This study was the first to investigate end-of-life decision making in overseas France and to compare with mainland France.

The frequency of medical decisions in this study illustrates the most likely chronology of decision-making when facing predictable death, where intensification of symptom treatment and treatment withholding come first, followed by treatment withdrawing and by sedation if previous measures were insufficient. Most decisions were made knowing their association with death, suggesting they were based on proper ethical deliberations and only a few are taken with the intention of hasten death. Deliberately hastening death with the use of drugs remained anecdotal.

High standards of care appear to be ensured in overseas France through a thorough adaptation of care provision to socio-cultural differences. The comparison with mainland data [7] shows similar decision making in both settings despite differences in provision of care, including a higher prevalence of home death, a lower provision of specialist and generalist palliative care, and younger physicians.

Local adaptation of care is yet performed in compliance with the law. The comparison with 2010 data shows continuous deep sedation until death, legally indicated in 2016 as a last recourse symptomatic end-of-life treatment, adds to the possible combinations of decisions without replacing other decisions.

Compared with mainland France [7], the most frequent main causes of deaths and their prevalence were similar in overseas France, yet with a doubled prevalence of endocrine, nutritional, and metabolic causes in overseas France.

Compared with Trinidad and Tobago [1], a comparable overseas setting, overseas France showed a higher prevalence of deaths due to cancer (26.6% vs. 18.6%) and a lower prevalence of deaths due to cardio-vascular diseases (24.6% vs. 31.9%), but with a sample restricted to home deaths in Trinidad and Tobago.

Home death was highly prevalent in overseas France and may have been slightly stimulated by hospital reorganisations to face the first outbreaks of the COVID-19 pandemic. Yet, place of death in France remained unchanged, unlike in other countries [13].

Like in mainland France [7], decisions were more frequently multiple – probably consecutive – than one single decision before death.

The decision to prolong life in the last month was less frequent in overseas France than in mainland France (34.8% [31.8–37.9%] and 37.0% [35.5–38.4], respectively). This result shows a statistically non significant difference in practices that is yet scientifically relevant as it suggests a trend towards a higher use of palliative care. More medical decisions seem to be made, with less situations left without medical decisions, which suggests a higher implication of physicians in end-of-life situations [14, 15]. In addition, less life-prolonging decisions suggest a trend toward more individually-tailored decisions at the end of life and thus better palliative care practices. Cancer deaths were less commonly associated with combinations involving life prolongation, as if the prediction of deaths prompted different responses.

Intensification of symptoms treatment was equally decided in overseas France and in French speaking part of Switzerland [16], with occurrences of 38.9% [35.7-42.0%] and 39.8% [36.8-42.9%] respectively.

Directly comparing our data on decision with those in Trinidad and Tobago was not possible due to major differences in follow-up time and inclusion criteria [1].

Decisions in overseas France are therefore close to those in territories with a similar socio-economic level like mainland France and Switzerland, despite different societal constructs and geopolitical histories.

Data on continuous deep sedation until death after the passing of the 2016 law [9] in France are scarce. The prevalence of continuous deep sedation until death was 13.3% [11.2-15.5%] in this study but is estimated 0.9% [0.6–1.2] in specialist palliative care services (hospital or home) [17]. This prevalence varies widely in Europe (2.5% in Denmark, 8.5% in Italy), with an increase to 18% in the Netherlands in 2015 [18]. Differences in prevalence may be explained by differences in definitions of continuous deep sedation until death [19], whose investigation should be completed by qualitative methods. Geographical isolation was no limit to law implementation into practice as showed by the frequency of continuous deep sedation until death only four years after the passing of the law.

Our study provides the largest report of medical decisions at the end of life in West Indies, with 723 questionnaires returned reporting (Guadeloupe, Martinique) on the last month of life in various settings and first in Indian Ocean (La Réunion). A previous study in Trinidad and Tobago reported on 96 home deaths [1]. In addition, one of the main strengths of our study is the robust methodology already used in the 2010 study in mainland France [8], with a complete anonymisation method that allows us to study illegal acts as well. Moreover, the design of data collection ensured a reduced recall bias with different waves of collection to reduce the time between death and sending the questionnaire. The multi-disciplinary scientific research team was key to combining medical and demographic approaches.

A limitation to this study is a lower response than in mainland France in 2010 (22% vs. 40%), due to the postal mail method, driven by the need to investigate potential illegal medical decisions. Response rate was maximised by defining a proper order of questions in the questionnaire, with demographics first [20], but was limited by the target population (physicians usually showing lower response rates than the public [21]), a high rate of physician turnover in some parts of overseas France, and the deliberate absence of financial compensation. The period of data collection, amidst the first COVID-19 outbreaks, may also have limited the availability of physicians for unrelated research despite extensions in the response periods. A general trend toward a lower response rate to surveys over time was previously described [22], but does not lead to a nonresponse bias.

The comparative aspect of this work is limited as it compares results 10 years apart of different territories with a changed legislation. It is difficult to disentangle the reasons for differences (cultural, changes of legislation, palliative care supply…). However, the results show no major difference expressing that health and end of life issues are managed differently in Mainland France in 2010 and overseas France in 2020.

The distribution of causes of deaths in our sample is close to the latest available [4, 5] statistics on causes of death of 2020, and suggest a marginal selection bias. The main challenge was to compare populations and decisions with the existing literature because of the wide disparity in methodology and presentation of results.