Most studies of the impact of SARS on health service utilization have been confined to a single facility. Our study, based on nationwide data, gave a complete and exact picture of hospice utilization. The hospice, also known as palliative care, has usually been considered as conservative and non-acute. One could well imagine that the use of hospice services would be compromised because of fears during uncertain times. Our findings support this view. Hospice inpatient utilization in Taiwan was indeed more sensitive to the emerging epidemic than general inpatient utilization. Not only was the reduction sharper, but also recovery to the previous level took longer.
Although it cannot be classed as acute care, hospice care is unlike elective surgery, which can put patients on a long waiting list. Hospice care is provided to terminally ill patients whose life expectancy is limited. During the SARS epidemic in Taiwan, many terminally ill patients might have been suspended or deterred from professional hospice care because well-organized patient transfers did not exist. Health consequences might not be a big issue among hospice patients. Instead, quality of care and humanity are the major concerns.
During the peak period of the SARS epidemic, i.e. May/June 2003, the number of admissions to the 15 hospice wards in our study decreased to 69% of those in the previous year, and the utilization in patient-day units to 54%. This meant that some patients might have been discharged earlier. It remains unknown whether the decrease of utilization was due to patients' voluntary decisions or to manoeuvring by the hospitals. According to our personal experiences, both patients' decisions and hospital policies could have played a role. On the patient's side, the family might have a major part in the decision about hospice care, because weak terminally ill patients are usually dependent on their families. The fear of SARS might have come from the family rather than the patient. Only interviews with the patients and their caregivers could give a satisfactory answer.
The aggregate utilization of the 15 hospice wards during the peak period of the SARS epidemic in Taiwan was lower than that during the Chinese New Year holidays (Figure 1). Like Christmas to Christians, the Chinese New Year holidays are important to most Chinese as occasions of family reunion. Usually hospital activities are kept to a minimum on these days. The daily curve of hospice utilization vividly illustrated the impact of the SARS epidemic.
As for the patients admitted to hospice wards during the SARS epidemic, the insurance claims used in our study could only offer three reliable variables (gender, sex and discharge diagnoses), which showed no differences. Because about 60% of cancer patients in Taiwan die at home , we could not calculate the survival time of every hospice patient as a proxy of disease severity from our closed datasets.
Although Taiwan had been declared a SARS-affected country, not every hospice ward was equally influenced by this epidemic. The impact was related to the locations and accreditation levels of the hospitals with hospice wards: the bigger the hospital, the stronger the SARS impact. Other features, e.g. the admission policy and mission of each hospital during the SARS epidemic, might have played a role. Hospices in middle Taiwan were most severely affected; only the hospice in eastern Taiwan was negligibly affected. The main island of Taiwan has 22 cities and counties. The probable cases of SARS appeared mainly in Taipei city (in northern Taiwan), Taipei county (northern), Taoyuan county (northern), Kaohsiung city (southern) and Kaohsiung county (southern) . On 26 March 2003, a resident of Hong Kong's Amoy Gardens flew to Taiwan and travelled to Taichung (in middle Taiwan). That man's brother in Taiwan became Taiwan's first SARS fatality . This might partly explain the impact of SARS in middle Taiwan (Figure 4).
Because the NHIRD prohibited any break of confidentiality, we did not continue the analysis hierarchically from geographic location to accreditation level. Instead, we displayed the aggregate data in location and level separately. However, hierarchical modelling could be used to examine the factors associated with reduced utilization and would help construct a real-time and on-line analytical processing system for nationwide surveillance during an outbreak of an infectious disease.
A striking feature of the SARS epidemic in Taiwan was that 56.3% of SARS patients were hospital-related . In retrospect, it was reasonable and preferable to reduce hospitalisations. A similar phenomenon was observed simultaneously in childbirth in Taiwan . Many deliveries were shifted from secondary and tertiary care hospitals to primary clinics. The question remains whether continuity and quality of care could be ensured. It is unlikely that patients shifted from the 15 hospices in the current study to the other six (unanalysed) hospices, which either quitted or joined the hospice program after the SARS epidemic. However, the possibility of patient flow to other non-hospice wards cannot be excluded. Our claims-based study could not determine whether the needs of patients with terminal illnesses were met during the epidemic, and threw no light on the psychological impact on patients and healthcare workers .