The HIStalk Advisory Panel is a group of hospital CIOs, hospital CMIOs, practicing physicians, and a few vendor executives who have volunteered to provide their thoughts on topical industry issues. I’ll seek their input every month or so on an important news developments and also ask the non-vendor members about their recent experience with vendors. E-mail me to suggest an issue for their consideration.
If you work for a provider organization (hospital, practice, etc.), you are welcome to join the panel. I am grateful to the HIStalk Advisory Panel members for their help in making HIStalk better.
What are the biggest lessons we’ve learned from cases where IT contributed to patient harm?
Common Themes Expressed
- System redundancy is sometimes poorly planned.
- Systems and system changes (especially those involving upgrades and application setup) are not adequately tested.
- IT systems management needs to be more formalized (change management, communication, quality assurance).
- System design should be user-centered and should make it easy for clinicians to do the right thing.
- User application training needs to be not only more comprehensive, but also tied to the workflow and job role changes that are involved.
- Clinicians are not represented in the IT governance process for changes that are seen by IT as purely infrastructure related.
- Clinicians need to take ownership of workflow analysis and get involved in IT projects that affect them and their patients.
- IT is specifically related to patient harm or patient safety – it’s an enabler of management and processes, whether good or bad. Technology is not a panacea.
- Clinicians can’t let the computer override their critical thinking, yet computer systems encourage them to.
- Need for better and more effective education; misuse and system workarounds for the sake of saving time, catching up, or general lack of change; poor IT change management (including maintenance, communication, etc. – all the ITIL stuff) – inadequate QA, communication of changes, poor / under maintenance.
- Testing, testing, testing.
- Redundancy (or lack thereof). Cerner’s recent cloud issues are a great learning lesson. I think that wireless is also an important lesson. Done right, it can save lives. Done cheaply, it can be deadly.
- Patient care and their flow through a hospital is so complex that no healthcare IT solution can completely avoid the unforeseen design flaw, non-intuitive workflow mistake, or inadvertent bug that ultimately harms a patient. Despite the overwhelming benefits of HIT solutions, they’re just as good as the humans that programmed them, and unfortunately, humans can err. Thankfully, with strong clinician input and deep “real world” testing, you can minimize the chances of these occurrences. One big difference with HIT solutions is that when something unforeseen and detrimental does happen, humans can quickly adapt and self-correct. Technology solutions are limited to whatever potential issues the developers and users have foreseen. This has the potential to lead to mistakes of wider consequence. Knight Capital’s near-bankruptcy due to its recent loss of $440M over two days due to a glitch in its trading software is an extreme example of this. Healthcare organizations need make sure their HIT vendors have strong clinician input into their solutions, ensure any solution is vetted thoroughly with “real-world” testing, and processes need to be in place to minimize the breadth of negative patient impact when the unforeseen finally does happen.
- There is a need for interoperability and data standards so that information collected across the patient care spectrum can be safely and securely made available to the proper healthcare providers. There is momentum on this front as groups form HIEs. There is still work to be done to ensure that a doctor or nurse treating a patient has all relevant information to help the patient, not harm them. From patient allergy and drug interaction data to proper condition diagnosis, the underlying up-to-date information and data needs to be made available at the point of care.
- No matter how much you think you’ve trained physicians on a system, they will figure out a way to circumvent the standard processes if they are hard to perform. We need to always make it "easy to do the right thing"… because if we make it easy to do the wrong thing, it will invariably happen.
- Lousy interfaces can and do kill patients. Hopefully not much, but it probably happens more than we realize. The system has to make it easy to recognize problems and, quite importantly, provide a feedback loop for providers to realize what they’ve done e.g. order entry feedback / confirmation / review.
- Assuming here that the harm can be directly attributed to IT and not to clinical practice, the biggest lessons learned from my experience relate specifically to how IT is integrated with clinical departments. I have seen IT make changes to a system that were not properly communicated, documented, and trained on; this led to a compromise in patient safety. I have seen changes in infrastructure, servers virtualized without communication, and suddenly the drug-to-drug allergy checking was not happening in the background unbeknownst to the clinician. These types of things can be catastrophic, and without proper governance in place, can lead to an increase in patient safety or patient harm issues.
- The biggest factors in IT contributions to patient harm are a lack of provider workflow analysis, a lack of proper training and support (could be related to either the trainer or trainee or both), and, rarely, improper testing and configuration of the system. Depending on your philosophy of ownership of these particular issues, IT contribution to the problem could be minimal or significant.
- There are very few documented cases in which problems with the technology specifically (e.g., software bugs) were linked to patient harm. In fact, the few studies that have been done do not separate the technology from the processes in which the technology was used, so in most cases we really don’t know whether the problem was with the technology, the processes, or how the technology was used in a particular set of processes. One might say that where there is significant unscheduled downtime and clinical processes have become dependent on electronic capabilities, that the down time might have the potential to contribute to patient harm. On the other hand, if there are inadequate downtime procedures and a hope that systems will never fail, is this a technology issue or a management issue?
- Similar to the reports we have learned from pilots of crashes or near misses on technically advanced aircraft, the information provided by their systems should be used to help guide decisions and monitor / verify expected outcomes. However, the information provided by systems should not override the learned skills and experiences of aviators. If they do, bad things can happen, as was the case in the Air France Flight 447 disaster.
- I’m not sure we’ve truly "learned" anything. What we should have learned is that you cannot put technology in place and expect it to eliminate patient harm. The process must be fixed before you add technology or you’ve just created another set of issues. Technology does what we program it to do, so if we (humans) don’t validate the technology that goes in place and provide (and execute) quality checks, we’ll continue to harm our patients. This isn’t news to the technology crowd, but we’re having a hard time communicating it to the clinicians. I have frequent conversations with clinicians that think technology is the panacea for our patient safety issues. They are not very happy with me when I point out the other potential issues or insist that workflows are done both pre- and post- conversion. They are even less happy when I insist that we (IT) don’t own these projects and must have significant engagement from the clinical area in order to be successful.
- Safety must be job one no matter what we do. Deploying a new healthcare IT solution requires a significant investment of time and energy on the part of many many people — including physicians, nurses, and all members of the care team, along with their administrative partners — to ensure we are making the environment as safe as it can be. It is critically important that we invest appropriately in testing and training in the spirit of this safety-focused partnership. Everyone in the equation should have an opportunity to "pull the cord" at any point if the product isn’t safe. But training is no longer enough. "Training" is no longer even accurate. We must focus our energy on workforce development. We must ensure that every member of the care team has every opportunity to learn how to perform their job in a new way in the presence of our new and emerging technologies. It isn’t enough to learn how to use a new system. It is critical that we learn how to do our work differently, in a way that is more efficient, more effective, more collaborative and more safe in the context of our new systems.
- To quote former Defense Secretary Donald Rumsfeld, "There are unknown unknowns." The shortcomings of health IT only come to light when someone has been harmed and then we generally create some sort of electronic or paper work around to make sure it doesn’t happen again. I think there will always be "unknown unknowns" when it comes to health IT. Hopefully, the number of these unknowns will decrease with time. Then there will be an acquisition that will create more unknown unknowns.
- Processes and Human Factors (training, effort on the end-user’s part, etc.) are the most important to "get right" with any new technology implementation. The technology is just a tool and is only as good as the people using it and the situation / processes in which it is used. An analogous situation would be a surgical scalpel – used appropriately in the right situation and the result is good, while used incorrectly or in the wrong situation and the result may be dire (e.g. wrong side or wrong site surgery).
- Certainly the recent Cerner system down experience–where one IT person made many hospitals’ data inaccessible, due to independently correcting something in production with no checks and balances. Users getting so used to computer systems providing information that they don’t question if the information is valid. Users who previously knew how to do drug calculations / titrations now relying on computer systems (once again, not checking the validity of data). Data conversions, especially EMR conversions– really still dependent on manual data abstraction in most instances. No one has really done a great job of electronically converting all data correctly.
- Healthcare is human. We need to provide tools that help clinicians provide better, more consistent care and provide them ‘actionable data’ to help avoid mistakes.
- Analysts assume that doctors will override anything that is wrong, and doctors assume that clinicians created the implied logic in the EHR. At our hospital, the model production system of our brand name EHR including a default route of IM for Lasix. This was promptly noticed and fixed. The far more subtle error of diltiazem IV dosing went unnoticed, and diltiazem was routinely then given in much lower doses than previously, until one of the physician champions went "hmm, that’s funny."
Which hospital uses of IT have driven the biggest improvement in patient outcomes?
Common Themes Expressed
- Hospitals need to define their quality goals, track their baseline quality, and then go after improvements.
- Real-time alerts and notifications can affect patient outcomes dramatically.
- Population health analytics can drive some of the biggest improvements beyond systems that just affect inpatient stays.
- Well-defined and closed areas have the most impressive IT-driven improvements: ED, pharmacy, and OR.
- Pharmacy-related IT has driven major patient care improvements: electronic medication administration record, barcode checking of drugs at the bedside, alerts for drug-drug interactions and other patient-specific problems.
- Telemedicine makes it possible to use hard to find expertise more broadly.
- PACS has dramatically changed how clinicians use diagnostic images and how radiologists work.
- Data analysis can pinpoint areas of potential improvement and allow ongoing monitoring.
- Technologies, even simple ones, that allow clinicians to communicate more effectively can have a significant patient impact.
- I would look to the bigger, more sophisticated systems – Trinity and Kaiser come to mind – for the success stories. These system have invested in “benefits realization” and track quality and benefits. All the improvements and benefits are there for mid-size and smaller providers, but they do not even baseline current performance and most don’t adequately go after actually tracking quality and benefits. Hence it becomes a big subjective conversation.
- While not yet a contributor to the biggest improvements in outcomes, Rothman Healthcare’s product is impressive as one that directly affects outcomes. By pulling from nursing notes and vital signs, it has a strong predictive capability of a patient’s impending state of health. It alerts caregivers to a potential change in status hours before they might otherwise pick up on it. The implications for quality of care, and therefore outcomes, are significant.
- Those organizations that are data driven and create actions around the work. They also are fearless in holding clinical staff accountable for their practice. Some organizations won’t fire clinical staff due to the notion of not being able to find replacements.
- Alerting, communication, and messaging is at the forefront of what HIT can do to improve outcomes. Access to data-driven alerting allows clinicians to be informed, combined with the interaction of a communication workflow solution, allows care team members to collaborate to provide a more informed response. Delivering information based on changes in data and allowing users to interact with those data and with other care givers — those two things improve outcomes in big ways.
- Some of the biggest improvements in healthcare come when patient outcomes are relatively improved across a population. Population Health Management (PHM) has become an important topic as providers and payers are moving toward compensation for outcomes. A key pillar to effectively managing and improving the healthcare outcomes of a population is predictive analytics — the ability to leverage historical data and care patterns to be able to rationally intervene in cases when a patient’s health may be
deteriorating. For example, by being able to predict the likelihood that a patient will develop Type II Diabetes based on historic and current clinical, pharmacy, and lab data allows a care coordinator to be alerted when the likelihood reaches a high threshold. The care coordinator may then work with the patient and other care providers (doctors, nurses) to develop a plan to properly manage or even delay the diabetes onset.
- Defined and somewhat closed settings (e.g. ED, pharmacy, OR) can produce very simple and dramatic data, but it’s hard to judge the potentially much broader and long-term effects of patient-centered medical homes + disease registries and population health management. I can say that "we" have seen incredible reductions in length of stay within the ED due to complete transparency/visibility of patient flow within the department.
- Bar coded administration of pharmaceuticals. Drug interaction alerts. Alerts from abnormal lab values.
- In my experience, I would say the pharmacy department. One of the best immediate workflow improvements related to pharmacy was when we went live with eMAR. A patient wait time for meds decreasing by at times over 20 minutes or more is substantial. This may not lead directly to outcomes, but it leads to a significant increase in patient satisfaction.
- Unquestionably, I have seen pharmacy usage as the biggest benefit producer – significantly shortened turnaround times improve patient clinical outcomes, better decision support alerts result in drastic reductions in adverse drug events (ADEs) which improve patient safety – you can see immediate results in these areas when technology is properly deployed.
- Reporting from the data collected that have allowed us to identify trends and opportunities for improvement.
- Ironically, I would guess that the biggest improvements in patient outcomes where technology can be linked have come from the simplest capabilities that IT brings: improved communications among providers. These are not very sophisticated improvements, but rather represent areas in which the technology itself can be said to have contributed to (enabled?) the potential for better outcomes. The question of how the data is actually used in decision making is not a technology issue but a cognitive one, and I would be suspicious of any study that claimed a primary contribution by IT to patient outcomes without a clear understanding as to how IT is actually used. from electronic notifications, and requirements to “sign off”; ability to actually read what someone wrote due to the elimination of handwritten notes (although dictation / transcription processes probably had a lot to do with this); ability to view electronic data from anywhere at any time, rather than hunt for a paper medical record.
- Even though it has been around for several years, electronic medication administration and recording (eMAR) at the bedside has by far had the biggest patient safety impact. It is the single most consistent (not anecdotal) source for improved outcomes with patients on a grand scale.
- This is often truly difficult to quantify or qualify, often because we are (or are forced to) utilize surrogate endpoints for patient outcomes rather than the outcomes themselves. An example would be core measures compliance – greater compliance is seen as success and benefit, although the actual outcomes of those patients may or may not be effected at all. However, one example that does seem to have fairly provable positive benefit for patients would be telemedicine-enabled neurologist evaluation for emergency department patients suffering possible stroke. Obviously such situations are not an ideal replacement for a neurologist being on site to perform the evaluation in person, but in reality that is not a possibility in many areas across the country (even some that are not that "rural"). In such cases, the telemedicine encounter / consultation replaces the only existing timely alternative – a consultative conversation over the phone between an ED physician and neurologist. In such cases, the technology allows for an appropriately trained specialist in the care of stroke to perform a more detailed "eyes-on" evaluation of the patient and provide improved medical decision making as a result of more and better patient information (e.g. stroke or no stroke, to give or not to give TPA). Due to the high risks associated with giving TPA in general, those cases where the TPA is NOT given when it otherwise might have been were it not for the direct telemedicine video evaluation are probably even more important than those cases where the decision is made to give the TPA.
- Laboratory systems have been around longer than many others, so often have more maturity re how data is processed and used. Clinical flowsheets for lab results are pretty standard now. Pharmacy systems also are fairly sophisticated, especially: drug / allergy checking, dosage calculations, and alerts.
- A good friend of mine is the chief of anesthesia and also the surgical suites medical director at a major health system in my home state. He is excited about the possibilities of having rich information available to him for analysis. He believes using tools such as regression analysis will allow his organization to target specific data points that will enable them to improve patient care. The ability to review surgery types, preparation processes, material usage and other important elements will enable them to adjust how, when, where and why adverse patient experiences happen. Without the use of information technology this type of analysis would be almost impossible in a meaningful timeframe. Now they will be able to complete analysis and create change in a fraction of the time it would take in the past. This reduction in reaction time will improve the outcomes for many patients.
- At MUSC and Indiana Heart Hospital, they are leveraging health IT to prepare heart failure patients more effectively prior to discharge — and they are seeing outstanding outcomes improvements in readmissions.
- PACS, without a doubt. Over the last 20 years, having images immediately available to clinicians has completely demystified radiology (to their detriment), and given bedside clinicians access to more information in a more visceral way (sorry). Can’t prove it, but I got anecdotes…