Publication Bias in Health Informatics: Results of a survey Elske Ammenwerth Innsbruck, Austria Nicolette de Keizer Amsterdam, The Netherlands UMIT Aim What percentage of IT evaluation studies does not result in internationally available publications? What are the reasons for not publishing the results of IT evaluation studies Methods - Survey E-mail based survey: 742 email adresses of: Which system, which aspect and how did you evaluate? Where was it published What were reasons not to publish (6 predefined categories or free text) EFMI/IMIA WG on assessment of HIS, AMIA WG on evaluation, Pubmed (paper 2001-2006 on evaluation of IT) 136 CIOs Results Response rate CIOs= 3.7% (5/136) Response rate= 18.3% (136/742) 141 respondents: 118 valid information on IT evaluations studies 19 no evaluation studies conducted 4 unclear/unusable answers Results Under-representation some non-Englishspeaking countries and commercial domain other 25% 4% 6% 5% USA 39% 7% AUS 5% DE 5% CA 8% NL 8% UK 10% 78% Academia IT management Industry Governmental instit unknown Results 217 evaluation studies, 213 publications 120 103 100 2 4 9 80 11 0 82 60 40 77 20 20 12 0 international publication local publication one two three internal publication four five no publication Total Results Source of publications of 103 evaluation studies Only other type of journal 13% Both medical and HI journal 3% Only HI journal 30% Only medical,nursing,p harmaceutical journal 29% Both HI journal and proceedings 6% Only HI proceedings 19% Results 40 37 35 33 30 25 20 “The results and systems are too specific to interest audience in other countries” “May “The publish results following were not validation. all positive. Probably not The [journal system name] administrators which seems and to be “Overwhelmed with ongoing tasks to closed stakeholders shop for those of theoutside systemofwould the implement EMR. Always a lack of time “The evaluations were for a academic not like communities to accept the ” done result.” when funding is coming fromthe outside.” government agency. “The setup (e.g. Once amount ofreports “The organization I work(ed) in wereinterviews) delivered, no activities wasfurther not robust enough.” prohibited publication (well, not were funded. The majority of authors expressly, but it was made clear that were consultants, not academics. 24 “Not subjected tobe our such a publication not Publication is not of would interest for welcome).” university ethics process, consultants.” therefore not published.” 17 15 15 11 10 4 5 0 Not of global interest Publication in preparation No time for publication Limited scientific quality Political corectness Only internal interest Other Discussion of methods Strong academic bias English-speaking countries bias No guarantee that provided information is complete or correct Results only illustrative, not representative for the IT community in health informatics Discussion of results Only half of the studies internationally published Limited number of HI journals or proceedings and large variation of medical journals Various reasons for non-publication but no strong hints for publication bias due to negative results Recommendations to increase published evaluation studies for EBHI Raise number of conducted studies Raise number of submitted papers GEP-HI Raise acceptance rate of submissions Rewards for evaluation, budget, outside IT project team STARE-HI Increase internationally available papers Motivate authors to publish, EVALDB (http://evaldb.umit.at )