published in the English language from dates of inception to January 2007. Patients were considered to have CHB if they had elevated ALT levels and active viral replication. Monotherapy, combination and sequential therapies were included. Among trials that met our inclusion criteria, we abstracted data describ- ing normalization of ALT, HBV DNA, sustained biochemical response, HBeAg seroconversion, histological improvement, drop-outs and adverse events. Intention-to-treat data were com- bined using a random-effects meta-analysis, with missing data considered as treatment failures. Outcomes were expressed as relative risks with 95% confidence intervals. RESULTS: The initial search yielded 2064 references, 127 were excluded due to inadequate blinding, allocation concealment, randomization and reporting of outcomes; 20 studies were included. Trials involved 5573 patients (4121 males, 1309 females), ranging in size from 200–814 patients. Mean age was 40.7. Eleven trials studied HBeAg-positive patients, four trials studied HBeAg-negative patients, and four trials studied both. Due to small numbers of trials for comparison led to pooling of HBeAg-positive and HBeAg-negative studies. No treatment was superior for all outcome measures. Monotherapy was superior to placebo. Com- parisons of single drugs favored treatment with ADF or ENT over LAM or PEG. LAM was superior to PEG with better clinical outcomes and fewer adverse events and patient dropouts. Com- bination and sequential treatments were not superior, however comparisons were limited by our one-year follow-up. CONCLU- SION: Monotherapy with ADF or ENT are the most attractive treatment options within the first year of treatment. Further research on combination and sequential therapies may provide better options but presently insufficient evidence exists to support this approach. PIN3 DATA MINING PHYSICIAN DECISION AND INVESTIGATING TREATMENT OPTIONS OF OSTEOMYELITIS Zahedi H University of Louisville, Louisville, KY, USA OBJECTIVE: The purpose of this study is to investigate treat- ment options of osteomyelitis based on physician decisions recorded in our dataset. METHODS: We want to find the fre- quency of a given input (code) for a variable, or more than one variable in health care data. Using the Thomson MedStat MarketScan data containing all patient claims for 40 million observations, the primary diagnosis code is given for each patient as well as fifteen possible secondary diagnoses. We use SAS Text Miner to demonstrate a simplified method to search these fifteen columns. We use ICD9 and CPT codes to find treatments for osteomyelitis. We also look for sequential treat- ments for recurrence of osteomyelitis. After filtering the data for Osteomyelitis, there are 18,721 observations in inpatients that contain 2661 patients, and 233,001 observations in outpatients with 78,957 patients. RESULTS: The difference between the number of observation and number of distinct patient IDs shows that most patients have a sequence of procedures during their treatment. After sorting the data by procedures, the most frequent (20%) is “Dorsal and dorsolumbar fusion, posterior technique”, second is “Excisional debridement of wound, infec- tion, or burn” (15%), third “Amputation of toe” (9%), and in forth place, “Revision of amputation stump” (7%). In the out- patient data, the most frequent procedure is code 86.59 (Closure of skin and subcutaneous tissue of other sites) with 4021 records out of 8711 records. We found that about 8 % of patients with osteomyelitis from inpatient data and about 0.3% from outpatient data had amputation. CONCLUSION: While amputation does not occur as often as debridement, we want to examine the sequence of treatments to see whether amputation follows a pattern of debridement. PIN4 ANTIMICROBIAL RESISTANCE PREVALENCE OF ENTEROCOCCI FROM BOGOTA, COLOMBIA HOSPITALS 2001–2006 Lemos EV 1 , Einarson TR 2 , Jimenez P 3 , Eslava J 1 , Gualtero S 1 1 National University of Colombia, Bogota, Cundinamarca, Colombia, 2 University ofToronto,Toronto, ON, Canada, 3 Instituto Nacional de Cancerologia, Bogota, Cundinamarca, Colombia OBJECTIVE: To determine antimicrobial resistance profiles for Enterococcus in 14 tertiary-level hospitals in Bogotá, Colombia. METHODS: Time series analyses were performed, as well as descriptive analyses of Enterococcus faecalis and Enterococcus faecium in hospitals belonging to the Bogotá Bacterial Resistance Control Group from January 1, 2001–January 1, 2007. We identified the presence and species of enterococcus according to anatomical site. RESULTS: During that period, a total of 5770 strains of Enterococci faecalis and 1259 of Enterococcus faecium were analyzed. Enterococcus was found in >20% of isolates from blood samples and abdominal fluid in critical care units (SUCI) and in >10% of samples from non-critical care services (SNUCI) in tertiary-level hospitals in Bogotá. Ampicillin was active against Enterococcus faecalis strains in Critical care units (SUCI) and other services (SNUCI), but >50% of E.faecium strains in SNUCI were resistant. Vancomycin-resistant strains occurred in >2% of E.faecalis strains identified in SUCI and >8% in SNUCI. The resistance trend in E.faecium to vancomycin in SNUCI was towards the low in the six years studied. CONCLUSION: This study confirms the worldwide trend towards an increase in infec- tions due to Enterococcus. The emerging pattern of antimicrobial resistance among such isolates is alarming. PIN5 INTEREST OF MULTI-CRITERIA MODELING APPROACH IN ASSESSMENT OF YELLOW FEVER EPIDEMIC RISK Beresniak A 1 , Briand S 2 , N’Guyen T 2 , Perea W 2 1 Data Mining International, Geneva, Switzerland, 2 World Health Organization, Geneva, Switzerland OBJECTIVE: The danger of widespread and intense epidemics of yellow fever (YF) in Africa has become very serious, requiring urgent immunization response. Because it is not possible to vac- cinate 100% of the adult population, the challenge is to prioritize immunization of the population at highest risk. An original risk assessment has been performed at the initiative of the World Health Organization, using modeling to enable countries to define populations currently at highest risk, which will be vacci- nated in priority. METHODS: Five exposure risk factors have been selected and collected at the district level in three African countries: Burkina Faso, Togo, Mali. The five indicators are: ecological risk zone, confirmed YF cases since 1960, suspected cases since 1960, number of years in which YF cases notified since 1960, district close to another district that has notified cases since 1960. A multi-criteria analysis based on multiple compo- nent analysis (MCA) has constructed a composite exposure indi- cator (CEi) from the five selected exposure risk factors. In reducing by mathematical projections the number of dimensions, MCA modeling synthesize complex data tables. RESULTS: For each of the three target countries, three analyses have been done for rural districts, urban districts and rural + urban districts. Four risk clusters have been determined from the lowest risk to the highest risks, allowing the construction of detailed YF risk maps in Burkina-Faso, Togo and Mali. These “YF risk assess- ment maps” present in four colors the four risk clusters at each Abstracts A93 brought to you by CORE View metadata, citation and similar papers at core.ac.uk provided by Elsevier - Publisher Connector