Professional Documents
Culture Documents
1 Department of Pediatrics, University of Toronto, Toronto, Ontario, Address for correspondence Shoo K. Lee, MBBS, FRCPC, PhD,
Canada Professor of Paediatrics, Department of Paediatrics, University of
2 Maternal-Infant Care Research Centre, Mount Sinai Hospital, Toronto, Toronto, 600 University Avenue, Rm 782, Toronto, ON, Canada M5G
Ontario, Canada 1X5 (e-mail: sklee@mtsinai.on.ca).
3 Department of Pediatrics, University of Alberta, Edmonton, Alberta
4 Department of Pediatrics, Queen’s University, Kingston, Ontario,
Canada
5 Department of Pediatrics, McGill University, Montreal, Quebec
6 Department of Pediatrics, Saint John Regional Hospital, Saint John,
New Brunswick, Canada
Abstract Objective Derive and validate a practical assessment of infant illness severity at
admission to neonatal intensive care units (NICUs).
Study Design Prospective study involving 17,075 infants admitted to 15 NICUs in
2006 to 2008. Logistic regression was used to derive a prediction model for mortality
comprising four empirically weighted items (temperature, blood pressure, respiratory
status, response to noxious stimuli). This Transport Risk Index of Physiologic Stability,
version II (TRIPS-II) was then validated for prediction of 7-day and total NICU mortality.
Results TRIPS-II discriminated 7-day (receiver operating curve [ROC] area, 0.90) and
total NICU mortality (ROC area, 0.87) from survival. Furthermore, there was a direct
association between changes in TRIPS-II at 12 and 24 hours and mortality. There was
Keywords good calibration across the full range of TRIPS-II scores and the gestational age at birth,
► neonates and addition of TRIPS-II improved performance of prediction models that use gesta-
► mortality tional age and baseline population risk variables.
► risk adjustment Conclusion TRIPS-II is a validated benchmarking tool for assessing infant illness
► outcomes severity at admission and for up to 24 hours after.
Neonatal illness severity scores such as the Clinical Risk Index riod may affect scores and limit their ability to truly measure
for Babies (CRIB)1 and the Score for Neonatal Acute Physiolo- illness severity at admission. In addition, CRIB is only validat-
gy, version II (SNAP-II)2 are important for risk adjustment and ed for infants with a birth weight of < 1,500 g, and neither
benchmarking outcomes in neonatal intensive care units SNAP-II nor CRIB is validated for sequential measurements to
(NICUs). However, both CRIB and SNAP-II have been criticized determine change in physiologic stability over time.
because they are derived from measurements taken over a The Transport Risk Index of Physiologic Stability (TRIPS)3 is
12-hour period. Consequently, interventions during this pe- a simple, practical, empirically weighted measure of neonatal
illness severity that comprises four physiologic measure- the growth charts established by Kramer et al.5 The SNAP-II
ments (temperature, respiratory status, blood pressure, re- neonatal illness severity score was calculated from six em-
sponse to noxious stimuli) that correlate with NICU mortality pirically weighted physiologic measurements made during
and can be measured within 1 minute without need for the first 12 hours of admission to the NICU, as previously
laboratory support. TRIPS has been validated for outborn described.2
infants of all birth weights requiring transport to a NICU,
and its predictive performance for mortality and major Derivation of TRIPS-II
morbidity in the NICU is equivalent to SNAP-II. However, Since the original TRIPS was designed to assess transport-
unlike SNAP-II, TRIPS is derived from measurements that are related effects on mortality, it was validated against 7-day
obtained within 1 minute, and therefore provides a true mortality in the NICU, as mortality after 7 days is more likely
reflection of stability at the exact time when it is measured. to reflect care in the NICU than care during transportation.3
The objective of this study was to determine whether However, with TRIPS-II we wished to determine whether a
TRIPS can be adapted to measure illness severity for both simple score based on physiology at this early stage could
inborn and outborn infants admitted to the NICU and wheth- accurately predict overall mortality in the NICU for both
er sequential measurement can give an accurate assessment inborn and outborn infants. We therefore validated TRIPS-II
of changes in physiologic stability over time. We compared against total NICU mortality, which is well defined and has
the predictive performance of the extended TRIPS version II been previously used to validate other measures of physio-
(TRIPS-II) with SNAP-II and assessed its ability to measure logic stability. All infants were followed until death or
neonatal illness severity sequentially over a 24-hour period. discharge.
Variable Definitions
Validation Procedures
Study variables were defined according to the Canadian
Neonatal Network Abstractor’s Manual. For GA at birth, the Predictive Performance and Criterion Validity
estimate was based on early prenatal ultrasound, time since We used data from the validation cohort and the area under
last menstruation, pediatric estimate, or obstetric estimate, in the receiver operating curve (ROC) technique of Hanley and
that order. An infant was defined as small for GA if the birth McNeil6 to assess performance of TRIPS-II measured at ad-
weight was less than the 3rd percentile for GA according to mission in predicting the criterion standard of NICU mortality
for infants of all GAs at birth, and for infants born at 32 weeks logic stability.2 We also compared TRIPS-II with GA alone, as
or less and more than 32 weeks GA separately. An ROC area of GA is predictive of mortality in the Canadian NICU popula-
1.0 indicates perfect discrimination whereas an area of 0.5 is tion.9 ROC areas derived from the same cases were compared
nondiscriminative. Previously published physiology-based using the method of Hanley and McNeil,6 with a p value of less
illness severity scores have included ROC areas between 0.7 than 0.05 indicating a significant difference between ROC
and 0.9.1,2,7 curves.
Abbreviation: TRIPS-II, Transport Risk Index of Physiologic Stability, version II. Note: Variables are significant if 95% confidence intervals do not cross 0.
significant differences between the two cohorts in terms of NICU mortality (e.g., the mortality log odds ratio for temper-
the physiologic measurements recorded. ature [<36.1°C or > 37.6°C versus 36.1°C to 37.6°C] was 0.52
[0.29 to 0.74]). In comparison to TRIPS,3 the component scores
TRIPS-II Variables and Scores Assigned in TRIPS-II assigned greater weight to respiratory status and
The TRIPS-II component items (temperature, respiratory reduced weight to blood pressure.
status, blood pressure, and response to noxious stimuli)
and the scores assigned to each item through derivation of Prediction of NICU Mortality
TRIPS-II using data collected at admission are shown The predictive performance of TRIPS-II for NICU mortality
in ►Table 2. The β-coefficient (95% confidence interval) of (►Table 3) was significantly (p < 0.05) better than GA at birth
these scores demonstrated that all four component variables but not significantly better than SNAP-II. Addition of TRIPS-II
were directly and significantly (p < 0.05) correlated with to the prediction model that included GA at birth plus other
Table 3 Outcome prediction (ROC areas) for TRIPS-II, SNAP-II, GA, GA þ variables, SNAP-II þ GA þ variables, and
TRIPS-II þ GA þ variables models
Abbreviations: GA, gestational age; NICU, neonatal intensive care unit; ROC, receiver operating curve; SNAP-II, Score for Neonatal Acute Physiology,
version II; TRIPS-II, Transport Risk Index of Physiologic Stability, version II. Notes: ROC areas for TRIPS-II are based on data measured on admission to the
NICU (within 15 min). Parentheses contain p values for Hosmer-Lemeshow goodness-of-fit statistic (p < 0.05 indicates poor goodness of fit).
a
Variables for mortality within 7 d of NICU admission and total NICU mortality: small for GA, 5-min Apgar < 7, cesarean section.
b
ROC areas significantly different (p < 0.05) from TRIPS-II for each outcome.
c
ROC areas significantly different (p < 0.05) between the GA þ variables model and the SNAP-II þGA þ variables model, for each outcome.
d
ROC areas significantly different (p < 0.05) between the GA þ variables model and the TRIPS-II þ GA þ variables model, for each outcome.
illness severity. Future studies should look into this further. It Acknowledgments
may also not be possible to fully extrapolate the TRIPS-II score This study was supported by Grant MOP-53115 from the
to populations in different countries/regions of the world due Canadian Institutes of Health Research. Additional funding
to differences in ethnic mix of the population and local was provided by the B.C. Children’s Hospital Foundation;
neonatal practices; the score should therefore be validated Calgary Regional Health Authority; Dalhousie University
in other countries. TRIPS-II is not intended for use in ethical Neonatal-Perinatal Medicine Research Fund; Division of
decision making and should not be used as a substitute for Neonatology, Children’s Hospital of Eastern Ontario; Child
clinical judgment. Health Program, Health Care Corporation of Saint John’s;
The Neonatology Program, Hospital for Sick Children;
Lawson Research Institute; Mount Sinai Hospital; Ontario
Conclusions
Ministry of Health and Long-Term Care; Saint Boniface
We have derived and validated a score of infant physiologic Hospital, Saint Joseph’s Health Centre; Sunnybrook Health
status that is empirically weighted, practical and easy to use, Sciences Centre; University of Saskatchewan Neonatal
and highly suitable for assessment of care following admission Research Fund; University of Western Ontario; Victoria
to NICUs. TRIPS-II is applicable to infants of all GAs and is a General Hospital; Winnipeg Health Sciences Centre. The
potentially useful tool to enhance performance review and Canadian Neonatal Network Coordinating Centre (Mater-
quality improvement initiatives. Validation in other populations nal-Infant Care Research Centre) is supported by the
should be undertaken to verify utility and generalizability. Ministry of Health and Long-term Care, Ontario, Canada.
Hospital, Regina, SK); Lajos Kovacs (Jewish General Hospi- index for babies) score: a tool for assessing initial neonatal risk and
comparing performance of neonatal intensive care units. Lancet
tal, Montreal, QC); Jean Lachapelle (Hôpital Maisonneuve-
1993;342:193–198
Rosemont, Montréal, QC); David S. C. Lee (St. Joseph’s Health 2 Richardson DK, Corcoran JD, Escobar GJ, Lee SK. SNAP-II and
Centre; London, ON); Douglas D. McMillan (IWK Health SNAPPE-II: Simplified newborn illness severity and mortality
Centre, Halifax, NS); Cecil Ojah (Saint John Regional Hospi- risk scores. J Pediatr 2001;138:92–100
tal, Saint John, NB); Abraham Peliowski (Royal Alexandra 3 Lee SK, Zupancic JA, Pendray M, et al. Transport risk index of
physiologic stability: a practical system for assessing infant trans-
Hospital, Edmonton, AB); Bruno Piedboeuf (Centre Hospi-
port care. J Pediatr 2001;139:220–226
talier Universitaire de Quebec, Sainte Foy, QC); Patricia Riley
4 Pollack MM, Ruttimann UE, Getson PR. Pediatric risk of mortality
(Montreal Children’s Hospital, Montreal, QC); Daniel (PRISM) score. Crit Care Med 1988;16:1110–1116
Faucher (Royal Victoria Hospital, Montreal, QC); Nicole 5 Kramer MS, Platt RW, Wen SW, et al. A new and improved
Rouvinez-Bouali (Children’s Hospital of Eastern Ontario, population-based Canadian reference for birth weight for gesta-
Ottawa, ON); Koravangattu Sankaran (Royal University tional age. Pediatrics 2001;108:E35
6 Hanley JA, McNeil BJ. A method of comparing the areas under
Hospital, Saskatoon, SK); Mary Seshia (Health Sciences
receiver operating characteristic curves derived from the same
Centre, Winnipeg, MB); Sandesh Shivananda (Hamilton cases. Radiology 1983;148:839–843
Health Sciences Centre, Hamilton, ON); Todd Sorokan (Royal 7 Chien LY, Whyte R, Thiessen P, et al. Snap-II predicts severe
Columbian Hospital, New Westminster, BC); Anne Synnes intraventricular hemorrhage and chronic lung disease in the
(Children’s and Women’s Health Centre of British Columbia, neonatal intensive care unit. J Perinatol 2002;22:26–30
Vancouver, BC); Herve Walti (Centre Hospitalier Universi- 8 Hosmer D, Lemeshow S. Assessing the fit of the model. In: Hosmer
D, Lemeshow S, eds. Applied Logistic Regression. New York, NY:
taire de Sherbrooke, Fleurimont, QC)
John Wiley and Sons; 1989:135–175
Canadian Neonatal Network Coordinating Center 9 Sankaran K, Chien LY, Walker R, Seshia M, Ohlsson A; Canadian
(Toronto, ON): Priscilla Chan, M.Sc.; Sarah A. De La Rue, Neonatal Network. Variations in mortality rates among Canadian
Ph.D.; Ruth Warre, Ph.D.; Xiang Y. Ye, M.Sc.; Woojin Yoon, neonatal intensive care units. CMAJ 2002;166:173–178
M.Sc.