6.3.1 STUDY I
To our knowledge, Study I represents the first systematic review to assess the effects of staff training on resident-related end-of-life care outcomes. As a result of increasing research activity in this area, a large number of studies with robust RCT designs and with clinically oriented research questions were found. Most of the studies we included were not included in previous related reviews (Low et al. 2015; Anstey et al. 2016); thus most of the contribution is novel. Our findings suggest that influencing resident-related outcomes is challenging, and positive changes in staff-related outcomes often fail to translate into resident-level outcomes.
The quality of the included studies was mostly adequate. The quality and quantity of the studies enabled clinically meaningful conclusions from the evidence. Also, the high number of participants allowed sufficient statistical power to evaluate outcomes such as hospitalizations. In addition, the outcome measures used for QoL, QoD and SWC were well validated. Thus, the lack of effectiveness was not due to problems in availability of studies, statistical power, insensitive or inappropriate outcome measures, or inappropriate statistical methods.
While clinically relevant, the included studies differed in perspectives, interventions and outcomes. This diversity meant that it was not possible to perform a meta-analysis on any outcomes. The focus of the review was on training, but most of the studies involved complex interventions where training played a varying but often limited role. This made the assessment of the training effects challenging. The used method for evaluating methodological quality of the included trials seemed to give a useful categorical evaluation. However, it was not optimized to evaluate clustered trials. Recently, instead of numerical scoring, tools that involve a more
81
narrative approach to evaluation of methodological quality, such as the RoB2 instrument developed by Cochrane collaborators, are gaining popularity (2021).
6.3.2 STUDIES II–IV
The trial showcased in Studies II–IV has several strengths. It was a rigorous, single-blind RCT with a typical LTCF population and a long-term follow-up. To prevent contamination of the intervention, we randomized the participants in facility clusters. This design allowed us to pair-match the wards according to case-mix. Two thoroughly trained and experienced research nurses blinded to the group assignment gathered the data to ensure reliability.
The drop-out rates were small.
The applied intervention was clear, well described, valid from a theoretical point of view and based on modern learning methods. Also, the intervention was designed to be feasible and easy to adapt and scale up. The thematic contents of the intervention were informed by way of a learning needs survey.
To our knowledge, this is the first educational trial consciously using constructive learning theory and adult-education theory to improve EOL care in LTCF setting. This approach of learner-centered training, activating learning methods and reflection, has previously been shown to result in effective learning in the medical field (Spencer and Jordan 1999; Pitkälä et al.
2014). Lack of predetermined aims allowed the staff members to set their own learning aims for quality improvement and care-process change. This was believed to enable better motivation and deeper learning. Decision-making in LTCFs as regards end-of-life care has been shown to be only partly based on factual knowledge and more strongly on the attitudes and culture in the facility (Lopez 2009; Abrahamson et al. 2014; Laging et al. 2015). Therefore, the training sessions encompassed emotion-evoking components and participants’ reflective exercises. Three recurring themes of important notions during the training were observed from the feedback: (1) avoiding hospital transfers, (2) promoting ACP and communication with relatives, and (3) practical EOL care and medication. The possible harmfulness of hospitalizations seemed to be a novel and intriguing notion to many of the participants. Reporting resident-related outcomes was another strength of the trial. Also, trials in EOL or palliative care rarely include economic analyses such as that in the present trial.
The outcomes used were valid and the collection of hospital-related outcomes was 100% complete. Our measures for symptoms, HRQoL, PWB, and proxy satisfaction were valid and appropriate, they focus on domains considered important by residents and proxies (Gonella et al. 2019a) and have been shown to be sensitive to change (Routasalo et al. 2009; Kiely et al. 2012;
Zwakhalen et al. 2012; Pitkälä et al. 2014; Hui et al. 2016).
Owing to practical considerations, the number of clusters in our trial was small. Even after pair-matching all wards, based on MDS characteristics, there were differences between the intervention and control groups. These differences were adjusted for in the analyses but this naturally decreased statistical power. Also, some of the control group wards had taken part in
82
another palliative care improvement project just before this trial started. This may well have diluted intervention effects and produced some of the unbalanced baseline findings. High staff turnover is likely to have diluted the effects of the intervention. As a majority of all residents were recruited, no major biases in resident selection are suspected. However, it has been shown that participants in trials tend to have better functioning and prognosis than real-life populations (Strandberg et al. 2001). Also, residents and proxies with existing conflicts with facility staff could be more prone to decline consent to participate in the trial. The mean length of stay before death in the LTCFs was known to be only two years. Unexpectedly, only one in four residents died during the first follow-up year, suggesting that our inclusion criteria failed in selecting only the residents with poorest prognosis. Thus, this non-validated method for resident selection might not be appropriate. Other investigators have had similar difficulties in resident selection (Agar et al. 2017).
It is possible that the chosen learner-centered and open-aimed approach was too demanding in this short intervention. While the intervention left freedom for learners to choose the necessary care process changes, implementing these changes might have benefitted from support during follow-up. Good motivation for the training was observed in the sessions.
However, it is unknown whether or not such motivation was sufficient to change actual attitudes and practices in daily work. Considering all the above, expecting long lasting effects from such intervention in this setting was could be seen as overly optimistic. In the trial we wanted to measure resident-related outcomes. However, a qualitative study examining the changes in staff’s attitudes, competencies and practices might have given more insight into the learning and implementation processes. Whether inadequate intervention contents or faults in implementation lead to a lack of intervention effects can only be determined by measuring staff- and practice-related outcomes (Curran et al. 2012; Vernooij-Dassen and Moniz-Cook 2014). It might be beneficial if investigators in a training intervention were to consider the different learning needs related to EOL care of different occupational groups rather than giving all staff members the same training (Kaasalainen et al. 2017).
Another limitation of the intervention is that it did not provide the facilities with any additional resources in implementation. This is in contrast to most previous trials in Study I. We wanted the intervention to be feasible and easily transferable into real-life settings. However, this might have compromised its effectiveness. As the interventionists were not present in or available to the wards after the initial training, the intervention did not include possibilities to re-train the staff, answer questions or to support cultural changes that might have had an impact on resident outcomes.
Regarding the choice of primary outcome measures, assessing QoL should be suited to this frail population. While the 15D instrument has shown sensitivity to change in response to interventions in previous studies (Pitkälä et al. 2014), it might not be the best possible instrument for this very frail, cognitively impaired population. Moreover, such dimensions as mobility, speech, vision, or hearing are not likely to be modified by any EOL care intervention. The use of dementia-specific QoL indicators such as QUALID or QUALIDEM, or EOL care quality measures (e.g., SM-EOLD, CAD-EOLD)
83
could have advantages over 15D. The general validity of the used secondary outcome measures can be considered good. However, the psychometric properties of the Finnish translations of ESAS, PAINAD or SWC-EOLD instruments have not been studied in detail. Also, the way in which most assessments of ESAS and 15D and PWB were carried out, using a proxy informant, differs greatly from their original uses. Of note, the observed mean ESAS total scores in this trial were markedly lower than those previously observed for cancer patients in different settings (10 vs. 20–50) (Nekolaichuk et al. 1999; Mercadante et al. 2019; Lee et al. 2020).