Discussion
Our hospital-wide implementation of the MOST order significantly improved both the frequency and quality of documentation of orders for the use or non-use of life-sustaining treatments.
The rate of hospital-based documentation of decisions about life-sustaining treatment is highly variable and the quality frequently poor15 16 leading to unwanted treatment, family distress, clinician burnout and higher costs.1 3–7 For example, Fraser Health Region, one of BC’s most populous health authorities reports MOST completion ranging from 29% to 75% depending on the site.17 Similarly, Alberta Health Services, reports Goals of Care Directive completion ranging from 45% to 89%.18 While it is impossible to compare study data to audit data directly, the degree of improvement we observed in documented orders for life-sustaining treatment (from 33% to 100% over 8 months) shows that the MOST order can be quickly and effectively implemented in the clinical setting. Further, an audit of months 3–6 after the project concluded (as part of a planned follow-up study) showed that improvements in this primary outcome had been sustained with a rate of 92% being observed (average of month 3–6).
We also observed an improvement in the quality of documentation in terms of reducing discordance. Standardising documentation with the MOST is associated with a substantial decrease in discordance between patient’s preferences for life-sustaining treatment and prescribed orders. Though measured differently, previously published multicentre studies of seriously ill hospitalised patients report high levels of discordance.19–22 To our knowledge, this QI study is the first to observe that a standard documentation tool is associated with a reduction in discordance between patients’ preferences and their prescribed orders, suggesting that a standard order set (MOST) is better than usual care.
The dramatic improvement in the rate and quality of EOL care documentation observed with MOST implementation is associated with a small, statistically significant improvement in satisfaction with decision-making, previously identified by Heyland et al as the highest priority for improvement of end-of-life care.8 9 True informed consent, however, requires a reasonable understanding of the risks and benefits of available healthcare options, thus, highlighting an important remaining target to improve EOL communication.
This study did not observe any difference in the use of acute care resources between patients with or without a MOST, as defined by the number of days patients are admitted to an acute care facility within 90 days of study inclusion. There is no standard way to measure acute care utilisation, limiting comparisons between studies. Available studies, however, show an inconsistent effect of ACP on acute care utilisation. For example, the SUPPORT trial23 found that facilitated discussion had no effect on reducing length of hospital stay, while Zheng and Edes reported decreased length of stay with professionally facilitated discussions about future medical care.5 24
Implementation lessons
This project encountered a number of barriers that are frequently experienced in QI studies, such as competing organisational priorities, change fatigue, clinical and administrative resistance and financial limitations. With an upcoming hospital transition, hospital administration and staff were already experiencing change fatigue. It was therefore not surprising that SJGH administration was hesitant to take on another large-scale clinical change. While the timing of our MOST implementation was suboptimal, it was mandated by VIHA and already being rolled out throughout the health authority. Failure to implement the MOST locally would have led to conflicting goals of care documentation between co-dependent hospitals, clinical uncertainty and potentially conflict. With health authority wide administrative support and local appreciation of the resulting clinical risk, SJGH agreed to implement the MOST locally. Ideally, we would not recommend implementing multiple significant changes at the same time. It was important that allied HCP resistance was managed by local clinical nurse educators, while physician concerns were addressed by the lead author (SK), who was locally considered to have clinical expertise. The financial administration of this project took significant resources and would not have been possible without external funding and continuous study optimisation.
Limitations
There are several limitations to this study. First, allocation to the MOST intervention was not randomised, making it more difficult to make inferences about causation. A single-centre randomised controlled trial, however, would be almost impossible to execute because contamination would be unavoidable, and a cluster randomised controlled trial was not feasible with available resources. Nevertheless, the significant improvement in ACP documentation is most likely due to MOST implementation, as opposed to other contemporaneous factors. Second, due to timeline and resource limitations, we were only able to collect baseline data for 1 month, limiting reliability of preintervention data. Third, to evaluate the impact of our intervention, we only included patients who consented to data collection. As half the eligible patients we approached did not consent to participate in our study, there is a possibility of a selection bias, limiting the generalisability of our findings to centres using administrative data. Fourth, because the MOST form and education strategy were simutaneously implemented, this study cannot differentiate each component’s contribution to the observed improvement. Fifth, discordance was measured by comparing a self-reported MOST category with chart documentation. We assumed that competent patients would be able to understand the simplified MOST form. This study’s strengths, however, include its tangible clinical results, use of validated tools and rigorous data collection processes.