Full text of DOI:10.1111/j.1553-2712.2001.tb00550.x

48
URGENT CARE
Batal et al. • PREDICTING URGENT CARE VISITS
Predicting Patient Visits to an Urgent Care Clinic
Using Calendar Variables
HOLLY
B
ATAL, MD, JEFF
TENCH, BS, SEAN MCMILLAN, BA,
J
ILL DAMS, BA, PHILLIP S. MEHLER, MD
A
Abstract. Objective: To develop a prediction equa- significant weather variables only minimally im-
tion for the number of patients seeking urgent care. proved the predictive ability (r2 = 0.786). The second-
Methods: In the first phase, daily patient volume phase final model was: daily patient volume = 57.2 ϩ
from February 1998 to January 1999 was matched 0.035 Newdate ϩ 52.0 Monday ϩ 44.2 Tuesday ϩ
with calendar and weather variables, and stepwise 39.2 Wednesday ϩ 30.2 Thursday ϩ 26.5 Friday ϩ
linear regression analysis was performed. This model 10.9 Saturday ϩ 12.2 February ϩ 3.9 March, which
was used to match staffing to patient volume. The accounted for 72.7% of the daily variation (p < 0.01).
effects were measured through patient complaint and The model predicted the patient volume in the vali-
‘‘left without being seen’’ rates. The second phase was dation set within Ϯ11%. When the first-phase model
undertaken to develop a model to account for the con- was used to predict patient volume and thus staffing,
tinual yearly increase in patient volume. For this the percentage of patients who left without being
phase daily patient volume from February 1998 to seen decreased by 18.5% and the number of patient
April 2000 was used; the patient volume from May complaints dropped by 30%. Conclusions: Use of a
2000 to July 2000 was used as a validation set. Re- prediction equation allowed for improved accuracy in
sults: First-phase prediction equation was: daily pa- staffing patterns with associated improvement in
tient volume = 66.2 ϩ 11.1 January ϩ 4.56 winter ϩ measures of patient satisfaction. Key words: fore-
47.2 Monday ϩ 37.3 Tuesday ϩ 35.6 Wednesday ϩ casting; utilization; walk-in clinics; regression anal-
28.2 Thursday ϩ 24.2 Friday ϩ 7.96 Saturday ϩ 10.1 ysis. ACADEMIC EMERGENCY MEDICINE 2001; 8:
day after a holiday. This equation accounted for 48–53
75.2% of daily patient volume (p < 0.01). Inclusion of
RATIONAL approach to the provision of ur- cessive length of wait.2 Patients are more likely to
A
gent care services must include considera-
tion of trends in the demand for these services. Ac-
cording to the National Ambulatory Medical Care
Survey, 17.6% of health care contacts take place at
sites that do not require an appointment; this
method of accessing care appears to be on the in-
crease.1 The ability to predict the number of pa-
tients seeking care on any given day in an urgent
care clinic is essential in order to allow for opti-
mization of staffing patterns. Matching staffing
levels to the variation in daily patient demand can
improve cost efficiency as well as patients’ satis-
faction with care by decreasing waiting times. The
major reason for dissatisfaction with emergency
care and patients’ leaving without being seen is ex-
leave without being seen as waiting times in-
crease, with more than a fourth of these patients
returning later for evaluation.3 Other studies have
also demonstrated this inverse relationship be-
tween waiting times and satisfaction with care.4–6
Simply attempting to divert patients with appar-
ently minor complaints out of the urgent care clinic
because of inadequate staff availability is fraught
with problems and may be an unsafe practice with
serious conditions being overlooked.7
To the best of our knowledge, few studies have
been conducted in an attempt to quantitate the re-
lationship between patient volume and calendar
and weather variables. A study of the use of a pe-
diatric emergency department (ED) found no sig-
nificant difference in the number of ED visits or
the number of resulting hospital admissions dur-
ing unfavorable weather vs favorable weather.8 An
older study examined patterns of ED use, includ-
ing ambulance use, in three Boston hospitals as
From the Urgent Care Clinic at Denver Health (HB, JA), the
University of Colorado, Denver (JT, SMcM), and Denver
Health and the University of Colorado Health Sciences Center,
Denver, CO (HB, PSM).
Received February 14, 2000; revision received August 15, 2000; they related to weather conditions.9 In this study,
accepted August 24, 2000.
there was a tendency in each of the three hospitals
Address for correspondence and reprints: Holly Batal, MD,
to have an increase in nonambulance patient traf-
Denver Health Medical Center, 777 Bannock Street, MC0107,
Denver, CO 80204. Fax: 303-436-3158; e-mail: hbatal@
dhha.org
fic during more favorable weather.
Only two previous studies have used calendar

ACADEMIC EMERGENCY MEDICINE • January 2001, Volume 8, Number 1
49
and weather variables to develop prediction for- 1999; in 2000 the clinic is expected to have ex-
mulas for patient volume, and only one used a val- ceeded 40,000 patient visits.
idation set of data to test their predictions.10,11 Us-
ing calendar variables, Diehl et al. were able to Study Protocol. In the initial phase of the study,
account for 74.5% of the variation in the square the number of patients treated daily from Febru-
root of daily visits; this number increased slightly ary 1, 1998, through January 31, 1999, was used.
to 77.3% when weather variables were also in- Daily patient volume was matched with calendar
cluded.11 Holleman et al. used a Veterans Admin- variables such as day of the week, month of the
istration hospital as the basis for their study.10 The year, season, holiday, and the day before and after
significant factors that predicted patient volume in a holiday. Designated holidays were Labor Day,
this model were calendar variables (season, week Memorial Day, Fourth of July, Martin Luther King
of month, day of week, holidays, and federal check Day, Thanksgiving, New Year’s Eve, New Year’s
delivery day) and weather variables (high temper- Day, and Christmas Eve. Christmas Day was the
ature and snowfall). They were able to use their only day of the year that the clinic was closed due
model to predict patient volume, with their model to historical data that showed a distinct paucity of
accounting for 84% of the daily variance. Within utilization that day. On three of the holidays,
their validation set, when their model was used to Thanksgiving, New Year’s Eve, and Christmas
predict staffing, it would have produced adequate Eve, the length of time the clinic was open was less
staffing on 67% of days, overstaffing on 15% of than 15.5 hours. For these days, the patient vol-
days, and understaffing on 18% of days. Using ume was corrected by extrapolating the actual vol-
these prior studies as a guide, we undertook this ume to what would have been expected for a 15.5-
project to identify the predictors of patient volume hour day. This was done by determining the
at the Denver Health Medical Center urgent care number of patients per hour actually seen and
clinic and to develop a prediction formula to project then multiplying by 15.5 to calculate the projected
the number of visits on future days. The ultimate number of patients who would have been seen in
goal was to allow for maximization of staffing to a normal-length day. Since a holiday can affect pa-
positively impact efficiency of patient care.
tient volume both on the day of and the days before
and after a holiday, we also included variables if a
given day was the day before a holiday, the day of
the holiday, or the day after the holiday.
METHODS
To account for the effect of season, we included
a variable ‘‘winter,’’ defined as November through
February. These months were chosen based on the
statistical distribution of patient volume. In addi-
tion to season, recent historical volumes may also
be an accurate predictor of current volume. That
is, the volume in the clinic tends to trend up and
Study Design. Stepwise linear regression was
performed on existing data to determine signifi-
cant variables for urgent care volume. A prediction
equation was developed and used to predict patient
volume presenting for urgent care, and validated
with an existing data set.
Study Setting and Population. Denver Health down over time, perhaps related to periodic out-
Medical Center is an urban safety-net public hos- breaks of upper respiratory infection or other com-
pital serving the indigent residents of Denver, Col- municable diseases or seasonal illnesses. To at-
orado. This fully integrated health care system tempt to capture this effect, we created an
comprises the state’s only Level 1 trauma center independent variable for ‘‘prior volume.’’ To create
and 11 outlying community clinics. Each of these this variable, we averaged the daily volume over
clinics provides primary care services for the resi- the prior seven-day period.
dents of the community, as well as urgent evalua-
The local climatological data during the study
tion of patients as availability allows. The urgent period were obtained from the National Weather
care center, known as the walk-in clinic (WIC), is Service. We considered maximum, minimum, and
located within the hospital and is open 15.5 hours average temperatures (in degrees Celsius) and pre-
per day, seven days a week. The WIC provides un- cipitation amount. Precipitation was recorded both
scheduled outpatient care for adult patients whose as total (water equivalent precipitation in inches)
clinical conditions are not of significant acuity to and as the amount of snowfall (also in inches).
warrant ED care but who have chosen not to access Whenever the National Weather Service listed pre-
primary care in community clinics. During the cipitation as ‘‘trace,’’ we assigned this a value of
clinic’s hours of operation patients either present 0.001 inches.
directly to the clinic for care, are referred from the
The second phase of the study used the daily
hospital’s network of outlying community clinics, patient volume from January 1998 to May 2000 to
or are triaged into the clinic from the ED. Approx- develop a prediction equation that captured the
imately 37,000 patient visits were recorded in significant calendar variables as well as the con-

50
URGENT CARE
Batal et al. • PREDICTING URGENT CARE VISITS
T
ABLE 1. Patient Volume by Day of the Week
tistics were calculated for all variables during each
step. Variables were deleted from the model se-
quentially as dictated by the values of these sta-
tistics, until all variables were significant at the
5% level. Because calendar variables are predeter-
mined, whereas weather variables are much less
predictable, we developed a parallel model that ex-
Average Daily
Day
Patient Volume
95% CI
Monday
Tuesday
Wednesday
Thursday
Friday
115
107
105
95
93
77
111, 118
103, 111
101, 108
91, 99
90, 96
74, 80
cluded weather variables.
A model without
Saturday
Sunday
weather variables is more practical for advance
calculation of day-to-day staffing needs. All analy-
ses were performed using Mini-tab software (Re-
lease 12, State College, PA).
69
66, 72
tinual rise in patient volume that we were expe-
riencing (an approximately 7–8% increase per year
since 1997). This new independent variable was la-
beled ‘‘Newdate.’’ Weather variables were not in-
cluded in this second-phase prediction equation de-
velopment since results of the first phase showed
that they added little.
Beginning September 1, 1999, we began to use
the equation developed in the first phase of the
study (inclusive of only calendar variables) to pre-
dict patient volume. Since the first-phase predic-
tion equation did not account for the continual
increase in patient volume that we were experi-
encing, 7% was added to the predicted values.
Since the average number of patients seen by each
provider class per hour (resident, nurse practi-
tioner, attending physician) has been consistent
and is well documented within our clinic, neces-
sary staffing patterns and provider mix were de-
termined by matching them to the predicted pa-
tient volume.
In order to test our second-phase model, we
used a validation set of data for the time period
May 1, 2000, to July 31, 2000. For each of these
days we calculated the predicted patient volume,
with comparison made with the actual number of
patients seen.
Accurate waiting time data are not historically
available within our clinic. In order to measure the
effects of prediction equation utilization, patient
complaint and ‘‘left without being seen’’ rates were
used as proxy measures. Patient complaint and
‘‘left without being seen’’ rates were calculated as
percentages of patient volume. These statistics are
internally kept, and methods of data collection
have been stable since January 1998.
In the second phase of the study, stepwise lin-
ear regression analysis was performed on data
from February 1, 1998, to May 31, 2000, with the
addition of a variable allowing us to account for the
continual increase in patient volume over time
(Newdate). All calendar variables were initially in-
cluded, and insignificant variables were eliminated
one at a time until all remaining variables were
significant at the p < 0.01 level. Analyses in this
second phase were performed using SAS software
(SAS Institute, Cary, NC).
RESULTS
As shown in Table 1, day of the week is the
greatest consistent predictor of patient volume.
The highest daily patient volume occurs on Mon-
day, with a fairly linear decrease until Sunday. On
any given day the patient volume shows a normal
distribution. Month of the year is much less pre-
dictive of patient volume (Table 2). However, vol-
ume is consistently lowest in the spring and sum-
mer months (April–August).
First-phase Model Development. In the first
phase of the study, the final regression equation
using all significant calendar and weather varia-
bles was: daily patient volume = 57.0 ϩ 12.0 Jan-
uary ϩ 6.63 winter ϩ 46.7 Monday ϩ 37.2 Tuesday
ϩ 36.0 Wednesday ϩ 29.4 Thursday ϩ 24.1 Friday
ϩ 8.67 Saturday ϩ 10.1 day after a holiday Ϫ 5.28
July Ϫ 5.57 August ϩ 0.147 maximum tempera-
ture (C) Ϫ 3.88 snowfall (inches). This equation ac-
counted for almost 79% of the daily patient varia-
bility (r2 = 0.786, p < 0.01).
When weather variables were excluded from
the model, the final regression equation became:
Data Analysis. In the first phase of the study, daily patient volume = 66.2 ϩ 11.1 January ϩ 4.56
using both weather and calendar variables, step- winter ϩ 47.2 Monday ϩ 37.3 Tuesday ϩ 35.6
wise linear regression analysis was performed on Wednesday ϩ 28.2 Thursday ϩ 24.2 Friday ϩ 7.96
data from February 1, 1998, to January 31, 1999, Saturday ϩ 10.1 day after a holiday. This equation
in order to determine how much additional vari- still accounted for approximately three-fourths of
ance we have accounted for by including different the variation in daily patient volume (r2 = 0.752, p
variables. We initially included all variables in the < 0.01). The coefficients before each variable deter-
equation, and eliminated insignificant variables mine the number of additional patients who will
one at a time. During model development, F sta- be expected on that day given that variable being

ACADEMIC EMERGENCY MEDICINE • January 2001, Volume 8, Number 1
51
T
ABLE 2. Daily Patient Volume by Month
present. For instance, if the day to be predicted is
a Monday in January that is not the day after a
holiday, the expected patient volume would be 129
(66.2 ϩ 11.1 ϩ 4.56 ϩ 47.2).
Comparison of the two models demonstrated
that inclusion of weather variables diminished the
Average Daily
Patient Volume
Month
95% CI
January
February
March
107
101
90
92
89
94
91
89
96
93
94
94
99, 115
93, 108
84, 96
86, 99
81, 97
87, 101
85, 98
83, 95
89, 104
86, 99
87, 102
86, 103
impact of day of the week and of winter season. April
May
June
July
August
September
However, it also included the months of July and
August as predictive of lower patient volume. The
final model, excluding the weather variables, con-
tained only nine easily measured and predicted
variables, in comparison with 13 variables in the October
November
December
alternative model.
The equation using only calendar variables
from the first phase of the study, with 7% added
on for the increase in subsequent yearly volume,
quential numbering of dates starting with a value
of 0 at the first date in the data set, February 1,
1998. For example, October 28, 2000, would have
1,000 as a value for Newdate, since it is the 1000th
date after February 1, 1998. Calculation of the
Durbin Watson statistic12 showed that there was
not significant autocorrelation among these varia-
bles during model development. In addition, the
final model was chosen based on Schwarz’s Baye-
sian criterion (SBC) statistics13 so as not to over fit
the data, which would have decreased the fore-
casting accuracy. This final equation accounted for
72.7% of the variation in daily patient volume and
had a p-value of <0.01.
Using the second-phase model to test our vali-
dation set of data (May 1, 2000, to July 31, 2000)
demonstrated that the model on average under-
predicted patient load only by an average of two
patients per day, with a standard deviation of Ϯ 11
patients. The mean difference between predicted
and actual patients was Ϫ3.0%, with a standard
deviation of 11.6%.
was then used to predict patient volume starting
in September 1999. Based on this, the staffing pat-
tern of providers was altered to allow for the pre-
dicted patient volume to be seen, given current
provider productivity. Model predictions for Sep-
tember allowed the clinic to be appropriately
staffed on 22 days (73%), understaffed on four days
(13.5%), and overstaffed on four days (13.5%). On
the four days that the clinic was understaffed, it
was only by an average of 0.9 patients/hour. This
small excess volume of patients can usually be eas-
ily managed by the providers who are working
throughout the day. For the days that the clinic
was overstaffed, it was only by an average of 1 pa-
tient/hour. This translated into an only 0.3 FTE
(full-time equivalent) per week that was not being
fully utilized.
The prediction equation was continuously used
to develop staffing patterns. It was expected that
waiting times would decrease as staffing was bet-
ter matched to patient volume. This would then
improve patient satisfaction as well as decrease
the number of patients who left without being
seen. When the time period June 1998 to August
1999 (prior to prediction utilization) was compared
DISCUSSION
with September 1999 to May 2000 (after prediction The weekly pattern of visits that we observed,
utilization), the percentage of patients who left highest on Monday and thereafter declining stead-
without being seen dropped from 2.71% to 2.21%, ily throughout the week, with the fewest patients
an approximately 18.5% decrease. In addition, the seen on Sunday, is consistent with that demon-
formal patient complaint rate dropped from strated in other studies.10,11,14 The study by Glass
0.208% to 0.145%, a 30% decrease.
and Friedman found a 42% difference in average
patient volume from the high on Monday to the low
Second-phase Model Development. In the sec- on Sunday.14 This is virtually identical to the 41%
ond phase of the study, a model was developed with difference that we detected. However, the monthly
the intent of providing the ability to continue to distribution of visits in our data differed from that
predict patient volume while accounting for the on- in other studies. Diehl et al. noted increasing pa-
going increase in patients seeking care. The final tient visits from May through August.11 The study
regression equation was: daily patient volume = by Glass and Friedman, which studied trends in
57.2 ϩ 0.035 Newdate ϩ 52.0 Monday ϩ 44.2 Tues- ED services, also found the monthly volume to be
day ϩ 39.2 Wednesday ϩ 30.2 Thursday ϩ 26.5 greatest in late summer and early fall and lowest
Friday ϩ 10.9 Saturday ϩ 12.2 February ϩ 3.9 throughout the winter, except for January.14 In
March. The value for Newdate was created by se- contrast, our peak months were November through

52
URGENT CARE
Batal et al. • PREDICTING URGENT CARE VISITS
February, similar to Holleman et al., who found between our study’s conclusions and that of others
that the winter months were the busiest, aside that have investigated these issues demonstrate
from December.10
the consistent and likely generalizability of the
We found an approximately 11% increase in pa- patterns patients show in presenting for care. In
tient volume after a holiday, identical to that in the addition, a model whereby routinely collected data
study by Holleman et al.10 The presumption is that are continuously used to develop projects for pro-
more patients present to the clinic the day after a cess improvement (such as staffing patterns) is ap-
holiday because they prefer to delay needed care plicable to all clinical settings. However, evidence
rather than disrupt their holiday time. Other ex- against generalizability is the study by Noble et
planations include illness resulting directly from al.,9 which showed discrepancies between institu-
holiday excess or family concern for previously ne- tions in the same city in regard to the effect of cal-
glected medical problems. Our study design did not endar and meteorological variables on patient vol-
allow us to distinguish between these alternate ex- ume. Thus, each institution needs to use the
planations.
Weather variables contributed little to our first- plicability with data from its own experience.
phase model, only marginally increasing the In order to further refine this model, it will be
overall principles expounded here, but hone its ap-
amount of variability that could accurately be ac- necessary to continually reevaluate the association
counted for. It is interesting that the two weather between calendar and weather variables and pa-
variables that were significant contributors to our tient volume. Future study in this area will also
model (daily high temperature and snowfall) were require focusing on uncovering additional deter-
the same variables that were significant in the minants of patient volume to be included in the
model of Holleman et al.10 Both this study and that model. These independent variables will need to be
of Diehl et al.11 conclude that weather-related var- easily measured far in advance to facilitate their
iables add little predictive value to the regression utilization in future predictions.
equation after consideration of calendar variables.
The next step will be to determine a model to
This is probably due to the strong association of predict the hourly variability in patient volume
adverse weather with winter season, which was during a given day. This would allow for additional
significant in both of our first-phase models. Our refinement of staffing patterns to meet patient
more streamlined model promotes ease of use of needs. In addition, models could be developed to
the prediction equation for determining staffing determine the projected need for various ancillary
patterns since adequate prediction can occur with- services, such as radiology. Increasing attention to
out actually checking weather forecasts.
the ability to predict patients’ presentation pat-
Although the studies by Holleman et al.10 and terns and their needs will allow for continued im-
Diehl et al.11 allude to using their derived equa- provement in patient satisfaction.
tions to determine staffing levels, to the best of our
knowledge, ours is the first published report to
CONCLUSIONS
demonstrate the applicability of this as well as
measure the effects of this on patients. Appropriate
staffing can theoretically decrease waiting times.
Prior ED studies have shown the relationship be-
tween increasing waiting times and the number of
patients who leave without being seen.3 Although
our study was not able to directly measure patient
waiting times, the accurate matching of provider
staffing patterns with predicted patient volume did
result in an 18.5% decrease in the number of pa-
tients who left without being seen (used as a sur-
rogate marker of waiting times). This is similar to
the results found in another study; reducing the
length of stay can decrease by more than half the
number of patients who leave an ED without being
seen.15 In addition, we were able to decrease by
almost one-third our patient complaints.
Our study demonstrates that careful consideration
of the relationship of calendar variables to patient
volume allows for fairly accurate prediction of fu-
ture patient volume. These predictions can be used
to optimize staffing patterns to allow timely access
for the burgeoning number of patients who tend to
receive their health care in these types of urgent
care settings.16 Emergency departments and ur-
gent care clinics currently serve as America’s
health care safety net of last resort17 and, as such,
need to have adequate staff capacity to care for all
who seek care. Accurate anticipation of this need
will allow for better use of available personnel re-
sources and improved efficiency of care.
References
L
IMITATIONS AND
FUTURE QUESTIONS
1. Schappert SM. National Ambulatory Medical Care Survey:
1991 summary. Vital Health Stat 13. 1994; 116:1–110.
2. Fernandes CMB, Dayor MR, Barry S, et al. Emergency de-
One limitation of this study is the generalizability
to other clinical facilities. However, the similarities partment patients who leave without seeing a physician: the

ACADEMIC EMERGENCY MEDICINE • January 2001, Volume 8, Number 1
53
Toronto Hospital experience. Ann Emerg Med. 1994; 24:1092– 10. Holleman DR, Bowling RL, Gathy C. Predicting daily vis-
6. its to a walk-in clinic and emergency department using cal-
3. Bindman AA, Grumbach K, Keane D, Rauch L, Luce JM. endar and weather data. J Gen Intern Med. 1996; 11:237–9.
Consequences of queuing for care at a public hospital emer- 11. Diehl AK, Morris MD, Mannis SA. Use of calendar and
gency department. JAMA. 1991; 266:1091–6.
weather data to predict walk-in attendance. South Med J.
4. Hall MF. Keys to patient satisfaction in the emergency de- 1981; 74:709–12.
partment: results of a multiple facility study. Hosp Health Serv 12. Durbin J, Watson GS. Testing for serial correlation in least
Admin. 1996; 4:515–32.
squares regression. Biometrika. 1951; 37:409–28.
5. Rhee K, Bind J. Perceptions and satisfaction with emer- 13. Schwarz G. Estimating the dimension of a model. Ann
gency department care. J Emerg Med. 1996; 14:679–83.
Stat. 1978; 6:461–4.
6. Thompson DA, Yarnola PR. Relating patient satisfaction to 14. Glass R, Friedman D. Trends in the demand for emergency
waiting time perceptions and expectations: the disconfirmation room services: the Mount Sinai Hospital. Mount Sinai J Med.
paradigm. Acad Emerg Med. 1995; 2:1057–62.
1977; 44:560–5.
7. Birnbaum A, Gallagher EJ, Utkewicz M, Gennis P, Carter 15. Fernandes CM, Price A, Christenson JM. Does reduced
W. Failure to validate a predictive model for refusal of care to length of stay decrease the number of emergency department
emergency department patients. Acad Emerg Med. 1994; 1: patients who leave without seeing a physician? J Emerg Med.
213–7.
1997; 15:397–9.
8. Attia MW, Edward R. Effect of weather on the number and 16. Rask KJ, Williams MV, Parker RM, McNagny SE. Obsta-
the nature of visits to a pediatric ED. Am J Emerg Med. 1998; cles predicting lack of a regular provider and delays in seeking
16:374–5.
care for patients in an urban public hospital. JAMA. 1994; 271:
9. Noble JH, La Montagne ME, Bellotti C, Wechsler H. Vari- 1931–3.
ations in visits to hospital emergency care facilities: ritualistic 17. Johnson LA, Derlet RM. Conflicts between managed care
and meteorological factors affecting supply and demand. Med organizations and emergency departments in Los Angeles.
Care. 1971; 9:415–27.
West J Med. 1996; 164:137–42.
ⅷ