Socio-demographic and disease related characteristics associated with unplanned emergency department visits by cancer patients: a retrospective cohort study

This study aimed to explore socio-demographic (patient) and disease related characteristics of cancer patients presenting to an ED within 28 days of receiving infusional, systemic anti-cancer therapy in a Day Oncology Unit (DOU), in order to identify those at greater risk of making an unplanned ED presentation, and to make recommendations for risk stratified models of follow up care.

For the purpose of our study, infusional, systemic anti-cancer therapy (SACT) was defined as: any type of treatment that uses drugs or other substances to travel through the bloodstream to identify and attack cancer cells, either by blocking or damaging certain enzymes, proteins, or other molecules involved in the growth and spread of cancer cells; or modulation of the immune system; or by stopping normal and cancer cells dividing or by directly causing cell death [24].

The objectives of the study were to explore socio-demographic and disease related characteristics associated with:

Andersen’s Behavioural Model of Health Service Utilisation was used to guide the exploration of factors that may be associated with ED utilisation in this study. The model has been widely used to explain variations in health service utilisation in community, acute and emergency care settings [25].

This study was conducted at a large, public tertiary hospital in Victoria, Australia which includes a dedicated cancer centre, with two oncology/hematology inpatient wards as well as a dedicated palliative care unit. The centre also consists of a radiotherapy unit and a DOU that has 18 treatment chairs and treats approximately 70 patients per month. The cancer centre services a complex case mix of patients and includes a large cancer clinical trials unit and provision for allogeneic hematopoietic stem cell transplantation. The hospital has a large Level II ED and trauma centre. The ED contains a Short Stay Unit (SSU) where patients may be admitted for further care following an ED presentation for a period of no longer than 24 h [26].

During business hours at the cancer centre patients who experience or have concerns about symptoms or side-effects of treatment are directed to contact a Clinical Nurse Consultant (hematological cancers, or patients receiving oral therapies or concurrent chemotherapy/radiotherapy in solid tumours) or call the DOU to speak to one of the senior oncology nurses. After hours, patients are directed to call the oncology inpatient wards for symptom support or advice. The DOU does not have a dedicated space for ad-hoc patient reviews, and commonly patients are provided with telephone advice, advised to attend their General Practitioner, or present to the ED.

Patients were eligible for inclusion if they received SACT in the DOU between 1 December 2014 and 30 November 2017. Patients treated with subcutaneous treatments or who received supportive therapies only (for example, zoledronic acid) were not eligible and therefore not included in the final dataset.

All episodes of care (that is, any attendance for SACT) in the DOU during the study period were linked to any ED presentation at the study site that fell within 28 days from the date of discharge from the DOU. Episodes of care that were not allocated an ICD-10 code were excluded. The final dataset included specific variables available from the Victorian Admitted Episodes Dataset (VAED) and the Victorian Emergency Minimum Dataset (VEMD). The VAED and VEMD are comprehensive datasets that collect details about the causes, effects and nature of illnesses as well as details on health service use for all public hospitals in the State of Victoria [27, 28].

Demographic and clinical variables collected included age, gender, postcode, country of birth, marital status, preferred language, interpreter required, Aboriginal or Torres Strait Islander status, usual accommodation, attendance source, time of triage, ED length of stay, mode of arrival to ED, triage category using the Australasian Triage Scale, ED discharge destination and inpatient length of stay. Cancer type was classified according to the International Classification of Diseases and Related Health Problems, 10th Revision, Australian Modification (ICD-10-AM) [29]. Data were screened for outliers such as those who made a high number of ED presentations, and who had an exceptionally long ED or inpatient length of stay recorded and were verified through a manual chart review.

Data were analyzed using IBM SPSS version 25.0 (Chicago, USA). The data analysis framework was developed to address the four research questions set for the study. Univariate analysis was performed, and continuous data reported as mean and standard deviation (SD) unless otherwise stated. Categorical data are reported as frequency and percentage. Country of birth, primary language and use of interpreter for in-hospital communication were highly correlated and thus, only country of birth was included in regression modelling.

Marital status was collapsed from six categories to two categories “married, de facto” and “single, divorced, widowed” representing those reporting having a significant other, or not. Cancer diagnoses were classified according to the ICD-10 coding and were collapsed from 14 categories into 10 with thyroid, bone and soft tissue, rare, gynecological, skin – non-melanoma and unknown primary grouped into ‘other’ cancers. Hematological cancers were used as the reference category, as they effect both males and females equally, encompass a variety of individual diagnoses and associated symptoms, are the largest tumor stream treated at the cancer center, and had the highest number of unplanned ED presentations.

The Socio-Economic Indexes for Areas (SEIFA) is a product developed by the Australian Bureau of Statistics which ranks areas in Australia according to relative socio-economic advantage and disadvantage [30]. SEIFA codes for all included patient records were ranked from lowest to highest and organized into quintiles. The top 20% represents those with the highest socioeconomic status, those that fall between 40 and 80% represent the middle bracket, and the bottom 40% represent those with the lowest socioeconomic status.

Destination from ED was grouped into two categories (discharged to place of primary residence or admitted to the inpatient area). Country of birth was collapsed in to two categories, born in Australia or born outside of Australia.

A binary outcome was defined for all multivariate logistic regression models as ED presentation/s “Yes/No”.

Adjusted odds ratios (AORs), 95% confidence intervals (95% CI) and p values are provided for variables associated with ED presentation and place of discharge. p values at < .05 were considered statistically significant. Effect size was calculated on AORs [31].

Patient demographics are shown in Table 1. Where individual demographics were used in analyses, age was taken at the first DOU episode of care. Where episodes were used for analyses, age was recalculated at each DOU episode of care accordingly.

Findings reported below with regard to characteristics associated with ED presentations are presented in accordance with Andersen’s Behavioural Model of Health Service Utilisation.

Being born outside of Australia was the only socio-demographic factor associated with ED presentation (AOR 1.322, 95% CI 1.109, 1.576). Patients diagnosed with lung cancer had more than three times the odds (AOR 3.727; 95% CI 2.781, 4.996) of making an unplanned ED presentation than patients diagnosed with hematological cancers (Table 2).

Patients diagnosed with genitourinary cancers had twice the odds (AOR 2.035, 95% CI 1.053, 3.934) of making two or more ED presentations within any given 28 day period after receiving cancer treatment. This was the only factor found to be independently related to making multiple ED presentations within any given 28 day period after receiving SACT (Table 3).

Being born outside of Australia was significantly associated with making frequent (four or more within an 12 month period) ED presentations within 28 days of receiving SACT (AOR 1.820, 95% CI 1.178, 2.810, p < .01) (Table 4).

Increasing age was significantly associated with being admitted to the inpatient area (AOR 1.014, 95% CI 1.004, 1.024, p < .01) for further care. Patients diagnosed with breast cancer (AOR 2.176, 95% CI 1.387, 3.416, p < .01) had more than twice the odds of being discharged home from the ED than being admitted to an inpatient area (Table 5).

Predisposing factors found to be associated with unplanned ED presentation in this study included being born outside of Australia, with over 50% (n = 548/1075) of patients in this study born outside of Australia presenting to the ED within 28 days of receiving SACT. A similar Australian study by Craike et al., reported 37.7% of patients treated in the DOU who were born in a non-English speaking country presented to the ED on at least one occasion [33]. In the present study, of the 1075 (40.8%) patients born outside of Australia, 300 (27.9%) recorded a primary language other than English, and of those 199 (66.3%) needed an interpreter for in-hospital communication. In a recent U.S. study by Ngai et al., a significant relationship was demonstrated between the general population of ED attenders with limited English proficiency, and making unplanned ED re-presentations within 72 h [34]. Our study found no association between English proficiency and frequent ED presentations, in both univariate and multivariate analysis. However, we found that limited English proficiency was associated with making at least one ED presentation after receiving SACT (p < .001), but not with making frequent ED presentations (data not shown). The significant association between being born outside of Australia and making an unplanned ED presentation in this study suggests that increased risk of ED presentation may not be based on language or communication barriers alone, but may be due to additional factors such as being unfamiliar with the Australian health care system. Stage of disease may also be an important factor.

A number of international studies have identified several factors that may be associated with ED presentation such as lower socio-economic status [13, 33, 35]. In our study no factors likely to increase likelihood of or mitigate against ED presentation were identified.

Disease related factors found to be associated with an ED presentation (head and neck, UGI, CRC, lung, skin or breast cancer diagnosis) may represent true symptom severity and need for health care but may also indicate levels of unmet supportive care need. In 2013, Aprile et al., reported that 21.6% of unplanned hospital presentations made by cancer patients were due to the patients’ desire for reassurance from their treating specialist [12]. Public use and perception of EDs were explored in an Australian study by Fitzgerald et al., which reported that the main reasons for ED use by the general population were perceived severity of illness, unavailability of alternative health services, and the perception that EDs were able to provide a higher standard of care [36]. More recently, Phillip et al., reported that patients with advanced cancer frequently presented to the ED as a means of accessing specialist oncology care, to receive treatment for worsening symptoms, or symptoms that they had been instructed to attend the ED for, such as a fever [4]. To the best of our knowledge, no studies have explored the decision-making process behind ED presentations for patients receiving active cancer treatment.

Results from this study have shown that patients who were of an older age were significantly more likely to be admitted to either the SSU or to an inpatient unit, than be discharged home following an ED presentation. This may suggest the presence of other comorbidities alongside a diagnosis of cancer, and thus a higher level of care complexity or clinical acuity [3, 37, 38]. This finding is consistent with international literature that has previously identified a significant relationship between older age and inpatient admission following an ED presentation [39, 40].

Patients diagnosed with breast cancer were more likely to be discharged home than be admitted to an inpatient area after an ED presentation. Previous studies have explored risk factors for inpatient admission following an ED presentation, but no studies were identified that explored the factors associated with being discharged home from the ED. Patients with early stage breast cancer are commonly treated with adjuvant chemotherapy and subsequently experience considerable treatment related toxicities and are known to make higher numbers of unplanned hospital visits [41, 42]. This may explain the higher frequency of ED presentations seen in this study, but also why ED presentation by such patients is unlikely to result in admission.

Patients with haematological cancers are known to have high rates of ED utilization due to the acuity and complexity of these diseases and their treatments [8, 37]. In this study, patients diagnosed with hematological cancers made the largest number of ED presentations, but which are comparatively low to the number of patients treated for hematological cancers in the DOU. At the study site patients with haematological cancers have timely access to specialist haematology nurse coordinators who provide telephone support during business hours. This may explain the low numbers of ED presentations by haematology patients in our study. Equivalent nurse coordinators are only available for patients diagnosed with solid tumours who are receiving oral anti-cancer therapies or concurrent chemotherapy and radiotherapy. Outside of these treatment modalities, patients only recourse is to contact the DOU and speak a senior oncology nurse where they are provided with telephone advice or advised to present to the ED for further assessment.

The limitations of the administrative databases make drawing conclusions about why certain cohorts are more likely to make unplanned ED presentations challenging. Important variables that may be associated with ED presentation such as stage of disease and those receiving multi-modality treatment are not currently captured in these datasets in Victoria, Australia.

This study was conducted at a single site and therefore may not generalizable to other healthcare settings. The models of care to support patients outside the hospital setting who experience cancer and treatment related toxicities, the case mix of patients treated at individual settings and the difference in health systems internationally means that risk factors for ED presentation may differ according to local structures.