A Briefing Note dated 22 November 2017 that was prepared by Prof. Penney Lewis of King's College London provides an excellent review of implementation of assisted dying regimes in Europe, the USA, and Canada. The full Briefing Note can be downloaded from here.

These conclusions on page 5 of that Briefing Note reflect a very positive evaluation for the systems for provision of VAD that were reviewed:

What is known about the effectiveness of safeguards?

An extensive body of empirical evidence exists relating to the safeguards and criteria outlined above, and how they operate in permissive regimes, with the most detail available in the Netherlands, Belgium, Oregon and Switzerland. The evidence from these jurisdictions suggests that the legal criteria that apply to an individual’s request for assisted dying are well respected: individuals who receive assisted dying do so on the basis of valid requests; third parties who assist individuals to die do not act unlawfully.

What is known about reporting?

Evidence of the effectiveness of the reporting requirement and the scrutiny of reported cases in the Netherlands, Belgium, Oregon and Switzerland is less consistent. There is no data on the reporting rate in Oregon. The reporting rate within the right to die organisations in Switzerland may be 100%. The reporting rate in the Netherlands rose when the RERCs were inserted as a buffer between physicians and the authorities, although the Swiss experience suggests that a buffer may not be needed to encourage reporting if the process leading up to the assistance involves several layers of administration involving a number of different actors coupled with few legal requirements.
The reporting rate is significantly higher (81% in 2015) in the Netherlands than in Belgium (53% in 2007) where legalisation occurred more recently. The reporting rate has risen over time in the Netherlands; it is not yet known whether this is the case in post-legalisation Belgium. “The major reason for failure to report [a case as euthanasia] is that the physician does not regard the course of action as a life-terminating act”. These unreported cases frequently involve the use of non- typical drugs to cause death (morphine rather than barbiturates and/or muscle relaxants which are typically used in euthanasia cases) and/or a very short life expectancy. The number of estimated deaths from euthanasia includes such cases, as it does not rely on doctors’ labelling of their own practice. Since almost all cases involving typical euthanasia drugs are reported, this inconsistent labelling now likely accounts for almost all unreported cases. This thesis is supported by anonymous data collected from physicians which indicates that consistently close to 100% of the acts termed by physicians as euthanasia and assisted suicide were reported.

What is known about vulnerable groups?

In 2007, researchers examined data from the Netherlands and Oregon in order to see if members of vulnerable groups were more likely to receive assistance in dying (either euthanasia or PAS). They examined the frequency of such assistance in ten groups of potentially vulnerable patients, defined by gender, age, ethnicity, educational and socio-economic status, illness and disability. They found “no evidence of heightened risk ... with the sole exception of people with AIDS.” It should be noted though that the lack of Oregon data on pre-existing disabilities weakens the force of this conclusion with respect to disability.
42 The researchers concluded that “the available data ... shows that people who died with a physician’s assistance were more likely to be members of groups enjoying comparative social, economic, educational, professional and other privileges.”

What is known about the frequency of end of life decisions?

Many of the empirical claims made about the practice of euthanasia and PAS under existing legal regimes misrepresent the data, take it out of context or neglect important comparisons with jurisdictions where these practices are prohibited. Chart 1 shows the percentage of all deaths in specific years that were cases of euthanasia (EUT), PAS or termination of life without request (TLWR). It combines data from a number of different anonymous prevalence surveys of doctors. All surveys were based on one originally designed by Dutch researchers. The relatively broad and overlapping confidence intervals suggest that fine comparisons should not be made between countries with the lowest percentages. As indicated, some comparisons are from different years. Although similar, the surveys are not identical. The percentage of deaths in which an end of life decision (ELD) is made varies across jurisdictions.

This evidence does not support the argument that there is a slippery slope between the legalisation of euthanasia (termination of life on request) and termination of life without request (TLWR).
The rates of TLWR vary. The evidence suggests that TLWR takes place in both permissive and non-permissive jurisdictions, with some of the highest rates in non-permissive jurisdictions (eg Australia in 1997), although rates of TLWR in some permissive jurisdictions are higher than in some non-permissive jurisdictions. TLWR occurs more frequently than euthanasia in all countries that have been surveyed except the Netherlands and Belgium. Rates of TLWR have decreased since legalisation in the permissive jurisdictions of the Netherlands and Belgium