Introduction
Explanation of Groups
Real Numbers
Conclusion
References
Revisions
Slide Download
A Public Health Approach to Aid-In-Dying
Introduction
From a public health perspective, the one test of a good policy is that it must help more people than it hurts. No policy is perfect--every treatment intervention can have side effects and unanticipated complications. Rare serious complications may only appear after large numbers of people have been given the treatment.
With this in mind, I decided to try running a few calculations to see if I could
figure out the ‘help-versus-harm’ ratio of assisted suicide--also called
‘medical-aid-in-dying.’
In order to do this I used the following logic and assumed the following:
1. The statutory safeguards would have some known error rate (the inverse of which is an accuracy rate).
2. The safeguards would divide people getting prescriptions into two groups, a qualified patient group and
a disqualified patient group. The disqualified patients are those who should not have been given a prescription due
to lack of decision-making capacity, coercion, or other disqualifying statutory conditions (like being a minor or under guardianship).
3. Both groups would decide to take (or not take) the medication at equal frequencies. This may not be actually
true, since disqualified people are disqualified for the very reason that they might be more prone to
undue influence in a decision to die. But, I’m using a conservative model that most favors the proponents of this policy.
Using the calculator below you can run various models of the ‘help-versus-harm’ scenario.
The safeguard accuracy number is how well you think the law works to distinguish between qualified and unqualified patients.
The ‘percent comforted’ number is the percentage of people--either qualified or disqualifed--who decide not
to take the medication. These are the people who are merely comforted by having a prescription--the real intent of the law.
This calculator generates the number of people who will be helped or comforted by the law versus the number who will
be harmed by a wrongful premature death. Play around with this for a bit. I'll tell you some of the real numbers later.
People who receive lethal prescriptions can be divided into one of four groups, as illustrated below. This explanation has been simplified by excluding the group of patients for whom we are missing data because they have been lost to followup. In most of these cases, we don‘t know if they are still alive, or the cause of death if they are dead.
Explanation of Comparison Groups
The ratio we need to think about is the comparative numbers in the green group relative to the
red group. This is the helped-to-harmed ratio (or the comforted-to-wrongful death) ratio. We don’t need to
consider the qualified dead group, because they haven’t been harmed--they got what they wanted. Similarly, we don’t
need to considered the accidentally comforted group because they also haven’t been harmed. They’re still alive
and may have derived some comfort from having the script. (They may have experienced psychological harm by being
coerced into accepting a lethal prescription, but again I’m assuming the case most favorable to proponents of these laws.)
The Real Numbers
As you can see, a safeguard accuracy of 80 will lead to one wrongful death for every four people comforted,
and any accuracy level below that point will lead to more than one wrongful death even if only half
the patients actually take the prescription. The proportion of wrongful deaths increases as safeguard accuracy drops.
Now for some real numbers. In the state of Oregon in 2015, 218 prescriptions were written and 132 people took them.
This gives us comforted percentage of about 40. In Washington the situation gets more complicated because we don’t know what
happened to 27 of the 170 people who got prescriptions in 2014. We know 126 took the medicine and died, while 17 didn't take the medication
and died of natural causes instead--presumably finding comfort in the script. Thus, Washington has only a 10% comforted rate--quite a
bit lower than Oregon (See Reference 1). Plug these numbers in and see how accurate the safeguards have to be to get only one wrongful death.
In order to have fewer people harmed than helped, the statutory safeguards have to be almost 90% accurate.
Now let's plug in some real safeguard numbers. This gets trickier because to date there has been no formal, methodologically sound
study of any of these safeguards. The country that reports safeguard data, which has a more rigorous review process than the U.S., is
Belgium. They claim an accuracy rate of 98%. BUT, at least 20% of their cases go unreported to review authorities--a safeguard
failure in itself. (Reference 3) So, the real safeguard number is really close to 80%--our calculated break-even point.
If you accept unreported cases as one measure of safeguard failure, then we do have some data to rely upon. An investigative report
of ten years of data in Washington and Oregon, done by the Des Moines Register in 2016, found that in 40% of reported cases
the reports were missing key data, giving us an accuracy rate of only 60%. (See Reference 2)
An accuracy rate this low virtually guarantees that more people are being harmed than helped by these laws now in the United States. The state of Oregon acknowledged this problem in the health department’s guidebook for healthcare providers. (See Reference 4) The Oregon safeguard procedures have not changed since 2008 when this guidebook was published.
The state of Colorado has reported results from the first year of their assisted suicide law. Nine of 69 cases were not reported by physicans, for an accuracy rate of 87%. Twenty-two cases had no written request, for an accuracy rate of 68%. Forty-two cases were missing the consultant's evaluation, for an accuracy rate of 39%. Only one patient received a mental health evaluation.
In spite of this clear failure to submit mandatory reports, all prescribing physicians attested that they followed the law. (See Reference 5)
Furthermore, for every incremental drop in safeguard accuracy there is an exponentially larger change in the ratio of wrongful deaths
to number comforted as Figure 1 illustrates below.
Figure 1. Small incremental decreases in safeguard accuracy cause a large change in the wrongful death ratio. This figure assumes 40% don't take the medication.
Conclusion
This policy model was designed using the best publicly available data, and is based on conservative assumptions that favor proponents of death-with-dignity or assisted suicide laws. It requires no a priori religious or ethical biases. It employs common public health principles: that wellness means providing the greatest possible benefit to everyone in the population, while balancing between burden and benefit. How well a given policy achieves this can only be determined by real data under real life conditions across diverse geographic, economic, and cultural landscapes. The difference between rates of lethal ingestion in Washington and Oregon could indicate disparities in access to health care generally, and mental health care specifically, and should warrant further study. More concerning, unknown ingestion status in about a quarter of those given prescriptions indicates a need for closer followup of these patients.
This demonstration illustrates the importance of safeguard accuracy, and also the importance of studying the causes of inaccuracy. Inter-rater differences in capacity assessments, prognostic variations, and prescriber bias could all affect this and have not yet been studied with regard to aid-in-dying or assisted suicide. Of course, this presumes the presence of a "gold standard" for each of these variables, which does not exist yet.
Within the "accidentally comforted" group there may be other reasons the lethal medication was not ingested. They may have had the good sense to tear up the script after they were coerced into asking for it. The medications may have been diverted for other purposes or thrown away. Also, psychological harm to this group was not considered in this model, nor was psychological harm to third parties such as minor-aged children of the patient, participating health care providers, and surviving spouses.
Presently there is no formal training or accepted methods for the detection of coercion. Although the law requires this to be assessed and ruled out, the evaluators I’ve spoken to have not been able to explain how they know a given patient hasn’t been coerced. They have relied only upon collateral sources who are cooperating and complicit in the process. This is because the patient controls the capacity evaluation; under current laws and bills, patients must give permission for the evaluator to contact people who might have information about coercion or lack of capacity. This is another potential safeguard deficit.
The purpose of this web site is to provide a preliminary mathematical model of the effect of aid-in-dying or assisted suicide laws. Calculation of harm-to-benefit ratios and wrongful death numbers should be considered seriously by policy makers and legislators.
Revision History:
| Date |
Revision |
| 2017/11/11 |
Site launched
|
| 2017/11/12 |
Real numbers updated to reflect 2016 data. Introduction edited to include mention of those lost to followup or for whom ingestion status is unknown. Conclusion also update to reflect this.
|
| 2018/12/4 |
Real numbers updated to include numbers from the Colorado 2018 annual report summarizing the first year of implementaton of their law.
|
| 2019/02/4 |
Updated reference number 4 to reflect the fact that the report was removed from the Oregon health department and OSHU web sites.
|
| 2019/10/21 |
Added slide download section and inserted within-page section navigation. |
Slide Downloads:
| Date |
Conference |
Download |
| 2019/10/24 |
Anorexia and Assisted Suicide: Coercion or Self-Determination?American Academy of Psychiatry and Law Annual ConferenceBaltimore, MD |
|
|
Mark Komrad MD |
PDF |
|
Annette Hanson MD |
PDF |
|
Angela Guarda MD |
PDF |
|
Patricia Westmoreland MD |
PDF |
| 2019/10/29 |
2019 Medical Ethics Conference: Medical Aid in Dying (MAID): The Discussion Has Arrived in VirginiaCarilion Clinic
Roanoke VA |
|
|
Annette Hanson MD |
PDF |
Must we?