My last post argued that from the perspective of gathering big data to improve predictive analytics for healthcare, we need to obligate patients to follow protocols. I emphasized that my personal opinion is in support of terminal-illness patients to make the rational choice that their remaining days may be better enjoyed by declining recommended procedures or even be choosing assisted suicide. I hope I have that option when my time comes. Despite this personal opinion, I offered another argument for why the patients should not have the option of assisted suicide. I attempted to show that the collection of new data to support better decisions in future health care requires us to observe the natural progression of the disease when the protocols are followed. I presented this argument earlier to suggest that we should not even permit the option to decline any recommended treatment.
These arguments used health care scenarios to illustrate a complain I have about the current trend of big data analytics to erode the role of an accountable human decision maker. I believe that accountable human decision-makers are essential to maintaining super-majority consent to be governed by smaller majority. The loss of this accountability will require implementation a more authoritarian government approach to keep the peace. Decision making by big data predictive analytics (a form of evidence-based decision making) inevitable obligates decisions to be made and obligates populations to cooperate. The health care example provides a vivid example where this obligation overrules the individual’s rational desire to avoid the additional suffering from attempts to extend the duration of their lives.
The evidence-collection rationale for obligating patients to accept treatment to live as long as possible has the same recommendation as religious rationale to allow a god to decide the time of death. There are many other justifications that lead the same conclusion of making the effort to extend life (postpone death) as long as possible. In the particular justification from big data analytics, the objective is to observe the new information about nature in order to improve future predictions.
As I presented in a number of earlier posts, I am bothered by this parallel of obligated decision-making and cooperation. The rationale for this obligation from big data is based on the evidence of incomprehensible volumes of data. This is similar to the religious rationale of incomprehensible religious texts. While the general population may understand pieces of the evidence, we need to defer to a tiny number of certified individuals to provide the definitive interpretation. Data-driven decision-making tends to obligate the implementation of analytic recommendations. Data-driven decision-making needs new data from what results from the recommendations in order to improve future analytic recommendations.
My motivation for writing the previous post was more about criticizing authoritarian tendency of big data, but I used concrete examples from healthcare as a vivid illustration of the extreme of denying a terminally ill patient to decline low-hope treatment or to have the option of assisted suicide. I agree with the patients in these examples: they should have these options and they have solid rational arguments for making their decisions. I want to support the option to decline treatment or even to end life early when there remains only few more months to live a life of rapidly declining quality. Unfortunately, the needs of science (or modern data science) provide an argument against allowing that option.
In the examples I presented in previous posts, the terminal illnesses typically involved some type of non-infectious internal disease such as cancer or an organ-destroying tumor. These illnesses affect only that one patient. Also, the rationale for declining treatment or ending life earlier is based on the notion that no available treatment offers a promising chance of survival.
Recent news provides a counter example the Ebola outbreak. The linked article offers some conflicting data indicating that the number of cases are expected to rise rapidly despite the recent data that shows recent week counts of new cases to be declining or holding steady. It also indicates that the fatality rate is estimated to be 70% although the documented data suggests closer to 50%. Even a steady number of new cases (not exponentially growing) with a 50% fatality rate is tragic, but for this discussion I’ll assume the worst case estimates of exponential increases of new cases with 70% fatalities.
Like terminal cancer, Ebola is also disease with very horrible consequences and with similarly poor prospects of survival. The difference is that Ebola is infectious. People can contract the disease by contacting others with the disease, with increasing risk as the patient’s conditions worsen.
This disease has similar or worse survival prospects compared with terminal cancer cases. However, for diseases like Ebola we do not grant the patient the options of declining treatment or of ending life early. At a minimum, we need to quarantine the patients, although “It would be horrifically unethical to suggest that we’re simply going to isolate people,” as this article quotes WHO Assistant Director-General Bruce Aylward. If the disease progresses to the final stages, we will attempt to provide as much care as possible to ease the suffering until death occurs naturally.
As noted earlier, I’m assuming Ebola is an epidemic that is currently out of control at least in certain countries in western Africa. We lack effective treatments to aid the sick and the affected African nations lack resources to stop the spread. Our current strategy to stop the exponential spread of the disease involve racing to find better cures and to prepare better treatment or disposal practices.
To assist these efforts to find better approaches to stop the spread of the disease, we need to evaluate our treatments and practices by observing the patients go through the available treatments. If they survive, their survival will provide evidence of good protocols. Also, as reported elsewhere, survivors can be a source of blood donation for transfusions that appear to help some cases. If the patient dies or the disease spreads despite our best efforts, then we can learn more about what is not working.
For Ebola, we are not permitting the option of declining treatment or the option of assisted suicide. The greater need of society to stop the spread of the disease obligates patients to accept every attempt to be cured. In the later horrific stages of the disease, I can’t imagine anyone declining offers of help. From reports of the disease, the eventual symptoms and fatality rates of Ebola are at least comparable to the worse cancers. But unlike cancer, it is imperative to make the effort to save every Ebola patient.
In my idle writing here, I have identified two categories of potentially (or even likely) fatal illnesses with no effective cure: infectious (like Ebola) and non-infectious (like cancer). The debate about the option to decline health care or to accept state-sanctioned suicide applies solely to non-infectious terminal and incurable illnesses. When these options are available, a patient can make a rational choice to take advantage of the certainty of the few remaining days of life that can be enjoyed instead of enduring life-disruptive (to say the least) treatments for little if any chance to survive to enjoy life later.
I previously argued that our quest to improve health care for all society may take precedence over the wishes of the individual. For cancers, we are steadily making progress in treatments. Each new case provides healthcare the opportunity to refine the treatments based on the experience from earlier cases. We make progress by observing what happens (good and bad) when we follow the best recommended approach based on what we currently know. That knowledge derives from analysis of data collected from previous cases. That data analysis will benefit from elimination of selection bias that occurs when individuals reject recommended treatments or opt to end life early.
It is cruel to impose on patients unwanted and harsh treatments until death occurs naturally. I have no expertise in healthcare or public health policies, but my goal of using these examples to illustrate unpleasant consequence of obligating obedience to automated decision-making as advocated by some promoters of big-data analytics.
I distinguish big data from clinical testing. Big data is statistical analysis of as-happened data exhaustively collected across the entire population, thus providing opportunities for statistical analytics (such as machine learning) to identify perhaps unexpected recommendations. In contrast, clinical testing involve carefully controlled experiments with some subgroup of the population that volunteers to participate in a study.
Legacy clinical approach permits discretion and voluntary actions such declining care. Big-data analytic approaches tend to forbid this discretion. As I discussed in earlier posts, the need to collect extensive and good data to support big data analytics eventually will obligate decision-makers to accept machine recommendations, and to obligate the population to cooperate (and forfeit their opportunity to address grievances). When big data analytics has access to comprehensive and high-quality data for entire populations, the evidence suggests the resulting recommendations can be superior to other forms of decision-making. The key to success of big-data approaches is to require participation in order to obtain more data unbiased by individual decision making.
I fear the consequences to social peace when we eliminate accountability from decision-makers by stripping them of their autonomy, and when we eliminate the options of the population to address their grievance when the recommendations result in injury. I suspect the younger generations have lesser fears of this consequence. Their confidence will be validated when big data solutions deliver more recognizable successes than devastating failures when we automate decision making at a policy level like healthcare. It seems inevitable we will experiment with automated evidence-based decision making even in healthcare.
Whether it is because I belong to an older generation, or because I’m naturally more pessimistic, I fear this experiment in automated decision making will fail. The future will deliver devastating failures not matter whether our decisions are accountable to people are to machines. When failures occur, human-accountability has the advantage of offering an opportunity to address grievances. The accountable human decision-maker can offer the opportunity to negotiate for a change in policy in response to bad consequences. That negotiation may involve replacing the human decision maker if he is unable to persuade the population that his methods were sound. Such opportunities are not available when machines make decisions. Instead, we must accept the bad consequences as an opportunity to collect data that will improve future decision making.
I described the big-data driven government as a form of autocracy and compared it to religious theocracies. Both involve a confidence that the source of wisdom is superior to human wisdom and also will tend to be benevolent toward humans in the long run. Big data analytics with benevolent goals is much like the loving God: it will reward us with paradise if only we can interpret and obey its recommendations.
Continuing the analogy of the theocracy, even among the believers in the source of superior wisdom, there remains the doubt that the human interpreters (the high priests, for example) correctly interpret the will of that intelligence. Theocracies typically have beneath its top decision-maker a larger community of qualified experts who argue extensively with each other. The same can happen in big data when different groups choose different approaches for selecting or rejecting data and corresponding algorithms for analytics and visualization. The same pool of data can present conflicting recommendations where only one can be acted upon. We claim to select the best possible recommendation based on statistical measures of various concepts of confidence. We reject the lesser confident recommendations.
The recommendation we choose will have consequences and we will demand that everyone accepts those consequences: that includes the dissenting data scientists and the overall affected population. The defense for this demand for accepting the consequence of recommendations is that the near-term hardships will provide valuable that that will reward us later with superior decision making. This eventual reward of a promised benefit despite near-term hardships is based on our confidence that we will eventually learn the right answers.
Certainly, human history of science (largely generalized as statistical practices) has uncovered some strong truths about how to control nature in beneficial ways. At times it seems that science already has explained everything in the universe and that the remaining challenges are just solving little isolated problems. For example, science understands how viruses operate and how they may be defeated, but each individual virus needs some work to figure out its particular weakness.
This brings me back to the current crisis of the outbreak and exponential spread of the Ebola virus. This is a virus. We understand viruses. It is a specific virus that we have studied and understand at least enough to assert scientific knowledge of how it propagates between humans. We are still trying to figure out how to treat this viral infection or ideally to immunize against future infections, but this is just a matter of applying known scientific principles on a specific example of this one virus.
The focus of this post is on the patient’s perspective instead of the scientific one. Consider a patient tested positive for Ebola immediately after experiencing early symptoms such as a mild fever. The patient knows the virus is in his body, and that it is inevitable that his condition will worsen substantially before it will get better. With current treatments, there is high probability that he will not survive especially in areas where the outbreak overwhelms local resources. From the perspective of the individual early-detected Ebola patient, the prospects of worsening quality of life with low chances of survival are similar or worse than the examples of the patients diagnosed with terminal cancers.
In the example of terminal cancer, we entertain the option of allowing the cancer patient to decline treatment, allowing the disease to take its natural course so he may enjoy his remaining days living as normally as possible. This option is not acceptable for the infectious disease. While the patient may be denied medical treatment by choice or by consequence of lacking treatment resources, we legitimately demand a quarantine even for when the current symptoms are tolerable and mild. The quarantine involves a loss of freedom. If they continue to live with family members, then society demands a quarantine for the family as well. The need to prevent the spread of the disease is greater than the desires of the patient to live a normal life before the disease turns worse.
Unlike cancer patients, we can not allow the infected patient to live his life freely. He must submit to treatment (at a minimum, this involved the quarantine).
There is the second option of state-sanctioned suicide. As mentioned the patient’s future prospects are similar for the recently diagnosed Ebola patient and the patient diagnosed with fatal and incurable cancer. There is an important difference in that state-sanctioned suicide for terminally ill requires a high certainty of death within a certain time period. In contrast, Ebola is survivable in at least 30% of the cases. A cancer with that kind of survival rate would not qualify for state-sanctioned suicide. The prospects though remain grim for both categories of patients. They are at least as likely to die as to survive, and their health is certainly going to worsen from their current conditions, even with treatment.
My tendency is to support an option of early suicide even for cancer patients when there is a low but significant chance of survival, such as this example of 30% survival rate. It is a tragedy to go through the progressive hardships of the diseases and its treatment only to end up dead any way. However, for cancer cases, we have many examples of grateful cancer survivors even when they experience significant degradation of life afterwards. The prospect of sometimes celebrating a successful treatment encourages to not sanction a suicide option. As noted, the US sanctioned suicide protocols require near certainty of eventual death within a few months.
For infectious diseases like Ebola, the argument is fundamentally different. First of all, the Ebola patient has far less time to make a decision. Debilitating symptoms occur only a few days after initial diagnoses. There is not enough time to come to as rational a decision about whether suicide is a good choice. The natural progression of the disease rapidly degrade ones ability to think rationally about suicide.
Ebola diagnosed patients face the prospect of a 30% survival rate after a certain period of severe illness that will subject the community with a high probability of infecting others. An option of an early exercise of suicide assures a certain death that may occur before the illness effects become severe. Also, an early suicide will end life when this virus is much less likely to infect others. We do not offer this choice for Ebola patients. Sanctioning suicide as a solution for an infectious disease seems profoundly unethical. Suicide eliminates the opportunity to survive the disease, and we have examples of grateful survivors.
The present problem is that we do not yet know how to contain this current outbreak. Lacking any effective treatments or immunizations, we have only the option of allowing the disease to take its natural course. According to news reports, the current trend suggest the disease infects 2 new people for everyone who gets it. The number of Ebola patients may double every few weeks (although as noted above, recent data is inconclusive that this is happening). Our current social measures of quarantine and sanitation (at least in West Africa, and arguably also in more affluent countries like USA) are ineffective in slowing the growth of the epidemic. Our medical options have no reliable remedies and even experimental unproven remedies cannot be produced in adequate quantities to make a difference.
We are facing a crisis where perhaps tens of thousands of healthy uninfected people today will be dead from Ebola in a few months. There does not appear to be anything we can do to prevent this outcome once we missed our sole opportunity by not containing this outbreak more effectively earlier when it involved a smaller number of individuals in very limited areas.
The unethical option of state-sanctioned suicide for patients diagnosed with Ebola might help slow the spread of the disease by reducing the exposure of the virus to others. Significantly, an early suicide and burial will prevent the virus to progress to the later stages where it is most infectious. Given the short time window available to make this decision, the suicide option may be state-encouraged in addition to state-sanctioned. This raises the objection of that a suicide patient is coerced into making this decision. To be effective, the decision to accept suicide would have to occur too quickly for rational contemplation by the patient. A rapid decision will likely require some external reassurance that this is the best answer. Once we introduce the coercion element, then we also allow the opportunity for abuse especially in territories with many local conflicts between competing subgroups.
Suicide for early detected Ebola patients is an awful suggestion. My purpose is not to advocate it, but instead to provide an extreme counter-example to the cancer victim who wishes to deny nature its natural progress for a terminal tumor. The earlier cancer examples permitting a state-sanctioned suicide for the personal benefit of the one patient at the expense of larger society learning from experimental treatments or how this particular case will progress naturally. I tend to support the decision to end life early before the cancer gets unbearable with certain death as an outcome. The counter-example of an Ebola victim is someone who is almost certainly going to experience a similarly terrible experience with the progress of the disease and will be as likely to die as a result.
We are not considering the option of early suicide for Ebola cases despite its potential to slow the spread of the disease, or even to stop the spread entirely and thus prevent future tragic infections. Currently, it might be the only option left to control the spread of the disease.
I mentioned ethical reasons for not permitting this option: given the short time available, the decision will likely require coercion and that will likely lead to corruption especially in very factious societies with weak central governments. In general, we publicly debate ethical considerations, but I have not encountered any public debate of this particular option.
One reason we don’t debate this option on ethical grounds is because we already dismiss this option on scientific grounds. Ebola is a virus. We have pretty good understanding of viruses, how they propagate, and what happens when they enter an organism. We also have a lot of information about this specific virus. Even though we do not currently know how to immunize against it, or even how to treat it effectively beyond a 30-50% survival rate, we are confident our science will eventually find a solution of finding a vaccine and finding cures with near 100% survival rates.
Ideally, that science will occur in experiments on non-human models, but we can also learn by trying the latest concepts to save current patients. Treating patients includes extensive documentation that can help in improving future treatments. Certainly, the current crisis is awful and certainly going to get much worse very quickly. However, we resign ourselves to only the option of managing the crisis by treating the current patients as effectively as we can until we inevitably will find a cure, a vaccine or some other discovery that will bring this epidemic back under control.
We are confident our science will find an answer. Inevitably, we will collect enough data to find a pattern that will lead to a solution that will put a stop to this crisis. We have science and statistics on our side, and these will never let us down. In the long run science and statistics will reward humanity with beneficial results. Thus we need to keep trying to save every single person who contracts this terrible disease. When we fail, the death will occur naturally.
Evidence of that confidence in that science is that we are not debating the ethical questions about an option to stop the epidemic through state sanctioned and encouraged suicide of early diagnosed Ebola patients. Ebola may be the nemesis of our science-inspired hubris. This may illustrate the dangers of automating decision-making based on data and algorithms alone and thus eliminate a potentially valuable role of an accountable human decision-maker to consider an option with tremendous ethical consequences.
Postscript: While writing the above, I read this post that presents the dangers of Ebola from the perspective of a practicing nurse who has experience with other infectious disease but not yet with Ebola. I welcome the long-form expression of a practical point of view. Most of this nurse’s post does not fit well with the theme of this post except for one comment: “When it comes to an epidemic of any sort, my first focus is on the patient, protecting and healing them, my second focus is on protecting the community.” This is a valid and respectable perspective for a nurse to take. But at some point an epidemic becomes so bad (such as the 1918-1919 influenza pandemic that killed 50-100 million people worldwide often with people dying within 24 hours) that a valid attitude may be to protect the community first and care for the patient second. In that pandemic of a century ago, communities were saved by refusing to allow people outside the community from entering (even if they were residents who were traveling). At some point, society confronts an ethically difficult decision of turning its back on the victims in order to prevent adding to their numbers. An option such as encouraged and sanctioned assisted suicide eventually must be considered when an epidemic becomes as severe as the 1918-1919 influenza pandemic.
Update 10/30/2014: Edited postscript to remove references to the nurse author as female.