis, and bleeding of unknowncause. Stigmata of recent bleeding, when found, andincompliance withthestudy criteria, wasblackbasehemorrhagicerosions.Onlyonepatientinthisgroup(0.17%) was treated endoscopically (Table 3); 28.4%were given blood transfusions. Of the total, 34.6%were discharged from the emergency departmentand, for those who were hospitalized, the averagelength of stay was similar to that for the group of patients admitted automatically because they metexplicit,predefinedcriteriafor admission. Recurrentbleeding (11 patients, 1.89%), emergency surgery(7 patients, 1.20%), or a combination of both (14patients, 2.41%) were significantly less frequent inthis group than in the group with pre-establishedadmission criteria, but mortality at 30 days was notsignificantly different between the two groups (2.1%vs. 2.8%).Inthegroupwithoutpredefinedadmissioncriteria(Table4),thelogisticregressionanalysisshowedthatthe following factors were associated independentlywithagreaterriskofrecurrentbleedingand/oremer-gency surgery: history of chronic alcoholism (oddsratio [OR] 3.26: 95% confidence interval [CI][0.81-13.19]) and of active malignancy in any organ (OR6.25: CI[1.09-35.99]), previous upper digestive tractsurgery (OR 11.99: CI[2.06-69.93]), associated wast-ingsyndrome(OR8.79CI[1.57-49.28]),moderatehe-modynamic compromise (OR 10.64: CI[1.25-90.99]),duodenal ulcer as the cause of UGIH (OR 3.26:CI[0.87-12.22]), and bleeding of unknown cause atEGD (OR 4.39: CI[0.86-22.45]). The discriminationcapacity of the model (C statistic 0.87) and calibra-tion (  p  value for Hosmer-Lemeshow goodness-of-fittest = 0.62)weregood.Thelogisticmodelwasusedtoderive the expected probability of adverse outcomesfor each patient. Probability values in the mostextreme cases, where either none or all of the riskfactors were present (with the exception of duodenal Table 3. Management and outcomes of patients with and without absolute pre-established criteria forhospitalization Patients without absolutepre-established criteria forhospitalization (n = 581)Patients with absolutepre-established criteria forhospitalization (n = 387)n % or mean 95% CI n % or mean 95% CIProcess Endoscopic treatment (%) 1 0.17 0.00 0.95 67 17.31 13.67 21.45Transfusions (%) 165 28.40 24.76 32.25 206 53.23 48.23 58.22Outpatient management (%) 201 34.60 30.72 38.62 0 0.00 — —Length of stay (if admitted) 380 4.35 3.87 4.84 387 4.81 4.44 5.19Outcomes Recurrent bleeding (%) 11 1.89 0.94 3.36 46 11.89 8.83 15.53Surgery (%) 7 1.20 0.48 2.46 37 9.56 6.82 12.93Recurrent bleeding or surgery (%) 14 2.41 1.32 4.01 57 14.73 11.34 18.65Death (%) 12 2.07 1.07 3.57 11 2.84 1.42 5.02 95% CI  , Confidence interval, 95% estimated by the binomial method. Table 4. Factors associated with adverse evolution (recurrent bleeding and/or surgery) in patientswithout pre-established absolute criteria for admission Coefficient  p  Multivariate odds ratio 95% CIMedical history Chronic alcoholism 1.183047 0.097 3.26 0.81-13.19Malignancy (any) 1.832802 0.040 6.25 1.09-35.99Upper digestive surgery 2.484597 0.006 11.99 2.06-69.93Wasting syndrome 2.173353 0.013 8.79 1.57-49.28Hemodynamic compromise Moderate 2.365175 0.031 10.64 1.25-90.99Endoscopy Duodenal ulcer (cause) 1.183553 0.079 3.26 0.87-12.22Undetermined 1.480833 0.075 4.39 0.86-22.45Constant 7.016262n = 581;  x  2 = 25.68;  p (  x  2 )<.001; R 2 = 0.19; C statistic = 0.8687;  x  2 Hosmer-Lemeshow = 2.63;  p (  x  2 ) = 0.6207. CI  , Confidence interval. VOLUME 59, NO. 7, 2004 GASTROINTESTINAL ENDOSCOPY   775 Upper bleeding outpatient management: risk score system P Almela, A Benages, S Peiro´  , et al.  ulcer as the cause of the UGIH, and bleeding of unknown cause at EGD, factors that logically wouldnevercoincide),were,respectively,closeto0(0.0009)and 1 (0.98). The prediction capacity of the model(Table 5) was evaluated by comparing the resultsactually observed with the results that had beenpredicted on the basis of the probability of eachpatient experiencing adverse outcomes, as derivedfrom the logistic model, and by using 0.03 as the cut-off point, which was chosen because forecastedobjectives indicated it would maximize sensitivity.With this cut-off point, the logistic model identified365 patients (of 581) predicted to have favorableoutcomes who were good candidates for outpatientfollow-up (sensitivity: 100%), even though 202 pa-tients whose clinical course ultimately turned out tobe favorable were hospitalized (specificity: 64%).Bycombiningthepredefinedcriteriaforadmissionwith the results generated with the model, a simpli-fied risk assessment score was developed for use asa supplementary instrument in deciding betweenhospitalization or outpatient management (Table 6).In this score, 3 points are given to each of the pre-established admission criteria, two points to the riskfactors with beta coefficients of more than two in thelogistic model (history of upper digestive surgery,associated wasting syndrome, and moderate hemo-dynamic compromise), because the presence of anyotherriskfactorwouldimplyaprobabilityofadverseclinical course higher than the cut-off point of 0.03,and one point for risk factors with a beta coefficientof less than two, because 3 of these factors mustbe present for the probability of an adverse clinicalcourse to be greater than 0.03. A patient score iscalculated by adding the points given for each factor. A score of less than 3 would indicate that there is noriskofanadverseoutcomeandthatthepatientcouldbe referred for outpatient follow-up.The results generated after applying the scale toa series of patients are shown in Table 7. Thesefindingssuggestthatslightlymorethanathirdofthe Table 5. Prediction values of the prognosticmodel for adverse evolution in patients withoutpre-established criteria for admission  Adverse evolution observed Yes No Total Adverse evolutionexpected Yes 14 202 216No 0 365 365Total 14 567 581Cut-off point: 0.03; sensitivity = 100% and specificity: 64.37%for a prevalence of adverse events of 2.4%; positive predictionvalue: 6.48%; negative prediction value: 100%. Table 6. Simplified score to support decisionof admission or outpatient management forpatients with UGIH not associated with portalhypertension WeightMedical history Severe heart failure* 3Recent acute myocardialinfarction*3Recent stroke* 3Severe coagulopathy(INR $ 1.5)*3Unsuitable socio-familyconditions*3History of upper digestivesurgery2Wasting syndrome 2 Active malignancy 1Chronic alcoholism 1HemodynamiccompromiseSevere or massive* 3Moderate 2Endoscopy Impossibility of performingupper endoscopy*3 Active bleeding, visible vessel,or adherent clot*3Duodenal ulcer(cause of bleeding)1Bleeding of unknown cause 1The score is calculated by adding the points for each factorpresented by the patient. Score of   $ 3 points indicate the needfor hospitalization; scores of <3 points indicate suitability of outpatient management. UGIH  , Upper GI hemorrhage;  INR , internationalnormalization ratio. (see Appendix for definitions.)*Pre-established criteria for admission. Table 7. Sensitivity and specificity of thesimplified score in view of adverse evolution(recurrent bleeding or surgery) and mortalityfor all the patients Recurrent bleeding surgery Yes No TotalDecision to admit Yes 71 532 603No 0 365 365Total 71 897 968Sensitivity = 100% and specificity = 29% for a prevalenceof adverse events of 7.33%; positive prediction value: 12%;negative prediction value: 100%.Overall mortality Yes No TotalDecision to admit Yes 18 585 603No 5 360 365Total 23 945 968Sensitivity = 78% and specificity:38% fora prevalence of deathof 2.38%; positive predictive value: 3%; negative predictivevalue: 99%. 776  GASTROINTESTINAL ENDOSCOPY VOLUME 59, NO. 7, 2004 P Almela, A Benages, S Peiro´  , et al. Upper bleeding outpatient management: risk score system
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks