Much of the focus on data science is on computing to retrieve results quickly from data stores involving large volume, velocity, and variety. The computing includes implementation of algorithms for analytics (involving statistics or other models) and visualization. As I mentioned in earlier posts (for example here and here), story-telling is a major contributor to the success of a data science project. Often huge investments in data-process even with captivating visualizations will fail because of the inability to present the conclusions in a form of a story that is easy to comprehend and where that all of the stakeholders agree on the same interpretation of that story. Story telling is an art that I described as being more of a talent for journalists than computer scientists.
I use the term story-telling deliberately to distinguish it from the deterministic or mathematical derivations from that that characterize analytics and visualizations. Story telling involves expressing a persuasive case for one interpretation through the creative use of language. This definition is the same as classical rhetoric, but story telling seems more appropriate to modern usage because it captures the creative elements to construct an engaging narrative that is entertaining as well as informative and persuasive.
The ultimate product of a data science project, the product that delivers the goods, is the story. Story telling does not necessarily require a huge amount of raw data. In fact, even for big-data projects, the story telling is built around the small data as presented in the ultimate refinements of analytic and visualization summaries. The story telling is about the pictures about the summary of data.
Story-telling underlies what I call hypothesis-discovery. Both concepts involve the same process that was first described to me as “listening to the data”. I like the visual of being in the audience of a lecturer but that lecturer is the data. Typically, a lecture is given in an isolated room cut off from external interfering noises and where the lecturer is given exclusive access to make his points. The point is that for the duration of the lecture, our attention is devoted to this one lecturer. Competing views will occur in other places and at other times. This is the image I have when tasked to listen to the data.
Listening to data does not involve direct contemplation of every single data point of big data. Instead it involves contemplation of the careful summaries presented in analytic summaries and visualizations. We listen to a small set of data that may be backed up by much large data sets.
With that sense in mind, I want to express some observations I have from listening to the data about the recent events in the Ebola cases occurring in Dallas, Texas. The data I’m listening to are news reports instead of database records. But I’m assuming the news reports are summaries of valid data about what is happening. The following are some things I hear when I listen to this data.
Video record lacking for care delivery
In an earlier post on this topic, I suggested that the protocols for protecting health care workers should include recorded videos of their entire experience from donning to doffing of personal protective equipment (PPE).
Even the Doctors Without Borders group following very diligent and well-trained practices are contracting the disease. We know the use of PPE can fail to prevent contamination.
Video recording of the entire period of performing duties related to serving an Ebola patient would be invaluable to identify the contamination event, where the contamination occurred, where the contamination came from, what path and duration exists between the patient and the provider. Apparently we do not have recordings even in well-equipped US hospitals. Our only tool for assessing the cause of an exposure is to perform interviews with the staff present at the time. A video recording would help tremendously.
Although video recording technology is cheap and widely used by consumers and by businesses for security purposes, there are likely to be more challenges in the hospital environment due to privacy requirements of HIPAA and the need to keep the camera contamination free. I do not expect it to be technically challenging to maintain clean cameras, although it will increase the cost of the cameras themselves. The real problem is probably storage of the video.
The video stream could be closed-circuit to avoid unauthorized distribution but there remains the challenge of storing the video content for long periods in a way that assures privacy will be protected. I am not familiar with Hospital IT systems, but I suspect their HIPAA compliant storage and retrieval capabilities are sized for smaller data sets of lab reports and patient records, plus some static images. Including extensive storing of videos recording each healthcare provider’s actions will probably require a major investment in new storage capacity. With today’s virtual computing (cloud computing) markets, this may be affordable if such providers were HIPAA compliant (if that is even possible with the outdated HIPAA law).
If there is a constraint of limited HIPAA compliant storage, the video recording could be limited to just cares involving a highly contagious disease like Ebola, and then even limited to the riskiest part of doffing of PPE. Although recording of doffing of PPE can identify breaches in protocols, it is unlikely to detect that that particular breach resulted in contamination. To learn that, we need to have videos of the actual duties in the care of the patient to determine what contaminant (if any) may be involved of that specific protocol breach. As I pointed out in the earlier post, we may not fully understand all the ways that Ebola can spread. Extensive videos of each health worker for the entire period with the patient can provide possible clues such as unexpectedly violent events that may find a way around the PPE.
In my non-health-care experience, I learned that if a measurement is required to solve a problem, then it is easiest to simply collect that measurement everywhere all the time. Trying to economize by having rules about when to turn on sensors and when to collect data is too difficult to formulate and compounds the workload of already busy workers. From a workload and workflow perspective, the easiest policy is to record everything all the time. Coincidentally, this provides the best opportunity to capture the critical events. The primary objection is probably the need for more budget for sufficient data storage, but in this era of big data technologies, this is increasingly affordable.
At some point risks of further treatment outweigh potential for success
In a recent post critical of inherent authoritarianism of evidence-based decision making, I made the absurd suggestion of offering the option of sanctioned and encouraged suicide for patients diagnosed with contagious diseases like Ebola. That absurd suggestion was to expand on my earlier thoughts on the rationale of allowing terminally ill (typically with cancer) patients to decline care or even to opt for sanctioned suicide by comparing these options involving comparably fatal and awful prospects of a disease like Ebola. A painless comfortable suicide may be an attractive option rather than attempt to fight through an infection, but we don’t consider that option at all.
Toward the end of the first post, I presented the scenario of a highly contagious disease during an out of control epidemic (or pandemic comparable to the Spanish Flu of 1919). In such a scenario where there is a high risk of currently healthy and disease free population contracting and dying from the disease and where there is no effective treatment or prevention, then the option of state-sanctioned and encouraged suicide of recently diagnosed patients may become the best option available to control the spread of the disease. In the specific case of Ebola, the virus continues to rapidly multiply so that the risk of infecting new victims increases as the disease progresses. An earlier suicide can provide a quick death and disposal before the body becomes highly contagious. I emphasize that this remains an absurd suggestion, but it may be the only effective option available to reduce the spread of the disease if it becomes a pandemic.
At the time I am writing this post, we are not yet experiencing an pandemic of Ebola. The above discussion is based on the observation that as the Ebola patient gets closer to death, the viral loads throughout the body become so large so that even the tiniest contact with anything from the body will have high probability of spreading the disease.
In the case of the index patient at Dallas, there was a point where it may have been more humane to perform illegal euthanasia. Ebola kills by destroying organs. Evidence of organ destruction was apparent due to the need to start dialysis and intubation and mechanical ventilation. Despite the low prospects of survival and heightened risks of contagion, the staff made the correct and ethical choices to escalate attempts save the life.
This additional heroic attempt to save the life ended up postponing death and lengthening the period of suffering. The additional time allowed the virus to continue to multiply and increase the risk of infecting the health workers. The risk of infection also increased by the performing difficult intubation procedure and the use of dialysis equipment. There is no evidence when the two nurses contracted the disease, but it seems likely to have occurred during this later stage of the disease.
For a disease like Ebola, we may be better off with a policy that defines a point where further care should stop. In effect this is an implicit DNR/DNI (do not resuscitate, do not intubate) when an contagious disease is involved. The risks may be too high to proceed with such efforts, and the prospects for recovery are too low. For the particular case of the index patient, the suffering could have ended earlier and perhaps we could have prevented the spread of this disease to the health workers.
Current medical practice and ethics demand the use of all resources to save a life. The above efforts to resuscitate the patient followed this ethical practice. I imagine that the health care workers did not even consider there was a choice.
For a contagious and dangerous disease like Ebola, there is a need to re-evaluate the ethics to go to such extraordinary efforts when the virus is still out of control in the body. We need a public debate about the ethics of limiting care for contagious only up to the point where resuscitation efforts are required. It may be more ethical to invoke an implicit approval for DNR/DNI in contagious disease scenarios. This could be standard practice without the need for the patient’s prior consent, or perhaps even contrary to the patient’s prior demand for such resuscitation if it became necessary. The risks of spread of the disease are too high, and the prospects for survival are too low. The best solution may be to not perform these extra efforts and allow the body to die sooner.
This is a tough ethical question. We should be debating this today before such a pandemic occurs. Having an implicit DNR/DNI for contagious diseases could help prevent or slow a pandemic. While we are at it, we could also debate the ethics of permitting sanctioned suicide or euthanasia for infected individuals before the disease kills them naturally.
Update 10/23/2013. This article claims that these ideas are being considered.
Practices for doffing protective garments need improvement
This is a more modest suggestion that occurred to me after watching this demo of how careful following CDC protocols for doffing PPE can lead to contamination. This demo involved chocolate sauce to leave visible traces where the contamination would occur.
For the two health workers in Dallas who contracted the disease, I suspect they would have immediately washed the locations of the contamination (or undergone more extensive decontamination) if the evidence of contamination were as obvious as the chocolate sauce smears in the demo. My suggestion is to amend the protocol for doffing PPE by first spraying it entirely with something of the consistency of chocolate sauce but instead include a disinfectant plus a dye that would glow when illuminated by ultraviolet light. Immediately after the doffing, they would enter a room illuminated with a blacklight (UV-A ultraviolet) to check for any glowing so they know where to immediately disinfect and clean those areas.
Additionally, if the health worker still contracts the disease despite not finding any evidence of contamination from doffing PPE, then we have a clue that the infection must have come from another route.
Nurses’ roles as medical front-line need more respect
Last week when I first heard of the diagnosis of Ebola transmitted to the first nurse at Dallas, I was disappointed to hear the immediate and confident declaration that the nurse must have made a mistake in following protocols. I had just finished writing this post that described this unfortunate tendency to substitute model-generated data for ignorance. In my mind, it was obviously too early to know how this transmission occurred, but CDC was so confident in their knowledge of the virus and the effectiveness of the protocols that the only possibility of transmission was from human error, in this case by a nurse. Since that time, there has been widespread criticism of this initial conclusion (one example is the previous discussion of the PPE doffing demo) and the director did come out later to clarify that there may have been other problems that can explain this contamination.
Later, we learned more details (not officially confirmed) from the National Nurses United about what was really happening in the treatment of the index patient. One particular theme in this presentation is the lack of respect afforded nurses during the actions of their duties. I think this is a serious problem for all of health care, but this particular case provides a vivid example that their intelligence and observations need more respect by both doctors and supervisors.
The above accounting identifies multiple instances of the nurses raising objections about the current conditions that appeared risky or inappropriate. The responses they received are consistent with stories I have heard in other contexts concerning nurses. Basically, they are directed to follow orders despite their objections.
Nurses deserve much more respect in medical decision making. While it is true that they are not as skilled or educated as doctors or senior supervisors, they do have extensive skills and training especially concerning how to protect patients and community. I consider nurses to serve a critical role as the intermediary between the patient and the rest of the medical apparatus. Nurses have two roles. One is to assist doctors in performing medical procedures (including monitoring and various routine tasks). The other is to look out for the interests of the patient and of the community. The latter role is in effect to be the representative of the patient during health procedures where the patient is likely less medically-informed as the nurse or often not even aware of what is happening.
We need more discussion about the respect and roles we give to nurses as they are a key part of the medical system and an extremely vital role when it comes to treating contagious diseases. As noted before, I have only a layperson’s knowledge of what happens in nursing. I only know what I read in news media, but what I read disturbs me greatly. Many news accounts of situations involving nurses expose disrespect for nurses as being unreliable or under-educated. The impression is that nurses require close supervision and direct orders for precise actions even when they demonstrate an ability to anticipate what those orders will be.
In this particular example, the nurses identified multiple practices that concerned them as being unsafe based on their knowledge and concern for risk. They were overruled and they obeyed with orders to do things they did not feel confident about.
Completely coincidentally and unrelated, I have been paying attention to recent news about recent campus rape laws being enacted in California, the yes-means-yes laws. These laws require proof of positive affirmation for each and every action involving intimate relations on campus. This law and its justification have nothing to do with nursing, but the idea of requiring freely delivered affirmation for each step may have application to nursing.
Nurses are well educated to know what is safe. Whenever they encounter a situation that they do not know for sure is safe, then they deserve the respect to have those concerns answered to their satisfaction. This could simply be something that they have not yet learned is safe as well as something they recognize as clearly unsafe. A policy of respect for the nurse’s intelligence and observation would require the nurse’s affirmation before she proceeds with any action. This could mean stopping the medical procedure entirely until the nurse can be convinced that the practice is safe or will be followed with appropriate actions to address the unsafe concerns.
The nurse should have the right to demand a satisfying explanation that answers the specific concerns raised. The example in the above link about covering exposed skin on neck with multiple layers of medical tape is not a satisfying answer. The nurse should be able to withhold affirmation for what seems unsafe and not be compelled to do something the nurse perceives as inadequate or unsafe.
The implication of my suggestion of granting nurses a right to refuse to do something is that this can halt a medical procedure with harmful consequences to the patient. I believe nurses in these situations are dedicated to their careers and to their patients. They are either appropriately trained for the assigned tasks or agree that to suspend their doubts for the opportunity to learn new skills. They are not going to withhold their affirmation of their directed tasks unless they are very uncomfortable with what is happening. When the conditions exceed their comfort level, they should be allowed to object to proceed.
I understand modern nursing practice demands obedience to follow orders from supervisors and more advanced specialists. In general medical situations, there are real risks for even slight delays in performing tasks in many medical situations. I do not doubt that many of even routine procedures involves unexpected emergency situations that require quick thinking and improvisation with something different. Allowing a nurse to interrupt in these cases may be risky.
My suggestion is to widely expect nurse’s affirmations of every assigned task throughout all healthcare. They should be trained well enough that that affirmation is would rarely need to be withheld.
I realize it may be impractical to implement such autonomy for most nursing care situations. However, the case of treating a patient with a contagious disease with pandemic risks could be treated separately from other healthcare duties. In such circumstances there are multiple patients to be concerned about: the present one receiving health care plus all of health workers and wider community who could become future patients of this disease if a mistake is made. In this specific type of case, my suggestion may be warranted. I think the case in Dallas demonstrates the risks of of infecting employed healthcare workers and of infecting other patients or the general population.
Assuming that the above statement by the National Nurses United is accurate, the nurses could have refuse care to the patient until the appropriate equipment was available or satisfactorily improvised. When confronting a dangerous contagious disease that could lead to an uncontrolled pandemic, the safety of the health workers and the non-infected community should take precedence over treating the patient.
Distinguish legal risks for responding to contagious from non-contagious diseases
Two of the above discussions suggest that in some cases health providers may need to place priority on protecting themselves and the community at the detriment of the patient. In the first case, there should be an option to stop saving a life when the condition degrades or the risk of contagion becomes too high. In the second case, there should be an option to halt medical procedures if there unresolved concerns within the team about the safety of some procedure especially in terms of excessive risk of spreading the disease. These do not appear to be options available currently because we have an ethical priority to do everything possible to save the patient. Changing this perspective will require public debates about whether in the case of highly contagious and dangerous diseases the ethical priority should be placed on protecting the healthcare worker and the wider community at the expense of caring for the patient.
If we do decide that there is a higher priority to prevent further infection than to save the patient, then we need to accept that there should be a different standard for legal claims concerning the patient’s outcome when a contagious disease is involved. Protecting the patient may mean not attempting to resuscitate a late-stage Ebola illness, or not attempting to aid the patient due to lack of adequate protections to all of the health workers. We need to accept that these decisions are justified and not subject to claims of negligence of malpractice.
In the case of the index Ebola patient in Dallas, there are some discussions to consider a lawsuit for not doing enough save his life. Despite the tremendous efforts performed to try to save the patient under difficult conditions, the claims are that there are other things that could have been done that might have saved the patient.
In my opinion, the legal liability should change when dealing with a contagious disease because more lives are at stake than just the patient.
In this case, health workers in some cases willingly and knowingly took risks they should not have been requested to take. Their actions focused on saving this one patient may have unknowingly risked the health of other patients and their communities. They should not be coerced into taking these risks out of concerns for ethical violations and malpractice lawsuits.
Part of the debate we should be having right now before a real pandemic occurs is whether we should assure health workers that they are not risking legal problems when they place priority on protecting their own health and the health of other patients and their communities over providing medical attention to the unfortunate patient suffering from a dangerous contagious disease. This will be a very difficult discussion to have but resolving it now may help prevent spreading an epidemic through caregiver practices.
Contrary experience challenges legitimacy of evidence based decision-making
During this period of first Ebola case to appear in this country, public officials expressed evidence-based assertions that were quickly discredited by actual experience. Soon after the assurance of the unlikelihood of someone becoming sick with Ebola after arriving in the country, we learned that this actually happened. Shortly after the assurance that our health care system makes it unlikely that health workers would catch the disease, we learn of two cases where this happened when treating just one patient. I heard assurances that our systems would prevent symptomatic Ebola patients from having close contact with general public, but this has apparently happened. Then there is the assurance that when caught early enough, the disease can be managed with low risk of fatality and perhaps even a quicker recovery. I sincerely hope that assurance is met with the two infected nurses.
The point I want to make here is that we need human decision makers to thoughtfully consider the current situation. I continue to believe the above assurances are well supported by the evidence that we understand how this disease behaves and how it is transmitted, and supported by the evidence that our health care system is very high quality. In my opinion, the pronouncements of these assurances lacked the human thoughtfulness of recognizing that there is an uneven preparation for handling this disease: some hospitals are better prepared than others, some protocols have not yet been fully defined and explained, some hospitals do not have well trained health professionals to handle this particular type of disease. Instead I heard was appeared to be prepared statements that were not challenged by the leadership.
I see this as an example of what I complained about in earlier posts of evidence-based decision making obviating the role of the human decision maker. Under evidence-base decision making, given the evidence available at the time the leadership has no choice but to follow the evidence-based recommendation. In contrast, we normally expect accountable decision-makers to exercise appropriately the option to be skeptical or to express doubts.
Some of the resulting controversy and increased public concern is a result of the appearance of no human accountability. While it is true the evidence supports each of the assertions, there should still be some room for doubts and skepticism by leadership. Leaders should demonstrate that they are prepared for the possibility that this instance may be the exception to the general rule.
The appearance (from my perspective) is that the leaders did not prepare for the possibility of contrary experiences. Someone who is personally accountable for making the statements is more likely to have expressed doubts and skepticism and prepared for contingencies. Instead, there appeared to be an over-reliance on the authority of the evidence. The resulting confidence was easily shattered when contrary evidence appeared.
A good part of the current public concerns about our preparation for this disease may have been avoided if at first the leaders were more modest (and realistic) about their confidence of being prepared for all aspects of health-care for this particular disease.
Human accountability instead of automated decision-making
This continues the previous point with a focus on the vital importance of human accountability even in a era of evidence-based decision making. Even with the assumption that we have solid knowledge about this disease, how its spread can be controlled, and how we can optimally treat it, we still need humans to be accountable for the decisions. As I discussed in earlier posts (such as here), we employ accountable human decision makers to think critically about the evidence and that includes asking follow-up questions based on doubts and skepticism that comes from the knowledge that the decision-maker will be personally accountable for the decision. This accountability is important to maintain social order: accountability is an essential component of democratic governments.
It is still early in this current event, but already we have seen CDC Director Thomas Frieden (rightly or wrongly) be the one accountable for the events of the past couple weeks. In a purely evidence-based decision-making government, our focus could be exclusively on that evidence. Instead, we have hearings where we expect the human leader to defend his prior decisions and to convince the audience that he understands the current situation and the best path to take going forward.
Having this type of human accountability is very important to calm the public’s concerns about whether the right people are in charge and that they are making good judgments that include skepticism of the data presented to them. It is possible the Ebola epidemic will get worse for this country. We demand to know that responsible people with good judgement have some control over the events. We need him to give his own explanation about what happened and how decisions will be made in the future. We want to know not only that he is using the best information available, but also he is appropriately doubtful and skeptical about that information.
Postscript 10/17/2014: I just read this article that expresses a different point of view of some of my latter points about accountability.