Integrative Palliative Care Service Model Improved End-of-Life Care and Overall Survival of Advanced Cancer Patients in Hong Kong: A Review of Ten-Year Territory-Wide Cohort

Patient selection

Patients who died from cancer in all local public hospitals managed by Hospital Authority (HA) of Hong Kong from four years (2006, 2009, 2012, and 2015) were retrospectively reviewed. One hundred patients were randomly selected at each time point from each of the seven service clusters (Hong Kong East, Hong Kong West, Kowloon East, Kowloon West, Kowloon Central, New Territories East, and New Territories West) of HA for detailed review and analysis. A total of 2800 patients was yielded, which formed a representative cohort of cancer death covering the whole geographical areas of Hong Kong.

Data collection

Records of these patients were retrieved from electronic patient record system. Data were collected in four parts through the review of detailed clinical records.

Part 1 consisted of demographic data, including age, gender, primary cancer, residence before death episode, date of diagnosis, and death.

Part 2 consisted of data of palliative coverage. Palliative care coverage was defined as specialist palliative care provided by either palliative physicians or oncologists in palliative wards, hospice, outpatient ambulatory palliative care clinic, or palliative care home care team. The service provision was documented in the territory-wide, highly unified electronic health record system of Hong Kong Hospital Authority. Hong Kong runs a unique combined oncology service with integrative palliative care service. ¹⁰ In all the six public cancer centers in Hong Kong, there were multidisciplinary palliative care teams providing comprehensive services. The teams were led by consultant oncologists with dual specialist qualifications of oncology and palliative medicine. One-third of practicing oncology specialists in Hong Kong also received palliative medicine training. The palliative care teams were under the same administrative department of oncology teams, which enabled the smooth transition of patients from active oncological treatment clinics to palliative care services. Besides the oncology departments, the palliative care physicians (under the College of Internal Medicine) also run hospice units and palliative clinics at various public hospital networks in Hong Kong. There was close collaboration between oncologists and palliative care physicians to provide complementary services to advanced cancer patients.

Part 3 consisted of data of death episode confined to the last 30 days of life. Data collected included place of death, prescription of strong opioids, use of chemotherapy and systemic anticancer treatment (including targeted therapy, immunotherapy, and hormonal therapy), ICU admission, documentation DNACPR, and performance of cardiopulmonary resuscitation (CPR) at the end of life.

Part 4 consisted of data of health care resource utilization during last six months of life. This included the number of specialist outpatient clinic (SOPC) attendances, accident and emergency (AED) visits, and the total duration of hospitals stay (acute, convalescence, and hospice stay).

Statistical analysis

Weighting adjustment was performed before analysis based on the population size of each HA service cluster. Normally distributed data were expressed as mean ± standard deviation (SD) and other continuous data were expressed as median. Continuous variables were compared by the Mann–Whitney U test. Categorical variables were compared using the chi-square test or Fisher's exact test if appropriate.

OS, which was calculated from the date of diagnosis of cancer to the date of death from any cause, was the prespecified survival endpoint in this study. Kaplan–Meier estimation of the survival outcome and log-rank tests were employed for unadjusted comparisons of survival difference between patients receiving palliative service and the counterpart.

Due to unbalanced baseline characteristics, there was an unavoidable selection bias influencing comparison of survival outcomes between patients receiving palliative care and the counterpart. Propensity score matching (PSM) analysis, which is a statistical method adjusting for potential baseline confounders, was used to minimize the impact of selection bias in this study. A logistic regression model was established to calculate the propensity scores. The use of palliative care service was considered a dependent variable, while age, gender, cancer types, use of chemotherapy, year of death, and hospital cluster were considered predefined covariates. Patients who received palliative care service were matched with those who did not in a basis of nearest neighbor matching at a 1:1 fixed ratio for survival analysis.

All tests were two tailed and a p < 0.05 was considered statistically significant. All of the analyses were performed using SPSS statistical software package version 25.0.

Ethical consideration

This study was supported by all seven service clusters of HA and was approved by the institutional review board (IRB) of all clusters (File Number UW 17-430).

Baseline characteristics

For the randomly selected cohort of 2800 patients, the mean and median age were 69.9 and 72 years (range 3–103 years; SD 13.8 years). The male to female ratio was 1.5:1. Lung cancer (32.5%), colorectal cancer (17.1%), and liver cancer (13.0%) were the most common primary cancer. The patients were further categorized into two groups based on palliative care coverage (palliative care group, n = 1770 and nonpalliative care group, n = 1030). Their baseline characteristics are listed in Table 1.

Palliative coverage

Over the past decade, the overall palliative care coverage in Hong Kong increased from 55.4% in 2006 to 68.9% in 2015 (p < 0.001). Significant increase in palliative care coverage was observed in two service clusters (p = 0.024 and p = 0.044, respectively). The performance of palliative care coverage of all of the seven service clusters over the past 10 years is demonstrated in Supplementary Figure S1.

Impact of palliative care provision on end-of-life outcomes

The impact of palliative care provision on end-of-life outcomes is shown in Table 2. Patients who had received palliative care service were less likely to have chemotherapy or systemic cancer treatment within 30 days of death (both p < 0.001). Fewer patients who had palliative service admitted to ICU within 30 days of death (0.9% vs. 4.3%, p < 0.001). The use of strong opioids was more common in palliative patients (45.8% vs. 27.1%, p < 0.001). Among all cancer deaths, 3.4% had CPR performed. Patients with palliative service had less CPR done (p < 0.001).

Across the decade, the end-of-life care outcomes on the rate of CPR and ICU admission were stable as demonstrated in Table 3. However, in the past 10 years, there was about 100% increase of the use of chemotherapy or systemic anticancer treatments in the last 30 days of life (3.0%–6.3% and 4.0%–8.0%, respectively, p < 0.02). This was observed in patients with and without palliative care service provision. Younger patients were also more likely to receive chemotherapy or systemic anticancer treatment in the last 30 days of life (p < 0.001) (Supplementary Table S1).

Survival outcomes

Among 2800 cancer patient deaths, the median OS was 3.87 months (95% confidence interval [CI] 3.55–4.19 months). PSM analysis was performed to adjust for potential baseline confounding variables, including age, gender, cancer types, use of chemotherapy, year of death, and hospital cluster. A cohort of 1958 patients was matched in which patients with and without palliative care service were in 1:1 ratio. Table 1 showed the details of the variables before and after PSM, noting that the impact of potential selection bias was reduced. Median OS in patients with palliative service was significantly better than those without palliative service (5.10 months vs. 1.96 months; 95% CI 4.52–5.68 months vs. 95% CI 1.66–2.27 months p < 0.001) (Fig. 1).

OPEN IN VIEWER

FIG. 1.

Overall survival of patients with and without palliative care.

Health care resources utilization

Utilization of health care services was different between palliative and nonpalliative care group. Patients in the palliative care group had more SOPC visits (p < 0.001) and longer hospital stay (p < 0.001) in the last six months of life, although the duration of last admission stay at acute general ward was shortened (p < 0.001) (Table 4). Across the decade, the average utilization of health care resources for advanced cancer patients (both palliative and nonpalliative groups) remained stable (Table 5).

Impact of end-of-life outcomes: DNACPR

DNACPR is an elective decision not to perform CPR when patients approach death after considering patients' best interests and wishes. DNACPR, as shown in our study, was commonly addressed across all settings in Hong Kong in the past decade. ¹¹ This was reflected by the fact that only 3.4% of patients in our cohort received CPR. Such low figures might be attributed to the promotion of DNACPR policies and guidelines in public hospitals. ¹⁰ It was encouraging to observe that palliative care patients had a lower CPR rate than those who did not receive the service. The difference was significant at 1.4% versus 6.7%. The finding was consistent with other palliative care series.^12–14

The use of strong opioid

Pain is one of the major concerns at end of life. Adequate cancer pain management helps maintain quality of life. ¹⁵ Nearly 40% of all cancer patients received strong opioids within the last six months of life. Patients having palliative care (45.9%) were more likely prescribed strong opioids, while most of their counterpart (27.2%) had no strong opioids prescribed. Difference in pain profile between groups and the tendency of nonpalliative physicians to prescribe weaker analgesics could explain this phenomenon. Other concerns and misconceptions such as fear of side effects and addiction might limit the decision of strong opioid prescription and cancer pain management. In Hong Kong, one-third of practicing clinical oncologists had received structured training on opioid use, so they were likely more confident in opioid prescription. ¹⁶

ICU admission

Admission to ICU within 30 days of death is one of the indicators of palliation. Lowering the admission to intensive care settings reduces the cancer patients' use of uncomfortable or painful interventions and maximizes the time for family gathering.^4,17 In our study, patients having palliative care had significantly lower rate of ICU admission. This was consistent with most of the other reported series. The positive impact may be due to the better communication and alignment on goal of care between families and oncology teams, such that the patients and families would not have unrealistic expectation on life-sustaining ICU care.

Aggressive treatment

Another indication of palliative care is the use of chemotherapy and other anticancer therapies, including targeted therapy, hormone therapy, and immunotherapy during the last stage of life. Having aggressive chemotherapy or other systemic anticancer treatments is likely harmful, costly, and not conducive to a peaceful end of life.^2,3 Patients with palliative care were less likely prescribed such treatments in the last 30 days of life as demonstrated. However, in the past 10 years, there was a significant increase in the use of chemotherapy or systemic anticancer treatments in end-of-life period. This was observed in patients with and without palliative care service provision. This was consistent with the advancement in oncology that patients were having many more options of anticancer treatments in recent years. This highlights the ever-increasing importance of palliative care in the era of rapid advancement in oncology treatments.

Impact on OS

We demonstrated in propensity score-matched analysis that the use of palliative care was associated with longer OS. The estimated OS benefit of three months was similar to other randomized control trials.^2,6–8 The mechanism for this OS benefit was not fully understood. Avoidance of overaggressive treatments, which might bring harm toward the end of life, could be a possible reason. Optimization of symptom controls by palliative care could also explain the phenomenon. Nonetheless, we expect our findings would help decision makers realize the need to expand the use of palliative care in the health system. It might enlighten cancer patients and their families to the benefits that palliative care could provide.

High utilization of health care resources

Heavy reliance of hospital-based service for palliative care was demonstrated. The high utilization of resources was consistent with the hospital-oriented health care settings of Hong Kong. In the last 6 months of life, advanced cancer patients spent an average of 4–6 weeks in hospital and no improvement was observed across the past 10 years. In addition to admission length, AED and SOPC visits were also the cases. These caused high burden to the patients and their caregivers, as well as the public health system.

Hospitalization is stressful to patients and their caregivers. It also takes away precious time from the families. Hence, strengthening of community care for reduction of unnecessary hospitalization is important. It was listed as one of the strategic directions of palliative care development in Hong Kong in 2017. ¹⁸

Enhancement of community care could be achieved through increasing day care and home care, support to residential care homes, and community partnership instead of solely relying on limited resources of local public hospitals. A sponsored pilot project for strengthening home care in Hong Kong, which was led by palliative physician, specialized palliative home care nurses, and trained health care assistants, successfully reduced the length of stay of the last admission by more than half (18.4–7 days). ¹⁹ Different nongovernment organizations are also exploring different models of community care to help patients stay at home for palliative care. Close collaboration between hospitals and community or home care units is the key for success of the programs. Long-term commitment in resource input and high-level strategic planning are eagerly awaited.

Study limitations and future directions

This study employed a retrospective design and did not provide patient-reported outcome (PRO) as well as quality-of-life data. PSM might reduce the bias, but was not comparable with prospective randomized trials. Nevertheless, the electronic patient database of HA was very comprehensive and accurate. Follow-up data were survival and health care resource utilization was near complete due to the relatively low geographical mobility of Hong Kong.

For future service development, research and service innovation are essential. In long run, service improvement in palliative care requires information feedback on the quality of care. Currently, there is a big information gap about the quality of care provided, as the palliative care outcomes studied were largely based on administrative data, including this study. The use of PRO measurements, empowered by the use of mobile technology and big data analysis, will make continuous quality monitoring possible. Finally, prospective clinical trials are irreplaceable for advancement of evidence-based palliative care.