Continuity of care and treatment intensity at the end of life in Swiss cancer patients

Abstract

Background

Continuity of care (COC) was shown to be associated with fewer hospitalizations. We aimed to evaluate whether COC was associated with intensive intervention(s) at the end of life (IEOL), a preference-sensitive outcome, in cancer patients.

Methods

The study is based on claims data of patients with incident use of anti-neoplastics in Switzerland. COC Index, Usual Provider Continuity score, Sequential Continuity index and Modified Modified Continuity Index were calculated based on consultations with the usual ambulatory care physician. Treatment intensity was evaluated in the last 6 months of life, and COC was evaluated in months 18–6 before death in those who died between 24 and 54 months after incident cancer. IEOL comprised life-sustaining interventions (cardiac catheterization, cardiac assistance device implantation, pulmonary artery wedge monitoring, cardiopulmonary resuscitation/cardiac conversion, gastrostomy, blood transfusion, dialysis, mechanical ventilator utilization and intravenous antibiotics) and measures specifically used in cancer patients (last dose of chemotherapy ≤14 days of death, a new chemotherapy regimen starting <30 days before death, ≥1 emergency visit in the last month of life, ≥1 hospital admission or spending >14 days in hospital in the last month of life and death in an acute-care hospital).

Results

All COC scores were inversely associated with the occurrence of an IEOL, as were older age, homecare nursing utilization and density of ambulatory care physicians. For COC Index, odds ratio was 0.55 (95% confidence interval 0.37–0.83).

Conclusions

COC scores were consistently and inversely related to IEOL. The study supports efforts to improve COC for cancer patients at their end of life.

Introduction

Due to the complexity of the disease and the common presence of additional chronic conditions, medical care of cancer patients requires multidisciplinary collaboration including specialist physicians, such as surgeons or oncologists, and general practitioners (GPs), typically located in different ambulatory practices or hospitals.¹ Transitions between these healthcare providers are highly vulnerable parts of the delivery of high quality and safe care, especially in decentralized healthcare systems such as those present in Switzerland. However, as much as 90% of Swiss decedents who died of cancer were transferred at least once in the last 6 months of life—mainly from home to an acute-care hospital—with a median number of three transitions.² Based on previous cross-sectional and longitudinal studies, better continuity of care (COC) was associated with fewer hospital admissions.^3–5 A recent Swiss study showed that primary care continuity was associated with lower healthcare costs, fewer hospitalizations and a lower risk of death in cancer patients.⁶ COC is therefore considered a marker of quality of care in cancer patients.^7,8 Given the close link between COC and shared decision-making, trustful communication and making patient preferences more explicit, it is reasonable to evaluate and quantify the relationship between COC and preference-sensitive outcomes like the use of intensive intervention(s) at the end of life (IEOL) and its drivers, as adverse or unintended outcomes triggered by a lack of continuity are avoidable.

This study aimed to evaluate the association between COC and IEOL. We hypothesized that COC was associated with a lower likelihood of IEOL, because COC improves coordination and therefore leads to care that is more in line with patient’s preferences and though likely less aggressive.⁹ The impact of COC on different outcomes is of great interest, as the number of patients with chronic diseases or multimorbidity is growing. Former research has shown that it is feasible to assess COC^4,6,10 and treatment intensity at the EOL^2,11,12 using routine claims and further administrative databases. To the best of our knowledge, this relationship has not been evaluated so far in Switzerland.

Methods

Design, data source and study population

The present study was conducted using health insurance claims data from the Helsana Group, covering about 1.4 million mandatory insured persons, largely representative of the Swiss population. The Helsana Group has a high market share in Switzerland. Every fifth death in 2014 was covered by the Helsana data set. Since the benefit basket in Switzerland is administered federally, including all appropriate and cost-effective inpatient and outpatient services, the coverage is independent of the insurance company. Therefore, data are not bound to specific groups of the Swiss population. Persons can opt for higher annual deductibles (ranging between 300 and 2500 Swiss Francs), resulting in lower monthly premiums. In addition to mandatory insurance, persons can choose supplementary hospital insurance, which covers further comfort such as single-room accommodation. In Switzerland, there is a free choice of physician, unless you opt for a managed care organization model, which leads to lower premiums.

Adult men and women with mandatory health insurance at the Helsana Group between 2014 and 2017 were considered. All patients with an incident prescription of an anti-neoplastic agent (based on the Anatomical Therapeutic Chemical Classification code L01) in the investigated year were included and followed up for a minimum of 2 years.^6,13 Incident cancer was defined by no prior use of any anti-neoplastic agent in the preceding 12 months. In order not to focus on the acute and very treatment-intense initial phase of cancer care, COC and outcome measures were only observed beginning 6 months after incident cancer.⁶ Treatment intensity was evaluated in the last 6 months of life and COC was evaluated in months 18–6 before death in those enrolees who died between 24 and 54 months after incident cancer. Patients with less than three consultations were excluded, because COC cannot be reliably assessed based on few consultations.^4,14 Of the 2978 decedents after incident cancer, 923 had to be excluded because they had less than three consultations within 1 year. The patients’ flow diagram is shown in the Supplementary material.

Measures

Patient characteristics were assessed at the time of the incident anti-neoplastic agent prescription: age (19–49, 50–69 and 70+ years), sex, language region (German, French or Italian), community character (urban, peri-urban and rural), the type of health insurance plan (supplementary hospital insurance, managed care model and deductible level), the year of incident cancer and patients’ co-morbidity. Twenty-one additional chronic conditions were identified based on the Anatomical Therapeutic Chemical classification system, using an updated version of the Pharmacy-based Cost Group model by Huber et al.¹³ Additionally, the following healthcare measures are provided per patient/year (all assessed simultaneously with COC): the number of primary care physician consultations (primary care physicians, general internists), the number of specialist consultations, as well as the number of different physicians contacted in both categories. Furthermore, homecare nursing utilization (yes/no), the proportion of patients admitted to a nursing-home and the median (mean) length of stay (in days) in a nursing home were evaluated. Information regarding the cantonal supply of care (density of beds in acute hospitals, nursing-home beds, home-care nurses and density of ambulatory care physicians) was derived from hospital statistics (KS), statistics on social medical institutions (SOMED) and statistics on homecare services (SPITEX), all administered by the Swiss Federal Statistical Office. Physician density information (GPs, oncologists and other specialists) of the corresponding year was provided by the Swiss Medical Association (FMH).

COC in the ambulatory setting was measured using the following four different indices which have been widely applied (see Supplementary material for details).^4,6,10,14,15

Continuity of Care Index (COCI): it weighs both the frequency of consultations with each physician and the dispersion of consultations between physicians.

Usual Provider Continuity index (UPC): it describes the proportion of consultations to the physician the patient visited most frequently.

Sequential Continuity index (SECON): it measures the number of consultations made to the physician whom the patient saw in the most recent visit.

Modified Modified Continuity Index (MMCI): it concentrates on the plurality of physicians consulted.

All indices can be applied in primary and specialty care and are, therefore, suitable for populations having chronic conditions such as cancer, who may experience continuity relationships with GPs as well as specialists. They all range from 0 (lowest) to 1 (highest COC Score). Consultations were defined as a visit with direct physician contact.⁶

Outcome variables

We defined intensity measures at the EOL according to Hanchate et al.¹⁶ as the occurrence of any of the following 10 intensive life-sustaining interventions: cardiac catheterization, implantation of a cardiac assistance device, pulmonary artery wedge monitoring, cardiopulmonary resuscitation or cardiac conversion, gastrostomy (for artificial nutrition), blood transfusion, dialysis, use of mechanical ventilators and intravenous antibiotics. Additionally, we used cancer-specific intensity measures based on the study by Earle et al.,^11,17 defined as the occurrence of any of the following indicators: last dose of chemotherapy within 14 days of death, a new chemotherapy regimen starting <30 days before death, at least one emergency visit in the last month of life, more than one hospital admission or spending >14 days in hospital in the last month of life and death in an acute-care hospital. The place of death was determined by the last claim received and further categorized into hospital, nursing-home or home/others.² Although these intensity measures have not been formally validated, they were repeatedly used in previous administrative data-based studies^16,18,19 as well as by the American Society of Clinical Oncology. The quality of these EOL intensity measures was regarded as good.¹²

Statistical analysis

Descriptive statistics were used to characterize the study population, stratified by treatment intensity at the EOL. Differences between the samples were calculated using Kruskal–Wallis test for continuous, Chi-squared test for categorical and Fisher’s exact test for dichotomous variables. Multiple logistic regression models were used to evaluate the association between the COC indices with the usual ambulatory care physician and IEOL. We calculated odds ratios (ORs) with their 95% confidence intervals (CI). Correlations and interactions were tested between various outcome and/or explanatory measures. Finally, we performed stepwise model selection using Akaike's Information Criterion.²⁰ Statistical significance was determined at the 0.05 level. All analyses were performed using R version 4.1.0.

Ethical approval

A formal request to the Ethics committee in the Canton of Zurich (Kantonale Ethikkommission Zürich) confirmed that no further ethics approval was needed as the study is in accordance with the national ethical and legal regulation (article 22 of the Swiss data protection law) and it falls outside the scope of the Swiss Federal Act on Research involving Human Beings (Human Research Act).

Results

The total study population comprised 2055 cancer patients. The median (interquartile range [IQR]) age of the study population was 77 (15) years and 48% of the patients were female (table 1). IEOL were found in 1492 (72.6%) patients. They were more common in men and in younger patients, as well as in those living in a canton with a lower density of ambulatory care physicians, presumably in rural regions.

Table 1.

Characteristics of the study population, by the occurrence of at least one intensive intervention at the end of life (n = 2055)

	Total	No intensive intervention	≥1 intensive intervention	P a
n (%)	2055	563 (27.4%)	1492 (72.6%)
Male sex	1068 (52.0%)	257 (45.6%)	811 (54.4%)	<0.001
Age				<0.001
19–49 years	58 (2.8%)	6 (1.1%)	52 (3.5%)
50–69 years	492 (23.9%)	75 (13.3%)	417 (27.9%)
70+ years	1505 (73.2%)	482 (85.6%)	1023 (68.6%)
Additional chronic conditions				0.671
0–1	335 (16.3%)	97 (17.2%)	238 (16.0%)
2–4	1074 (52.3%)	286 (50.8%)	788 (52.8%)
5+	646 (31.4%)	180 (32.0%)	466 (31.2%)
Language region				0.139
German	1578 (76.8%)	448 (79.6%)	1130 (75.7%)
French	305 (14.8%)	70 (12.4%)	235 (15.8%)
Italian	172 (8.4%)	45 (8.0%)	127 (8.5%)
Community character				0.808
Rural	268 (13.0%)	77 (13.7%)	191 (12.8%)
Peri-urban	401 (19.5%)	106 (18.8%)	295 (19.8%)
Urban	1386 (67.4%)	380 (67.5%)	1006 (67.4%)
Managed care	914 (44.5%)	231 (41.0%)	683 (45.8%)	0.059
High deductible	145 (7.1%)	33 (5.9%)	112 (7.5%)	0.210
Suppl. hospital insurance	542 (26.4%)	144 (25.6%)	398 (26.7%)	0.653
Cantonal supply of care, mean (median density)
Acute hospital beds	2.9 (3.0)	2.9 (3.0)	2.9 (3.0)	0.076
Nursing-home beds	65.6 (70.7)	66.2 (70.8)	65.4 (70.3)	0.054
Homecare nurses	2.3 (2.1)	2.3 (2.1)	2.3 (2.1)	0.972
Ambulatory care physicians	8.4 (8.5)	8.5 (8.9)	8.3 (8.3)	0.035

^a P-values were calculated using Kruskal–Wallis test for continuous, Chi-squared test for categorical and Fisher’s exact test for dichotomous variables.

The most frequent IEOL were cumulative hospital admissions or long hospital stays, death in an acute hospital as well as cumulative emergency visits (table 2). Regarding procedure-specific interventions, blood transfusions in the last 6 months before death, start of a new chemotherapy regimen in the last month before death, and the use of intravenous antibiotics in the last 6 months before death were most frequently recorded. Patients with at least one IEOL more often had face-to-face consultations, were less often admitted to a nursing-home and spent fewer days in nursing-homes (table 3). Differences in relation to the number of different physicians visited during the observation period were small.

Table 2.

Number of patients with intensive interventions at the end of life (n = 2055)

≥1 intensive intervention	1492 (72.6%)
Cardiac catheterization	23 (1.1%)
Cardiac assistance device	8 (0.4%)
Pulmonary artery wedge monitoring	2 (0.1%)
Resuscitation/cardiac conversion	36 (1.8%)
Gastrostomy	25 (1.2%)
Blood transfusion	479 (23.3%)
Dialysis	38 (1.8%)
Mechanical ventilation	51 (2.5%)
I.v. antibiotics	71 (3.5%)
Last dose of chemotherapy within 14 days of death	1 (0.0%)
Starting a new chemotherapy regimen ≤30 days before death	204 (9.9%)
>1 emergency visit in the last month of life	258 (12.6%)
>1 hospital admission or spending >14 days in hospital in the last month of life	1219 (59.3%)
Death in an acute-care hospital	1129 (54.9%)

Table 3.

Healthcare utilization in months 18–6 before death, by the subsequent occurrence of at least one intensive intervention (n = 2055)

Median (IQR)	Total	No intensive intervention	≥1 intensive intervention	P a
	2055	563 (27.4%)	1492 (72.6%)
No. of consultations with GP	8 (4–13)	6 (2–11)	8 (4–14)	<0.001
Mean (SD)	10 (10)	8 (8)	10 (11)
No. of consultations with specialist	5 (1–12)	4 (1–9)	5 (1–12)	<0.001
Mean (SD)	9 (12)	7 (10)	9 (12)
No. of different GPs	1 (1–2)	1 (1–2)	1 (1–2)	0.010
No. of different specialists	2 (1–4)	2 (1–3)	2 (1–4)	<0.001
Homecare nursing utilization, n (%)	693 (33.7%)	207 (36.8%)	486 (32.6%)	0.082
Nursing-home admission, n (%)	315 (15.3%)	206 (36.6%)	109 (7.3%)	<0.001
Days in nursing-home	0 (0–0)	0 (0–182)	0 (0–0)	<0.001
Mean (SD)	34 (99)	92 (147)	13 (61)
Ambulatory continuity of care
COCI	0.50 (0.35–0.78)	0.57 (0.40–0.85)	0.49 (0.33–0.75)	<0.001
Mean (SD)	0.56 (0.27)	0.61 (0.26)	0.55 (0.27)
UPC	0.70 (0.53–0.89)	0.75 (0.57–0.92)	0.67 (0.50–0.88)	<0.001
Mean (SD)	0.70 (0.21)	0.74 (0.20)	0.69 (0.21)
SECON	0.63 (0.43–0.87)	0.71 (0.50–0.90)	0.62 (0.40–0.85)	<0.001
Mean (SD)	0.64 (0.27)	0.69 (0.25)	0.62 (0.27)
MMCI	0.86 (0.77–0.93)	0.88 (0.80–0.94)	0.86 (0.75–0.92)	<0.001
Mean (SD)	0.86 (0.21)	0.86 (0.12)	0.83 (0.14)

COCI, Continuity of Care Index; GP, general practitioners; MMCI, Modified Modified Continuity Index; SECON; Sequential Continuity index; UPC, Usual Provider Continuity index.

^a P-values were calculated using Kruskal–Wallis test for continuous and Fisher’s exact test for dichotomous variables.

Unadjusted COC indices were significantly higher in patients with no IEOL (table 3). This holds true for all four ambulatory COC scores. The healthcare providers determined most frequently as the usual provider in the ambulatory setting for the COC measures were general internal practitioners (72.1%), medical practitioners (10.4%) and oncologists (5.7%). Regarding the association between ambulatory COC and single interventions at the EOL, bivariate analyses revealed that patients with a higher COCI were less likely to start a new chemotherapy regimen starting <30 days before death (P = 0.006), to have more than one emergency visit in the last month of life (P = 0.031), to have more than one hospital admission or spending >14 days in hospital in the last month of life (P = 0.008), and they were less likely to die in an acute-care hospital (P < 0.001). Highly similar results were found for the other COC scores, except for SECON and starting a new chemotherapy regimen as well as for MMCI and cumulative emergency visits, where differences between both samples did not reach statistical significance.

The results of the multiple logistic regression model showed a constant, inverse relationship between all ambulatory COC scores and the occurrence of at least one IEOL. For COCI, the OR was 0.55 (95% CI 0.37–0.83, P = 0.004, figure 1 and Supplementary table S1). Scaled to correspond incremental one-tenth increases, an increase of the COCI score by 0.1 reduced the odds of an IEOL by 5.8 (95% CI 1.9–9.5). However, due to the wide CIs, these results need to be interpreted with caution. Patients aged 70+, when compared with patients aged 19–49 years, and patients using homecare nursing services had a lower odds of an IEOL (OR = 0.31, 95% CI 0.12–0.68, P = 0.007 and OR = 0.65, 95% CI 0.52–0.81, P < 0.001, respectively). Besides, the number of days in a nursing-home and the cantonal density of ambulatory care physicians were inversely associated with an IEOL (OR = 0.89, 95% CI 0.81–0.97, P = 0.006 and OR = 0.99, 95% CI 0.99–0.99, P < 0.001, respectively). In contrast, patients living in the French-speaking when compared with the German-speaking regions of Switzerland, and those having two to four additional chronic conditions, when compared with patients with one at the most, had a higher odds of an IEOL (OR = 1.74, 95% CI 1.24–2.46, P = 0.002 and OR = 1.56, 95% CI 1.16–2.09, P = 0.003, respectively). The cantonal density of beds in acute-care hospitals was also positively associated with intensive intervention occurrence (OR = 1.37, 95% CI 1.01–1.89, P = 0.048). Density measures stratified by physician specialty (GP or oncologist) did not yield statistically significant associations neither did health insurance covariates nor urbanity. Therefore, these covariates were not part of the regression model.

Figure 1.

Multiple logistic regression model on the occurrence of at least one intensive intervention at the end of life (n = 2055).

Regression analyses of the association between all other COC measures and at least one intensive intervention hardly altered the results of all the covariates and yielded similar ORs: OR = 0.46 (95% CI 0.27–0.77, P = 0.003) for UPC, OR = 0.54 (95% CI 0.36–0.82, P = 0.004) for SECON and OR = 0.25 (95% CI 0.10–0.59, P = 0.002) for MMCI.

In a sensitivity-analysis, we divided the COCI into three categories, based on their mean values (0–0.3, 0.4–0.6 and 0.7–1), leading to a nearly equal distribution of patients. In the multiple regression analysis, patients in the medium COCI category and patients in the highest COCI category had a lower odds for an IEOL (OR = 0.73, 95% CI 0.56–0.95, P = 0.002 and OR = 0.72, 95% CI 0.55–0.95, P = 0.021, respectively) compared with patients with a COCI below 0.4. The density of beds in acute hospitals was the only covariate that changed noticeably. It was no longer significantly associated with intensive interventions.

Discussion

This is the first study—to our knowledge—investigating the link between COC and a variety of IEOL, a very frequent outcome in Swiss cancer patients. We observed IEOL in 73% of all cancer patients. Almost 55% of all patients died in an acute-care hospital. Nearly 10% started a new chemotherapy regimen in the last month before death. Although, according to the recommendations by the European Society for Medical Oncology, chemotherapy should not be used in the last weeks of life in adult cancer patients.²¹ Our findings are in line with our previous findings, where 9% of cancer patients started a new chemotherapy regimen in the last month before death and 56% died in an acute-care hospital.² Comparable findings were observed internationally. In Portugal, 14% of patients with advanced solid tumours have started a new chemotherapy regimen within 30 days of death.²² Analyses in four Canadian provinces revealed that 54% of cancer decedents died in-hospital.²³ The review by Langton et al. showed that chemotherapy was delivered to 1–19% of cancer patients in the last 14 days and that almost 40% received chemotherapy or life-sustaining interventions (cardiopulmonary resuscitation, intubation and/or mechanical ventilation) in their last month of life.²⁴ The proportion of people with at least one hospitalization in the last year of life ranged from 54% to 76% across 15 European countries, whereby hospitalizations were comparably more likely in patients dying of cancer.²⁵ Multiple emergency visits were found in 9–18% of cancer patients in previous studies compared with 13% in our study using the same time span.^11,26

Our results show a robust, consistent and inverse association between ambulatory COC indices and the subsequent occurrence of at least one IEOL. Previous studies investigating the impact of COC on end-of-life care in cancer patients have shown that COC was associated with reduced acute care.^5,27,28 Increasing COC was associated in a dose–response relationship with decreased odds of in-hospital death, hospitalizations and emergency department visits in the last 2 weeks of life in Canadian cancer patients.⁵ Similarly, patients with advanced cancer who died in-hospital had lower odds of having received higher COC in another Canadian study.²⁸ Findings are also in line with previous research in elderly patients with ≥3 chronic conditions, where a higher COCI was associated with a decreased risk of inpatient admissions and emergency department visits.¹⁴ A more recent study in patients with end-stage renal disease also found reduced costs for inpatient and emergency visits at the EOL, and a decreased utilization of mechanical ventilation in those patients with higher COCI scores.¹⁰ In contrast, a study of Medicare patients with advanced cancer did not find an association with primary care physician continuity and EOL care intensity.²⁹ However, they only looked at primary care continuity. In our study, GPs were regarded as the usual provider in less than three-fourth of all cases. Presumably, more often specialists, especially oncologists or geriatricians, than GPs are involved in the treatment process of cancer patients at the EOL when compared with other stages of life. The MMCI seems to have a stronger lowering effect on IEOL in our study. This measure does not consider the number of consultations with a usual provider but rather the plurality of physicians consulted. It is plausible that lower numbers of providers are indicative for fewer IEOL in a sample of cancer patients with a high disease burden.

Older age, using homecare nursing services, number of days in a nursing-home and cantonal density of ambulatory care physicians were inversely associated with at least one intensive intervention, while having two or more chronic conditions and living in the French-speaking regions of Switzerland were significantly and positively associated with at least one IEOL. Likewise, Canadian patients with increased homecare nursing services had lower odds of a hospitalization at the EOL.³⁰ Higher treatment intensity or higher rates of in-hospital death in French-speaking regions have been found previously in Switzerland.²

Limitations

Due to the retrospective design of the present study, no causality can be established between COC and treatment intensity. Moreover, some anti-neoplastic drugs might also be used in non-cancer diagnoses, such as Morbus Crohn. We were also unable to discriminate the different levels of disease severity or stage of disease by means of our data. Patients with a more severe disease course might be more likely to regularly visit an oncologist and—at the same time—receive a lower COC score. Additionally, these patients might have more intensive treatments or a higher likelihood of a hospitalization for other conditions. However, by skipping the first 6 months of initial cancer care for the calculation of our COC and outcome measures, we tried to minimize this bias. Further factors possibly explaining the relationship between COC and IEOL, such as the living conditions, could not be considered.

Implications

Between 2013 and 2017, the annual incidence of new cancer diagnosis in Switzerland amounted to 23 100 in men and 19 650 in women; the increase in the number of new cancer diagnosis of 8.5%, compared with the years 2008–12, was due to the demographic development, since age-standardized rates hardly changed.³¹ Healthcare patterns in cancer patients are therefore of great interest. Many patients with serious illnesses, especially cancer patients, experience treatment and interventions that are inconsistent with their preferences, and treatment does not meet their needs.³² As opposed to differences in the distribution of population factors such as sociodemographic characteristics, health service provision can principally be modified and might therefore have an important impact on cancer care in Switzerland. The question of how EOL can be supported as appropriate as possible in patients with chronic diseases has increasingly been focused on in research and practice.³³ Therefore, more intensive cancer care at the EOL was not always regarded as better care.³⁴ While some differences in IEOL undoubtedly are justified, others might not be advisable for cancer patients at the EOL and indicate over- or underuse of treatment options.^35,36

Our results indicate that a higher COC index was associated with fewer IEOL, such as the start of a new chemotherapy regimen or cumulative emergency visits in the last month of life. But whether an intensive intervention is appropriate and fully complies with the patients’ needs and preferences is unclear. Yet, the usual providers know their patients and their patients’ needs over a long term and should, therefore, be able to optimize and coordinate their treatment and anticipate deteriorations, thereby reducing unnecessary healthcare utilization of acute hospital services. Furthermore, cultural aspects may have an influence on IEOL. Eventually, nurses, palliative care teams and/or geriatricians may support the GP or usual care physician in improving healthcare and reducing unwarranted IEOL. Palliative care utilization was shown to be inversely related to using life-sustaining procedures, such as blood transfusion and cardiopulmonary resuscitation, among colorectal cancer patients.³⁷ Various studies have shown that early use of palliative care may, among others, result in lower in-hospital mortality, prolonged life expectancy and less pain in dying patients.^38–40

Conclusion

Higher COC scores in the ambulatory setting were consistently and inversely associated with at least one IEOL in the present study. In line with previous research, this study supports efforts and initiatives to improve and foster (interpersonal) COC for cancer patients or patients with chronic diseases at their EOL, for example, by the development of models of integrated service delivery for EOL care.^10,17,28 Investigating the association between COC and preference-sensitive outcomes like IEOL and searching for underlying factors potentially contributing to unwarranted variations in chronic care is important to improve healthcare for cancer patients and to optimize resource use at the same time. Further prospective research is needed to investigate whether interventions to increase COC in cancer patients reduce IEOL and what proportion of these intensive treatments has been performed contrary to patient preferences.

Supplementary data

Supplementary data are available at EURPUB online.

Funding

This study was funded by the Swiss Cancer Research Foundation and the Swiss Cancer League [grant number: HSR-4944-11-2019].

Disclaimer

All authors gave approval for the final version of the manuscript and agree to be accountable for all aspects of the work.

Conflicts of interest: None declared.

Data availability

The dataset analysed during the current study is not publicly available because it is part of the confidential Helsana health insurance claims database. However, additional information is available from the corresponding author on reasonable request.

Key points.

• This study examined whether continuity of care was associated with preference-sensitive outcomes like intensive interventions at the end of life in cancer patients.

• Higher continuity of care was consistently and inversely associated with an intensive intervention at the end of life.

• The study supports efforts and initiatives to improve continuity of care for cancer patients at their end of life.