Cancer Economics
Incidence of Emergency Visits Among Oncology Patients Receiving
Outpatient Chemotherapy: Implications for Care in a Capitated Market
Charles G. Martin, PhD; Kathryn F. Kennedy, LMSW-AP; Linda S.
Elting, DrPH; Ellen Manzullo, MD, FACP; Carmen P. Escalante, MD, FACP; and Edward B.
Rubenstein, MD, FACP
The University of Texas M.D.Anderson Cancer Center, Houston, Texas.
Introduction
The shift in providing medical care from the hospital setting to the outpatient setting
has been fueled by diagnosis-related groups, technological advances, and changes in the
healthcare environment as a consequence of managed care initiatives. It is estimated that
85% to 95% of cancer care occurs in the outpatient setting and that most chemotherapy
treatments are administered outside the hospital in freestanding cancer centers, community
oncology offices, comprehensive cancer centers, and ambulatory infusion suites.[1] All
cancer patients are expected to receive treatment and care on an ambulatory basis at some
point during the course of their treatment.[2]
Although patients receiving ambulatory chemotherapy may develop acute nausea and
vomiting, most adverse events from chemotherapy occur days to weeks following treatment.
In the future, the medical and financial responsibilities for these downstream events will
likely become those of the provider. The number of cancer patients who will have an
emergency event while receiving outpatient chemotherapy is not known, and risk factors for
these events have not been examined systematically.
The increasing number of cancer diagnoses and the extent to which treatment is provided
on an outpatient basis are factors that emphasize the importance of identifying subgroups
of patients at high risk for subsequent emergency care. Therefore, a prospective study was
conducted at our institution to estimate the incidence of emergency visits made by
outpatients receiving scheduled chemotherapy and to identify risk factors associated with
such visits.
Methods
Study Site
The University of Texas M. D. Anderson Cancer Center is a 518-bed, acute-care,
comprehensive cancer center in Houston, Texas. During fiscal year 1992-1993, a total of
586,979 outpatient visits were reported.[3] During 1993, approximately 60,000 patients
received chemotherapy on an outpatient basis.[4]
The Ambulatory Treatment Center at our center is a multidisciplinary unit that provides
ambulatory hospital-based care, outpatient chemotherapy, and supportive care. This
facility consists of a short-term infusion therapy unit (for treatments requiring less
than two hours) and a long-term infusion therapy unit (for treatments requiring 24 hours
or less). A 24-hour-per-day acute-care unit functions as the emergency room and is the
focus of this report.
Selection of Patients
All patients included in the study fulfilled the following criteria: (1) They were
officially registered patients, (2) they were currently receiving outpatient chemotherapy,
(3) they had at least two outpatient chemotherapy appointments scheduled during the 90-day
study period, (4) they had no procedures requiring a scheduled inpatient admission during
the study period, and (5) they were at least 16 years of age.
Time at Risk
The Index Appointment was the first scheduled appointment, and the Follow-up
Appointment was the second scheduled appointment during the 90-day period. The outcome
measure - an emergency visit - was defined as an urgent, unplanned visit (ie, unscheduled
outpatient services provided to patients whose conditions required immediate care). Time
at risk for an emergency visit was defined as the number of days from the Index
Appointment to the date of an emergency visit, date of hospitalization, date of death, or
date of the Follow-up Appointment, whichever occurred first. No time-at-risk analysis was
conducted for subsequent emergency visits.
Study Variables
The hospital is served by a clinical computing system that maintains information
pertaining to resource scheduling, medical records, referral and census information,
biographic and demographic data, employment and insurance information, and laboratory and
pathology data. All data pertaining to this study were retrieved through this system.
Twenty randomly selected medical records were reviewed, with an accuracy rate of 100%.
Demographic variables included age at Index Appointment and the patient's gender. Race
was categorized as American Indian, Asian, black, Hispanic white, and non-Hispanic white.
Clinical variables included the type and site of cancer, as well as the length of time
from registration to the Index Appointment. Diagnostic categories, based on ICD9-CM
codes,[5] were either grouped according to disease-site clinics or combined where multiple
diagnoses were treated in a given clinic area.
Analysis
The incidence rate of one or more emergency visits was analyzed in relation to the
demographic and clinical variables examined in this study. Proportional hazards regression
analyses were used to investigate factors associated with the risk of an emergency
visit.[6,7] These models were used to derive rate ratios (the ratio of the rate of an
emergency event in patients with a given factor to that in patients without the factor).
Rate ratios greater than 1.0 suggest an increased risk for an emergency visit, whereas
rates less than 1.0 suggest a protective factor (ie, a reduced risk).
Each prognostic factor was examined univariately. A multivariate model was then used to
evaluate the effect of each factor separately on the rate of an emergency visit. This
model reflected the increased or decreased risk of an emergency visit based on each
independent variable (prognostic factor) under investigation.
Results
The demographic and clinical characteristics of the study population are shown in Table
1. Of the 1,283 patients, 70% were between 40 and 69 years of age, 56% were women, and 76%
were non-Hispanic white patients. Patients diagnosed with breast cancer (24%) and
gastrointestinal or genitourinary cancer (19%) predominated. Length of time from
registration to the Index Appointment ranged from less than one month to 36 years.
Approximately 40% of the study population had been registered less than four months, and
44% had been registered longer than six months.
The mean (± standard deviation) length of time at risk was 27.6 ± 7.6 days for the
1,238 patients who had no emergency visits compared with 24.1 ±19.6 days for the 45
patients who had an emergency visit (P=.19). Of these 45 patients, 10 (22%) had a
second emergency visit, three (7%) had a third visit, and two (4%) had a fourth visit
during the study period.
As shown in Table 2, the overall incidence rate of an emergency visit was 10 visits per
100 patients per month (10/100/month). Patients 30 to 39 years of age had the highest rate
of emergency visits (18/100/month), whereas patients older than 69 years of age had the
lowest rate (5/100/month) (P<.05). A higher incidence rate of emergency visits
was also found for patients diagnosed with melanoma (32/100/month) compared with
6/100/month for patients diagnosed with cancer from an unknown primary (P<.05)
or with breast cancer (P=.12). The lowest incidence rate was found for those
diagnosed with leukemia (3/100/month). The incidence of emergency visits was inversely
associated with the length of time registered at the hospital. Emergency visits were
29/100/month for patients registered less than one month. This rate decreased to
6/100/month for patients who were registered longer than 12 months (P<.01).
Presenting problems and disposition
of patients who had emergency visits are shown in Table 3. Fever (27% of patients) was the
most common reason for the first emergency visit, followed by pain (18%), nausea (18%),
weakness (9%), and other reasons (9%). This last category consisted of two patients who
presented with headache and two who presented with abdominal pain. The distribution of
problems resulting in subsequent emergency visits approximated those of the initial visit.
Reasons for a second emergency visit were fever (30%), pain (20%), and nausea
(20%).Presenting problems that resulted in a third and/or fourth emergency visit included
fever, pain, nausea, bleeding, and other reasons, with each category representing 20%.
Most patients (71%) were discharged after the initial emergency visit; this figure
decreased to 60% after the second emergency visit. The frequency of hospitalization
following the emergency visit, however, increased with each subsequent emergency visit
(27%, 30%, 40%, and 40% following the first, second, third, and fourth visits,
respectively).
Prognostic Factors
Prognostic factors were examined separately, with emergency visits as the outcome
measure in the proportional hazards models (Table 4). Univariately, an increased incidence
of subsequent emergency visits was significantly associated with a diagnosis of melanoma
or a diagnosis of lymphoma, Hodgkin's disease, or myeloma (P<.01). The rate of
emergency visits, however, decreased significantly for patients registered three months or
longer (P<.01). Age, sex, and race were not univariate predictors.
The results presented in the multivariate model included all prognostic factors in the
study, using selected categories to compare the remaining groups for each variable. The
referent categories included patients less than 30 years of age compared with those 30
years of age or older, women compared with men, non-Hispanic white race compared with all
other races, breast cancer patients compared with other cancer patients, and patients
registered for less than one month compared with those registered one month or longer.
Patients diagnosed with melanoma, as well as those with lymphoma, Hodgkin's disease, or
myeloma, were four times more likely to have an emergency visit (P<.05) than
were breast cancer patients. In addition, emergency visits were less likely to be reported
by patients registered seven to 12 months or by those registered longer than 12 months (P<.01).
Age, sex, and race were not significant predictors of emergency events in the multivariate
model, although at least one emergency visit was three times as likely as younger patients
for those 30 to 39 years of age and twice as likely for those 50 to 59 years of age.
Prognosis According to Prior Emergency and Subsequent Visits
Patients who had an emergency visit 30 days prior to the Index Appointment (n=47) were
four times more likely to have an emergency visit during the study period (P<.01).
Emergency visits that occurred more than 90 days prior to the Index Appointment, however,
were not significant univariate or multivariate predictors of subsequent emergency visits.
The multivariate model, which included all prognostic variables as well as an emergency
visit 30 days prior, demonstrated that a diagnosis of lymphoma, Hodgkin's disease, or
myeloma (hazard ratio = 3.63), a diagnosis of melanoma (hazard ratio = 3.58), registration
of six to 12 months (hazard ratio = .23), or registration of longer than 12 months (hazard
ratio = .35) remained statistically significant (P<.05).
A similar pattern was found in subsequent emergency visits. In the multivariate model,
a diagnosis of lymphoma, Hodgkin's disease, or myeloma, as well as patient registration of
either six to 12 months or longer than 12 months, remained statistically significant
factors for two emergency visits (P<.01) and for three or more emergency visits
(P<.05). An additional prognostic variable specifically associated with an
increased risk for two subsequent emergency visits, however, was found for patients 30 to
39 years of age (P<.05).
Discussion
The overall incidence rate of 10 visits per 100 patients per month (10/100/month) can
be used as a baseline rate for capitation in a managed care environment. However, the data
indicate that population and clinical demographics can affect capitation calculations. A
younger patient population and one with newer patients could have an incidence rate two to
three times baseline. In this study, patients 30 to 39 years of age had an incidence rate
of 18/100/month, and the incidence rate for those registered less than one month was
29/100/month. While a new hospital or one with an aggressive marketing strategy would have
higher emergency costs, one with an older population in terms of both age and chemotherapy
experience would have lower costs. Likewise, cancer diagnosis can have a similar effect.
Melanoma patients have a higher rate and breast cancer patients a lower rate. An
institution serving breast cancer patients, particularly with adjuvant therapy, would have
a lower incidence of emergency visits and lower costs.
The most plausible explanation for the high rate of emergency visits for patients with
melanoma is the combination chemotherapy administered as part of our clinical
investigations. Patients with melanoma are commonly referred to our center for aggressive
treatment regimens. During the period of this study, the regimen consisted of 20 mg/m2
of cisplatin plus 2 mg/m2 of vinblastine daily for four days, and 800 mg/m2
of dacarbazine on day 1. Some patients received this regimen plus a-interferon and
interleukin-2.
Demographic and clinical prognostic
factors can distinguish patients at greater risk of having an emergency visit while
receiving outpatient treatment for cancer and may alert the health care team to
populations needing interventions during chemotherapy to avoid subsequent emergencies.
Patients registered less than seven months were at greatest risk of having an emergency
visit and generated 95% of the emergency visits in this study. These patients are usually
scheduled for more frequent clinic visits and more aggressive therapy. Newer patients are
likely to be less informed about and less experienced in dealing with the side effects of
chemotherapy. Pain, nausea, and weakness accounted for 45% of emergency visits in this
study. Experienced patients may have adapted to symptoms that send less experienced
patients to the emergency room. In addition, patients who had an emergency visit within 30
days prior to entrance to the study were four times more likely to have another emergency
visit. Patient education, explicit instructions about signs or symptoms of true
emergencies, and telephone follow-up may reduce the incidence of emergency visits for new
patients.
Fever, which is often a symptom requiring an emergency visit, accounted for 27% of the
emergency visits in our study. Although fever may not be preventable, treatment strategies
such as outpatient antibiotics for febrile neutropenia can reduce the cost of the common
downstream clinical event.
Future studies should further define the role of prognostic factors. For example, the
low incidence rate of emergency visits by older patients may indicate that age is a
surrogate for cancer diagnosis, extent of disease, or chemotherapy regimen. Type of
therapy, whether adjuvant, curative, or palliative, also may be a predictive factor.
Prospective studies that follow larger numbers of patients are needed to develop an
accurate prognostic model. Additional variables such as comorbidity, extent of disease,
and medications (including chemotherapy, antiemetics, and pain medications) may be more
important factors than population demographics. Further studies are needed to determine if
interventions such as patient education during chemotherapy or as part of emergency
treatment can reduce the incidence of subsequent emergency visits. Also, analyses that
focus on resource use and cost would provide useful data for care in capitated oncology
markets.
References
- Sandrik K Oncology: who's managing outpatient programs? Hospitals. 1990;64:32-37.
- Goldberg J. Preface: Challenges of Ambulatory Medicine, Care and Research in Cancer
in the 90's. Satellite Symposium, London, September 6, 1989.
- 1992-1993 progress report. Conquest. 1994;8:20.
- Patient care. Conquest. 1994;8:6.
- Commission on Professional and Hospital Activities. The International Classification of
Diseases, 9th revision, Clinical
- Modification. Ann Arbor, Mich: Edwards Bros; 1980.
- STATA Reference Manual: Version 3.0. Santa Monica, Calif: Computing Resource Center;
1992.
- Bailar JC III, Mosteller F, eds. Medical Uses of Statistics. 2nd ed. Boston,
Mass: NEJM Books; 1992.
Back to Cancer Control
Journal Volume 3 Number 5