Encouraging innovation and access to health technologies has emerged as an important strategy for combating infectious diseases, internationally. Infectious diseases HIV and hepatitis B and C are contemporary public health problems, especially among vulnerable populations such as People who Inject Drugs (PWID). PWIDs carry a disproportionate burden of infectious diseases compared to the general population. In 2011 there was an HIV infection outbreak among PWID, resulting to approximately one out of six PWIDs living in Athens area to be already infected with the virus. In most developed countries where mortality from communicable diseases has declined significantly, the concern is focused on preventing the emergence of new outbreaks. Early detection of epidemic outbreaks can reduce the size of the epidemic and related morbidity and mortality. In the contemporary era of electronic health (eHealth), the development of automated systems of early and reliable detection of epidemic outbreaks is at the heart of relevant research.
The primary objective of the project was the development of an integrated automatic system of early and reliable detection of epidemic outbreaks and the development of an algorithm for assessing individual risk of infection, focusing on HIV and hepatitis B and C and their application to PWIDs addressed to Organization against Drugs (OKANA).
Within this project, software for an automated system to detect epidemic outbreaks and for estimating personalized risk of infection have been developed. The software for the early epidemic detection was based on mathematical models, while a new innovative model combining a hidden Markov Model with a model based on prediction interval was developed and assessed. Alternative software based on neural networks was also developed. To estimate personalized risk of infection, techniques of “Deep” Machine Learning were used. For the development of the two software, training data from PWIDs registered to OKANA programs as well as simulated data were used.
For the training, but also for running in real time the two developed software/instruments, an integrated information system (IIS) for patients’ registration and personalized monitoring was developed. Within IIS data form various sources (lab tests, behavioral questionnaires, harm reduction questionnaires, substitution therapy, clinical examination) have been unified while new applications, such as specific forms recording detailed information on HIV, HBV, HCV infections, platforms to collect aggregated data from harm reduction programs and data on new questionnaires recently adopted have been developed.. The various matrices of the IIS communicate between them, to produce specific reports as well as alerts facilitating the optimal monitoring of PWIDs registered to OKANA. All data can be extracted in a format that allows further statistical data analysis. To facilitate data entry, several techniques were developed and implemented in the IIS, such as clever scanning, speech-to-text trained to medical conditions/terminology and automated uploaded data from pdf files. Finaly, protocols for monitoring and linking PWIDs were developed, based on which relevant alerts was set-up to notify medical and nursing staff.
The project was implemented by the collaboration of the Laboratory of Hygiene, Epidemiology and Medical Statistics, Medical School of the National and Kapodistrian University of Athens (LHEMS-NKUA), OKANA and the Greek technology company InDigital S.A.. LHEMS-NKUA has great experience in implementing epidemiological studies, mainly on hepatitis B, C and HIV, and in developing innovative biostatistical methods. OKANA has great experience in designing, promoting and implementing prevention policies and treatment of PWIDs. InDigital S.A. has long experience in the field of IT and development of electronic applications. The project was co-ordinated by the LHEMS-NKUA, which undertook the responsibility of the development of the needed mathematical models. OKANA undertook responsibility of the development of appropriate protocols for personalized medical monitoring and the definition and collection of the necessary data for product development and final testing. InDigital undertook the responsibility of the development of an IIS for recording data and individual monitoring, and to implement the automatic detection system for epidemic outbreaks and the tool for assessment individual risk of infection on the IIS.
The developed technological systems could be used for the design, allocation and resources’ transfer to activities addressed to the needs of PWIDs for hepatitis B, C and HIV, in national programs, but they can also be easily modified and exploited in other populations (migrants, Roma, prisoners, general population) and organizations, as well as in overseas markets. Additionally, the separate sub-products that have been developed, can be exploited as separate commercial products in corresponding markets.
In the installed IIS, 5 new questionnaires, among which the one on harm reduction, have been integrated. The later collects data from ΟΚΑΝΑ street work actions and refers to active, non-linked to any program, drug users and thus it reflects their current trends in both drug use and, in general, life-style. Such data allow early detection of behavioral changes and thus can contribute to predict rather than detect possible new epidemics with huge public health benefits, as they can help preventing an epidemic by focused intensification of harm reduction programs. The development of a software for predicting personalized risk of infection was necessarily based on the already available data. These were limited and, at the same time, have many missing. It would be desideratum to continue the collaboration of the three partners aiming to: validate the developed software/instruments to new data, develop an instrument/software for predicting rather than detecting epidemics, optimize IIS’ functionality based on comments/remarks from day-to-day use, scientific utilization of the collected data.
Key words
Epidemic outbreak detection system, personalized risk of infection estimation, infectious diseases, PWIDs, integrated information system, “deep” machine learning