More older adults died at their preferred place after implementation of a transmural care pathway for older adults at the end of life: a before-after study

The pathway contained five key elements. These were:

A GP specialized in palliative care and district nurses were already part of the hospital-based palliative care team. This involvement increased during the implementation, meaning the primary care professionals of the palliative care team attended meetings more often and had more contact with other members. Individual care plans were formulated for patients and discussed with professionals who attended the multidisciplinary team meeting, to which the patient’s own GP and community nurses were invited. If the patient’s own GP could not attend to the meeting, a warm handover was conducted by phone. During team meetings, further involvement was determined according to a colour code (green, orange or red), which represented a ‘complexity assessment’ based on symptom burden, patient needs, and whether the patients’ own GP could provide the necessary care. Green indicated low complexity with no further input needed from the palliative care team; orange indicated disease progression and/or inadequate symptom management; and red indicated a highly complex and unstable situation. The patient, GP, and community nurse all received a paper or electronic copy of the care plan. The transmural palliative care team visited the patient at home after hospital discharge at least once, and more often if needed based on the colour code for complexity.

We collected data from all first-time palliative care team consultations conducted during the study period. A list of patients who received a consultation during this time period was provided and data were collected from electronic patient records. Data included patient characteristics, consultation characteristics, readmissions, and mortality.

Primary outcomes were;

Secondary outcomes were changes in transmural collaboration: presence of the patient’s own primary healthcare professional (GP, district nurse, hospice resident) at team meetings, handover by phone, and follow-up home visits.

Independent variables were age, gender, the main diagnosis (malignant or non-malignant), WHO/ECOG performance status, estimated life-expectancy, admission ward, preferred place of death, place of death, time until death after consultation, and reason for consultation.

We analyzed the association between implementation of the pathway and the dependent variables. The implementation was divided in three phases: 1) pre-implementation phase: one year period before implementation started (Feb 2017 – Jan 2018), 2) during/shortly after implementation phase: six months period of the implementation and six months after implementation (Feb 2018 – Jan 2019), and 3) long term after implementation phase: one year period that started six months after the implementation ended (Feb 2019 – Jan 2020). The pre-implementation phase was used as a reference category. We analyzed changes on the individual patient level, and in every implementation phase, data from new patients were collected. Consequently, patients in the different implementation phases were not the same and data were not dependent. We performed multivariable logistic regression analysis to calculate and report adjusted odds ratios (OR) with corresponding 95% confidence intervals (CI). For each outcome, the included patients were tested for differences between implementation groups. Variables with statistically significant differences were discussed to assess whether these were relevant co-variables and if so, included in multivariable logistic regression analysis.

We used the STROBE guidelines [20] to report our results.

Patients had a mean age of 71.6 years (SD 12.6) and 50.4% were male. Most patients were diagnosed with cancer (71.3%), however, the proportion of consultations for patients with non-malignant diseases increased after implementation (pre-implementation: 20.8%, during/short-term: 32.7%, long-term: 31.5%). The majority of patients had a WHO/ECOG performance status of 3 (39.9%) or 4 (26.3%) and the estimated life expectancy was less than three months (30.5%). Most consultations were for patients admitted to the department of pulmonology or cardiology (36.9%) Almost half of all consultations (43.4%) were requested for advance care planning and/or guidance in the upcoming future. After implementation, the number of consultations for advance care planning did not change, but the number of consultations for advice in the dying phase decreased. The preferred place of death was discussed less often during/shortly after implementation, but more often long term after implementation (pre-implementation: 54.2%, during/shortly after implementation: 24.2%, long term after implementation: 58.2%) (Table 3).

There were 638 patients included to analyze changes in transmural collaboration. Attendance of the patient’s own primary care professionals (GP and/or district nurse) at team meetings was increased during/shortly after implementation, but decreased long term after implementation (pre-implementation: 49.1%, during/shortly after implementation: 71.3%, long term after implementation: 30.5%). Attendance of the patients’ own GP at team meetings decreased after implementation (pre-implementation 38.7%, during/shortly after implementation: 30.9%, long term after implementation: 15.8%). Warm handovers were done a little more often after implementation (pre-implementation: 34.6%, during/shortly after implementation: 38.9%, long term after implementation: 34.7%). Pre-implementation, one follow-up home visit was done, during/shortly after implementation as well as long term after implementation seven visits to the patients home were made (Table 3).

Patient who did not die during the hospital admission at which the palliative care team was first consulted were included in the logistic regression analysis for unplanned hospital admission, leaving 522 patients included in this analysis, of which 23.2% experienced an unplanned hospital admission within six months after consultation (Additional file 1: Appendix 1). We found statistically significant differences between implementation groups for the factors: malignant/non-malignant diseases, admission department, reason for consultation and actual place of death. Because the actual place of death was not considered as a confounder, but as a consequence of a hospital admission, we included the reason for consultation, malignant/non-malignant diseases, and admission department in the multivariable logistic regression analysis. Adjusted logistic regression analysis showed no statistically significant effects of the implementation (during/shortly after implementation adjusted OR: 1.23; 95% CI: 0.67–2.26; p-value 0.51; long-term after implementation adjusted OR: 0.79; 95% CI: 0.44–1.41; p-value: 0.42) (Table 4).

Most patients died within one year after consultation (88%) with a median of 18.5 days. We collected data on the preferred and actual place of death for 208 patients (Additional file 2: Appendix 2). Of these patients, 58.2% died at their preferred place. Within the sample of 208 patients, we found statistically significant differences between implementation groups for admission department, reason for consultation, preferred place of death. All three variables were included in multivariable logistic regression analysis. During/shortly after implementation, we found no statistically significant change in death at place of preference (adjusted OR: 2.12; 95% CI: 0.84–5.35; p-value: 0.11). Long term after implementation, the pathway was positively associated with death at place of preference (adjusted OR: 3.14; 95% CI: 1.49–6.62; p-value: 0.003) (Table 5).

Our study only showed improvement in transmural collaboration during/shortly after implementation, indicating the team had a more hospital-based approach. At least one of the patients’ own primary healthcare professionals attended more meetings during/shortly after implementation, but not on the long term. This could be caused by practical issues such as travel time [13], and the COVID-19 pandemic which could have made it less likely to visit other healthcare organizations. Moreover, the division between primary and secondary care, and perceptions on were palliative care should be provided [21] probably influenced the involvement of primary care professionals, especially GPs.

While a systematic review of Saunders et al. [6] suggested that palliative care involvement during hospital admission might decrease readmission rates, we found no association between implementation and (re)admission, in line with results of another transmural care pathway [22]. Members of the palliative care team felt responsible for the patients under discussion, and patients and GPs expected the palliative care team to become the main contact [13]. Consequently, patients and GPs might have had the tendency to reach out the hospital sooner in case of problems at home, leading to more hospital admissions. The pathway aimed to avoid hospital admission with the following key elements: [23] recognizing the approach of death, discussions about treatment wishes and symptoms, and monitoring through follow-up home visits. However, the transmural palliative care team still got involved mostly in the last weeks to months of life, limiting the time to discuss and prepare for the approach of death and to fix problems such as limited support systems, small houses with high stairs, and limited home care [24, 25]. These problems cannot be fixed by individual hospitals or professionals and require large scale improvement strategies on the social or (community) political level. In addition, the number of performed home-visits was low, which could indicate insufficient monitoring after discharge. The number of consultations before and after implementation requested for advance care planning remained equal, however, it could be possible that non-specialist performed more advance care planning discussions themselves after the educational program. This had no effect on the prevention of hospital admissions, but might have contributed to the effect on death at place of preference.

Overall, half of the patients in our study died at their place of preference, which is in line with previous research [26]. Our results showed that long term after implementation was associated positively with death at place of preference. A systematic review comparing different transmural palliative care approaches revealed that a hospital-based approach was most effective in facilitating death at home [27], which could explain why we found an effect on death at place of preference. We hypothesize that patient-level effects are not the result of individual elements of the pathway, but of education, increased attention for palliative care and the pathway as a whole including a change of culture in the hospital. This is supported by the increased number of consultations for specialist palliative care. There were also more requests for patients with non-malignant diseases. This demonstrates the importance of gaining knowledge about recognition of palliative care needs, which is considered to be difficult especially for patients with non-malignant diseases [14]. Also, long term after implementation the preferred place of death was discussed more often, which might indicate more attention for patients preferences. Since unplanned hospital admissions were not reduced, we think that for some patients, these admissions eventually made death at the place of preference possible because of better symptom management and time for arranging care at home, in care homes, and in hospices [28]. We found that on long term after implementation, especially the number of patient preferring and actually dying in hospices increased. This could be caused by an overall increase in hospice care [29], but could also be caused by the involvement of the palliative care team, [30] which might have led to more awareness of hospice care among patients.

A strength of this study is that we collected data from all new consultations conducted with the palliative care team before, during, and after implementation of the pathway, which provides insight into the short-term and long-term effects of implementation.

This study also has limitations, and therefore the interpretation of the changes we found should be made with caution. A limitation of this study is the retrospective nature of the study design. Because of that, changes in patient-related outcomes could have also been caused by other factors, which we did not measure. The patients that were recruited during the pilot study were screened for SPICT criteria and the surprise question while other patients were not. This could have caused differences between these groups, however, only eight patients were recruited during the pilot study [13]. We do acknowledge that a randomized controlled trial is the gold standard approach to evaluating effectiveness [31]. However, given the complex and time-consuming nature of improving transmural palliative care [32], we felt a before-after design would be the best alternative [33]. All data were collected from electronic patient records, which were based on reports made by the patients’ caregivers and members of the palliative care team. Consequently, we could not collect data about conversations and contacts that were not reported in the electronic patient record, which might have caused an underestimation of the transmural collaboration. Also, we could not collect data of hospital admissions in other hospitals. Considering the sample size, we could not include all possible relevant factors in the multivariable logistic regression analysis, but only included statistically significant factors. Future studies should involve prospective longitudinal data collection and collect data on symptom management, patient satisfaction with care, the number of care transitions and other patient-related outcomes.

In conclusion, our study shows that implementation of the pathway did not result in improvement of all patient-related outcomes and transmural collaboration. The pathway did not affect unplanned hospital admissions, but more patients died at their preferred place after implementation. The attention and awareness for in-hospital palliative care increased, but did not improve transmural collaboration on long term. For some patients, the hospital admissions might have helped in facilitating death at place of preference.