relevant patient data and putting it into a standardized format for eventual analysis by the MediClass program. This task has proven to be a major challenge, since each health system has idiosyncratic forms of encounter data. For HIT 2, we chose to use a clinical document architecture (CDA) written in HTML format. The HTML format is required by the Java processing design of MediClass. It can also be exported as a WORD document for the team of MediClass evaluators to use for the validity study coding.

To test the validity of the MediClass program, we have trained medical record abstractors at each site in the use of a common coding protocol. To date they have coded 500 primary care encounters and we are currently in the process of comparing their coding to that done by the MediClass program. The results of this test will be used to refine the MediClass program and produce the final version which will be used in the second phase of HIT 2. This second phase of the study will include a clinical trial testing the effect of feedback on the delivery of tobacco cessation services by primary care physicians.

Baseline data collection for the provider feedback study will cover the first six months of 2004 and providers will be randomized in July. We will also be sending questionnaire to patients of study providers. We will use the questionnaire results to gauge the delivery of the 5A's as perceived by the patients and use it to evaluate the effectiveness of our feedback study in changing provider behavior.

Sabina Smith, KPNW

Collaborating CRN sites include KPSC (Ann Geiger and Shelley Enger), GHC (Diana Buist), Health Partners (FeiFei Wei), HFHS (Marianne Ulcickas Yood), and Fallon (Terry Field, CRN principal investigator). Lovelace Health System (Floyd Frost) from the HMO Research Network is the sixth site. The project is using a historical cohort design to study over 2000 older women for 10 years to answer treatment effective- ness questions among those with early stage disease. To maximize study efficiency, all eligible stage II, minority group, and 80+ year old women are being enrolled; a sampling strategy is being used for Stage I, white, and younger women.

mobile, a single person provided onsite training for all sites. The common data collection problems that are usually associated with project startup were minimized because project resources were "front loaded" early in the data collection planning process. Meticulous attention was paid to achieving consensus on inclusion/ exclusion criteria and data elements/ coding; the interface between SEER/ Tumor Registry sites and those without; procedure manual creation; extensive system pre-testing; and indepth on site training. Because of these efforts, roll-out of the DCS was smooth. The few problems encountered have been minor and most have been attributable to the user rather than the system. Due to the geographic separation of the project sites, an important challenge is managing these technical issues remotely. We are using the existing CRN internet technology to transfer data files between the study sites and BUMC securely, eliminating the need for physical transfer of data, thus reducing potential for loss and inefficiency. Data are transferred to BUMC on a monthly basis via the CRN website. As of the end of March, data have been collected on 581 subjects (27% of the total). We have also modified the DCS by creating a "stripped down" version which has automated our quality control strategies. We are assessing intra- and inter-rater reliability electronically and have automated data comparisons and transfer, thereby providing additional project efficiencies.