Early Warning Scores Generated in DevelopedHealthcare Settings Are Not Sufficient at Predicting Early
Mortality in Blantyre, Malawi: A Prospective Cohort
Study
India Wheeler1*, Charlotte Price2, Alice Sitch2, Peter Banda3, John Kellett4, Mulinda Nyirenda3,
Jamie Rylance5
1 College of Medical and Dental Sciences, University of Birmingham, Birmingham, United Kingdom, 2 Department of Public Health, Epidemiology and Biostatistics,
University of Birmingham, Birmingham, United Kingdom, 3 College of Medicine, University of Malawi, Blantyre, Malawi, 4 Department of Medicine, Nenagh Hospital,
County Tipperary, Ireland, 5 Liverpool School of Tropical Medicine, Liverpool, United Kingdom
Abstract
Aim: Early warning scores (EWS) are widely used in well-resourced healthcare settings to identify patients at risk of mortality.
The Modified Early Warning Score (MEWS) is a well-known EWS used comprehensively in the United Kingdom. The HOTEL
score (Hypotension, Oxygen saturation, Temperature, ECG abnormality, Loss of independence) was developed and tested in
a European cohort; however, its validity is unknown in resource limited settings. This study compared the performance of
both scores and suggested modifications to enhance accuracy.
Methods: A prospective cohort study of adults ($18 yrs) admitted to medical wards at a Malawian hospital. Primary
outcome was mortality within three days. Performance of MEWS and HOTEL were assessed using ROC analysis. Logistic
regression analysis identified important predictors of mortality and from this a new score was defined.
Results: Three-hundred-and-two patients were included. Fifty-one (16.9%) died within three days of admission. With a cutpoint $2, the HOTEL score had sensitivity 70.6% (95% CI: 56.2 to 82.5) and specificity 59.4% (95% CI: 53.0 to 65.5), and was
superior to MEWS (cut-point $5); sensitivity: 58.8% (95% CI: 44.2 to 72.4), specificity: 56.2% (95% CI: 49.8 to 62.4). The new
score, dubbed TOTAL (Tachypnoea, Oxygen saturation, Temperature, Alert, Loss of independence), showed slight
improvement with a cut-point $2; sensitivity 76.5% (95% CI: 62.5 to 87.2) and specificity 67.3% (95% CI: 61.1 to 73.1).
Conclusion: Using an EWS generated in developed healthcare systems in resource limited settings results in loss of
sensitivity and specificity. A score based on predictors of mortality specific to the Malawian population showed enhanced
accuracy but not enough to warrant clinical use. Despite an assumption of common physiological responses, disease and
population differences seem to strongly determine the performance of EWS. Local validation and impact assessment of
these scores should precede their adoption in resource limited settings.
Citation: Wheeler I, Price C, Sitch A, Banda P, Kellett J, et al. (2013) Early Warning Scores Generated in Developed Healthcare Settings Are Not Sufficient at
Predicting Early Mortality in Blantyre, Malawi: A Prospective Cohort Study. PLoS ONE 8(3): e59830. doi:10.1371/journal.pone.0059830
Editor: Jorge I F Salluh, D’or Institute of Research and Education, Brazil
Received November 11, 2012; Accepted February 19, 2013; Published March 29, 2013
Copyright: ß 2013 Wheeler et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Funding for the research was granted by the University of Birmingham. The funders had no role in study design, data collection and analysis, decision
to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: indi_wheeler@hotmail.com
settings [7,8] might significantly impact on the high mortality
observed in these areas of great disease burden [9,10].
The Early Warning Score and its modified counterpart,
MEWS, predict the need for hospital admission and in-hospital
mortality in European cohorts [2,11]. In South Africa, an
abbreviated MEWS assessment was used to predict patient
deterioration in an emergency department [12]. However, the
low positive predictive value of the score in this context limited its
usefulness [6]. Early Warning Scores (EWS) incorporate physiological measurements which do predict outcome [6,12,13]
although the addition of other simple clinical parameters might
further improve the sensitivity and specificity of these scores.
Introduction
Adverse outcomes in patients are often preceded by abnormal
physiological signs [1]. Early warning scores are composite scores
of these abnormalities which correlate with outcome. These tools
aid physicians in identifying the critically ill in order to direct
timely medical intervention [2]. The use of emergency assessment
tools in developed healthcare systems is relatively well documented
[2,3,4], as demonstrated by an agreement by the Royal College of
Physicians for the universal adoption of a Nationalised Early
Warning Score in the United Kingdom [5]. Conversely, their
performance is not well validated in resource limited settings [6].
Addressing the low rate of adoption of triage systems in such
PLOS ONE | www.plosone.org
1
March 2013 | Volume 8 | Issue 3 | e59830
A Comparison of Early Warning Scores in Malawi
The HOTEL score (Hypotension, Oxygen saturation, low
Temperature, ECG abnormalities, Loss of independence) [4]
presents an alternative to MEWS. This score is simple and easy to
calculate compared to the graded responses utilised in other early
warning assessment scores [14], making its use appropriate in an
emergency setting. The inclusion of ECG alongside the other
routine observations was shown to be particularly predictive of
mortality in an Irish cohort [4]. In sub-Saharan Africa its use as
part of scoring systems is not well validated. In light of the
emerging prevalence of ischaemic heart disease as a leading cause
of death in sub-Saharan Africa [15,16] this merits investigation.
The primary aim of this study was to compare the performances
of the HOTEL score and MEWS in a Malawian population. The
most important predictors of mortality in this population were also
identified and a modified assessment score suggested.
Statistical Analysis
The primary outcome of interest was death within 72 hours of
admission (i.e. time to death #3 days). 72 hour mortality is an
objective and clinically relevant endpoint [17,18] and this timeframe represents a compromise between the short term mortality
used in the original HOTEL study (15 mins –24 hours) and the
longer term mortality of one month used in other studies.
Descriptive statistics were used to investigate patient characteristics in the two outcome groups (dead or alive) and included
means (with standard deviations, SD), medians (with interquartile
ranges, IQR) and frequencies with percentages. Chi-squared tests
were used to compare proportions between groups. Means and
medians were compared using t-tests and Mann-Whitney U tests
respectively.
Individual HOTEL and MEWS scores were calculated
retrospectively for each participant and the distributions of the
scores in the two outcome groups were visualised using bar charts.
To calculate the HOTEL score, patients received a value of +1 for
each of five present abnormal measures; systolic BP,100 mmHg,
oxygen saturation,90%, temperature,35uC, ECG abnormality,
loss of independence (inability to stand unaided). MEWS was
calculated as shown in Table 1. Sensitivities and specificities with
95% confidence intervals were calculated for each cut-point of the
HOTEL and MEWS scores and receiver operating characteristic
(ROC) curves plotted. The area under the ROC curves
(AUROC), with 95% confidence intervals, were obtained to
evaluate the accuracy of HOTEL and MEWS at predicting inhospital mortality within three days. The positive predictive value
(PPV) and negative predictive value (NPV), with 95% confidence
intervals, were calculated for an optimal cut-point of each score.
A multivariable logistic regression model was built to investigate
important predictors of death in this population. Prior to fitting the
main model, univariable logistic regression models were used to
investigate the association between each variable and the outcome,
death within three days. A model was then fitted containing all
variables with p,0.25 from the univariable analyses. The model
was refined using backward elimination to remove variables one at
a time, based on Wald statistics, with a cut-point of p = 0.05. In
order to produce an easily calculated scoring system, continuous
variables were dichotomised at levels used in the HOTEL score
[4], or in previous research; for example, tachypnoea (.30 breaths
min21, bpm) is a commonly used predictor of mortality [2,12].
Conscious level was condensed to ‘alert or abnormal’; abnormal
being ‘V’, ‘P’ or ‘U’ from the AVPU score as previously suggested
[12]. Apparent model performance was assessed using the
Hosmer-Lemeshow goodness of fit test [19] and the c-statistic,
which is equivalent to the AUROC for binary outcomes. An
optimism adjusted c-statistic was obtained using bootstrap
validation of the model with 1000 bootstrap samples.
Regression coefficients from the final model were used to
generate a new score coined the TOTAL score. A ROC curve was
constructed for TOTAL and the area under the curve, with 95%
confidence interval, was obtained to evaluate its accuracy at
predicting in-hospital mortality within three days. Its performance
was compared to both HOTEL and MEWS.
Statistical analyses were carried out using IBM SPSS Statistics
version 19 and R version 2.10.1.
Methods
Ethical Considerations
The University of Malawi College of Medicine Research and
Ethics Committee (COMREC) and the Internal Ethics Review
Committee at the University of Birmingham, UK, approved the
study and waived the need for individual informed consent. The
use of verbal consent was also approved.
Design and Setting
A prospective observational cohort study of all adult medical
patients admitted to the Queen Elizabeth Central Hospital in
Blantyre, Malawi, between February 8th and March 9th 2012.
Patients aged over 18 years were recruited from the Adult
Emergency and Trauma Centre which operates as an admission
unit to the largest service and referral facility in Malawi, admitting
a total of 150,000 patients annually.
Practical constraints limited recruitment to weekdays 0800–
1700.
Data Collection
Data were pseudoanonymised by study number. Demographic
and physiological details were collected and recorded on the study
proforma by the primary investigator at enrolment. The primary
investigator and other healthcare professionals who collected
physiological information had received formal training to do so.
The following information was collected: time and date of
admission, sex, age (years), reported HIV status (positive, negative,
unknown), blood pressure (AU941 fCA,Rossmax Ltd, UK)
(mmHg), percutaneous oxygen saturation (ANAPULSE 100,Ana
Wiz Ltd, UK) (%), axillary temperature (uC), respiratory rate
(breaths counted over one minute), mid upper arm circumference
(cm), 12 lead electrocardiogram (CP100, Welch Allyn, UK), loss of
independence and conscious level using the AVPU score (A for
‘alert’, V for ‘response to vocal stimuli’, P for ‘response to painful
stimuli’, U for‘unconscious’). Loss of independence was defined as
the inability to stand unaided (yes or no). Mid upper arm
circumference was measured half-way between the acromion of
the shoulder and olecranon process of the elbow using a 150 cm
tape measure. ECGs were checked by two independent researchers who classified them as normal or abnormal (see Appendix S1).
ECGs containing sinus tachycardia or bradycardia in isolation
were recorded as normal in accordance with the original HOTEL
study [4]. Discrepancies between researchers were revisited and a
final status agreed upon. Patients were excluded if an ECG trace
was not obtained within one hour of admission.
The date and cause of death were ascertained from medical
notes.
PLOS ONE | www.plosone.org
Results
During the study, 361 patient observations were taken.
Nineteen patient records (5.3%) were excluded due to nonadmission to the medical wards and 11 patient records (3.0%) due
to missing data on one or more variables. A further 29 records
2
March 2013 | Volume 8 | Issue 3 | e59830
A Comparison of Early Warning Scores in Malawi
Table 1. Calculation of the Modified Early Warning Score (MEWS).
Systolic blood pressure (mmHg)
3
2
,70
1
0
1
2
3
70–80
81–100
101–199
Heart rate (bpm)
,40
40–50
51–100
101–110
111–129
$130
Respiratory rate (bpm)
,9
9–14
15–20
21–29
$30
Temperature (uC)
,35
35–38.4
Reacting to Voice
Reacting to Pain
Alert
AVPU score
$200
$38.5
Unresponsive
Each component of MEWS has an associated score ranging from 0 to 3, based on the degree of derangement of the parameter. The total score is the sum of each
component: the maximum possible score is 14.
doi:10.1371/journal.pone.0059830.t001
(8.0%) were excluded due to patients being discharged less than 3
days after admission, thus excluding those who may have died at
home within this period. There were 302 patient records (83.7%)
in the final analyses. The flow of patients through the study is
shown in Figure S1.
The mean age was 39.5 years (SD 15.9 years). A total of 155
(51.3%) patients were male and 180 (59.6%) patients were known
to be HIV positive. Fifty-one patients (16.9%) died #3 days after
admission; the median time from admission to death was 2 days
(IQR 1 to 5 days). The most common cause of death was due to a
respiratory complaint (n = 19, 37.3%).
mortality in this population, clinical usefulness is limited at these
levels.
Predicting Mortality within Three Days of Admission
In the univariable analyses (Table 4); low temperature (,35uC)
was associated with a 9.15-fold increase in odds of inpatient
mortality (95%CI: 2.86 to 29.30; p,0.001), thus making it one of
the strongest predictors. Reduced conscious level was also very
predictive with an unadjusted odds ratio (OR) of 3.87 (95% CI:
1.95 to 7.69; p,0.001) when compared to fully alert. Also strongly
associated with in-patient mortality were hypoxia (oxygen
saturation ,90%) (OR 3.92; 95% CI: 2.00 to 7.71; p,0.001),
tachypnoea (respiratory rate .30 bpm) (OR 2.26; 95%CI: 1.23 to
4.17; p = 0.009) and the inability to stand unaided (OR 4.75; 95%
CI: 2.06 to 10.95; p,0.001). ECG abnormality showed a weaker,
but significant, association with mortality (OR 2.08; 95%CI: 1.12
to 3.88; p = 0.020), but known HIV positive status did not (OR
0.91; 95%CI: 0.44 to 1.86; p = 0.790).
The final multivariable logistic model contained five variables
(see Table 5): respiratory rate.30 bpm (OR 2.14; 95% CI: 1.04 to
4.42; p = 0.039), oxygen saturation,90% (OR 3.22; 95% CI: 1.47
to 7.05; p = 0.004), temperature,35uC (OR 10.33; 95% CI: 2.90
to 36.86; p,0.001), reduced conscious level (OR 3.25 95% CI:
1.47 to 7.16; p = 0.003) and loss of independence (OR 3.24; 95%
CI: 1.30 to 8.07; p = 0.012). The Hosmer-Lemeshow goodness of
fit test demonstrated good calibration (p = 0.860). The model also
had good discrimination with an apparent c-statistic of 0.794 and
an optimism adjusted c-statistic of 0.778 after using bootstrap
validation. This suggests that it performs well at distinguishing
patients who died from those who did not.
The five parameters from the final regression model were used
to describe a new score, coined TOTAL (Tachypnoea, Oxygen
saturation, Temperature, Alert and Loss of independence). The
scores for each parameter were obtained by rounding the
associated regression coefficient to the nearest integer; each
abnormal reading scores +1 except temperature which is allocated
+2 points due to its larger regression coefficient. The TOTAL
score ranges from 0 to 6 and the distribution of the scores in the
two outcome groups (dead and alive) is shown in Figure S4.
Physiological Parameters on Admission
Table 2 presents the demographic and physiological characteristics on admission by outcome. Patients who died had significantly
lower oxygen saturation (p,0.001), lower temperature (mean
difference: 20.51; 95% CI: 20.91 to 20.11; p = 0.012) and higher
respiratory rate (mean difference: 6.02; 95% CI: 2.60 to 9.45;
p = 0.001). Patients who died more commonly had reduced
conscious level (p,0.001), were less frequently able to stand
unaided (p,0.001) and more likely to have an ECG abnormality
(p = 0.019). There was no significant difference in systolic blood
pressure (mean difference: 6.61; 95% CI: 22.12 to 15.34;
p = 0.137).
Comparison of HOTEL and MEWS in Predicting Mortality
within a Malawian Population
Figure S2 illustrates the frequency distributions of the HOTEL
and MEWS scores in the two outcome groups (dead and alive).
Most frequent HOTEL score was 1 for those who survived
(n = 101, 40.2%), and 2 for patients who died (n = 18, 35.3%). The
only patient who had a HOTEL score of 5 died within 24 hours.
The most frequent MEWS score was 4 for those who survived
(n = 46, 18.3%) and 6 for patients who died (n = 10, 19.6%).
Table 3 shows the sensitivities and specificities with 95%
confidence intervals for HOTEL and MEWS at select cut-points.
Optimal discrimination using HOTEL is found at a score $2
which gives sensitivity 70.6% (95% CI: 56.2 to 82.5) and specificity
59.4% (95% CI: 53.0 to 65.5). Using the optimal cut-point of score
$5, MEWS performs less well with sensitivity 58.8% (95% CI:
44.2 to 72.4) and specificity 56.2% (95% CI: 49.8 to 62.4). At these
thresholds, MEWS has positive predictive value (PPV) 21.4%
(95% CI: 14.9 to 29.2) and NPV 87.0% (95% CI: 80.9 to 91.8),
while HOTEL is a better discriminator; PPV 26.1% (95% CI:
19.0 to 34.2) and NPV 90.9% (95% CI: 85.4 to 94.8). The
performance accuracy is further illustrated by the ROC curves
(Figure S3). AUROCs were: MEWS = 0.59 (95% CI: 0.51 to 0.68)
and HOTEL = 0.70 (95% CI: 0.62 to 0.78). While it is evident that
HOTEL is more accurate than MEWS at predicting in-patient
PLOS ONE | www.plosone.org
Performance Accuracy of TOTAL
The performances of MEWS, HOTEL and TOTAL are
compared in Table 2. The optimum cut-point for TOTAL ($2),
gives sensitivity and specificity 76.5% (95% CI: 62.5 to 87.2) and
67.3% (95% CI: 61.1 to 73.1) respectively, with corresponding
PPV 32.2% (95% CI: 24.0 to 41.3) and NPV 93.4% (95%CI: 88.7
to 96.5). The TOTAL score outperforms both HOTEL and
MEWS in terms of sensitivity and specificity, as illustrated by the
3
March 2013 | Volume 8 | Issue 3 | e59830
A Comparison of Early Warning Scores in Malawi
Table 2. Patient characteristics on admission by outcome group; n = 302.
Characteristic
All (n = 302)
Dead (n = 51)
Alive (n = 251)
p-value
Age (years); mean (SD)
39.5 (15.9)
Sex*; male (n, %)
155 (51.3)
39.4 (14.2)
39.6 (16.3)
0.930
29 (56.9)
126 (50.2)
Positive
0.753
180 (59.6)
28 (54.9)
152 (60.6)
0.566
Negative in last year
77 (25.5)
13 (25.5)
64 (25.5)
Unknown
45 (14.9)
10 (19.6)
35 (13.9)
Systolic blood pressure (mmHg);
mean (SD)
114.0 (28.9)
114.0 (28.9)
112.8 (26.9)
0.137
Oxygen saturation#; median (IQR)
96 (92 to 98)
92 (80 to 97)
97 (93 to 98)
,0.001
Axillary temperature (uC); mean (SD)
37.0 (1.3)
36.5 (1.5)
37.1 (1.3)
0.012
Respiratory rate (breaths per min);
mean (SD)
29.2 (11.5)
34.2 (13.5)
28.2 (10.9)
0.001
Pulse rate (beats per min);
mean (SD)
105.0 (25.6)
99.5 (27.9)
106.2 (25.0)
0.087
Mid upper arm circumference (cm); mean (SD)
22.8 (3.8)
22.1 (4.0)
23.0 (3.7)
0.115
ECG abnormal*; n(%)
89 (29.5)
22 (43.1)
67 (26.7)
0.019
Loss of independence*; n(%)
187 (61.9)
44 (86.3)
143 (57.0)
,0.001
Alert
253 (83.8)
33 (64.7)
220 (87.6)
,0.001
Voice
33 (10.9)
10 (19.6)
23 (9.2)
Pain
9 (3.0)
5 (9.8)
4 (1.6)
Unresponsive
7 (2.4)
3 (5.9)
4 (1.6)
HIV status* (n, %):
Conscious level*; n(%):
SD, standard deviation; IQR, interquartile range.
t-tests were used to test differences between the alive and dead groups unless otherwise indicated;
*Chi-squared test;
#
Mann-Whitney U test.
doi:10.1371/journal.pone.0059830.t002
Table 3. Sensitivities and specificities (in percentages) with 95% confidence intervals for MEWS, HOTEL and TOTAL at select cutpoints (n = 302).
Score (range)
Cut-point
Frequency
%
Sensitivity (95% CI)
Specificity (95% CI)
MEWS (0 to 14)
$1
295
97.7
98.0
(89.6, 100.0)
2.4
(0.9, 5.1)
$2
268
88.7
92.2
(81.1, 97.8)
12.0
(8.2, 16.6)
$3
234
77.5
86.3
(73.7, 94.3)
24.3
(19.2, 30.1)
$4
192
63.6
70.6
(56.2, 82.5)
37.8
(31.8, 44.2)
$5
140
46.4
58.8
(44.2, 72.4)
56.2
(49.8, 62.4)
$6
93
30.8
45.1
(31.1, 59.7)
72.1
(66.1, 77.6)
$7
56
18.5
25.5
(14.3, 39.6)
82.9
(77.6, 87.3)
$8
30
9.9
13.7
(5.7, 26.3)
90.8
(86.6, 94.1)
$1
251
83.1
94.1
(83.8, 98.8)
19.1
(14.4, 24.5)
$2
138
45.7
70.6
(56.2, 82.5)
59.4
(53.0, 65.5)
$3
44
14.6
35.3
(22.4, 49.9)
89.6
(85.2, 93.1)
$4
8
2.6
11.8
(4.4, 23.9)
99.2
(97.2, 99.9)
$5
1
0.3
2.0
(0.5, 10.4)
100.0
(98.5, 100.0)
$1
236
78.1
98.0
(89.6, 100.0)
25.9
(20.6, 31.8)
$2
121
40.1
76.5
(62.5, 87.2)
67.3
(61.1, 73.1)
$3
49
16.2
43.1
(29.3, 57.8)
89.2
(84.7, 92.8)
$4
11
3.6
19.6
(9.8, 33.1)
99.6
(97.8, 100.0)
$5
1
0.3
2.0
(0.5, 10.4)
100.0
(98.5, 100.0)
$6
1
0.3
2.0
(0.5, 10.4)
100.0
(98.5, 100.0)
HOTEL (0 to 5)
TOTAL (0 to 6)
doi:10.1371/journal.pone.0059830.t003
PLOS ONE | www.plosone.org
4
March 2013 | Volume 8 | Issue 3 | e59830
A Comparison of Early Warning Scores in Malawi
Table 4. Unadjusted odds ratios for demographic and physiological variables; n = 302.
Univariable Analysis
Variable
OR
95% CI
p-value
Age (years)
HOTEL
MEWS
TOTAL
1.00
(0.98, 1.02)
0.930
Sex(reference: female)
1.31
(0.71, 2.40)
0.390
Positive
0.91
(0.44,1.86)
0.790
Unknown
1.41
(0.56, 3.54)
0.470
0.93
(0.86, 1.02)
0.120
HIV status (reference: negative):
Mid Upper Arm Circumference (cm)
Pulse rate (.120 beats per min)
Systolic Blood Pressure (,100 mmHg)
3
ECG abnormality
3
Respiratory rate (.30 breaths per min)
3
1.22
(0.62, 2.43)
0.570
3
0.89
(0.47, 1.70)
0.730
3
Oxygen saturation (,90%)
3
Temperature (,35uC)
3
Conscious level (reference: alert):
2.08
(1.12, 3.88)
0.020
3
2.26
(1.23, 4.17)
0.009
3
3.92
(2.00, 7.71)
,0.001
3
3
9.15
(2.86,9.30)
,0.001
3
3
Abnormal
Loss of independence
3
3
3.87
(1.95, 7.69)
,0.001
4.75
(2.06,10.95)
,0.001
doi:10.1371/journal.pone.0059830.t004
inclusion of respiratory rate is perhaps to be expected, as its lack of
value in previous research [4] was suggested to be due to
inaccurate recordings by the research team. Surprisingly, HIV
status was not identified as an important predictor of death. This is
perhaps because although the frequency of ailments such as
pneumonia is higher in seropositive individuals, this does not
necessarily translate into an increased case fatality rate if
appropriate treatment is available [20].
The importance of conscious level and loss of independence as
predictors of mortality [6,12] is particularly useful in this context
as both observations can be recorded without equipment and can
be ascertained quickly and objectively in the emergency department by individuals with minimal training [6].
In contrast, some previously identified important predictors of
death were found to be less important in this population.
Hypotension, for example, has been shown to be an independent
predictor of in-hospital mortality [4,6,12,21] and an important
trigger for Intensive Care Unit admission [22], but our study
found no evidence of this. This could be explained by the small
scale of the study, or because hypotension in this context may be
related to treatable disease processes, such as malaria or bacterial
ROC curves (Figure S3). The AUROC for TOTAL was 0.78
(95% CI: 0.71 to 0.85).
Discussion
Previous research has shown the benefits of comprehensive
triage assessment [2,4,6,9–12]. This study compared two early
warning assessment scores, HOTEL and MEWS, and found some
evidence that HOTEL is superior to MEWS in a Malawian
population, but also that further modifications may lead to modest
improvements in predictive accuracy.
The average age of participants in this study was considerably
younger than those reported in studies using HOTEL and MEWS
in developed healthcare environments, where almost no patients
under the age of 50 died [1,2,4]. This situation was also observed
in research conducted in study settings similar to Malawi and is
likely to be explained by the lower life expectancies in these
countries [10].
The factors most associated with in-patient mortality within
three days were tachypnoea, hypoxia, low temperature, deterioration in conscious level and inability to stand unaided. The
Table 5. Adjusted odds ratios for demographic and physiological variables used in the TOTAL score; n = 302.
Variable
Coefficient
(95% CI)
OR
(95% CI)
p-value
Score*
1
Respiratory rate (.30 breaths per min)
0.76
(0.04, 1.49)
2.14
(1.04, 4.42)
0.039
Oxygen saturation (,90%)
1.17
(0.38, 1.95)
3.22
(1.47, 7.05)
0.004
1
Temperature (,35uC)
2.34
(1.06, 3.61)
10.33
(2.90, 6.86)
,0.001
2
Conscious level (reference: alert)
1.18
(0.39, 1.97)
3.25
(1.47, 7.16)
0.003
1
Loss of independence
Abnormal
1.18
(0.26, 2.09)
3.24
(1.30, 8.07)
0.012
1
Intercept
23.48
(24.39, 22.58)
2
2
,0.001
2
*The score for each component of TOTAL was obtained by rounding each regression coefficient to the nearest integer value. The TOTAL score ranges from 0 to 6.
doi:10.1371/journal.pone.0059830.t005
PLOS ONE | www.plosone.org
5
March 2013 | Volume 8 | Issue 3 | e59830
A Comparison of Early Warning Scores in Malawi
well-performing scoring system could help with this. Therefore, we
suggest that local data collection, looking at individual physiological predictors, should occur in the locale intended for EWS
implementation before such systems are adopted. This should
occur alongside impact assessment studies to investigate whether
the use of a prognostic score results in an improvement in doctors’
decision making and ameliorated patient outcome [28]. This will
encourage an evidence based practice approach to emergency
medicine in under-resourced settings, and will enhance our
understanding of the physiology of disease in different populations.
infection [6]. Secondly, hypotension is a relatively late sign of
physiological derangement [23] and may be missed without serial
measurements. However, a recent study in Uganda found an
association between a reduced mean arterial pressure and
mortality [24], and this relationship could also apply in our
population.
Similarly, a normal ECG tracing has previously been documented as a strong predictor of survival in a European cohort [4].
In this study, abnormal ECG was associated with increased risk of
mortality in the univariable analysis, but its predictive ability was
lost after adjusting for the confounding effects of other variables.
Further analysis of the effect of individual ECG abnormalities on
patient outcome may be useful, although a differing prevalence of
ischaemic heart disease may explain this discrepancy.
These findings demonstrate that the usefulness of early warning
scores depends on the situation in which they are used.
It is perhaps unsurprising that the HOTEL score, which was
shown to accurately identify at risk patients in an Irish cohort [4]
(AUROC: 0.85, 95% CI: 0.75 to 0.96), did not achieve the same
predictive accuracy in this setting (AUROC: 0.70. 95% CI: 0.62 to
0.78) and the low PPV (26.1%) precludes its adoption into clinical
practice. Interestingly, MEWS, which is relied on extensively in
well-resourced healthcare settings [2,13,14], had a weaker
performance in the Malawian population compared to HOTEL.
This may in part be due to the inclusion of oxygen saturation in
the HOTEL score which supports calls for its increased clinical use
due to its ease of application and low cost.
This study presents evidence for the use of population-specific
predictors of mortality in adults. In comparison to HOTEL and
MEWS, TOTAL had a higher specificity allowing resources to be
more effectively managed. This is of critical importance to the
sustainable adoption of risk scores in low income countries [25].
The balance between identifying high risk individuals yet not
overburdening health services is essential in this setting and
ultimately determines the suitability of such a tool. Some
assessment tools such as the Integrated Management of Childhood
Illness [26], which is utilised in the community, work on a ‘rule
out’ basis, identifying those who do not have a defined illness and
treating all others [27]. In contrast, the MEWS, HOTEL and
TOTAL scores, when applied to medical patients are ‘rule in’
systems relying on high positive predictive value; they identify the
patients most in need of immediate intervention. In our cohort,
TOTAL had a PPV of 32.2%, which was higher than HOTEL
and MEWS. Although a higher PPV is desirable, as it stands
TOTAL would identify more than half of the total patient
population at high risk and, on average, for every three patients
identified as high risk, one death would be expected.
Our findings suggest that risk scores work best in the
populations from which they were derived. We have shown that
adopting an early warning score generated in a developed
healthcare setting into a contrasting environment is likely to
impact negatively on performance, despite the assumption that
physiological responses to disease are common to all patients. This
apparent discrepancy between predictors of mortality in different
populations could be explained by a number of theories. For
example, specific disease processes may impact more on outcome
than the physiological effect they exert and physiological responses
to disease may be highly dependent on age as our patients were
considerably younger than the population studied in the previous
HOTEL cohort.
In our hospital, we feel that there is considerable room for
improvement in the service despite limited resources. This might
be through a better focus of the available personnel and improved
efficiency in identifying ill patients to reduce time to treatment. A
PLOS ONE | www.plosone.org
Limitations
As observations were only taken within office hours, patients
admitted during the night and weekends were not represented, yet
previous research has demonstrated that admissions during these
times carry a greater risk of mortality [29,30]. Patients were also
excluded from analysis if they had been discharged less than three
days after admission, resulting in roughly 10% of patient records
being discarded. However, since this group of patients had
unknown outcomes, removing them prior to analysis ensured
accuracy in the statistical calculations and models.
This study was fairly small and since the TOTAL score was
developed on this dataset, its reported accuracy is likely to be
optimistic. We compensated for this by using bootstrap validation
to provide an optimism adjusted estimate of performance, but
external validation on larger unseen datasets is required in order to
evaluate its true predictive accuracy in the Malawian population.
Finally, while the TOTAL score demonstrates a modest
improvement in high risk patient identification, the incremental
benefit over MEWS and HOTEL is not enough for robust triage
or resource allocation in clinical use. Instead, further clinical
parameters, with possible inclusion of laboratory tests results such
as rapid diagnostic tests for malaria, offer the potential to
considerably improve usefulness of EWS, and should be investigated in this context.
The need for cost effective comprehensive adult emergency
assessment systems to be implemented in low income settings
remains a priority. We have demonstrated that EWS generated in
developed healthcare systems do not have the same predictive
accuracy in differing populations. Modified scores, such as the
TOTAL score, tailored to the unique predictors of that
population, show an improved performance and could also have
the potential for adoption in similar patient populations. However,
we appreciate that, currently, TOTAL is a not a clinically useful
tool and further investigation of different clinical observations and
studies on larger datasets are necessary in order to create an
accurate assessment tool with real clinical use.
Supporting Information
Figure S1
Flow of patients through the study; n = 302.
(TIF)
Figure S2 Distribution of the HOTEL and MEWS scores
across the two outcome groups (alive and dead); n = 302.
(TIF)
Figure S3 Receiver operator characteristic (ROC)
Curves for the HOTEL, MEWS and TOTAL scores
(n = 302). The solid line shows the line of no discrimination
where the test is no better than chance.
(TIF)
Figure S4 Distribution of the TOTAL scores across the
two outcome groups (dead and alive); n = 302.
(TIF)
6
March 2013 | Volume 8 | Issue 3 | e59830
A Comparison of Early Warning Scores in Malawi
(DOCX)
grateful for the help of Dr Opio and his team in Uganda for providing their
helpful correspondence of the research they are carrying out.
Acknowledgments
Author Contributions
Thanks to the AETC staff for all their help, Mrs Julie Shore and Dr Lesley
Roberts for their support in conducting this research; to the University of
Birmingham and to the Birmingham Primary Care Research Trust who
kindly supplied the ECG machine used in this study. Moreover we are very
Provided expertise and advice: JK. Organised ‘on site’ materials and
resources: MN PB. Conceived and designed the experiments: IW CP AS
MN JR. Performed the experiments: IW. Analyzed the data: IW CP AS.
Contributed reagents/materials/analysis tools: JK MN PB. Wrote the
paper: IW JR CP AS.
Appendix S1 Classification table of ECG abnormalities.
References
1. Goldhill R (2001) The critically ill: following your MEWS. QJM; 94 (10): 507–
510.
2. Subbe CP, Kruger M, Rutherford P, Gemmel L (2001) Validation of a Modified
Early Warning Score in medical admissions. QJM; 94 (10): 521–526.
3. Prytherch D, Smith G, Schmidt P, Featherstone P (2010) ViEWS-Towards a
national early warning score for detecting adult inpatient deterioration.
Resuscitation 81 932–937.
4. Kellett J, Deane B, Gleeson M (2008) Derivation and validation of a score based
on Hypotension, Oxygen saturation, low Temperature, ECG changes and Loss
of independence (HOTEL) that predicts early mortality between 15 min and
24 hr after admission to an acute medical unit. Resuscitation; 78: 52–58.
5. Royal College of Physicians (2012) National Early Warning Score. Standardising
the assessment of acute-illness severity in the NHS. LondonRCP: 2012.
6. Rylance J, Baker T, Mushi E, Mashage D (2009) Use of an early warning score
and the ability to walk predicts mortality in medical patients admitted to
hospitals in Tanzania. Trans R Soc Trop Med Hyg; 103 (8): 790–794.
7. Nolan T, Angos P, Cunha AJ, Muhe L, Qazi S, et al (2001) Quality of Hospital
Care for seriously ill children in less-developed countries. Lancet; 357: 106–110.
8. Dunser M, Baelani WI, Gambold L (2006) A review and analysis of intensive
care medicine in the least developed countries. Crit Care Med; 34: 1234–1242.
9. Roberston M, Molyneux E (2001) Triage in the developing world- can it be
done? Arch Dis Child; 85: 208–213.
10. Wallis L, Gottshcalk SB, Wood D, Bruijns S, de Vries S, et al (2006) The Cape
Triage Score- a triage system for South Africa. S Afri Med J; 96 (1): 53–56.
11. Cei M, Bartolomei C, Mumoli N (2009) In-hospital mortality and morbidity of
elderly medical patients can be predicted at admission by the Modified Early
Warning Score: a prospective study. Int J ClinPrac; 63 (4): 591–5.
12. Burch VC, Tarr G, Morroni C (2008) Modified early warning score predicts the
need for hospital admission and in-hospital mortality. Emerg Med J; 25: 674–
678.
13. Subbe CP, Slater A, Menon D, Gemmell (2006) Validation of physiological
scoring systems in the accident emergency department. Emerg Med J; 23: 841–
845.
14. Subbe CP, Hibbs R, Williams E Rutherford P, Gemmel L (2002) ASSIST: a
screening tool for critically ill patients on general medical wards. Intensive Care
Med; 28 (supplement 1): S21.
15. Stewart S, Wilkinson D, Becker A, Askew D, Ntyintyane L, et al (2006) Mapping
the emergence of heart disease in a black urban population in Africa: the Heart
of Soweto Study. Int J Cardiol; 108 (1): 101–108.
16. Mensah GA (2008) Ischaemic heart disease in Africa. Heart. Jul; 94(7): 836–43.
17. Yeguiayan JM, Garrigue D, Binquet C, Jacquot C, Duranteau J, et al (2011)
Medical pre-hospital management reduces mortality in severe blunt trauma: a
prospective epidemiological study. Critical Care, 15: R34.
PLOS ONE | www.plosone.org
18. Eriksson EA, Barletta JF, Figueroa BE, Bonnell BW, Sloffer CA, et al (2012) The
first 72 hours of brain tissue oxygenation predicts patient survival with traumatic
brain injury. J Trauma Acute Care Surg. May;72(5): 1345–9.
19. Steyerberg EW (2009) Clinical Prediction Models. A Practical Approach to
Development, Validation and Updating Series: Statistics for Biology and Health.
New York: NY. Springer.
20. Sowden E, Carmicheal AJ (2004) Autoimmune inflammatory disorders, systemic
corticosteroids and pneumocystis pneumonia: A strategy for prevention. BMC
Infectious Diseases, 4: 42.
21. Buist M, Bernard S, Nguyen TV, Moore G, Anderson J (2004) Association
between clinically abnormal observations and subsequent in-hospital mortality: a
prospective study. Resuscitation; 62: 137–41.
22. Kennedy M, Joyce N, Howell MD, Lawrence Mottley J, Shapiro NI (2010)
Identifying infected emergency department patients admitted to the hospital
ward at risk of clinical deterioration and intensive care unit transfer. AcadEmerg
Med;17(10): 1080–5.
23. Harrison GA, Jacques TC, Kilborn G, Mc Laws ML (2005) The prevalence of
recordings of the signs of critical conditions and emergency responses in hospital
wards–the SOCCER study. Resuscitation; 65 (2): 149–157.
24. Jacob ST, Moore CC, Banura P, Pinkerton R, Meya D, et al (2009) Severe
Sepsis in Two Ugandan Hospitals: a Prospective Observational Study of
Management and Outcomes in a Predominantly HIV-1 Infected Population.
PLoS ONE; 4(11). e7782.
25. Baker T (2009) Critical care in low-income countries Trop Med Int Health; 14:
143–148.
26. Gove S, Tamburlini G, Molyneux E, Whitesell P, Campbell H (1999)
Development and technical basis of simplified guidelines for emergency triage
assessment and treatment in developing countries. WHO Integrated Management of Childhood Illness (IMCI) Referral Care Project. Arch Dis Child: 81:
473–7.
27. Tulloch J (1999) Integrated approach to child health in developing countries.
Lancet 354: Suppl 2S: 16–S: 20.
28. Moons K, Altman DG, Vergouwe Y, Royston P (2009) Prognosis and prognostic
research: application and impact of prognostic models in clinical practice. BMJ;
338: b606.
29. Mabiala-Babela JR, Senga P (2009) Night time attendance at the Paediatric
Emergency Room of the University Hospital Centre in Brazzaville, Congo. Med
Trop; 69(3): 281–5.
30. Wujtewicz MA, Suszyńska-Mosiewicz A, Sawicka W, Piankowski A, DylczykSommer A (2011) Does the time of admission to ITU affect mortality?
AnestezjolIntensTer; 43(4): 230–3.
7
March 2013 | Volume 8 | Issue 3 | e59830
© 2013 Wheeler et al. This is an open-access article distributed under the
terms of the Creative Commons Attribution License:
https://creativecommons.org/licenses/by/4.0/ (the “License”), which permits
unrestricted use, distribution, and reproduction in any medium, provided the
original author and source are credited. Notwithstanding the ProQuest Terms
and Conditions, you may use this content in accordance with the terms of the
License.
Hey Tutor,
This job includes an assignment and a discussion post. The discussion post should be submitted on
Wednesday. It is a one-page document with single spacing. This does not include the reference page.
The assignment should have 4 pages different from the reference page. You must cit the book. Chapters
3 and 4 are pasted in this word document. Also make use of other sources. You must carefully follow the
instructions and scoring guides for both the assignment and the discussion post. The discussion post
must be submitted on Wednesday and the assignment must be submitted before Friday this week.
Applying Big Data Analytics to Quality Assurance
Introduction
Consider how big data can impact health care decision making. It is important to
note that big data collection must be accurate and done in way that meets needs
of the organization. Big data collection is a big task and health care organizations
must have the personnel with the skills to analyze and transform the data.
Successful data analysis and transformation is proven to improve quality and
patient safety. Now take a moment to review the following videos on how data
are used.
Why Big Data Is About Making Better Decisions.
• How Big Data Could Transform The Health Care Industry.
After watching the videos, consider how these concepts can improve quality and
safety. Next, research published data from a credible source such as CMS,
AHRQ, or other health-related databases. (Refer to your unit readings from the
Capella library for additional information on databases). Review the data being
gathered and consider how these data can be applied to health care
organizations to improve processes that provide quality assurance.
•
Instructions
Apply data generated by performance outcome measurement systems for the
purposes of decision support, risk adjustments, and benchmarking. To be
successful in this assignment, you need to complete the following:
•
•
Watch the two videos regarding big data.
Research existing databases and consider how big data are gathered in health
care.
Research how organizations use benchmarking to help meet stakeholder
needs.
• Research 2–3 types of data from the databases you reviewed that are
typically used in a health care organization.
In your paper:
•
Define and describe the types of data typical to health care organizations.
Describe how using data in benchmarking helps organizations improve
quality and patient safety.
• Present current trends and best practices from scholarly literature and
credible industry sources on how to apply data to organizational practices.
• Recommend leadership approaches for ensuring performance improvements
in a health care setting based on benchmarking data.
Make sure your paper follows appropriate APA formatting and has 3–5 current
peer-reviewed references.
•
•
Pay attention to the critical elements that form part of the grading criteria for this
assignment:
• Demonstrate an understanding of the concept of big data in health care by
defining the types of data typically gathered and analyzed in health care
settings.
• Determine the types of data needed for quality improvement in health care
settings.
• Apply current best practices in generating meaningful outputs from data to
improve performance.
• Analyze current trends in data collection and analysis to promote patient
safety.
• Apply proven leadership approaches to interpreting benchmark results and
applying data analysis to quality improvement.
Writing Requirements
Your paper should meet the following requirements:
Length: 4–5 double-spaced pages (excluding the cover page and references
list). Include page numbers, headings, and running headers.
• References: 3–5 current peer-reviewed references.
• Formatting: Use current APA style and formatting, paying particular attention
to citations and references.
• Font and font size: Times New Roman, 12 point.
Review the Applying Big Data Analytics to Quality Assurance Scoring Guide to
ensure you understand the grading criteria for this assignment.
•
Submit your Word document as an attachment to the assignment area.
Note: Your instructor may also use the Writing Feedback Tool to provide
feedback on your writing. In the tool, click the linked resources for helpful writing
information.
Discussion Topic:
Improving Quality Through Data Use
Research and discuss with your peers the types of data typically gathered by
health care organizations to help improve quality, reduce risk, and support
decision making.
Response Guidelines
Review the posts of your peers and respond according to the Faculty
Expectations Message guidelines, using one of the following approaches:
Identify knowledge gaps or unknowns that were not considered in your
peer’s post.
• Identify an assumption on which the post seems to be based, and pose a
useful alternative or contrasting approach, based on a different assumption.
• Ask a probing question.
• Elaborate on a particular point.
If you are responding with a personal perspective or an example from your
workplace experience, you are not required to cite a source. However, if you offer
an alternative viewpoint or refer to the ideas or work of others in your response to
your peer, it must be supported with an outside source, cited using current APA
style.
•
Other Links:
•
•
View the video How Big Data Could Transform The Health Care
Industry | Transcript from Booz Allen Hamilton. This video focuses on how
data drive decision making in health care.
View the video Why Big Data Is About Making Better
Decisions | Transcript from cmispeakers. This video discusses the use of big
data in industry as a whole by demonstrating data use at airports. However,
think about how the health care industry uses the same concepts.
Book Reference:
Transforming Health Care Management: Integrating Technology Strategies Written by Ivan J. Barrich
Chapter 3:
COMPUTER USE AND USER REQUIREMENTS
Widespread evidence of pervasive computer use is readily apparent in virtually all healthcare delivery
organizations. For example, in hospital settings, admissions departments process information about
incoming patients. Using computer termi- nals or personal computer networks, hospital representatives
enter and/or update information about each patient (patient’s name and other demographic
information, insurance information, and a reason to be admitted to the hospital). This informa- tion is
stored in the Hospital’s Information System (HIS) to be retrieved as needed.
Group practices or health maintenance organizations electronically submit insurance claims for each
patient’s bill by using computer systems specifically designed for their practice or business. Medical
secretaries and administrative assistants schedule appointments, transcribe medical reports, or use
office com- puter systems for medical practice accounting. As the pace of digital transforma- tion
quickens in patient care delivery processes, exciting opportunities exist in these provider organizations
and in other businesses providing products and ser- vices related to healthcare delivery.
Workers entering this workforce must be prepared to demonstrate an opera- tional understanding of
computer systems of all types. For example, medical tran- scriptionists operate word processors
integrated with voice recognition software to prepare and edit documentation of care delivery dictated
by clinicians. Nursing staff monitor patients’ vital signs using wireless Personal Digital Assistants (PDAs),
smart cellular telephones, or bedside computer terminals. Radiology tech- nologists use advanced
imaging technology to capture, analyze, and communicate information gain without invasive diagnostic
and therapeutic patient procedures. Large research institutions employ many professionals with a
variety of skills who are making new discoveries and advancing patient care options.
Healthcare team members work together, researching and delivering solutions to life-threatening
problems in specialties such as neurosurgery, endocrinology, and cardiology. Computerized technologies
of all types are used to facilitate patient specimen and data analyses and to rapidly produce information
to support research
ctivity of these teams. By decreasing time devoted to necessary but increasingly burdensome
documentation and other administrative tasks, physicians are devot- ing more time to patients by
practicing medicine with computer assistance equipped with more user-friendly technology. Federal
mandates require digital creation, processing and transmission of insurance claims, and automated
posting of paid claims to each patient’s account.
Medical claims companies need qualified computer-literate workers with spe- cialized medical training
to analyze patient claims and documentation necessary to process and adjudicate medical insurance
claims. Nurses are frequently employed to review medical records, collecting data archived for future
utiliza- tion analyses. Specialized support services augment healthcare professionals (e.g., companies
developing, producing, and distributing pharmaceuticals or other spe- cialized equipment, such as
hospital beds and wheelchairs). These businesses employ administrative and sales staff to market and
distribute products. More intensive use of ITs of all types increases productivity and reduces cost.
Special- ized professionals need computer-literate support staff with combined medical backgrounds
and IT training.
IT functions include acquiring, processing, and storing enormous amounts of information. As
technologies have become ubiquitous, IT is increasingly com- monplace in hospitals, clinics, laboratories,
imaging centers, research facilities, and physicians’ offices. Unfortunately, significant technology
investment has not been and is not well integrated or interoperable. Current healthcare processes do
not effectively support patient-centered, efficient, quality-driven, and affordable healthcare delivery.
PATIENTS DESERVE MORE, NOT JUST MORE TECHNOLOGY
Despite per capita expenditures that are consistently the highest in the world, America’s health status is
currently unacceptable, reflective of inefficient and inef- fective care delivery systems and processes.
Figure 3.1 is a bar graph showing that America has the highest healthcare expenditures as a percent of
gross national product among developed nations. Figure 3.2 is a bar graph that highlights Amer- ica’s
healthcare expenditures as the highest per capita among developed nations. Figure 3.3 is a scatter
diagram comparing male life expectancy to annual per capita health expenditures. America is in the
lower right quadrant, reflecting significantly higher expenditures and lower life expectancy outcomes
(America spends the most per capita for health care and experiences the lowest male life expectancy
among these developed countries). Life expectancy is but one of a number of health sta- tus indicators
that are internationally accepted benchmarks used to demonstrate regional, state, or national
healthcare delivery effectiveness.
limited work process analysis and reengineering which would otherwise have resulted in efficient
interfaces, seamless integration, and interoperability among key systems throughout most provider
organizations.
To proficiently participate on any project team, aspiring IT professionals need to understand
fundamental software and hardware functions, input and interface requirements, and output
specifications. Through collaboration, team members gain a required appreciation of how a proposed
solution will influence workflow effectiveness, efficiency, staffing requirements and timeliness, cost, and
functions and feature trade-offs. IT professionals must be able to analyze, interpret, and pre- sent
relevant operations research. These analyses should necessarily involve user, clinician, and patient
interactions with each module of any system application. Benchmarking techniques are critical and
applicable to proficiently evaluate tech- nical hardware and software performance as well as user
efficiency and overall care delivery process effectiveness.
SYSTEM FUNDAMENTALS AND COMPUTER APPLICATIONS
These reflections illustrate the need to gain an essential understanding and appre- ciation of how any
computer works. Such learning begins with a fundamental def- inition of a computer: an electronic
device capable of facilitating data input and performing manipulation, storage, calculation, comparison,
communication, trans- formation, formatting, reporting, and presentation.
Computers and computer systems are classified in different ways for different purposes, including size,
type, function, and feature. Traditionally, size and prossing capabilities were predominant characteristics of supercomputers, main- frame computers,
minicomputers, microcomputers (i.e., desktop, laptop, PDAs, and, more recently, smart telephones).
Technology trends reflect accelerated development of faster, more powerful multiprocessors with more
dense and greater memory capacity, and smaller size that consume less power and generate less heat.
Supercomputers are very fast frequently networked systems capable of pro- cessing extraordinarily
complex applications in healthcare research, robotics, national defense, weather forecasting, and
artificial intelligence.
Mainframe computers are large systems capable of processing massive vol- umes of data. Initially, these
computers were platforms for many legacy Hospital Information Systems (HIS). Originally, these
computers supported administrative and financial applications in large hospitals (e.g., patient
registration, scheduling, and claims processing), and, later, clinical applications. Mainframe computers
have been characterized as those with faster processing speeds than either micro- or minicomputers.
They also had much greater storage capacity, although more recent advances have diminished these
distinctions. Costs of mainframes vary widely depending on their processing speeds and storage
capacity. Historically, most providers used in-house mainframes for HIS and vender hosting of shared
sys- tems. These shared systems were configured with on-site minicomputers linked to mainframe
systems supporting financial and clinical application in large remote data centers that operated for
hundreds of provider organizations.
Minicomputers are characterized by larger storage capacity, faster processing speeds than
microcomputers, but with a smaller size than mainframe systems. Many departmental systems were
and still use minicomputers with fast processing speeds interfaced to clinical instrumentation and other
networked technology. Many mid-sized and small hospitals and large physician group practices use minicomputers for core applications. Some legacy systems are hybrid minicomputer and networked
solutions.
Microcomputers (i.e., personal computers [PCs]) are in widespread use in vir- tually every healthcare
organization and venue and are quite capable of functions previously relegated to legacy mainframe and
mini-systems. They are ubiquitous, versatile, flexible, and quite commonplace. They may be network
servers and clinical instrument interfaces and are common replacements for legacy video- terminals.
Based on Moore’s Law as explained in detail by Dale, Null, et al., addi- tional resources increased
processing power, faster speeds, and more storage options are growing at an extraordinary pace with
phenomenal numbers of tran- sistors packed on each microchip. Computing power, speed, and storage
capacity that was only available with early mainframe systems are now common in lap- tops, notebooks,
PDAs, and smart telephones. As IT research continues to develop faster, smaller, and less expensive
options, this pace of innovation accelerates with
multiprocessor devices, nonsilicon processors, photon and biologic-based com- puting with optical
circuits replacing electron-based computing.
Although it is sometimes difficult to distinguish between hardware issues and software issues, computer
design algorithms (mathematic, logical, or algebraic expressions) are software programs written in a
“high-level” fourth-generation or higher computer language. As such, a computer program is a
structured set of sequenced algorithms, in which a “nested” algorithm produces an algorithm and
another algorithm that runs that algorithm and so forth until instructions are exe- cuted as a series of 1’s
and 0’s at the machine level. An algorithm at the machine level operates as an electronic device.
Because of this hierarchal structure, algo- rithms are best understood as “nested” instructions to
perform ever more funda- mental tasks that might otherwise have been manually executed by a user.
Thus computers are implementations of algorithms that execute other algo- rithms, and this chain of
algorithms leads to this principle of equivalence:
Anything that can be done with software can also be done with hardware, and anything that can be
done with hardware can also be done with software. Even more fundamentally, anything done by a
computer’s hardware and software can also be done by a human or a group of humans if time,
precision, and accuracy are not limiting factors.
Although this principle does not explicitly address the speed with which these equivalent tasks are
completed, hardware implementations are almost always faster and software executions are usually
faster than manual processing.
More fundamentally, data input must be represented in language that a computer can understand and
process. Each keystroke must be translated into machine lan- guage before processing, with information
processed as a binary digit, or a bit. When an electrical charge is present, a bit has a value of 1. When no
electrical impulse is present, a bit is 0. Bits are grouped to make a byte, representing a sin- gle machine
language character.
Data are collected, processed, and stored to meet process, function, application, or program
requirements. Data processed by computer systems are controlled by programs directing this digital
transformation into useful information for analytic action. Information results from data or facts that are
collected and processed. Data are input to a computer’s system memory component via input devices
and
Information systems are groups of interoperable components, appli- cations, and associated
technologies working together to completprocessed in a system’s central processing unit. Information
output is used, pro- duced, or processed further through other computer programs or applications after
data processing completion.
A data processing cycle results as information is stored in secondary storage for immediate action or for
later use. Output may become input for further processing operations. Identical data may be collected
and processed by various other pro- grams, applications, or computers to produce different information
depending on organizational requirements.
Facts, or data, are organized in a variety of ways to provide useful information. For example, when a
patient receives a procedure as ordered by a surgeon, supplies are consumed. Procedures and supply
items are charged and will appear as a charge with a clinical procedure terminology (CPT) code on this
patient’s final bill. Based on this charging process for supplies consumed, a purchase order is generated
via a materials management application. This is then transformed into a SKU-coded purchase order that
is transmitted to a preferred supply vendor’s system, which replenishes inventory for this item. This
process is a materials management life cycle.
This example details how identical data or facts reflect different informational needs within and
between provider departments and external organizations. Each digital module involved in processing
these discrete data is defined as functions performed within a data processing cycle (order entry,
surgical scheduling, billing, accounts receivable, and materials management).
A basic understanding of fundamental computer hardware components and key functions of any
computing device (mainframes, minicomputers, desktops, laptops, handheld PDAs, and smart cellular
telephones) is required working knowledge for all IT professionals. Any computing device consists of
three essential components.
The first component is a processor, which interprets and executes programs by supervising and
managing all operations, communicating through circuitry that oversees execution of commands to
begin, perform arithmetic or logical opera- tions, and terminate each operation or task upon execution:
•
Arithmetic operations are mathematical calculations of addition, subtrac- tion, multiplication,
and division.
•
Logical operations involve three essential comparisons based on decisions when values are
greater than, less than, or equal to each other.
The second component is memory, which stores programs and data:
•
Primary memory stores data and program instructions, without performing any logical
operations. Different terms may identify primary memory, such as primary storage, memory, main, or
internal memory.
IT:
COMPONENTS, FUNCTIONS,
AND FEA TURES
••
•
••
Memory can be random access memory (RAM) or read-only memory (ROM). RAM is volatile memory,
and stored data are lost if electrical power is interrupted. Cache memory is buffer storage of frequently
used data blocks that are clos- est to a Central Processing Unit (CPU).
Cache is volatile memory and may be virtual memory. Virtual memory is a physical extension of main
memory that is designated on another storage device to “enlarge” cache beyond physical limitations to
concurrently run more processes.
Shared memory residing with multiple processors may be accessed as a sin- gle entity, enabling
inexpensive “supercomputing capability.” ROM is not volatile and includes basic instructions to start or
“boot” com- puter operations. ROM can be firmware with hardware and software char- acteristics.
Memory is measured in bytes, kilobytes, megabytes, or gigabytes. Data and program instructions are
stored in primary storage and moved as necessary for processing. Computer memory is hierarchically
organized (i.e., larger, slower, and less costly as well as smaller, faster, but more costly memory). Most
computers include some combination of cache, main memory, and secondary memory (e.g., disk and/or
tape drives, optical disks, digital video disks), as successive hierarchi- cal layers.
The third component is a mechanism that transfers data to and from an exter- nal environment:
•
Hardware is any physical component used to process data (e.g., CPU, mem- ory, keyboard,
mouse, display, buses, printers, displays). Hardware includes devices for input, output, processing, and
storage of data. Examples of peripheral devices are monitors, disk drives, compact disks or digital video
recorders, keyboards, scanners, and printers.
•
Software is a set of programmed instructions directing system hardware com- ponents to
complete data processing tasks. •
Systems software controls system operations by directing,
commanding,
scheduling, and confirming data processing functions. Systems software includes operating systems and
specialized programs such as utilities, programs, or translators.
•
Applications software includes programs performing specific data pro- cessing functions, usually
for end-user functions.
These three key elements are required in all systems to perform data process- ing and transform data
into useful information.
Fundamental System Terminology and Operations
51 More advanced memory and processor
relationships include
•
Data flow computers that enable data to drive computation • Neural networks that learn to
execute computations via arrays of processing
elements which continue until computations are completed • Biologic, optical, and quantum
computers developing computing concepts
(see Chapter 15)
A processing cycle includes all data collection, through input, and processing task to various types of
output, as well as all associated manual and computerized functions and tasks.
A clock is a vital system component that keeps all system processing synchro- nized. Clocks
simultaneously send electrical pulses to key components, ensuring that data will be where expected and
instructions executed as expected. These pul- sations are frequencies emitted each second measured in
cycles per second, or hertz. When system clocks generate millions of pulses per second, they operate in
the gigahertz range. As of this writing, processors in typical desktop and laptop com- puters are
operating in the gigahertz range, generating billions of pulses per second.
Frequency ratings are benchmarks of microprocessor speed and are determi- nants of overall system
performance. Similarly, networking speed reflects time devoted to file-sharing or communicating with
peripheral devices (e.g., scanners or printers).
A simple analogy is a class with each student functioning as a computer with these three computer
components: a brain as a processor, class notes as memory, and pen and paper representing transfer
mechanisms. Instructor-to-student, student- to-instructor, and student-to-student interactions during a
class discussion are anal- ogous to a transaction transported on a network. When an instructor poses a
question, this question is recognized and documented (stored) by each student and discussed among
the class. When concluded with a correct response to the instruc- tor’s question, this process is
representative of a “data processing” cycle.
FUNDAMENTAL SYSTEM TERMINOLOGY AND OPERATIONS
A basic understanding of measurement terminology, although specific to computer applications, is
fundamental knowledge required of all transformationalists because automated applications function in
an analytic context of process efficiency, process capability, and workflow. The National Institute of
Standards and Tech- nology approved standard names and symbols to differentiate binary and decimal
prefixes early in the IT evolution. Table 3.1 highlights key prefixes and symbols.
For example, a small word processing file may be 50 Kb in size, a graphic image, a 4-Mb object, a flash or
hard disk drive 2 Gb in size, or the distance
between transistors on a microchip 48 nm apart. Processing speed or microchip cycle times are
expressed in terms of fractions of a second (e.g., thousandths, mil- lionths, billionths, or trillionths as
fractional prefixes on the right side of Table 3.1). Exponents are reciprocals of these prefixes depicted on
the left side of Table 3.1. As a simple example, if an operation requires a microsecond to complete, then
a million operations occur within a second.
An understanding of these fundamental computer operations is essential. IT professionals must
understand and demonstrate multidisciplinary competency to effectively facilitate team leadership and
be able to communicate these concepts to application users and senior executives.
Effective digital transformation of healthcare delivery processes depends on IT professionals to lead
team discussion and collaborate about system capabilities, strengths, and limitations. Mastering basic
computer applications and dependent issues require given appropriate attention, deliberation, and
decision-making for results to produce effective process redesign and achieve project objectives.
Healthcare IT professionals must meet these challenges of digital transforma- tion of care delivery with
demonstrated computer literacy and competency. Vir- tually all healthcare diagnostic, therapeutic, or
treatment protocols, applications, policies, and procedures are computerized and increasingly
dependent on infor- mation systems at an ever accelerating pace. All IT professionals must be computer
literate, be able to define what a computer is, know how to operate a computer to perform professional
and administrative tasks, and appreciate the social and ethi- cal impacts that IT has on individuals and
society.
In summary, a data processing cycle refers to each stage of data transformation of information. The
input stage, often referred to as data entry, occurs when data are first introduced into a computer
system. Processing refers to data manipulation that occurs in the CPU to produce output as data or,
more importantly, as information. Storage is retention of data or information in either a volatile or
permanent form.
Processing may occur in different time frames. Off-line processing occurs at intervals, and information is
not always current or readily available in real time. Online systems are real-time systems in which
processing occurs immediately so that data input are current and virtually occur simultaneously.
Data can be processed in many ways to produce desired output. Information system operations produce
information by processing data, including calculations, input, output, query, classifications, sorting,
updating, summarizing, storage, and retrieval. Online processing is real-time interactive or transaction
data processing without any delay, as processing occurs immediately upon the availability of input and
output transactions. Processing of input transactions and processes typically occur in transaction input
sequence order. Frequent feedback to users with “front- end edits” produces more accurate data input
because input is validated online in real-time communication with a CPU.
Online processing occurs through a direct connection between terminals and CPU. Off-line processing
systems do not provide direct communication with a CPU. Online systems are connected to a CPU via
“online” terminals for data input or for receiving output. These online terminals may be a personal
computer (PC), a video display terminal, a cathode ray tube (CRT), a “client” or “dumb terminal,” or
another such device.
Batch processing is an off-line application or system that stores data for pro- cessing at periodic
intervals. For example, a payroll system using batch process- ing may collect information available on
hours worked for all employees in advance of processing paychecks and associated reports at the end of
a pay period. Batch processing is useful for certain applications, such as payroll or billing sys- tems, when
information does not need to be available on a real-time basis. Batch processing is relatively inexpensive
and an efficient method for processing large data sets.
Data processing, aggregation, and organization reflect data transformation into meaningful information
using basic data processing operations. Brief descriptions of these fundamental data processing
operations follow:
•
Calculation transforms data into information through mathematical calcu- lations. For example,
monthly patient billing statement totals can be added together to provide an accurate annual
assessment of patient revenues.
•
Input requires data entry via a variety of hardware devices into a system’s CPU arithmetic–logic
unit where processing operations are executed.
IT:
COMPONENTS, FUNCTIONS,
AND FEA TURES
•
••
•••
••
•
Regardless of the input device used, data input must occur before other data processing operations can
occur. Many input operations are by data entry. Output generates information by transforming
organized data into a useable form (a letter or a monthly revenue report). Output may be digital, hard
copy, or both that can be printed or written to a compact disk or other more per- manent media. A
digital or “soft copy” is information displayed on a com- puter monitor and is not a permanent copy
because it is only temporally retained in volatile memory. Output operations are performed by a variety
of hardware devices.
Query or inquiry is a request for information performed by keying a com- mand or a key word associated
with information to be accessed. Classification is grouping data as being meaningful and useful.
Classifying simply organizes data into categories (e.g., alphabetical or male and female designations).
Numerical codes often represent classifications (e.g., a numer- ical classification system might be CPT
codes, which are numerical codes representing medical procedures).
Sorting is a specialized classification operation that arranges data in a sequence or specified order.
Common sorts include alphabetical or numeri- cal listings, first to last, high to low, small to large, and so
on. Update is a change in stored data to reflect more current conditions (updating, editing, or revising
involves adding, deleting, or modifying existing informa- tion). For example, making an update to a
patient’s previous address in a HIS. Indexing creates a file-based key field while leaving original data
intact by creating a pointer field within an indexed file to establish, maintain record sequence, and
reestablish an original position of a record in an original file. Fields include both a record number and a
pointer in each original file. An end of file indicator means that there are no more records. For example,
to create an indexed file patient zip codes are in ascending order fields to be sorted, and after sorting
record numbers and pointer fields change position. An original file is maintained in proper sequence
through the pointer field. Summary reporting includes counting and totaling fields within files to provide counts, subtotals, and totals.
Exception or ad-hoc reporting selects records within a file that reflect a unique characteristic to be
studied. Patient care information systems are designed so that authorized users may specify and
execute these reports as needed with a “report writer.”
Detail reporting includes listings of every record within a file, including cat- egory subtotals and totals.
System documentation typically outlines a report’s content and format (line spacing, tab settings, and
any unusual fea- tures). Fields to be calculated are usually specified along with calculation instructions.
Database constructs or database architectures reflect structure, format, and
organization of data storage within each database of an HIS application:
•
Relational databases use existing relationships among and between individ- ual records by
identifying key record field(s).
•
Hierarchical relationships are established among records as a “tree struc- ture” (a top level
record is identified as a “parent” record and is related to lower level “child” records).
•
Networked databases use network structures to relate records when database fields reside
physically or virtually on different processors within an inte- grated delivery system (IDS) network.
As IT evolved from relatively simple general-purpose calculating devices into increasingly sophisticated
applications, a standardized hierarchy of different lev- els of software and functions became necessary
to understand and explain system architecture at different levels. Each level in this hierarchy defines
“semantic gaps” between high-level programming applications and physical attributes of computer
components (microchip circuits, buses, gates, and wires). These levels are num- bered from 0 to 6,
where 0 represents machine-level operations and executable programs with which IT professionals must
be familiar. Table 3.2 summarizes these seven layers.
HEPIS (HEALTHCARE E-PATIENT INFORMATION SYSTEM)
The remaining sections of this chapter describe system modules, applications, and functions in a
narrative summary. Some combination or frequently updated ver- sions of these applications operate in
healthcare provider organizations. Table 3.3 highlights these capabilities at an application level, and
Figure 3.4 depicts com- mon interactions and interfaces of these modules as a high-level schematic. In
the remaining chapter narrative more detailed descriptions of each application are given, including
typical functions and features. In the context of these descrip- tions, as well as elsewhere throughout
this book, a comprehensive legacy HIS alludes to various characteristics, functions, and features of the
Healthcare E-Patient Information System (HEPIS).
A digitally transformed information system implies this minimal set of IT capabilities, functions, and
features. An IT configuration of these or compara- ble functions and features is necessary to provide
effective, patient-centric, and state-of-the-art healthcare information management technology to a
digital transformation of any provider organization. Capabilities, applications func- tions, and features in
this comprehensive but not all-inclusive example describe a typical provider information system
currently operational in hundreds of
provider organizations throughout the nation. This level of system functions and features is required to
support effective digital transformation of care deliv- ery processes throughout a typical healthcare
provider organization. Specific functions and features have been available as turnkey solutions from
numerous commercial vendors for many years. These specific applications, functions, and features are
presented to explain system capabilities in sufficient descriptive detail of operational or proposed
functions and features (required functional
content for a request for proposal [RFP] to replace an outdated legacy system). Although obviously very
rich in technical capabilities, this system specifica- tion, including these applications as described, is
available to any qualified healthcare provider organization at nominal cost.
HEPIS supports a true longitudinal healthcare record, including data from both in-network and out-of
network sources. HEPIS supports research and population analyses and facilitates patient access to data
and sharing of information to improve data quality, consistency, and integrity in a securely networked
environment throughout multiple provider entities within an IDS.
HEPIS serves as an operational clinical repository (e.g., a cohesive portfolio of clinical information),
incorporating both in-network and out-of network provider data that resides on one or more
independent platforms to be used by clinicians and
HEPIS (Healthcare E-Patient Information System)
59
other personnel to facilitate longitudinal patient-centric care. Data in HEPIS is organized in a format
supporting timely and effective care delivery in virtually any venue, regardless of patient or provider
location or a patient’s clinical infor- mation. These HEPIS applications, modules, functions, and features
transparently support patient-centric care delivery processes using industry-standard interfaces within
an interoperable integrated network environment.
HEPIS’s “e-Health” application functions support a variety of customer- friendly patient-centric
capabilities, such as prescription refills, appointment sched- uling, and online forms completion, as well
as patient and provider access to health records, online health assessment tools, and high-quality realtime health infor- mation. These capabilities are representative of digitally transformed care delivery.
HEPIS provides significant additional benefits beyond seamless automated sup- port of care delivery. It
also provides information to support demographic research and population analyses, facilitating patient
access to data and information shar- ing. HEPIS enhances IDS capabilities by improving data quality and
security and minimizing mundane administrative costs.
Admission, Discharge, Transfer, and Registration
This comprehensive array of functions and features is dedicated to administrative functions for patient
admission, discharge, transfer, and registration. Admission, discharge, transfer (ADT) modules provide
functions to be used throughout each patient’s acute inpatient, long-term care, and/or outpatient stay.
Registration, enroll- ment, and eligibility applications are designed as a single IDS-wide data system and
demographic database that supports comprehensive registration and eligibil- ity verification, thereby
ensuring information consistently and accessibly in all other applications in the HEPIS network.
Automated Medical Information Exchange
This module facilitates electronic interchange of patient information among and between providers and
facilities and retains timely, accurate, and comprehensive audit trails of information exchange. This
module includes several key compo- nents, such as provider- and facility-specific administrative options,
applications, and components that enable efficient daily reporting and maintenance.
Authorization/Subscription
HEPIS utilities support identification and tracking of transactions by authorized personnel to perform
various actions on clinical documents (e.g., signing, cosign- ing, and enhancing text and image definition
requirements).
IT:
COMPONENTS, FUNCTIONS,
AND FEA TURES
Care Management
This HEPIS application offers care providers inquiry capability to view pertinent information about
multiple patients on a single screen. With this function, users can view multiple patients requiring
attention:
•
Clinician dashboard is a table of patients for whom clinicians have unac- knowledged results or
event notifications (admissions, discharges, or unscheduled clinic visits, unsigned documents, or
uncompleted tasks).
•
Nurse dashboard provides a similar list of patients for whom nurses have unacknowledged
results, unviewed events, uncompleted tasks or text orders, unverified orders, or recent vital signs.
•
Query tools enable authorized users to create reports based on most current patient data via
defined user-created custom reports.
•
Sign list enables authorized users to approve multiple actions for multiple patients. For example,
clinicians can sign a discharge summary for one patient and notes for another. Predefined reports
include abnormal results, consultation status, incomplete orders, and recently scheduled activities.
Case Registries
Registries contain key demographic and clinical indices for patients identified by provider with specified
diseases and/or disorders based on primary, secondary, or tertiary International Classification of
Disease, Version 9 (ICD-9) and Version 10 (ICD-10) coded diseases and disorders or CPT coded
procedures. These registries extract from HEPIS databases (e.g., pharmacy and laboratory) and provide
key clinical information. Data are used to track and optimize clinical care of patients receiving care by
various providers in multiple venues.
Clinical Monitoring System
This module enables users to capture and monitor patient data as required by var- ious quality
management initiatives. Typically, patient data are automatically cap- tured from an existing database
and edited to track specific events. Statistical data are maintained for each patient scanned, determining
the number of patients meet- ing monitoring criteria, by selected data trending for user-specified time
frames (hourly, daily, weekly, monthly).
Clinical Procedures
This application uses HL7 messaging to pass final patient results between com- mercial clinical
information systems and displayed patients’ test results. Discrete
HEPIS (Healthcare E-Patient Information System)
61
report data are stored in HEPIS databases to share data and images with other HEPIS modules
(consult/request tracking, text integration utilities, patient care encounters, and HEPIS imaging
packages). Clinicians may document findings and complete final procedure reports generated by medical
devices. These key func- tions enable clinicians to enter, review, and continuously update order-related
information for any patient. Patient allergies or adverse reactions to medications are recorded, as well
as requests for and tracking consultations status, progress notes, diagnoses, treatments, and discharge
summaries for each patient.
Integration with other HEPIS applications facilitates accurate and timely record- keeping and compliance
with clinical guidelines and patient medical record require- ments. Authorized personnel maintain
comprehensive patient records enabling clinicians, managers, and quality assurance staff to review and
analyze data to directly support clinical decision-making. Adverse reaction tracking is a common and
consistent data structure to detect and monitor adverse reaction data. Options expedite data entry and
real-time validation and references used by both HEPIS and external systems to report adverse drug
reaction data to the U.S. Food and Drug Administration (FDA).
Clinical Reminders
Clinicians can provide and receive appropriate and timely patient treatment reminders to assist clinical
decision-making and educate providers about appro- priate care options. Electronic clinical reminders
improve documentation and follow-up, enabling authorized providers to view test or evaluation results
and track and document care as ordered and delivered. This capability reduces duplicate doc- umenting
activities, assists in targeting patients with particular diagnoses and pro- cedures using site-defined
criteria, and assists in compliance with clinical care guidelines. Consultation and request tracking
enables clinicians to efficiently order and track real-time status of consultations and procedures from
other providers or services within each clinical setting in any in-network venue.
Health Summary
Health summaries are clinically oriented structured reports that extract and abstract data from HEPIS
databases and use a standard format to display information and report information. Health summaries
are printed or displayed for individual patients or groups of patients. Displayed data include userdefined health-related information such as demographic data, allergies, active medical problems, and
test results. Data and summary information are integrated with clinical HEPIS appli- cations and
modules.
Health summary functions export components to authorized staff who can remotely view patient data.
Clinical reminders are integrated with health summary
IT:
COMPONENTS, FUNCTIONS,
AND FEA TURES
functions to inform providers with timely information about each patient’s health maintenance
schedules. Clinical providers collaborate with authorized technical staff coordinators to configure
customized schedules based on local and national clinical guidelines. Health summary components such
as past progress notes are available to display new interdisciplinary progress notes and relevant entries.
Home Care
This is an optional module designed with secure remote entry and verification of patient-related data
from home care providers delivering care in patient homes or other in-network facilities. This module’s
design uses a relational database structure, incorporating wireless security to ensure data integrity and
access accountability.
Imaging, Radiology, and Nuclear Medicine Applications
These function-rich applications provide seamless real-time imaging by automat- ing diagnostic and
therapeutic services. Examples of key features include patient registration for examinations, order entry
and service requests, image processing and interpretation, automatic request tracking, result recording
and reporting, and online report verification, display, and printing of recurring ad-hoc statistics and
management reports.
•
Remote order entry functions support chain order fulfillment production. •
HEPIS
applications and associated databases integrate clinical, administra- tive, and business processes (e.g.,
contracting and acquisition management, order fulfillment, distribution management, finance, and
equipment life cycle
management). •
Extensive order tracking, serialized device registration, and service history
are available. • Scheduling automates appointment and scheduling processes, including procedure setup and maintenance, enrollment, scheduling, and resources. Mes- sages are displayed as an
appointment is scheduled (e.g., notification that an appointment is an overbooking, conflicts reporting,
etc.).
•
Imaging applications enhance and expedite medical decision-making by delivering integrated
multimedia patient information to each clinician’s desk- top. Imaging transmits high-quality image data
for many specialties (cardi- ology, pulmonary, and gastrointestinal medicine, pathology, radiology,
hematology, and nuclear medicine).
Imaging processes textual reports as scanned documents and electrocardio- grams. These HEPIS imaging
capabilities are integrated and interoperable with other applications, providing a comprehensive
electronic patient record
HEPIS (Healthcare E-Patient Information System)
63
that is accessible to authorized clinicians any time and anywhere. Imaging improves patient care and
enhances both quality and timeliness of clinician communication, daily conferences, educational
seminars, and patient rounds.
•
Diagnostic imaging displays are used for enhanced digital film-less inter- pretation of radiology
studies and for radiology workflow management. Imaging includes a variety of core components.
•
Core infrastructure includes components used to capture, store, and display images captured
from video cameras, digital cameras, document and color scanners, x-ray scanners, sonograms, and files
created by commercial cap- ture software. Images are directly acquired from Digital Imaging and Communications in Medicine (DICOM)–compliant devices such as computed tomography, magnetic
resonance imaging, positron emission tomography, and single photon emission computed tomography
and other additional dig- ital radiology equipment. Captured images are accessible throughout all venues with a compliant graphic user interface on clinician desktops. Associated images are automatically
accessed when viewing a procedure or examination report or progress note.
•
DICOM text gateways provide patient and order information to medical devices (e.g., computed
tomographs and digital radiography systems to expedite each examination performed). Data provided
by these DICOM text gateways comply with all industry-wide DICOM modality work-list stan- dards.
These DICOM text gateways enable patient and communication of order information to any commercial
picture archiving and communications system (PACS) within an IDS network as well as to out-of-network
clini- cians. DICOM image gateways securely receive images from external PACS systems or acquisition
devices. Image gateways transfer images to DICOM-compliant storage devices for display, printing, or
teleradiology purposes:
•
Enabling online availability of all information in an electronic patient record, including
immediate availability of critical documents to author- ized personnel (e.g., advance directives and
informed consent forms) at any time in addition to handwritten papers, drawings, signed documents,
and medical correspondence
•
Linking to and scanning of paper-based patient information to electronic patient records such
that all patient information is expeditiously available and easily retrievable from a single source
•
Potential savings due to reduced medical records staffing requirements and filing costs as well
as minimizing lost or misfiled medical chart infor- mation
•
Sustainable benefits include decreased retrieval and delivery functions, reduced volume of
paper records, and rapid information accessibility for clinicians
IT:
COMPONENTS, FUNCTIONS,
AND FEA TURES
Background processing manages image storage on network devices. An imag- ing database links images
and electronic patient records, including magnetic stor- age via redundant array of independent disks
and optical jukebox storage for archives.
Intake and Output
This application design stores patient intake and output information for each hos- pital stay or
outpatient visit. This application is not service specific and is inte- grated with appropriate HEPIS
modules.
Incident Reporting
This module supports organizational policies by compiling data on patient inci- dents and organizing
data into defined categories for reporting and tracking at each provider facility level and for
transmission to quality assurance databases for cor- porate office review and tracking.
Laboratory Applications
HEPIS’ clinical laboratory applications provide laboratory data and formatted information to clinicians as
required to support specimen collection; analyze, report, interpret, and evaluate clinical implications;
and provide high-quality, effi- cient, cost-effective timely patient care. These modules enable clinical,
technical, and administrative personnel to effectively manage increasing automation and robotics driven
by heavier workload to be completed faster with fewer resources while continuing to enhance
laboratory processes and reporting. Modules support clinical laboratory, chemistry, hematology,
immunology, and microbiology. Instru- ment interfaces are industry standard for virtually all routine
high-volume proce- dures. Web-enabled links are available via secure networks to expedite logistics to
and reporting from reference laboratories at any location.
HEPIS’ Surgical/Anatomic Pathology Laboratory module provides automated record-keeping and
reporting for surgical pathology and anatomic pathology, his- tology, cytology, electron microscopy, and
autopsy. These functions enable staff to improve quality management, increase productivity, use and
provide compre- hensive research and reporting capabilities, and facilitate workload management and
statistical reporting.
Blood banking functions use data tied primarily to a donor, a patient, or a unit of blood/blood
component. Information about a blood donation or a donation attempt includes demographics and
potential risk history of each blood donor. Sim- ilarly, accurate and timely information about each unit of
blood or blood compoHEPIS (Healthcare E-Patient Information System)
65
nent in inventory requires effective real-time unit identification and location. Transfusion and testing
information involves patient sampling and associated blood or blood component administration.
Nutrition and Food Service
Nutrition and food service applications integrate automation of clinical nutrition, food management,
and management reporting functions. HEPIS supports activi- ties such as nutrition screening, nutrition
assessment, diet order entry, tube feed- ing and supplemental feeding orders, patient preferences, diet
pattern calculations, nutrient analysis of meals, consultant reporting, and quality of care tracking. Food
management functions include total automation of food acquisition, production activities, service,
distribution, inventory, cost management, recipe expansion, menu and recipe nutrient analysis, meal
and diet pattern development and imple- mentation, diet card and tray ticket printing, and service
quality tracking. Meals served, cost per meal, tube feeding cost, supplemental feeding cost, staffing,
encounter data, information query and summary capabilities, and annual manage- ment reporting are
also reported.
Patient Representative Applications
Patient-centric and patient representative design ensures that providers and facil- ities respond to
unique patient needs while tracking and trending patient com- pliments and complaints. This module
measures the facility’s types of complaints as they relate to effective customer service standards. These
functions collect and categorize complaints and compliments with components that identify opportunities to meet and exceed each patient’s expectations. Issue codes provide tracking capabilities for
complaints and trends of specific complaints. Customer service standards can be tailored to include
unique issue and reliability studies. These studies drive initiatives to improve patient perception tracking
and extract specific data for each provider, department, nursing unit, outpatient clinic, prod- uct line, or
service.
From registration or admission eligibility determination through discharge by online real-time data
transmission to optimize reimbursement for care with com- prehensive DRG and resource utilization
group (RUG) functions and features, these functions support collection of patient information, including
demograph- ics, employment, insurance, medical history and associated data. Other modules (e.g.,
laboratory, pharmacy, radiology, and nursing) use information gathered through various ADT options.
These features optimize operational efficiency while limiting unauthorized user access to specified
sensitive patient records. A patient sensitivity function permits a level of security to be assigned to
certain records
HEPIS (Healthcare E-Patient Information System)
67 out-of-network facilities, recharging records
to other borrowers, and flagging missing records.
Patient Identification Card
This application identifies patients with a color photograph and encrypted patient demographics to
initiate care and service processes in all in-network care venues within the IDS. A print option provides
labels with a patient’s key identifying information on bar-coded labels affixed to medical record forms.
This color pho- tograph uses nationally standardized patient image capture software (PICS) on a CCOWenabled workstation.
Population Health
This application is an optional module that computerizes, tracks, and aggregates data for all venues. An
automated assessment capability captures and reports var- ious longitudinal aspects of care provided
(e.g., care efficiency, care outcomes, and care quality). These reports
•
Enable clinicians to determine differences in disease frequency between individuals and among
selected cohorts in a population
•
Provide information to support clinical guideline development • Support health screening
guidelines for patients and populations •
Collect and report workload data, preventive screening,
outcome measurement, and provider profiling data
Problem List
Problem list capabilities include documentation and tracking of patient problems. These functions
provide clinicians with current and historical views of healthcare problems across clinical specialties for
each identified problem traceable through- out HEPIS for treatment, test results, and outcomes.
Resident Assessment/Minimum Data Set
This long-term care module provides a standardized, comprehensive, accurate, and reproducible patient
assessment tool as a fundamental baseline as a resident’s plan of care is developed. This module
streamlines data collection processes with acute care, rehabilitation, and skilled nursing facilities as
required by the CMS using a nationally standardized minimum data set (MDS). Versions are defined by
CMS to receive Medicare or Medicaid reimbursement using standard HL7 messaging.
IT:
COMPONENTS, FUNCTIONS,
AND FEA TURES
In addition, this module provides a structure that meets the JCAHO long-term care accreditation
standards and provides opportunities and data comparison capabili- ties for resident outcomes. This
module incorporates the nationally standardized HL7 gateway interfacing to import and export data
triggered by specific MDS responses in multidisciplinary care plans. The MDS generates version-specific
RUGs and a variety of quality indicator reports.
Pharmacy Applications
Pharmacy functions are seamlessly integrated throughout the IDS and are com- plemented with
extensive comprehensive HEPIS applications. This module’s key functions are as follows:
•
Automatic pharmacy replenishment/nursing unit stock tracks drug distribu- tion and inventory
management throughout each facility, venue, and IDS services. Key capabilities include inventory
management for clinical care locations and drug crash carts.
•
Medication administration capabilit…