Saving Patient Ryan|Can Advanced Electronic Medical Records Make Patient Care Safer? Muhammad Zia Hydari Tepper School of Business, Carnegie Mellon University, [email protected] Rahul Telang Heinz College, Carnegie Mellon University, [email protected] William M. Marella Pennsylvania Patient Safety Authority, [email protected] Patient safety is one of the foremost problems in US healthcare, affecting hundreds of thousands of patients and costing tens of billions of dollars every year. Advanced electronic medical records (EMRs) are widely expected to improve patient safety, but the evidence of advanced EMRs' impact on patient safety is inconclu- sive. A key challenge to evaluating EMRs' impact on safety has been the lack of reliable and comprehensive data. We overcome this challenge by constructing a panel of Pennsylvania hospitals over 2005{2012 using data from several sources. In particular, we source confidential patient safety data from the Pennsylvania Patient Safety Authority (PSA). Since mid-2004, Pennsylvania state law has mandated that hospitals report a broad range of patient safety events to the PSA. Using a differences-in-differences identification strategy, we find that advanced EMRs lead to a 27 percent decline in patient safety events. This overall decline is driven by declines in several important subcategories|30 percent decline in events due to medication errors and 25 percent decline in events due to complications. Our results hold against a number of robustness checks, including, but not limited to, falsification test with non-clinical IT and falsification test with a sub- category of events that is not expected to benefit from advanced EMRs. Overall, we provide evidence to policy makers, hospital administrators, and other stakeholders that hospitals' adoption of advanced EMRs improves patient safety. Acknowledgements : We thank Idris Adjerid, Bernard Black, Erik Brynjolfsson, Chris Forman, John Gardner, Martin Gaynor, Avi Goldfarb, Shane Greenstein, Tat Koon Koh, Rema Padman, Michael D. Smith, Jonathan Treadwell, and Catherine Tucker for helpful suggestions. We also thank seminar participants at the National Bureau of Economic Research Summer Workshop; The Office of the National Coordinator for Health Information Technology (ONC) Summer of Data Meeting; ONC's National Health IT Week; and the Pennsylvania Patient Safety Authority Board Meeting. We are responsible for all opinions and errors. 1 Hydari, Telang, Marella: Advanced EMRs and Patient Safety 2 Essay 1 We cannot look at insurance coverage, medical costs, quality of care and information technology as separate issues. -Paul O'Neill, 72nd US Secretary of the Treasury (New York Times 2007). 1. Introduction Patient safety is one of the foremost problems in US healthcare. According to the landmark Institute of Medicine (IOM)1 report To Err is Human: Building a Safer Health System, 44,000 to 98,000 people die each year in US hospitals from preventable medical errors (see IOM 2000, Chap. 2). In addition, hundreds of thousands of other patients survive after being harmed or after having faced the risk of harm from medical care. Moreover, patient safety events cost tens of billions of dollars to society (Bos et al. 2011, Seabury et al. 2012, Cheeks 2013). Health information technology (IT) is widely considered part of the solution to improving the safety of healthcare in the United States. For instance, the IOM report Health IT and Patient Safety notes: \One strategy the nation has turned to for safer, more effective care is the widespread use of health information technologies" (IOM 2012, p. 1). The question of interest is whether hospitals' adoption of health IT has matched expectations and improved patient safety. Despite the importance of this question to policy makers, hospital administrators, patients, and other stakeholders, the IOM concluded from a review of more than 200 research articles: \. current literature is inconclusive regarding the overall impact of health IT on patient safety" (IOM 2011, slide 22). The IOM report and other experts note that existing research on this topic suffers from limited samples (one or few prominent hospitals), weak methodology, and conflict of interest due to researchers' financial ties to the health IT industry. Furthermore, systematic reviews of the existing literature do not suggest a general pattern of impact of health IT on patient safety| with some studies suggesting a positive impact and other studies suggesting mixed or no impact. Press reports, while acknowledging the potential of health IT in improving patient safety, have highlighted unfortunate incidents such as the death of a woman and a baby boy which may have 1 The Institute of Medicine is one of the four organizations that comprise the National Academies. Hydari, Telang, Marella: Advanced EMRs and Patient Safety Essay 1 3 been health IT-induced (Rowland 2014). Thus, the overall impact of health IT on patient safety remains an unsettled empirical question. We contribute to the literature on the value of health IT generally and to the question of health IT's impact on patient safety specifically by constructing a novel data set and using rigorous methods. A key challenge to evaluating IT's impact on safety has been the lack of reliable and com- prehensive data. We overcome this challenge by constructing a panel of Pennsylvania hospitals over 2005{2012 using data from several sources. In particular, we source confidential patient safety data from the Pennsylvania Patient Safety Authority (PSA). Since mid-2004, Pennsylvania state law has mandated that hospitals report all patient safety events to the PSA. These data allow us to test the impact of hospitals' adoption of advanced electronic medical records (EMRs) on patient safety events while controlling for hospital fixed effects, year fixed effects, hospital size, hospital teaching status, and other hospital-level covariates (we also control for county-level covariates, including population, median household income, and unemployment). Using a differences-in-differences iden- tification strategy, we find that advanced EMRs lead to a 27 percent decline in patient safety events. This overall decline is driven by declines in several important subcategories|30 percent decline in events due to medication errors and 25 percent decline in events due to complications. Thus, our study offers evidence to hospital managers and policy makers of substantial improve- ments in patient safety due to the adoption of advanced EMRs. Despite financial incentives from the federal government, advanced EMR adoption in the US was 67 percent in 2012.2 Our results suggest that further efforts to foster the adoption of advanced EMRs will make the resulting patient safety benefits more universal. 2. Health IT, Electronic Medical Records, and Patient Safety Health IT is an all-encompassing term for computer and communication technologies used by healthcare providers. Although many IT applications play a role in the overall improvement of care quality and patient safety, EMRs play a particularly salient role and thus EMRs are widely studied by multiple disciplines. However, precisely defining EMRs is difficult because EMRs continue to 2 Source: HIMSS Hydari, Telang, Marella: Advanced EMRs and Patient Safety 4 Essay 1 evolve. Table1 lists applications that the Healthcare Information and Management Systems Society (HIMSS) deems to be part of \Electronic Medical Records" category. Table 1 Electronic Medical Records Component Applications Category Applications Electronic Medical Record Business Intelligence - Clinical Clinical Data Repository (CDR) Clinical Decision Support System (CDSS) Computerized Practitioner Order Entry (CPOE) Order Entry (Includes Order Communications) Patient Portal Physician Documentation (PD) Physician Portal This table lists all application components categorized by HIMSS as Electronic Medical Records in HIMSS dataset 2012. Table2 provides a synopsis of four EMR applications: Clinical Data Repository (CDR), Clinical Decision Support System (CDSS), Computerized Provider Order Entry (CPOE), and Physician Documentation (PD). Dranove et al.(2014) define basic EMR as CDR or CDSS and advanced EMR as CPOE or PD. CDR and CDSS are baseline EMR applications but may not form part of the physician workflows. In contrast, CPOE and PD are integrated into physician workflows and may also have the most clinical impact (Jha et al. 2010, p. 1952; Jha et al. 2009b, pp. 1634{1635). The remaining applications in Table1 have either been supplanted (e.g. Order Entry by CPOE), are too new (e.g. Business Intelligence), or considered less consequential for patient safety. Henceforth, we primarily focus on advanced EMR applications. 2.1. Patient Safety Great Britain House of Commons Health Committee(2009) defines patient safety as \freedom, as far as possible, from harm, or risk of harm, caused by medical management (as opposed to harm caused by the natural course of the patient's original illness or condition)." Until recently, the medical community viewed medical errors and concomitant harm either as unavoidable side effects of modern medicine or the result of medical treatment by incompetent providers. Leape (1994) argued forcefully that many errors are preventable and many are \evidence of system flaws not character flaws." The publication of To Err is Human (IOM 2000) catapulted the patient safety movement into the medical mainstream. The goal of the patient safety movement is to Hydari, Telang, Marella: Advanced EMRs and Patient Safety Essay 1 5 Table 2 CDR, CDSS,