Combination anti-HIV drugs (HAART) have dramatically reduced the morbidity and mortality due to HIV and AIDS in the developed world. A major shortcoming with HAART is the development of drug resistance. Genotypic resistance testing is regularly used in clinical practice, producing clinically useful but complex data. Misinterpreting or overlooking such data can lead to adverse outcomes. The advent of electronic medical records presents an opportunity to better integrate complex genotypic testing data into the management of HIV infection, increasing the safety and effectiveness of HAART. This project will create a prototype for future integration of pharmacogenomic information into the electronic medical record for the purpose of improving patient safety and quality of care using HIV genotype as a model system. This proposal has two aims. First is to design and implement an automated system that will integrate HIV genotypic testing results with corresponding patient medication data within an electronic medical record system in order to reduce antiretroviral prescribing errors and improving antiretroviral drug selection. This will be accomplished by developing an automated process to store genotypic testing data within the electronic medical record, creating and implementing """"""""expert rules"""""""" to recognize instances of an incorrect or sub-optimal antiretroviral drug selection in the context of genotypic data, creating a mechanism to alert care providers when the expert rules are triggered, and validating the rules.
The second aim i s to assess the efficacy and usability of this system in a community-based, Ryan White-funded outpatient setting serving a predominantly urban, minority, and low-income population. Data to be measured and analyzed are the frequency of rules triggered, provider responses to alerts, and clinical outcomes measured by HIV viral load. Clinical outcomes will be statistically analyzed as success or failure by two different definitions of success: viral load of <400 and decrease in log (viral load) of _>0.5. Both analyses will look at correlation of success or failure with the presence or absence of an alert and the provider response to alerts. Analyses will be carried out using Fisher's Exact Test. System usabili b providers will also be assessed qualitatively.
