Penalizing Hospitals For Being Unsafe

Ashish JhaAdverse events – when bad things happen to patients because of what we as medical professionals do – are a leading cause of suffering and death in the U.S. and globally.  Indeed, as I have written before, patient safety is a major issue in American healthcare, and one that has gotten far too little attention. Tens of thousands of Americans die needlessly because of preventable infections, medication errors, surgical mishaps, and so forth.  As I wrote previously, according to Office of Inspector General (OIG), when an older American walks into a hospital, he or she has about a 1 in 4 chance of suffering some sort of injury during their stay.  Many of these are debilitating, life-threatening, or even fatal.  Things are not much better for younger Americans.

Given the magnitude of the problem, many of us have decried the surprising lack of attention and focus on this issue from policymakers.  Well, things are changing – and while some of that change is good, some of it worries me.  Congress, as part of the Affordable Care Act, required Centers for Medicare and Medicaid Services (CMS) to penalize hospitals that had high rates of “HACs” – Hospital Acquired Conditions.  CMS has done the best it can, putting together a combination of infections (as identified through clinical surveillance and reported to the CDC) and other complications (as identified through the Patient Safety Indicators, or PSIs).  PSIs are useful – they use algorithms to identify complications coded in the billing data that hospitals send to CMS.  However, there are three potential problems with PSIs:  hospitals vary in how hard they look for complications, they vary in how diligently they code complications, and finally, although PSIs are risk-adjusted, their risk-adjustment is not very good — and sicker patients generally have more complications.

So, HACs are imperfect – but the bottom line is, every metric is imperfect.  Are HACs particularly imperfect?  Are the problems with HACs worse than with other measures?  I think we have some reason to be concerned.

HACs – Who Gets Penalized?

Our team was asked by Jordan Rau of Kaiser Health News to run the numbers.  He sent along a database that listed CMS’s calculation of the HAC score for every hospital, and the worst 25% that were likely to get penalized. So, we ran some numbers, looking at characteristics of hospitals that do and do not get penalized:


These are bivariate relationships – that is, major teaching hospitals were 2.9 times more likely to be penalized than non-teaching hospitals.  This does not simultaneously adjust for the other characteristics because as a policy matter, it’s the unadjusted value that matters.  If you want to understand to what degree academic hospitals are being penalized because they also happen to be large, then you need multivariate analyses – and therefore, we went ahead and ran a multivariable model – and even in the multivariable model (logistic model with each of the above variables in the model), the results are qualitatively similar although not all the differences remain statistically significant.

What Does This Mean?

So how should we interpret these data?  A simple way to think about it is this:  who is getting penalized?  Large, urban, public, teaching hospitals in the Northeast with lots of poor patients.  Who is not getting penalized?  Small, rural, for-profit hospitals in the South.  Here are the data from the multivariable model:  The chances that a large, urban, public, major teaching hospital that has lots of poor patients (i.e. top quartile of DSH Index) will get the HAC penalty?  62%.  The chances that a small, rural, for-profit, non-teaching hospital in the south with very few poor patients will get the penalty? 9%.

Is that a problem?  You could make the argument that these large, Northeastern teaching hospitals are terrible places to get care – while the hospitals that are really doing it well are the small, rural, for-profit hospitals in the south. May be. I suspect this is much more about the underlying patient population and vigilance than actual safety.  Beth Isarel Deaconess Medical Center (BIDMC) in Boston is one of the very few hospitals in the country with exceptionally low mortality rates across all three publicly reported conditions and a hospital that I have written about as having great leadership and a laser focus attention on quality. And yet, it is being penalized as being one of the hospitals with, according to the HAC metric, a poor record on safety.  So is Brigham and Women’s (though I’m affiliated there, so watch my bias) – a pioneer in patient safety whose chief quality and safety officer is David Bates, one of nation’s foremost safety gurus.  So are the Cleveland Clinic and Barnes Jewish, RWJF Medical Center, LDS Hospital in Salt Lake, and Indiana University Hospital, to name a few.

So what are we to do?  Is this just whining that our metrics aren’t perfect?  Don’t we have to do something to move the needle on patient safety?  Absolutely.  But, we are missing a great opportunity to do something much more useful.  Patient safety as a field has been stuck.  It’s been 15 years since the IOM’s To Err is Human report came out – and by all counts, progress has been painstakingly slow.  Therefore, I am completely on board with the sentiment behind Congressional intent and CMS’s efforts.  We have to do something – but I think we should do something a little different.

If you look across the safety landscape, one thing becomes clear:  when we have good measures, we make progress.  We have made modest improvements in hospital acquired infections – because of tremendous work by the CDC (and their clinically-based National Hospital Surveillance Network) that collects good data on patient safety and feeds it back to hospitals. We have also made some progress on surgical complications, partly because a group of hospitals are willing to collect high quality data, and feed it back to their institutions.  But the rest of the field of patient safety?  Not so much.  What we need are good measures.  And, luckily, there is still a window of opportunity if we are willing to make patient safety a priority.

How to Move Forward

This gets us to the actual solution:  harnessing the power of meaningful use in the Electronic Health Records incentive program.  We need clinically-based, high quality patient safety metrics.  Electronic health records can capture these far more effectively than billing codes can.  The federal government is giving out billions of dollars to doctors and hospitals that “meaningfully use” certified EHRs.  A couple of years ago, David Classen and I wrote a piece in NEJM that outlined how the federal government, if it wanted to be serious about patient safety, could require, that EHR systems measure, track, and feed back patient safety events as part of certification and requirements for meaningful use.  The technology is there.  Lots of companies have developed adverse event monitoring tools.  It just requires someone to decide that improving patient safety is important – and that clinically-based metrics are useful.

So here we are – HACs.  Well intentioned – and a step forward, I think, in the effort to make healthcare better.  Everyone I know thinks HACs have important limitations – but reasonable people disagree over whether their flaws make them unusable for financial incentives or not.  The good news is that all of us can agree that we can do much better.  And now is the time to do it.

15 replies »

  1. We thank Dr. Jha for his positive comments regarding our healthcare-associated infection measures that form a component of the overall patient safety landscape. However, we need to continue to move forward in creating measures that are more imbedded in pathways of clinical care and thereby less ‘game-able’ and better risk adjusted. We also agree with Dr. Jha that the way forward is to capture the power of the EHR by reusing clinical data that have been recorded in structured, electronic form. However, we would also add the importance of deep subject matter expertise that leverages understanding of pathogenesis and clinical care pathways to create measures that are non-subjective, respond well to implementation of prevention strategies (which presupposes the condition is preventable), and still remain clinically meaningful. As an example, we recently worked closely with the critical care community to develop a ‘ventilator-associated event (VAE)’ measure to replace our ‘ventilator-associated pneumonia’ pneumonia which did not perform reliably. While VAE is not currently reported to CDC electronically, we have other measures that utilize our Clinical Document Architecture (CDA) to report electronically such as MRSA bloodstream infections and Clostridium difficile LabID events and already many hospitals are reporting these events electronically via the CDA. Given the major threat to patient safety posed by antibiotic resistance as well as C. difficille infections that are associated with antibiotic use, we have developed an important process and outcome measurement system known as the ‘antibiotic use and resistance (AUR) module’. In contrast to previous measures we have developed for NHSN, the AUR module will be the first that only accepts data from hospitals electronically via the CDA.

    L. Clifford McDonald, MD,
    Senior Adviser for Science and Integrity
    Division of Healthcare Quality Promotion
    Centers for Disease Control and Prevention

  2. This is a very important topic and deserves more shares and discussion on higher level. Medical errors have claimed lives of many people not just in US, but all around the world. There is a critical need for us to come up with a plan to at least reduced the number of such cases. Penalizing hospitals is a good move.

  3. There is a very real problem with confounding all adverse events or HAC’s with the term “injury”. Read this: http://skepticalscalpel.blogspot.com/2014/06/1-in-5-elderly-us-patients-injured-by.html
    When critically ill patients need mechanical ventilation for a prolonged period of time, some of them will get pneumonia. The incidence will never be zero. But what would the mortality be if they were not mechanically ventilated in the first place? Probably close to 100%. Of patients with acute problems such as intestinal perforation due to diverticulitis, some will have underlying coronary artery disease or cerebral vascular disease, and will have perioperative complications due to those underlying disease processes. The obvious solution to preventing hospital falls is never to let patients out of bed, which of course would be associated with its own set of complications. Adverse events should be minimized as far as we can, but it is unreasonable to penalize hospitals or physicians for caring for critically ill patients, which essentially what is happening.

    My experience with EHRs is that reams of “data” of suspect accuracy are copied and pasted over and over, and few people have time to edit and correct because of production pressures. The idea that EHRs are the solution to anything, other than perhaps correct reproduction of vital sign flowsheets, seems laughable.

    Industry–airline, automotive, whatever–knows that a target of 100% safety can never be reached, as much as we would like it to be. The question is, what are the safety targets that will benefit the most people, and what is the cost of achieving them?

    Whether you’re dealing with doctors and nurses and hospitals, or trying to train a seal at Sea World, penalties are hardly ever the answer. Teaching, helping, and creating a system that rewards good ideas and excellent performance are much more likely to be effective. The ACA, with its idea that standardized protocols and fewer physicians will solve everything, isn’t going in that direction.

  4. Ten of thousands of people shouldn’t need to be dying due to the lack of good patient safety in healthcare. People come into hospitals to get the proper treatment they need for a sickness that has brought them there. When respectable procedures are made, we will see advancement. I believe the CDC has been doing fairly well in protecting hospitals. If everyone in a hospital worked together and followed all the guidelines and took proper procedures, such lack of safety can be significantly reduced. I strongly support the affordable care act’s decision in punishing healthcare facilities with Hospital Acquired Conditions. Maybe this way some sort of revolution will be made in order to keep the hospital open.

  5. Great article. I’m all for “harnessing the power of meaningful use in the Electronic Health Records incentive program”. However, what makes adverse event reporting suitable for being built into an EMR?

    It’s hard for a software package to do all these things well. The authors of “Escaping the EHR trap” write, “Just as consumers select and manage myriad technologies — Facebook for social networking, Twitter for microblogging, Google for search, iTunes for music — so should physicians.”

    Sure, reporting is easier when built into the same interface, and you may derive some information about adverse events from the notes, but EHRs are already cumbersome doing the things Meaningful Use already defines: documentation and order entry.

  6. I see trouble here.

    First, as you showed, choice of data is critical to have true meaningful results. The unfortunate thing is that the folks choosing the data points are often not right in what they choose. Bad data will potentially mark relatively good providers as being deficient, which will do nothing to build trust in these systems, much less increase actual quality.

    Second, the EHR systems are not always systemic tools (i.e. One patient doesn’t usually have all doctors using a single EHR product), but rather individual tools for the providers. Meaningful use checks are sent to providers, not hospital systems. The problems of HAC’s, at least as you reported, are systemic problems. They reflect on coordination of care, on inpatient care problems, or on poor communication or follow-up. In other words, bad outcomes are usually not the fault of a single individual, but rather a system that is wrong. So using EHR to measure “quality” is not looking at the right level, and is probably going to falsely assign blame to individuals. Again, this will do nothing for true quality measures and will undermine trust issues (which are already pretty lousy).

    Finally, it seems like this would involve a lot of self-reporting. Putting notes in the patient charts are an activity done in addition to care, they are not care itself. There is a false equation of care to care documentation which this would only amplify. EHR’s are not generally used by clinicians as a tool for care, but rather are increasingly a place where doctors are increasingly devoting time that would/should be given to the patient for actual care. In other words, EHR documentation is, in some ways, anti-care. It is done after the fact, and done with an enthusiasm equal to that of doing the dishes or doing taxes. Far too many pundits and EHR proponents (of which I was once a card-carrying member) ignore the most important thing about EHR: doctors (and nurses) hate to document. It’s a task we realize we must do, and we do because it can help care, but it is an afterthought and is always secondary to the care itself. Penalizing docs for documentation is not the same as penalizing them for care, and doing so will build on the already high resentment docs feel toward “the man” who is measuring them.

    So how else do payors and government agencies measure quality so as to influence it in a positive way? How about rewarding those docs who help patients avoid procedures, hospitalizations, unnecessary medications, etc? How about realizing that payors are at fault, by rewarding sickness and procedures with more money; and that to turn the eye on doctors who are simply doing what they’ve been trained to do by the payment system (do more things and have sicker patients) is simply shifting blame. Payors give financial incentive to keep people sick and do unnecessary procedures, yet now we are going to make the docs out to be the bad guys for following those incentives? Small wonder why more docs are giving up.

  7. Ashish, I like the idea of using the EHR as a reporting tool, but I think that the problems raised in some of the other comments complicate the picture. As you and others have written, EHRs themselves are often problematic. It’s not just that the EHR itself may be the proximate cause of an adverse event, but where does the data go, in what timeframe, for access by whom, and how to handle the whining, complaining, and regulatory footdragging by hospitals that want to challenge, either administratively or in court, how their data has been represented? If the EHR becomes the aggregation and reporting tool, how do you prevent gaming of the inputs?

    Ideally, data such as these should populate scoreboards, perhaps at CMS’s website, where any consumer can look up scores and compare hospitals. But, the reality is that hospitals themselves really don’t want to see this happen, even though hospital leaders often say otherwise.

    “Not a provider” has it right: hospitals are dangerous and the best way to deal with a hospital is to avoid it all costs. And, if you or a family member does end up in one, do as my karate instructor says, “become a controlled beast as soon as you enter [the ring], because no one can protect you except you.”

  8. GY — hospitals already spend a lot of money reviewing charts. We did a study years ago that found that it would take a pharmacist about 5 to10 hours a week (real $ but not undoable) to review all the possible adverse drug event alerts and track down which ones are real — and that would not only save lives but also $. Same for other types of adverse events.

  9. Thank you Datum for your comment. There are a lot of tools that allow for automated monitoring of adverse events — they vary in their utility but most that I have seen are a big step forward from where we are today (either nothing or just billing data).

  10. Inexpensive care v Quality care v Safe care… Not possible. Just teach consumers that hospitals are dangerous and to stay healthy so they don’t have to go to one.

  11. There is a problem. You failed to consider the EHR device as a cause for the neglect and errors.

    After all, with the $$$$$ billions spent, adverse events, never events, falls, and overall outcomes do not show any ROI.

    Any discussion about MU and safety should issue a disclaimer: EHR devices running all care have not been assessed for safety, and may be the cause of new errors.

  12. My god, this is an important post!

    The key to the righting the meaningful use program is to rethink how we define “meaningful.” This much-maligned program can do the good its well-intended designers intended if we aim it in the right direction.

    Requiring EMRs to track patient safety is an important step towards giving this program teeth. We already have the technology we need. Why are we waiting?

    Actually don’t answer that question, everybody knows why. Please tell us more about how this works, instead!